Jump to ContentJump to Main Navigation
Essays on Reference, Language, and Mind$

Keith Donnellan, Joseph Almog, and Paolo Leonardi

Print publication date: 2012

Print ISBN-13: 9780199857999

Published to Oxford Scholarship Online: May 2015

DOI: 10.1093/acprof:osobl/9780199857999.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2019. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in OSO for personal use.  Subscriber: null; date: 17 September 2019

Speaker Reference, Descriptions, and Anaphora

Speaker Reference, Descriptions, and Anaphora

(p.115) [5] Speaker Reference, Descriptions, and Anaphora
Essays on Reference, Language, and Mind

Keith Donnellan

Joseph Almog

Paolo Leonardi

Oxford University Press

Abstract and Keywords

This chapter discusses the connection between “speaker reference” and “semantic reference”. The distinction between the two uses of definite descriptions have importance beyond a contribution to the understanding of how definite descriptions work and the problems it seems to pose for the theories of Bertrand Russell and Peter Strawson on that matter. The chapter assesses whether there are two uses of definite descriptions in the sense of two semantic functions, in one of which the description conveys speaker reference and in the other not; or if definite descriptions are used in two kinds of circumstances, in one of which there is an accompanying phenomenon of speaker reference though it has no effect on the semantic reference of the description.

Keywords:   speaker reference, semantic reference, definite descriptions, Bertrand Russell, Peter Strawson

People refer and expressions refer. Let us call these phenomena “speaker reference” and “semantic reference” respectively.1 What connection exists between the two? The question has importance for how we theorize about various referring expressions, demonstratives, for example. And it is fairly crucial for what significance should be attached to a distinction I proposed some time back between what I called two uses of definite descriptions, the referential and the attributive.2 At the time I thought of the distinction as having importance beyond a contribution to the understanding of how definite descriptions work and the problems it seemed to pose for the theories of Russell and Strawson on that matter, precisely because it apparently showed the necessity of bringing in speaker reference for an explanation of the semantic reference of certain expressions.

While the referential/attributive distinction proves, I believe, to appeal to our intuitions, vagueness about the role of speaker reference threatens its significance. Are there two uses of definite descriptions in the sense of two semantic functions in one of which the description (p.116) conveys speaker reference and in the other not? Or is it rather that definite descriptions are used in two kinds of circumstances, in one of which there is an accompanying phenomenon of speaker reference though it has no effect on the semantic reference of the description? If the latter, it is not clear what importance we should attach to the distinction in the philosophy of language. It would not, for example, seem to have a bearing on the correctness or incorrectness of a semantic analysis of sentences containing definite descriptions such as Russell gives us.

I want to investigate this matter further in this paper and I offer certain arguments derived from a consideration of the phenomenon of anaphora to show that speaker reference cannot be divorced from semantic reference.


It would no doubt be enlightening were I to begin by saying what speaker reference consists in. We naturally gravitate to such expressions as “what the speaker had in mind” that echo locutions from ordinary speech—“Whom do you have in mind?” But what is it to have someone or something in mind? Is it, for example, to possess a body of descriptions that identify a particular person or thing? I will not attempt, however, any general answer to this important problem. Instead, I will rely on what I think is the indisputable fact that there is something corresponding to these locutions and go on to ask where it fits into the theory of reference. I believe, in fact, that even those who find little use for a notion of speaker reference acknowledge that the phenomenon occurs. Here, for example, is what Peter Geach says about it in Reference and Generality:

Personal reference—i.e. reference corresponding to the verb “refer” as predicated of persons rather than of expressions (p.117) [I take it this is what I am calling “speaker reference”]—is of negligible importance for logic; and I mention it only to get it out of the way. Let me take an example: Smith says indignantly to his wife, “The fat old humbug we saw yesterday has just been made a full professor!” His wife may know whom he refers to, and will consider herself misinformed if and only if that person has not been made a full professor. But the actual expression “the fat old humbug we saw yesterday” will refer to somebody only if Mr. and Mrs. Smith did meet someone rightly describable as a fat old humbug on the day before Smith's indignant remark; if this is not so, then Smith's actual words will not have conveyed true information, even if what Mrs. Smith gathered from them was true.3

Whatever we finally decide about the nature of the distinction, I believe we have enough raw intuitions about it often to be able to tell, given a sufficiently rich description of the circumstances, whether a particular example falls on the referential or the attributive side. Geach's description of this speech act pretty clearly makes it an example falling on the referential side.

Geach tells us that speaker reference is of “negligible importance for logic.” I think we can fairly substitute for “logic,” “semantics.” In any event the specific conclusion we are to draw from the illustration is clear: the speaker's reference in no way determines the semantic reference of the expression he used. The two referents may, of course, be identical, but they may also diverge. The person Mr. Smith has in mind, the one he wants to inform his wife has been made full professor, may not in fact fit the description Smith used, may not be a fat old humbug whom they met the day before. In that event, while Mr. Smith can be said to have referred to whoever it was he had in mind, the uttered description did not. If it referred at all, it would be someone who did (uniquely I suppose) fit the (p.118) description—even someone whom Mr. Smith has forgotten entirely and who never enters the heads of him or his wife during their conversation. Moreover, as Geach makes clear in the passage following the one quoted, the truth or falsity of what was said by Smith, as opposed to what he may have wanted to convey to his wife, depends in that event not on the properties of the person Mr. Smith has in mind to inform his wife about but upon the properties of that person who fits the uttered description. In this way Smith's words have, so to speak, a semantic life of their own. I believe that Geach intends the divorce of speaker and semantic reference to be a quite general across-the-board matter. And I think he is not alone among philosophers in supposing this to be so.4

In Geach's example we may suppose that Mr. Smith believes that a certain person he and his wife met the day before is a fat old humbug; he believes, that is, that a particular person he has in mind satisfies the description uttered. Perhaps we should add that this is likely an active belief—one that he has currently before his mind. Could this be all that speaker reference in such a case amounts to? If it is, then the referential/attributive distinction would presumably come to no more than this, that sometimes a speaker uses a definite description with an accompanying (active) belief about somebody or something that it fits the description and sometimes with no such accompaniment. Construed thusly, the distinction would be real enough, but, to be sure, of no interest for semantics and of little interest, one would think, even for a total theory of speech acts. A myriad of things accompany every speech act—beliefs, desires, itches, and, of course, various external circumstances. A division of speech acts or of a particular kind of speech act based merely on cooccurrence or lack of it between the speech act and some other event promises little for the theory of language. Even some connection between the accompanying item and, say, the content of the (p.119) utterance will not by itself warrant much attention. Some utterances of the form “All X's are Y's,” for instance, are undoubtedly accompanied by an active belief of the speaker's to the effect that the world contains too many X's and others are not. The distinction is real enough, but unlikely to find a place in any treatise on the philosophy of language.


The account of speaker reference just given, however, is obviously too thin. In the first place, it ignores the speaker's intentions toward his audience with respect to what he has in mind. In Geach's example, Mr. Smith intends his wife to recall some particular person met by them the day before and through that, together with his statement, to become informed of a fact about that person. Moreover, it is surely by his having produced the description “the fat old humbug we met yesterday” together, perhaps, with the circumstances of the utterance, that he intends she shall recollect that person. Such intentions can be crucial for the existence of speaker reference. Suppose in some appropriate setting I say, “The strongest man in the world can lift at least 450 lbs.” I might have as my grounds for stating this some general considerations about the limits of human strength with no belief about anyone in particular that he is the strongest man in the world. In such an event there is, of course, nothing we can identify as speaker reference. But my grounds might still be those general considerations while I happen, in fact, to believe about some particular person, Vladimir Jones, say, that he is the strongest man. The addition of this belief should not make us talk of speaker reference. Suppose, though, that my grounds for my statement are that I believe of Vladimir that he is the strongest and I believe he can (p.120) lift 450 lbs. Still, those are my grounds and if I do not expect nor intend that my audience shall recognize that I want to talk about Vladimir and to become informed about his strength, we have no reason to say that referred to Vladimir. What I have been describing, of course, is a case of what I would call an attributive use of a definite description. And what I am saying comes to this, that the referential/attributive distinction and the presence or absence of speaker reference should be thought of as based on such speaker intentions toward his audience or the lack of them—not on whether the speaker believes or not about someone or something that he or it fits the description.5 (Later I will give a more fundamental criterion for speaker reference that will require modifications in what is said here about speaker intentions for a certain range of cases, but for the sort of example Geach gives us it will still be correct.)

The focus of interest of both speaker and audience in the sort of situation Geach describes surely lies in the speaker's reference. And we have in language various resources for asking questions about what the speaker refers to. The speaker's audience will not infrequently fail to recognize what he has in mind to talk about just from the uttered description and the context. It then becomes appropriate to ask him for more information. This may take the form of a simple request—“Whom do you mean?” “What are you talking about?”; or a request that the speaker provide further particulars in order to distinguish among several things that seem to fit the description he used—“Which fat old humbug? We met a couple of them yesterday”; or a question asking whether the speaker's referent fits also some further description—“Do you mean the boring old geezer with the goatee?” It also is not unnatural, we might note, to speak as if the speaker's reference were connected to the expression uttered, for example, “Whom do you mean by ‘the fat old humbug we met yesterday’?”

(p.121) The very form of these questions makes it hard to deny that they concern speaker reference. And in fact they are only appropriate when the speaker did intend to convey a reference. In the example given above in which when uttering “The strongest man in the world can lift 450 lbs.,” the speaker had no intention of having his audience recognize him as referring to anyone in particular, it would have been out of place for his audience to ask him, for example, “Whom do you mean?” or “Are you talking about the Russian who won the gold medal?” or “By ‘the strongest man in the world’ are you referring to Vladimir Jones?” To do so would be to misread the situation and the speaker in turn may correct the misimpression by saying, for example, “I don’t mean anyone in particular, just whoever is the strongest.” Now if these questions were not about the speaker's reference but about the denotation of the description (whatever satisfied it uniquely), then there would be no reason why they should be appropriate only in the presence of speaker reference.


If the audience questions discussed in the last section are questions about the speaker's reference, as they certainly seem to be, I believe we can also show that in the circumstances in which they are appropriate, audience responses can also contain expressions whose referents are determined by what the original speaker had in mind; that here, at least, we have semantic reference determined by speaker reference. And that, I believe, was what Geach wished to deny took place when he said that we could conveniently ignore speaker reference when pursuing logic or, as I would rather say, semantics. But before showing this I want to introduce a notion I will have occasion to use in what follows.

(p.122) In recent work Charles Chastain6 has pointed to the importance for the theory of reference of what he calls “anaphoric chains.” By these he means sequences of singular terms that are such that if one of them refers to something, they all do. A pronoun with an antecedent in the same sentence provides the simplest example:

  1. (1) Mary is coming to town and John will meet her.

The pronoun “her” is coreferential with the name “Mary,” and the two form what Chastain calls an anaphoric chain. There can be more than one chain in a single sentence, of course:

  1. (2) John went to meet Tom and he waited for him almost two hours.

Here “he” is linked with “John” and “him” with “Tom.” If we ask how we know which pronouns belong to which chains, I imagine that at least in some instances it has to do with rules of syntax discoverable by linguists—perhaps something to do with order of occurrence or possibly with a linking of subject to subject and object to object.

The phenomenon I want to exploit is not confined to the limits of single sentences. Chains can pass over sentence boundaries:

  1. (3) John went to meet Tom. Does anyone know whether he has brought him back yet?

Here the pronouns of the second sentence take as antecedents names in the first. Or, to be more precise, they would do so on a natural interpretation and within the same stretch of discourse. If the sentences are uttered on widely separated occasions, for example, or if the utterance of the second sentence is accompanied by (p.123) demonstrations giving the pronouns independent reference, then no such linkage will probably exist.

More important, chains can pass over speaker boundaries. In (3) the second sentence could have been uttered by someone other than the person who uttered the first. If that person is among the first speaker's audience, then the pronouns of the second sentence may be parasitic on the names used by the first speaker forming anaphoric chains with them. We will make use of this fact in what follows.

In general, there will be one member of an anaphoric chain that determines the referent of each of the other members and without which the other members would be left dangling. In the examples so far given proper names have served this function and pronouns have been parasitic on them. I will assume, for simplicity, that this determining member of the chain occurs first and call it the antecedent.7


Let us say that a definite description is uttered in a “referential context” when speaker reference exists relative to it. So far, all this will mean is that the speaker intends to refer to something and intends his audience to recognize his reference in part through his having used that definite description. Thus in introducing this terminology we are not begging the issue, for example, as against Geach. Similarly, a definite description will be uttered in an attributive context when speaker reference relative to it is absent. Definite descriptions in attributive contexts can serve as the antecedent in an anaphoric chain. Imagine the following to be uttered in an attributive context relative to the definite description that forms the subject of the first sentence: (p.124)

  1. (4) The strongest man in the world can lift at least 450 lbs. He can also win a tug-of-war with a jackass.

In a suitable context the pronoun “he” in the second sentence forms an anaphoric chain with the definite description in the first; they both refer to the same thing if they refer at all. And what they both would refer to is undoubtedly the denotation of the definite description—that person, in this example, who uniquely fits the description. The pronoun, in fact, can be regarded as what Geach called a pronoun of laziness8—simply a means of avoiding repetition of the description. Now let us return to Geach's example, which, we are assuming, contains a definite description in a referential context, and extend the dialogue a bit.

  1. (5) [Mr. Smith] The fat old humbug we met yesterday has just been made a full professor. He must have bamboozled the committee.

We seem once more to have an anaphoric chain. Certainly there is some link between the pronoun in the second sentence and the definite description in the first. And one would suppose they are coreferential, if they refer at all. If so, what do they both refer to? On the view that speaker reference and semantic reference are to be kept in their separate realms, the answer is clear: they both refer to the denotation of the description, if it has one. What this means is that if the person Mr. Smith has in mind, his reference, does not in fact fit the description, “the fat old humbug we met yesterday,” what his second sentence expresses will be true just in case there was a (unique) fat old humbug he and his wife met the day before and who bamboozled the committee. And this would be so even if that person never enters into the Smiths’ heads. If semantic reference is (p.125) to be kept untainted by speaker reference, the pronouns in such anaphoric chains must be kept pure also.

But now recall that an audience may ask questions about the speaker's reference when it exists. So the conversation between the Smiths might have proceeded in this way:

  1. (6) [Mr. Smith] The fat old humbug we met yesterday has just been made a full professor.

    [Mrs. Smith] Do you mean the funny little man with the goatee?

But her query, which we have argued concerns Mr. Smith's reference, could have been cast into a sentence employing a pronoun:

  1. (7) [Mrs. Smith] Is he the funny little man with the goatee?

If this is just another way of posing the same question, as I think it is, it too is a question about the speaker's reference. But then the pronoun “he” must refer to the referent of the speaker—to whatever person Mr. Smith was referring to.

An even more striking way of getting this result comes from considering the fact that an audience will sometimes disagree about the applicability of the description the speaker used. And such disagreement can hardly be a question of whether the description applies to its own denotation. Thus in the Smiths’ conversation, Mrs. Smith might have said at a certain point:

  1. (8) [Mrs. Smith] I don’t think he's fat; he's just large boned. And as for his being a humbug, he seemed quite genuine and above board.

(p.126) It would make nonsense of Mrs. Smith's comments to suppose that the third-person pronouns they contain are “pronouns of laziness” standing in for Mr. Smith's original description or that their referent is to be the denotation, if it has one, of that description. Surely it is Mr. Smith's reference that is in question and which determines the referent of these pronouns.

The Smiths’ dialogue might have gone like this:

[mr. smith]

  • The fat old humbug we met yesterday has just been made a full professor. He must have bamboozled the committee.
  • [mrs. smith]

  • Is he the one with the funny goatee?
  • [mr. smith]

  • He's the one I mean.
  • [mrs. smith]

  • I don’t think we met him yesterday. Wasn’t it Friday?
  • [mr. smith]

  • I think you’re right. He was coming from a faculty meeting, so it must have been a weekday.
  • We have an initial definite description followed by a string of pronouns in subsequent utterances. Some of those surely must take as their referent the man Mr. Smith has in mind, the speaker reference rather than whoever is denoted by the description, supposing it does have a denotation.

    Now the (third-person) pronouns in this discourse fragment seem to form an anaphoric chain and the initial definite description seems to be the antecedent. If this were so, it would follow that the speaker's reference determines the semantic reference throughout.

    I suppose that it would be possible to maintain that despite these data the truth or falsity of the utterance containing the definite description depends upon the properties of the denotation of the description, if it has one, and that its semantic referent is its (p.127) denotation—even if this is not so for subsequent pronouns that seem on the surface to be anaphorically linked with it. There would certainly be complications for such a view. In the conversation as recorded Mr. Smith said “He must have bamboozled the committee” immediately following his utterance containing the definite description. Is this occurrence of the pronoun “he” anaphorically linked with the definite description? If so, would it no longer be had he produced this utterance after the Smiths had agreed that the original description was inapplicable?

    In any event, some ground will have to be ceded. It will not be possible to set aside speaker reference as of no importance in the determination of semantic reference. For if speaker reference does not determine semantic reference in certain instances of the use of definite descriptions, it does for subsequent pronouns in some stretches of discourse containing them. The referential/attributive distinction, rested on the notion of speaker reference, will have semantic importance because it will mark the dichotomy between occurrences of definite descriptions that can initiate strings of pronouns whose reference depends upon the speaker's reference and those that do not.


    I now want to turn to some further facts about definite descriptions and anaphoric chains that bear on the topic of speaker reference. Philosophers have often contrasted definite and indefinite descriptions. Definite descriptions, it is said, are used to speak of some one particular thing while indefinite descriptions are not. In the recent essay already mentioned Charles Chastain9 has challenged this as a universal rule. Without for the moment assessing this interesting (p.128) conclusion of his, I want to examine some of the facts that he uses in arriving at it.

    It is certainly true that sometimes a use of an indefinite description in no way involves a reference to a particular thing while a shift to the definite article in the same context would indicate such a reference. We can correctly answer the question “Have you ever seen an elephant?” by saying, “Yes, I have seen an elephant” even if one has in fact seen many elephants. And in so answering, no one of the elephants seen has been singled out. But a particular elephant is presumably referred to if the question is “Have you seen the elephant?” and the answer is “Yes, I have seen the elephant.”

    The definite article indicates in some sense that a particular reference is being made,10 but how does that get accomplished? We know that according to Russell's theory sentences containing definite descriptions (where they are what he called “primary” occurrences) assert in part that one thing uniquely has the properties ascribed by the description. Thus, provided that this part of what is asserted is true, some one particular thing gets singled out. This contrasts with the Russellian treatment of indefinite descriptions according to which sentences containing these expressions assert the existence of members in a certain class but do not, relative to the occurrences of indefinite descriptions, assert that there is a unique member.

    There are two well-known difficulties with this treatment of definite descriptions as far as ordinary speech goes. One is that the definite article can accept plural noun phrases—“I have seen the elephants.” The second is that definite descriptions perhaps most frequently found in ordinary discourse are too “indefinite” to allow us to suppose that their users intend to assert that some particular thing alone has the properties they ascribe. There are a large number of elephants in the world even today and most people who might (p.129) have occasion to use the sentence “I have seen the elephant” are surely aware of this.

    It is not clear how an account along strictly Russellian lines would get over the first of these difficulties, but a view suggested by Zeno Vendler11 might provide a beginning. He says of the assertion “I know the men who fought in Korea” that it would imply that “ . . .  in some sense or other, I know all those men.” And “It transpires, then, that the definite article marks the speaker's intention to exhaust the range determined by [the description].”12 Singular definite descriptions would be a limiting case where it is asserted or implied that the range is limited to a single item. (Later I will try to show that this account fails.)

    The second difficulty remains even on Vendler's view. That the descriptive content of many of the definite descriptions we actually utter is too meager to suppose that we mean to imply that they fit something uniquely has frequently not been taken seriously enough. The usual attempt at saving the situation is well known. The idea is that in ordinary speech we rely on attendant circumstances to supply implicit qualifications on the description actually uttered. So, “I have seen the elephant” uttered at the San Diego Zoo, with speaker and audience aware of their location, succeeds in being an utterance about a particular elephant because the speaker presumes that his audience will take his description as qualified by the restriction “at the San Diego Zoo” or some other qualification such as “at this zoo” that does the same work. This is not without plausibility for a large number of cases, but I believe it will not work as a general rule. Interestingly enough, nowhere do we find more intransigent examples than among the sorts of uses of definite descriptions investigated by Vendler that led him to the view of the definite article just mentioned.

    Definite descriptions not only serve as antecedents in anaphoric chains, they also can act as later links in such a chain. Chastain's (p.130) essay points to the importance of such occurrences of definite descriptions for the theory of reference. Now, interestingly, given the great contrast that is supposed to exist between definite and indefinite descriptions, the antecedents of such anaphoric chains turn out to be sentences containing indefinite descriptions. Consider the following discourse fragment:

    1. (9) A man carne to the office today. He tried to sell me an encyclopedia.

    The pronoun beginning the second sentence seems clearly to be connected in some fashion to the first sentence, though whether anaphorically linked in particular to the indefinite description “a man” is another question.13 The pronoun can be replaced without any shift in what we would understand as being said by a definite description. In fact, as both Chastain and Vendler point out, a definite rather than an indefinite description is required to preserve the anaphoric linkage. We cannot read the following discourse fragment as saying that the same man who carne to the office tried to sell the speaker an encyclopedia; in fact, it is most naturally read as implying that there were two distinct men:

    1. (10) A man carne to the office today. A man tried to sell me an encyclopedia.

    The required definite description can be formed from the preceding sentence in one of two ways, either from a generic noun obtained from the indefinite description or from such a generic noun modified by adjectives or a restrictive relative clause obtained from what was predicated in the preceding sentence: (p.131)

    1. (9a) A man carne to the office today. The man tried to sell me an encyclopedia.

    2. (9b) A man carne to the office today. The man who carne to the office today tried to sell me an encyclopedia.

    3. (11) A man carne to the office today carrying a huge suitcase. It contained an encyclopedia.

    4. (11a) A man carne to the office today carrying a huge suitcase. The suitcase contained an encyclopedia.

    5. (11b) A man carne to the office today carrying a huge suitcase. The huge suitcase carried by the man who carne to the office contained an encyclopedia.

    In some of these cases repetition of information makes the discourse sound like the awkward language of a children's first reader; the equivalences nevertheless hold.

    There are several interesting features of this use of definite descriptions. One is that having used an indefinite description we immediately become entitled to use a definite description. This should give one pause before accepting the usual account of indefinite descriptions. In fact the usual account simply will not fit the use of the indefinite description here, as is suggested by trying to substitute for the initiating sentence in our examples one that explicitly asserts what that account says should be asserted:

    1. (12)* At least one man carne to the office today. The man . . . (He . . .)

    2. (13)* One or more men carne to the office today. The man . . . (He . . .)

    The trouble seems to be that the pronoun or definite description in the second sentence can no longer have an anaphoric link with the (p.132) first sentence.14 If the sentences containing the indefinite description(s) in our examples do not assert merely that there is at least one thing having certain properties, nor obviously do they assert that there is exactly one thing having certain properties.

    One would like to say that the initial sentences in these examples, the ones containing the indefinite description, serve to introduce a particular thing, a man, a suitcase, or whatever, and that this is what justifies the subsequent use of a pronoun or a definite description. But in what sense is a particular thing introduced?

    One initially attractive idea is that the introducing sentence as a whole, not just the indefinite description, serves, if everything goes right, to identify an individual by providing a unique description. (It will be recalled that a subsequent definite description linked with the introducing sentence can be formed from the sentence as a whole by means of a restrictive relative clause—“The man who carne to the office. . . .” from “A man carne to the office.”) There are many cases in which this has plausibility. Sometimes the introducing sentence yields a description that by its very nature could be satisfied by one thing at most, as when the description contains a superlative or a definite ranking in some ordering:

    1. (14) This set contains a least member. The least member of this set. . . .

    2. (15) A rank amateur carne in first in the Podunk Open today. The rank amateur who carne in first. . . .

    The speaker may assume certain background information available to the audience that would limit the things fitting the description to a single one:

    1. (16) I believe she has a husband. He must be kind to her.

    (p.133) A background assumption that one can only have one husband (at a time) may operate. Finally, we may be able to view a speaker as sometimes implying (or asserting) that only one thing has the properties ascribed by the description:

    1. (17) A loose connection is causing the hum in the television. The loose connection must be in the audio section.

    But the problem is that in many cases we cannot suppose that the speaker believes or intends the description plus any background assumptions to pick out something uniquely. Our initial examples show this. (9), (9a), or (9b) might naturally begin an anecdote about an event at one's office told to friends who know little or nothing about what goes on there. Even if one used the fully expanded description, as in (9b), “A man carne to the office today. The man who carne to the office today tried to sell me an encyclopedia,” the speaker is surely not committed to, nor does he intend to suggest that just one man carne to the office that day. Nor would he suppose that there are background assumptions shared by the audience that would allow them to recognize a particular man. There is a particular man presumably about which the speaker is talking, but that person is not identified by the descriptions used plus the circumstances of utterance. An introducing sentence, in fact, may supply no more than a generic noun:

    1. (18) Once there was a king. The king rode a white horse.

    This might begin a historical anecdote about a particular king. But which king is not determined by the descriptive content and the audience need not be presumed to have background assumptions that would narrow down the range.

    (p.134) Similar examples employing plural definite descriptions can easily be constructed:

    1. (19) Some men carne to the office today. They wanted to sell me a philosophical computer.

    2. (19a) Some men carne to the office today. The men. . . .

    Several groups of men may have come to the office today, although only one is being talked about. In suitable circumstances the speaker obviously may have no presumption that his audience can from any shared assumptions, contextual clues, etc., supply further qualifications that distinguish the group talked about from the others.

    These occurrences of definite descriptions in very prosaic discourse fragments show, it seems to me, that by using the definite article a speaker need not be signaling any intention to “exhaust the range” of the description, as Vendler suggests, nor even to exhaust the range of an augmented description that his audience could be presumed to be capable of deriving from the context of utterance. Not only does this seem to show that Vendler's suggestion must be mistaken, but it indicates that the usual way of attempting to save a Russellian account of definite descriptions for ordinary discourse in the face of the indefiniteness of the descriptive content found so frequently cannot succeed either.

    Yet in these examples some particular person or persons are being talked about and the definite descriptions and pronouns seem surely to have particular semantic referents. If the descriptive content of the uttered descriptions even augmented by background assumptions, etc., are insufficient to determine the referents, how is this possible? My answer will not be unexpected. The speaker having some person or persons in mind to talk about can provide the needed definiteness. Once more, then, we have a series of instances in which speaker (p.135) reference appears necessary to provide semantic reference. Here, not just to provide the right reference, but to allow for reference at all.


    With the examples of the last section we have now considered two ways in which definite descriptions can be introduced into a discourse: they may simply appear tout court, as did Mr. Smith's “the fat old humbug we met yesterday,” in which case they may be followed by pronouns apparently anaphorically linked to them; or they may appear after a sentence containing an indefinite description that in some fashion serves as an introduction, as in the examples of the last section. In the latter case they may be replaced by pronouns. In both cases we can distinguish between referential and attributive contexts depending upon whether or not there is speaker reference. But it may have been already noticed that there has been an important difference in the way in which I described the speaker's attitude toward his audience in referential contexts between the two. In the first part of the paper, where we were considering definite descriptions appearing tout court in a stretch of discourse, when there was speaker reference I said that the speaker intended his audience to recognize, partly by means of the description used, what his reference was to. And this, I believe, is a correct account, for instance, of the Mr. Smith example. But in the examples of the last section, where the definite description appears after an introducing sentence containing an indefinite description, the speaker need not be assumed to have any such intention toward his audience even where I claimed there was speaker reference. In telling the anecdote about what happened at the office beginning with “A man carne to the office today” and continuing, for example, with “The man . . . ,” (p.136) the speaker need not expect nor intend his audience to recognize anyone as the subject of the story. What then makes such an example a case of speaker reference at all?

    Of course I have argued that in such examples some particular person or thing is being talked about and the pronouns and definite descriptions that occur subsequent to the introducing sentence and linked with it seem surely to refer to that person or thing. But I believe we can put the point more precisely in terms of what the speaker intends concerning the truth conditions of his utterances: that he intends that truth or falsity shall be a function, in part, of the properties of the person or thing he has in mind.

    Suppose that Woodward and Bernstein in their account of their investigation of the Nixon White House had said at a certain point:

    1. (20) We now had a telephone call from a man high in the inner circle. He asked us to meet him at a certain suburban garage where he would give us confirmation of some of our conjectures. We later decided to give the man the code name “Deep Throat.”15

    Woodward and Bernstein never, of course, reveal to their readers who this high official is, and while some have tried to guess his identity and some may be certain of it, the authors did not intend that there should be recognition. The truth value of what they would have expressed in (20), however, depends upon the properties of the person they are writing about. Suppose that the man they had in mind did not ask them to meet him at a garage but at a certain bus stop. Then the second sentence in (20) expresses a falsehood. Suppose they never decided to give him the famous code name, but put this tidbit in the book because it seemed amusing. Then the third sentence expresses a falsehood.

    (p.137) Let us complicate this a bit by introducing a second mystery person. Suppose that the man that Woodward and Bernstein had in mind when writing the passage was not ever referred to by them as “Deep Throat,” but that in fact they had given this code name to a second informant who entered into their investigations. So someone did possess the property attributed to “the man” in the third sentence. Still, I believe it is clear that that would not save the sentence from expressing a falsehood. The second informant was not being referred to here.

    This determination of truth value by the properties of the speaker's referent extends also to the initiating sentence. If the man Woodward and Bernstein are writing about did not call them, but, say, left them a note in Woodward's copy of the New York Times, the first sentence in (20) would express a falsehood. And it would not be saved if some man high in the inner circles did in fact call them at the time in question.

    We have then a distinction among the cases in which definite descriptions and pronouns anaphorically linked are preceded by an introducing sentence containing an indefinite description comparable to the referential/attributive distinction. The distinction rests on a notion of speaker reference, but one we were able to explicate in terms of the truth conditions of the utterances. It would be nice were we now able to do this across the board. The barrier in the way of giving such an account for definite descriptions introduced tout court, as in Geach's example of the conversation between the Smiths, however, is that we did not show in any conclusive fashion that the semantic reference of such a definite description is determined by the speaker's reference, when such exists, although we did, I believe, show that pronouns apparently anaphorically linked to it were. And to apply without further ado the present account of speaker reference would obviously beg the question. We have, however, been (p.138) able in the last two sections to link speaker reference with the semantic reference of certain occurrences of definite descriptions, so these expressions are not slipping through our net altogether.


    A diesis to be found in somewhat different forms in both Chastain's and Vendler's essays would, if correct, allow us to apply the results of the last two sections directly to the case of definite descriptions introduced tout court. For both hold that the occurrence of definite descriptions in anaphoric chains initiated by a sentence containing an indefinite description is fundamental, and that when we find one introduced into a discourse apparently on its own hook, either it is not a genuine singular term or we should view it as derived from an understood, presupposed, or deleted initiating sentence containing an indefinite description. What we have been treating as two different sorts of cases are really not distinct.

    If we accept their view, we could give the following sort of argument. Suppose, for example, that we are talking to someone we have reason to think will be able to single out a particular person we have in mind and we say:

    1. (21) The man who carne to the office today tried to sell me an encyclopedia.

    Now on the Vendler/Chastain view the definite description here really stands as a member of an anaphoric chain, the antecedent of which is either presupposed or in a covert discourse. The fully explicit discourse fragment from which this should be viewed as derived would be: (p.139)

    1. (22) A man carne to the office today. The man who carne to the office today tried to sell me an encyclopedia.

    But we have already argued concerning (22) that the truth conditions of these utterances would depend upon the properties of the speaker's referent. We should therefore have to say the same thing about (21). And this, of course, would be directly contrary to the view represented at the beginning by Geach that speaker reference has nothing to do with the truth conditions of the utterance.

    I am naturally not antagonistic toward the Vendler/Chastain view, but I am also not sure, aside perhaps for some considerations about simplicity and unification of the treatment of definite descriptions, that there is an argument that shows that the fundamental grammatical construction is that of a definite description anaphorically linked to an antecedent sentence containing an indefinite description. Nevertheless, I believe we can give something like the argument just mentioned.

    First let me say something about how I view these two constructions—the definite description introduced tout court and the definite description anaphorically linked. In referential contexts, those where speaker reference is present, the choice of which construction to use is, I believe, a matter of the speaker's expectations and intentions toward his audience: does he expect and intend that they will recognize who or what he has in mind? If he does, then he will use a definite description with no further introduction; if not, he will begin with an introduction via an indefinite description. What the latter does, so to speak, is to announce that the speaker intends to speak about a particular thing or particular things following under a certain description—for example, he intends to speak of a particular king or a particular man or particular men who carne to the office. There is no implication in most cases that he will speak of everything falling under (p.140) the description. Having done this, he can then go on to use a definite description or a pronoun to refer to what he wants to talk about. Where the speaker intends and expects his audience to be able to recognize what he speaks about from the description used (plus attendant circumstances), such an introduction is otiose. In fact, it would often be downright misleading because it would strongly suggest that the speaker did not believe his audience to be in a position to recognize this reference. Suppose we have just been talking to our colleague Joe who had with him an obviously close acquaintance. After they leave I say to you, “Joe's friend seemed a bit daffy.” It would be no doubt obvious whom I was talking about. Had I said instead, however, “Joe has a friend. He seemed a bit daffy,” I think you would be puzzled because my use of the introductory sentence suggests that you will not recognize my reference and that conflicts with the supposition that I am referring to a person we both have just met.

    This account does not imply that the one construction is grammatically derived from the other, although it is consistent with that possibility. But if it does represent the difference between the two as far as why one might be used in some circumstances and the other in other circumstances, it also suggests that there should be no real difference in truth conditions or semantic references. Consider the indignant Mr. Smith again. Suppose he has two friends at the university, one of whom is a close confidant and familiar with the politics of the situation, the other of whom has just returned from a lengthy sabbatical in the nether regions of central Asia. To the second he says:

    1. (23) A guy in the English Department has been getting chummy with the Dean. He, the guy in the English Department who has been chummy with the Dean, just got promoted to full professor. Shows what things have come to since you were away.

    (p.141) About such a case we argued that a particular person is being referred to but that the particularity cannot be expected to result from the descriptions actually employed applying to one person alone (the speaker may know full well that several members of the English Department have been getting chummy with the Dean), nor from background assumptions (the speaker does not expect that his newly returned colleague will possess facts that together with the speaker's descriptions will enable him to isolate a particular individual). Hence, we must go to the fact that the speaker has a particular person in mind, to speaker reference, to obtain the particularity. And it is properties of that individual that determine the truth or falsity of what the speaker uttered. Now to his more au courant colleague, the speaker might say,

    1. (24) The guy in the English Department who has been getting chummy with the Dean has just gotten promoted to full professor. Shows what things are coming to.

    In this sort of case we were able to argue that subsequent pronouns anaphorically linked with the definite description have as referent the speaker's referent. In particular, Smith's colleague might reply:

    1. (25) You are wrong about him. I’ve investigated and he is far from chummy with the Dean—in fact they had a quarrel just last week.

    And the pronouns here could hardly have as a referent whatever member of the English Department has been getting chummy with the Dean, since there is an explicit denial of that description.

    But now if the only factor that affects Mr. Smith's shift from using an introductory indefinite description to introducing the definite (p.142) description tout court is his expectation or lack of it about whether his audience will be able to recognize his reference, there can be no reason why the semantic facts about reference in the two cases should be different. And we can then, after all, apply the arguments that show semantic reference to be determined by speaker reference from the one sort of case to the other.


    I have not touched upon the question of syntactic or semantic ambiguity. Are sentences containing definite descriptions that can be used in either referential or attributive contexts ambiguous? Are there two senses of the definite article? I have not done so in the first place because it seemed to me I could argue for the semantic significance of speaker reference and the referential/attributive distinction without, directly at any rate, tackling the problem of ambiguity. Secondly, the problem of ambiguity may require, for a solution, that another question I have put to one side, what it is to have something in mind, be answered first. And thirdly, we may need to work out more clearly than we have just what we mean to attribute when we speak of semantic or syntactic ambiguity, and that lies beyond the scope of this paper.

    It might be thought, however, that if the position of this paper were correct that an ambiguity in the definite article would at least be suggested and that it is intuitively very implausible to suppose such an ambiguity. So that, until the question of ambiguity is resolved, a real doubt remains about whether that position can be correct. To this worry I will point out here that the situation of the definite article, given the position of this paper, remains in no worse shape in regards to ambiguity than several other operators and (p.143) connectives in natural languages already are after philosophical scrutiny. We have already seen that sentences containing indefinite descriptions, while they may often be used to assert existential generalizations, are also used to introduce particular individuals, as in some of the examples in this paper where the sentence initiates an anaphoric chain. The same sort of worry about ambiguity should arise here. The “if . . . then_____” construction provided another example. We know that such a construction (in the simplest form of declarative sentence) can be used to express what we might call a true conditional, but also can be used to assert the proposition expressed by the “consequent” where the “antecedent” expresses something like a condition the speaker presupposes for the asserted proposition to be of interest or significance in the circumstances. So I might tell someone to whom I am lending my car for a trip, “If it snows in the morning, there will be chains in the trunk.” This might be used to assert a true conditional; I might, for example, be informing my audience that I will see to it that there are chains in the trunk on condition that I find it snowing in the morning. But I might also be asserting that there will be chains in the trunk, this being of interest, of course, only if it is snowing. (Similarly, Austin's “There are cookies in the cupboard, if you want some” would normally be taken as an assertion of the “consequent,” but might be used by one's fairy godmother, who makes things true providing one wants them to be, to assert a true conditional.) I am not certain that just because of these data we should declare an ambiguity in the “if . . . then_____” construction, although there certainly is a semantically interesting distinction to be drawn.

    If we abstain from passing a ruling about ambiguity, can we draw any conclusions about the Russellian theory of definite descriptions? I am inclined to say that it cannot provide the proper analysis for the referential context in the sense of telling us what proposition (p.144) is expressed. To fully sustain this would, I believe, require saying something more about what it is to have an individual in mind, for I can foresee the possibility of trying to obtain a Russellian proposition from descriptions in the speaker's mind. Let me then just say something about one way that may immediately occur to one attempting to obtain a Russellian proposition in the referential case.

    The main problem brought up in the paper for a Russellian analysis in referential contexts was that of providing a unique denotation when neither speaker nor audience could be expected to believe that the description actually uttered was true of just one individual. It was argued that the usual way of trying to handle this problem—suggesting that the context of utterance would supply further distinguishing descriptions—would not in general work. It was at that point that we turned to what the speaker had in mind. But it might be thought that we could treat this as a special case of context supplying additional descriptions by supposing that what is implicit in such cases is just the additional description “which I, the speaker, have in mind,” or some close approximation. In the Geach example, Mr. Smith could be represented as having said,

    1. (26) The fat old humbug we met yesterday [and whom I have in mind] has just been made a full professor.

    What occurs in brackets was not uttered by Mr. Smith, but intended to be supplied by the audience, Mrs. Smith, by its recognition of what Mr. Smith is up to. (26), in turn, would be analyzed, would be taken as expressing a proposition, along Russellian lines.

    The problem I envisage with this suggestion is that it gives us the wrong proposition. The truth value of the Russellian proposition expressed by (26) would be partly a function of whom the speaker had in mind. If the Smiths met two fat old humbugs yesterday, one of whom (p.145) got promoted and the other of whom did not, the Russellian proposition expressed by (26) would be true or false depending upon which Mr. Smith had in mind. But, I should like to say, whom Mr. Smith had in mind determines what proposition he expressed, not whether the proposition he expressed in any case is true or not. To put it shortly, while Mrs. Smith might say, “You could have prevented that by expressing your views to the committee,” it would be absurd for her to say, “You could have prevented that by having someone else in mind.”

    Much more needs to be said, of course, about the topics mentioned in this last section. My aim, however, has been to argue that whatever the final view about them, we cannot divorce speaker reference from semantic reference.16


    First published in Syntax and Semantics, Vol. 9. Pragmatics. P. Cole, ed., New York: Academic Press, 1978, pp. 47–68.

    (1.) The research for this paper was done in part under a grant from the National Endowment for the Humanities and while a Fellow at the Center for Advanced Study in the Behavioral Sciences.

    (2.) “Reference and Definite Descriptions,” Philosophical Review 75 (1966): 281–304 and “Putting Humpty Dumpty Together Again,” Philosophical Review 77 (1968): 203–215.

    (3.) Reference and Generality (Ithaca, 1962), p. 8.

    (4.) Cf. Saul Kripke, “Naming and Necessity,” in Semantics of Natural Language, eds. D. Davidson and G. Harmon (Dordrecht, 1972), fn. 3, p. 343.

    (5.) As stated this is somewhat misleading. As will be seen in Section VI, I take such a speaker's intention toward his audience to be only a sufficient, not a necessary, condition for speaker reference.

    (6.) “Reference and Context,” Language, Mind, and Knowledge, ed. K. Gunderson, Minnesota Studies in the Philosophy of Science, Vol. VII (Minneapolis, 1975), pp. 194–269. As will be obvious, I am especially indebted to this essay.

    (7.) “His smile is John's best feature,” for example, is an instance where the determining member occurs later.

    (8.) Reference and Generality, p. 124f.

    (9.) “Reference and Context.”

    (10.) In this paper I use the term “reference” and variants on it in connection with definite descriptions both in referential contexts and in attributive. This is not (p.146) intended to imply that there may not be a big difference between the two situations, as “Reference and Definite Descriptions” suggested.

    (11.) “Singular Terms,” in Linguistics in Philosophy (Ithaca, 1967), Ch. 2.

    (12.) “Singular Terms,” pp. 50–51.

    (13.) Chastain's view is that it is. He thus holds that “a man” in the first sentence of the discourse fragment would be a singular term with the same referent as the pronoun in the second. I am not wholly convinced of this. In some sense the first sentence introduces a particular person and, as I will argue, the truth value of what it expresses depends upon the properties of a particular man. I thus agree with Chastain that the sentence does not express an existential generalization of the sort we are often given as the reading of sentence containing indefinite descriptions. But I am not sure that that is enough to show that the indefinite description itself is a singular term. My reluctance stems from wondering whether it is not possible for a sentence to introduce, so to speak, an individual that subsequent pronouns refer to without itself containing an expression that refers to the individual. Perhaps the following would be such a case: “I had steak and lobster at Delmonico's before the play. It was a wonderful meal.”

    (14.) The anomaly of the preceding discourse fragments is not the main argument for the conclusion. That comes from a consideration of the truth conditions as will be seen later.

    (15.) All the President's Men (New York, 1975). The example, of course, is not an actual quotation.

    (16.) Too late for treatment in this paper, Saul Kripke's paper “Speaker Reference and Semantic Reference” (this volume, pp. 6–27) was published in Midwest Studies in Philosophy 2 (1977): 255–276. My paper was then in the final stages of production for a different publication (Syntax and Semantics, Vol. 9, edited by Peter Cole, Academic Press, 1978). The two papers now appear together in the present volume. I had heard a version of Kripke's paper given as a talk, but having no written copy to consult nor transcript of the talk, I did not want to rely on memory. Kripke's paper deals with the same topic as mine and comes to an apparently contrary conclusion, although I believe he is more concerned with the question of ambiguity than I have been here.