Can definite descriptions be used as proper names?

Full Text

Introduction

The problem to be discussed goes back to the distinction between referential and attributive uses of definite descriptions made famous by Donnellan (1966). According to Donnellan, when a speaker uses a definite description referentially, she intends to talk about a certain object or person and utters the description to enable the audience to single out that object or person. The speaker’s intention identifies the referent prior to, and independently of, her choosing a definite description as a means of pointing at it for the audience. This makes it possible that the speaker, using a definite description referentially, can succeed in referring to an entity that does not satisfy the description. For instance, this takes place when the speaker’s belief that the entity she wants to talk about fits the description she uses to point at it happens to be false. On the contrary, if a speaker uses a definite description attributively, there is no independent identification of the entity to be spoken of – what she intends to talk about is exactly whatever/whoever fits the description (Ibid., p. 286; Donnellan 1968, p. 206).

Russell (Russell 1905) offered a quantificational semantic analysis of sentences containing definite descriptions in the subject position. According to Russell, when we utter ‘The F is P’, we express the proposition that there is exactly one F and it is P. Many philosophers, including Donnellan, accept, with or without qualifications, the Russellian analysis for attributive descriptions. (For brevity, I will hereafter refer to attributively/referentially used definite descriptions as attributive/referential descriptions.) I accept it, too. Granted this analysis for attributive descriptions, the problem is how to account for referential descriptions. Basically, there are two approaches to this problem: the semantic and the pragmatic ones. Semantic theories, implementing the semantic approach, posit for referential descriptions a meaning other than the quantificational one, thus treating definite descriptions as semantically ambiguous (Devitt 1981; Wettstein 1981; Reimer 1998; Amaral 2008; Abbott 2010). Pragmatic theories, representing the pragmatic approach, defend the Russellian analysis as the unitary semantic analysis, valid both for attributive and referential descriptions (Kripke 1977, Bach 1981, Neale 1990). Pragmatic theorists account for the specificity of referential use utilizing the Gricean (Grice 1989) distinction between what is literally said and what is communicated in a speech act. Since the Gricean distinction is motivated independently, the pragmatic approach has, ceteris paribus, the advantage of parsimony over the semantic approach, for the latter postulates two meanings for definite descriptions, while the former postulates one. Therefore, semantic theories must provide special reasons for imposing a referential meaning on definite descriptions¹.

In section xxii of Demonstratives, Kaplan (Kaplan 1989, pp. 558-563) has adumbrated a semantic theory construing referential descriptions by analogy with proper names. More recently Marti (Marti 1995; 2003; 2008; 2015) has developed Kaplan’s proposal in detail. I will refer to Kaplan-Marti theory as KM. According to KM, when a definite description is used referentially, its attributive semantic content loses semantic relevance, and the description becomes a directly referential expression of the same kind as proper names in the Millian sense². Both Kaplan and Marti claim that their semantic construal of referential descriptions explicates Donnellan’s notion of referential use and try to explain the phenomenon Donnellan held to be illustrative of referential use, namely the possibility of referring by misdescription. My aim here is a polemic one: I will argue that the evidence KM adduces to back up the semantic treatment of referential descriptions is inappropriate with respect to all cases of referential use except one, and that the only case KM provides a plausible account for does not fit Donnellan’s notion of referential use. This means that KM has no advantage for the pragmatic account of referential use in Donnellan’s sense.

The pragmatic theory that I accept has been elaborated by Kripke (Kripke 1977). It accounts for referential descriptions as follows. Suppose that a speaker utters ‘The F is P’ using ‘the F’ to refer to an object or person O. Usually, in such a speech act, the speaker presupposes that O is the only F and communicates the singular proposition that O is P, but what she expresses is the general proposition that there is just one F and it is P³. On this view, what accounts for the specificity of referential use of definite descriptions is the speaker’s intention to speak of a particular entity, and what accounts for the phenomenon of referring by misdescription is the falsity of the speaker’s presupposition or other pragmatic factors.

The premises I share with Kaplan and Marti are:

1) The Russellian analysis of the meaning of attributive descriptions, according to which an attributively used definite description expresses a certain attribute and denotes the entity that uniquely satisfies it⁴.

2) Donnellan’s idea that the reference made in the course of referential use of a definite description does not depend on the attributive content of the description. This independence enables the phenomenon KM tries to explain, namely the phenomenon of successful referring by misdescription.

3) The Millian notion of a proper name, according to which the only semantic content of an ordinary⁵ proper name is the reference to its bearer. This means, inter alia, that the semantic link of a proper name to its bearer is not mediated by any descriptive content.

A note on the notion of direct reference, which plays an important role in what follows, should be made here. Kaplan and Marti give different definitions of direct reference:  Marti offers a ‘strong’ conception of direct reference that Kaplan does not use. Kaplan defines a directly referring expression as an expression that contributes a particular entity to the proposition expressed (Kaplan 1978). Marti calls this notion of direct reference the ‘weak’ notion and distinguishes it from the ‘strong’, or ‘Millian’, notion on that an expression is directly referring if it is immediately linked to its referent (Marti 1995, p. 280; 2008, p. 47). As Marti (Marti 1995, p. 281; 2008, p. 48) stresses, Kaplan’s notion of direct reference and her notion of strong direct reference have different extensions: for instance, indexicals on the Kaplanian view are directly referring in the weak sense but are not in the strong sense, for the link of an indexical to its referent is mediated by its Kaplanian character. This difference between Kaplan’s and Marti’s views on direct reference is irrelevant to my criticism. What matters is that both Kaplan and Marti consider referential descriptions a type of proper name and that both hold that the semantic link of a proper name to its bearer is not mediated by any descriptive content. This makes proper names directly referential expressions both on Kaplan’s ‘weak’ and Marti’s ‘strong’ notion. Hereafter, I will apply the term direct reference to proper names as compatible with both notions.

In what follows, I expose KM’s construal of referential descriptions (section 1), propose an amendment to Marti’s terminology (section 2), define the standard case of referential use of definite description (section 3), and present my objection to KM with regard to the standard case (section 4).

1. KM’s construal of referential use

According to KM, referential use of definite descriptions is a special case of unconventional use. Kaplan (Kaplan 1989, p. 559) distinguishes two types of unconventional use of an expression: when the speaker relies on its conventional meaning but mistakenly applies it to an entity that does not fit its conventional meaning, and when she endows it with a new meaning. Consider Kripke’s Smith/Jones scenario (Kripke 1977, p. 263), in which the speaker refers to Smith using the name ‘Jones’. If she does so because she mistakes Smith for Jones, as in Kripke, then she relies on the conventional meaning of ‘Jones’, and we have the unconventionality of the first type. But the speaker may also use ‘Jones’ intending to refer to the man over there, regardless of the conventional meaning of ‘Jones’. So the speaker acts in Marti’s variation of the scenario in which ‘she has forgotten Smith’s name, or doesn’t know his name or, quite simply, she does not care what Smith’s name is’ (Marti 2008, p. 54). In this case, we have the unconventionality of the second type: the speaker intends to give ‘Jones’ a new meaning rather than to utilize a preexisting meaning. The first type of unconventional use can only be accounted for in pragmatic terms, while the second type is semantically relevant.

In KM, referential use of definite descriptions is a type of semantically relevant unconventional use that Kaplan calls dubbing (Kaplan 1973, p. 501). Dubbing is an event or historical process through which a new proper name is introduced into a language. In some cases, expressions of definite descriptions – I will call them descriptive expressions – are used as vehicles of the proper names introduced. It would be convenient to have terms for dubbings of this sort – dubbings that turn descriptive expressions into proper names – and for proper names introduced in this way. I will call the former d-dubbings and the latter d-names. Introducing a nickname into a language is an example of d-dubbing. Since Sidney Crosby has been called ‘the Kid’, the descriptive expression ‘the kid’ can be used both as a definite description and a proper name.

It is clear that d-dubbings make descriptive expressions semantically ambiguous. A d-dubbing semantically neutralizes the attributes the descriptive expression in question expresses when it is used as a definite description (hereafter descriptive attributes): they lose their semantic relevance⁶ (though they may remain pragmatically relevant) so that the reference occurs independently of whether or not the descriptive attributes hold for the object referred to⁷. Then, as Marti puts it, ‘the speaker, quite simply, couldn’t care less whether the attributes apply or not’ (Marti 2008, p. 53).  Let me call this point a neutralization claim.

Another feature characteristic of referential use of definite descriptions in KM’s view can be defined in terms of Kaplan’s distinction between the permanent meaning, pertaining to an expression in constant use, and the nonce-meaning given to an expression, such as ‘a fleeting “Hi-ya, Beautiful!”’ (Kaplan 1989, p. 561) for purposes of the moment. On KM, referential use of a definite description is a d-dubbing in that the speaker does not intend to introduce a stable name into the language she speaks, aiming instead ‘at attaching a nonce-sense to the expression’ (Ibid., p. 561). Of course, an expression applied to an object in a speech act of this kind can subsequently ‘stick’ to the object so dubbed and thus become a stable name, but if it does not then we can say that a throwaway name⁸ has been generated. The latter case is precisely the referential use of definite descriptions in KM’s view. Thus, it can be defined as a d-dubbing, in that a descriptive expression is used as a throwaway proper name.

Since this definition renders the referential use of definite descriptions as a type of d-dubbing, it presupposes the neutralization claim with respect to referential descriptions. This is why Marti thinks that this definition provides the needed account of the characteristic mark of referential use – the possibility of pointing by misdescription. On the other hand, the differencia specifica provided by this definition – being a nonce-use – distinguishes referential descriptions from stable proper names of descriptive origin, such as ‘the Kid’, ‘The Holy Roman Empire’ etc. (Marti 2008, p. 53).

So far so good. What raises a problem for KM is, I think, the fact that the attributive descriptive content of a referential description can be pragmatically relevant to the speaker’s purposes. Both Kaplan and Marti admit this. Moreover, as I will show below, Marti tends to consider the pragmatic relevance of descriptive attributes to be important for the speaker’s success in establishing the referential link between the description used and the object referred to. But neither Kaplan nor Marti notices that this affects the validity of the neutralization claim with regard to nonce-use of definite descriptions in referential function. I will develop this point in section 4.

2. A terminological note

Before I present my criticism, I would like to propose an amendment to Marti’s terminology motivated by the fact that some of her formulations contain, in my view, confusion. Consider the following passage (partially quoted above):

What is characteristic of a referential use of a definite description is, according to Donnellan, the fact that the description can refer to something that does not satisfy the attributes expressed by the description. (Marti 2008, p. 45)

What is wrong in this formulation? Remember that Marti considers the neutralization of the attributive semantic content of definite descriptions an essential feature of referential use. If she is right then referentially used descriptions refer directly and express no attributes at all! This makes this formulation, taken literally, contradictory.

Here is another example:

What determines successful reference with the use of “the F,” as it was the case with proper names, is not the referent’s satisfaction of a qualitative profile associated semantically with the expression in question, precisely because the expression, “the F” in this case, is being used as a genuine device of reference. (Marti 1995, p. 283)

In this passage, we have the same contradiction: if an expression functions as ‘a genuine device of reference’, then no ‘qualitative profile’ can be semantically associated with it. If there is such thing as referential use of definite descriptions on Marti’s view, definite descriptions have different semantic contents in different uses. Then it would be incorrect to speak of the attributive semantic content of definite descriptions (‘qualitative profile associated semantically’ with them) simpliciter, without qualification to the effect that definite descriptions have such content only when used attributively. And it would be misleading to speak of the attributive semantic content of a definite description in discussing its referential use.

To preclude terminological confusion, I suggest speaking of two uses of descriptive expressions rather than descriptions – the referential use, in which descriptive expressions function, according to KM, as throwaway proper names, and the attributive use, in which they function as definite descriptions. Utilizing these terms, we can rewrite the claim quoted in the beginning of this section like this:

What is characteristic of a referential use of a descriptive expression E is… the fact that if E is used referentially (as a throwaway name) it can refer to something that does not satisfy the attributes expressed by E when it is used attributively (as a definite description).

In what follows, I will use terms in this way.

3. The standard case

In this section, I discuss a certain type of what KM considers referential use of definite descriptions. Speech acts of this type are much discussed, therefore I suggest calling them the standard case. As noted above, both Kaplan and Marti admit that the descriptive attributes associated with a definite description, thought they are semantically neutralized when the description is used referentially, nonetheless can be used some way in the same speech act, thus gaining a pragmatic relevance. For instance, Kaplan, discussing the ‘the pure referential use’ of a descriptive expression, says:

When a definite description is in question, again the speaker does not intend to give the expression its conventional meaning although he may intend to make use of the conventional meaning in conveying who it is that is being referred to or for some other purpose. (Kaplan 1989, p. 560)

Speech acts that I call the standard case have two features in addition to ones fixed in KM’s definition of referential use. One of them is mentioned in the passage quoted: the speaker makes use of descriptive attributes in order to convey what entity she wants to talk about. The other feature is that the speaker uses the descriptive expression in a quite straightforward way – not ironically, not metaphorically, etc. These two features together mean that the speaker chooses the descriptive expression the F to point at the object/person O because she believes that O is the sole F and relies on her audience’s ability to recognize O as the sole F.

According to KM, the standard case has the following features:

1) It is a type of d-dubbing, i.e. the speaker turns a descriptive expression E into a proper name of an object/person O. The speech act in question semantically neutralizes the attribute F that constitutes the semantic meaning of E in contexts where E is used attributively.

2) E is used as a throwaway name: the act of establishing E as a proper name of O is the only act of so using E.

3) The speaker uses E to point at O for her audience.

4) The speaker chooses E as a device of referring to O because she believes O to be the only F and expects that her audience will be able to see this, too. This makes F pragmatically relevant.

The last characteristic of the standard case is especially important for me, and it seems to be of importance for Marti as well. Let me dwell on this point at some length. Discussing a scenario in which the speaker uses the description the man drinking a martini to refer to a certain person named Aston-Martin⁹, Marti writes: ‘The attribute drinking a martini, plus a number of other factors… play an important role in helping establish the referential connection …’ (Marti 2008, p. 53), though (she adds) they do not determine the reference. She does not specify to what extent she finds the role of descriptive attributes in establishing the referential connection ‘important’. Does its importance go so far that the discrepancy between the descriptive attributes and the object the speaker tries to refer to makes her try unsuccessful? Marti gives no definite answer but adduces two illustrations of referential failure that suggest that the answer must be affirmative.

Both illustrations are variants of the scenario mentioned in that the speaker tries to point at Aston-Martin by uttering ‘the man drinking a martini’. In the first variant, ‘Aston-Martin is going around with his hands in his pockets at a teetotalers’ party’ (Ibid., p. 55). Marti holds that in such a situation the speaker’s attempt to refer to Aston-Martin cannot be successful, even if the speaker has him in mind. The obvious cause of this failure is Aston-Martin’s not fitting the description in question. In the second, ‘someone else who is very salient and right in front of the speaker and hearer has a martini glass in his hand’ (Ibid., p. 55). Marti finds it ‘at the very least arguable…that the speaker makes a claim about the salient martini-glass holder, even if he does not have him in mind’ (Ibid., pp. 55-56). In this case, even if the speaker intends to refer to Aston-Martin, she does refer – unwillingly – to a salient man who happens to fit the description used.

Marti adduces these examples to back up the claim that having in mind is not the reference determiner. What is important for me in them is that they show that, in the standard case, the speaker believes that the entity she tries to point at satisfies the attributes semantically expressed by the description she uses, and expects that her audience will share this belief. Indeed, in the first example, the speaker does not succeed in referring to Aston-Martin because of his apparently not satisfying the attribute in question. This suggests that an entity’s fitting the descriptive attribute in the speaker’s view is necessary for success in establishing the referential link between the description in question and that entity. In the second example, a man’s apparently fitting the descriptive attribute (plus his salience for the speaker) seems to be enough for establishing the referential link to him independent of the speaker’s intention. This suggests that if an entity apparently fits the descriptive attribute, it is sufficient – granted that the speaker is acquainted with it – for the description to refer to it. I do not claim that Marti is committed to the view that an entity’s apparently fitting the descriptive attribute(s) is the necessary and sufficient condition of successful establishing the referential link in the general case. What I want to stress is that her examples show the high importance of this factor in the standard case of referential use of definite descriptions¹⁰.

4. Criticism

Now I am in a position to present my criticism¹¹. As we saw earlier, KM ascribes to a descriptive expression as it is used in a standard case a double function of pointing at an item and naming it. Insofar as it is used as a device of pointing, it functions as a definite description with its conventional attributive semantic content. Insofar as it is used as a device of naming, it functions as a throwaway proper name. According to KM, in such a double use descriptive attributes are semantically neutralized although they continue being pragmatically relevant. My objection to KM is that it tries to substantiate the neutralization claim with respect to the standard case relying on inappropriate evidence. I will argue that the datum KM relies on can takes place only in the course of a constant practice of using an expression; it simply cannot occur within a nonce-use.

Marti illustrates the neutralization claim by discussing Mill’s example with ‘Dartmouth’ (since ‘Dartmouth’ can express the attribute of being at the mouth of the River Dart, it can be considered on par with definite descriptions):

Dartmouth is at the mouth of the Dart, and this fact played a very important role in the decision to call the city ‘Dartmouth’. But had the river changed course, it would not be incorrect to continue to apply ‘Dartmouth’ to Dartmouth. In fact, were we to discover that Dartmouth never was near the river, the name would not fail to name the city. (Marti 2008, p. 47)

In this counter-factual scenario, people believed, at the moment of dubbing, Dartmouth to be at the mouth of the River Dart, and this inter alia explains why they so named the city. (This makes that dubbing akin to what I consider the standard case. The difference is that ‘Dartmouth’ turned out to be a stable name.) Subsequently the river changed its course and people discovered that Dartmouth does not satisfy the attribute of being at the mouth of the River Dart anymore. Nevertheless, they did continue to apply the expression ‘Dartmouth’ to the city. This allows stating that ‘Dartmouth’ became a directly referential device so that the descriptive content it might have in other usages has no semantic relevance in the usage in question. As we see, the evidence needed to support the neutralization claim with regard to cases like the case of Dartmouth is the practice of applying an expression E to an object O despite the fact that people have discovered that E’s attributive semantic content does not hold for O. To gain this evidence, it is not enough to observe the dubbing as a single speech act – it is also necessary to observe the subsequent use of the expression in question. Thus, this type of evidence is specific to cases of introducing stable names.

But what about the standard case, in which, allegedly, a throwaway name is introduced? Imagine that the dubbing was the first and last use of ‘Dartmouth’ – that ‘Dartmouth’ did not happen to become a stable name. Then the speech act in question instantiates the standard case, and the speakers used the expression ‘Dartmouth’ because they 1) believed Dartmouth to be at the mouth of the River Dart, 2) intended to use this attribute to point at Dartmouth, and 3) semantically associated this attribute with ‘Dartmouth’. It is quite natural to account for their choice of ‘Dartmouth’ by supposing that they utilized its attributive semantic content, thus using it qua definite description. Since there was, as assumed, no subsequent use of ‘Dartmouth’, we have no data supporting the neutralization claim and, consequently, no reason to suppose that ‘Dartmouth’ was used as a proper name rather than a definite description.

I know of no appropriate datum adduced by Kaplan or Marti in supporting the neutralization claim with respect to the standard case¹². I think the argument of the previous paragraph shows that there is no phenomenon of the sort they need. Marti is aware that examples she discusses are ones of using stable names rather than throwaway names, but she finds this irrelevant: ‘whether or not ‘Dartmouth’ and ‘The Holy Roman Empire’ do become stable referential devices (i.e. names) for a city and a country is a story that is independent of their being referential devices that do refer to the city and the country in that first introductory use, or first few uses’ (Ibid., p. 52). For sure, Marti is right in that if a descriptive expression was once used referentially, its subsequent becoming or not becoming a stable name is an independent story. But what allows stating that it was so used? I contend that KM has no answer to this question. What I find wrong in KM with regard to the standard case is that it tries to account for using throwaway names relying on evidence that is specific to cases of introducing and subsequent using stable names.

Conclusion

I have examined two points in KM: the semantic construal of referential use of definite descriptions and the claim to having explicated Donnellan’s notion of referential use. KM’s semantic construal of referential use includes two tenets:

• if a definite description (strictly speaking, a descriptive expression) is used referentially, its attributive semantic content loses any semantic relevance;
• a referential description (referentially used descriptive expression) functions as a throwaway proper name.

My objection to KM was to the effect that KM tries to ground (1) relying on evidence that (2) excludes. I argued that if the attributive semantic content of a referential description has any relevance at all, only subsequent use can support the claim that its relevance has changed the semantic character to a purely pragmatic one. But the subsequent use is possible only for stable names, not for throwaway names. This outcome allows concluding that KM has no advantage over the pragmatic account of referential use in Donnellan’s sense, and that the pragmatic account is preferable for reasons of parsimony.

 

¹ Donnellan himself is not quite clear about this dilemma. On one hand, he expresses his preference for the pragmatic approach (Donnellan 1966, p. 297). On the other, he contends that Russell’s analysis of sentences containing definite descriptions ‘ignores altogether the referential use’ (Ibid., p. 297), and the most natural way of understanding this claim is to suppose that he ascribes to his distinction a semantic import. I leave the exegetic question as to what position toward the dilemma of semantic and pragmatic approaches, if any, can be attributed to Donnellan, beyond consideration.

² The view outlined in Demonstratives xxii essentially differs from the view Kaplan offered in Dthat (Kaplan 1978). In Dthat, Kaplan construes referential descriptions as a type of demonstratives rather than proper names.  This view is irrelevant to KM.

³ If ‘the F’ is an incomplete definite description, an appropriate completion of the description or an appropriate restriction of quantifier domain is presupposed. This holds for the speaker’s presupposition that O is the only F as well.

⁴ This view is compatible both with Russellian quantificational and Strawsonian presuppositional analysis of definite descriptions.

⁵ I distinguish ordinary proper names from descriptive names. See n. 7

⁶ This is often signaled by the fact that a description, as Strawson (Strawson 1950, p. 338) puts it, ‘ha[s] grown capital letters’.

⁷ This is why we must distinguish d-names, such as nicknames, and descriptive names like ‘Neptune’ when it was introduced by Leverrier. Names of the latter sort are abbreviations for definite descriptions (‘Neptune’ was introduced by Leverrier as the abbreviation for ‘the cause of perturbations in Uranus’s orbit’) and denote their objects (if any) rather than refer to them. Reimer shows that descriptive names can turn into ordinary (directly referential) ones in the course of their use, as happened with ‘Neptune’ after astronomers became empirically acquainted with Neptune (Reimer 2004). In this case, descriptive names lose their attributive content and become d-names.

⁸ Marti attributes the expression ‘throw away name’ to Ruth Barcan Marcus (Marti 2008, p. 54).

⁹ A fusion of two examples by Donnellan (Donnellan 1966, p. 287; 1970, p. 349).

¹⁰ It is opportune here to remember Marti’s characterization of referential use cited above: ‘…the speaker, quite simply, couldn’t care less whether the attributes apply or not’ (Marti 2008, p. 53). Now it should become clear that this claim does not hold for the standard case. As we see, in the standard case – in particular, in Marti’s examples just discussed – the speaker cannot be so light-hearted about the attributes’ applying or not applying to the entity she tries to refer to, for it influences the success of her try. It seems to me that what Marti has in mind here is that the speaker feels no concern for the semantic relevance of attributes though possibly finds them pragmatically relevant. I will argue below that this claim needs evidence KM fails to provide.

¹¹ A short version of this criticism was presented in (Borisov, 2023).

¹² In section 1, I mentioned Marti’s version of Smith/Jones scenario, in which the speaker uses the name ‘Jones’ as a throwaway name, without taking into account whether ‘Jones’ is the conventional name of the man she is referring to or not (Ibid., p. 54). Marti is right in claiming that this speech acts illustrates the neutralization of the semantic meaning of ‘Jones’. But it does not instantiate the standard case because, granted the Millian notion of proper names, ‘Jones’ has no attributive semantic content.

About the authors

Evgeniy V. Borisov

Institute of Philosophy and Law of the Siberian Branch of the Russian Academy of Sciences

Author for correspondence.
Email: borisov.evgeny@gmail.com
ORCID iD: 0000-0001-6587-9616

doctor of philosophy, chief researcher of the Institute of Philosophy and Law of the Siberian Branch of the Russian Academy of Sciences

Russian Federation, 8, Nikolaeva Str., Novosibirsk, 630090, Russian Federation

References

Supplementary files