1 Introduction

It is fairly widely accepted that Saul Kripke, Keith Donnellan, and others showed in the 1960s–1980s that proper names, in particular uses by speakers, can refer to things even though the speakers do not associate sufficient information with their uses of the names to identify those things.Footnote 1 This contradicted the then-prevailing view [associated most strongly with Frege (1892)] that in order to refer to a thing with a name, a speaker must in some way associate the name with a condition satisfied only by that thing. I will call this view Referent Identification. Against Referent Identification, Kripke and Donnellan pointed to the fact that people succeed in referring to things with names even when they possess very little, and no uniquely identifying, information about the things. Just to rehearse one of Kripke’s examples: someone who has heard of Richard Feynman may refer to him by her use of the name ‘Feynman’ even if the only information she associates with the name is something like ‘a physicist’ (1980: 81).

At least in its Fregean form, Referent Identification is the product of two other views. The first is a parallel requirement on thinking of a thing: in order to think of a thing, a thinker must be able to identify which thing she is thinking of by means of a condition satisfied only by that thing. I will call this requirement Object of Thought Identification. The second is the view that language is thoroughly enmeshed with thinking, in such a way that a proper name as used by a speaker on a given occasion refers to a particular thing in virtue of the use of the name being an externalization of the speaker’s thinking of that thing. I will call this view Cognitive Priority. Both Frege and Bertrand Russell (who is also often associated with Referent IdentificationFootnote 2) seem to have accepted Cognitive Priority.Footnote 3

In this paper, I am going to take it for granted that the Kripke–Donnellan challenge shows that Referent Identification is false. I am interested in which of the contributing views should be given up, conditional on accepting this. Should one give up the metaphysical thesis that names in particular uses refer to things in virtue of speakers thinking of those things (reject Cognitive Priority)? Or should one give up the epistemic thesis that thinking of things requires a satisfactional means of identifying which thing one is thinking of (reject Object of Thought Identification)? Or should one reject both Cognitive Priority and Object of Thought Identification? My answer will be that we should reject only Object of Thought Identification, not Cognitive Priority.

The reason for this is that Cognitive Priority is the most promising basis for answering a question made pressing by the rejection of Referent Identification. This is the question of what determines (metaphysically) the reference of proper names, as used on particular occasions. I will call this “Kripke’s Question,” because Kripke thought it was a central challenge for anyone rejecting Referent Identification. At the beginning of Naming and Necessity he says, “The basic problem for any view such as Mill’s [or Kripke’s own view], is how we can determine what the referent of a name, as used by a given speaker, is”. And a few lines later: “…if there is not such a descriptive content to the name [i.e., if names are not short for definite descriptions, as on the view Kripke attributes to Frege and Russell], then how do people ever use names to refer to things at all?”Footnote 4 Kripke’s subsequent discussion makes clear that the question is not how we determine reference in the epistemic sense (i.e., how we figure out or identify what is referred to), but how reference is determined in the metaphysical sense (i.e., in virtue of what a particular thing is referred to).

The Fregean view, Kripke thought, offered at least the start of an answer. Although Kripke does not present it this way, this answer comes from converting the epistemic thesis of Referent Identification into a metaphysical thesis that I will call Cognitive Satisfactionism. According to Cognitive Satisfactionism, not only is it necessary for reference that a speaker has a means of identifying her referent, but this means of identifying the referent is also what makes the identified thing be the referent. The fact that the speaker would in principle identify her referent as whatever satisfies a certain condition makes it the case that the thing that would be so identified is the referent. However, if the Kripke–Donnellan challenge is correct, names, as used, may refer to particular things in the absence of speakers having any means of singling out those things. Thus, it cannot be in virtue of such identifying conditions that names, as used, in general refer to what they do.

Having rejected Cognitive Satisfactionism, one could still answer Kripke’s Question by relying on Cognitive Priority. One could hold that in the Feynman case (for instance) the speaker is able to think of Richard Feynman despite having no means of identifying him, and her use of the name ‘Feynman’ refers to Richard Feynman in virtue of being generated by her thinking of Richard Feynman. This is the general form of answer to Kripke’s Question that I favor. Such an answer pushes Kripke’s Question back to the level of thinking: in virtue of what is a particular thing thought of, if the thinker lacks a means of singling it out? This is a large and difficult question. My aim in this paper is not to tackle it but to show that we need to tackle it if we want to answer Kripke’s Question.

The more modest aim nonetheless opposes a prominent, perhaps even orthodox, outlook on Kripke’s Question. Part of this outlook, which I endorse, is acceptance that Cognitive Satisfactionism is false, for the kinds of reasons advanced by Kripke and Donnellan. Another part, which I do not endorse, is the expectation that the answer to Kripke’s Question lies instead in developing the claim—often regarded as a truism—that the reference of proper names as used on particular occasions is determined by linguistic convention. I will call this second part of the orthodox outlook the Conventional Stance.

The Conventional Stance rejects Cognitive Priority. It holds that at least part of the problem with Cognitive Satisfactionism is that it does not take account of the independence of linguistic reference from thought. As Genoveva Martí puts it, language should be seen as a conventional institution with “a life that goes beyond the actions, interactions, mental states, intentions, and goings-on in the minds of [its] members” (2015: 85). Accordingly, words in speakers’ mouths have a life of their own, and may refer to particular things independently of what those speakers are thinking of. Thinking of things often causes utterances that refer, and the reference of such utterances is often to the very same things that were thought of. But linguistic reference is not to a particular thing in virtue of the referring utterance being caused by (or otherwise connected with) the speaker’s thinking of that thing. Rather, reference is to a particular thing in virtue of the utterance’s connection to a linguistic convention centered on that thing.

Philosophers who embrace the Kripke–Donnellan conclusions about names typically also adopt the Conventional Stance. This is evident from the frequency with which they appeal to Kripke’s (1977) distinction between speaker’s reference and semantic reference.Footnote 5 Kripke’s distinction reflects the Conventional Stance: of the two species of reference, semantic reference is advertised as the specifically linguistic relation and is, by definition, given by convention. Despite the popularity of endorsing the Kripke–Donnellan conclusions and taking up the Conventional Stance, I will try to show that this combined outlook is not as tenable as it is often assumed to be.

The rest of the paper is divided into five sections. Sections 2 and 3 are devoted to a critical examination of what I take to be the two main options for developing the Conventional Stance. Section 2 examines the Inheritance Chain Approach, which comes from Kripke’s own suggestion for answering Kripke’s Question. This development of the Conventional Stance focuses on the idea that the reference of an utterance can be inherited directly from previous utterances to which it is causally linked (i.e., not via the thinking of the speaker). The problem for this approach, I find, is that the usual processes by which name-utterances are produced do not look like reference-preserving kinds of processes.

Section 3 examines the prospects for Unchained Conventional Determination approaches. These are versions of the Conventional Stance on which particular uses of names refer in virtue of their connection to existing practices or conventions, without these connections involving chains of uses. I argue that the most promising type of Unchained Conventional Determination approach faces problems parallel to those Kripke and Donnellan raised against Referent Identification. Thus, those who find the Kripke–Donnellan challenge compelling should be reluctant to take this route.

In Sect. 4, I argue that the Conventional Stance is not supported by a need to account for a distinction between actual or conventional or utterance reference, on the one hand, and merely intended or speaker’s reference, on the other. In Sect. 5, I shore up the argument of Sect. 4 by arguing that the referring of names need not be prior to thinking of things even in cases of referring with names in total ignorance of the referent. Finally, Sect. 6 concludes by pointing out the direction in which the Kripke–Donnellan challenge pushes us: toward the rejection of epistemic requirements on singular thinking.

2 The Inheritance Chain Approach

Kripke gave the inspiration for the Inheritance Chain Approach in the following passage:

Someone, let’s say, a baby, is born; his parents call him by a certain name. They talk about him to their friends. Other people meet him. Through various sorts of talk the name is spread from link to link as if by a chain. A speaker who is on the far end of this chain, who has heard about, say Richard Feynman, in the market place or elsewhere, may be referring to Richard Feynman even though he can’t remember from whom he first heard of Feynman or from whom he ever heard of Feynman. (1980, p. 91)

It should be noted that this passage is not in conflict with Cognitive Priority. It might be that what happens when reference is passed along as if by a chain is as follows. First, a speaker refers by ‘Feynman’ to Feynman in virtue of her use of ‘Feynman’ being generated by her thinking of, or having in mind, Feynman. Second, her referring act causes someone else to think of Feynman, which eventually (perhaps some time later) causes this new speaker also to refer to Feynman by ‘Feynman’. Like the first use, this second use refers to Feynman in virtue of having been generated by the speaker’s thinking of Feynman.Footnote 6

To instead reject Cognitive Priority while taking up the picture presented in the Feynman passage is to hold that the links between uses in the chain preserve the referential features of earlier uses independently of what later speakers have in mind to refer to. This kind of connection is often described as “reference borrowing”.Footnote 7 For a speaker to borrow the reference of a previous use of a name would be for her use of the name to refer to whatever the previous use referred to, without the need for her to think of that thing herself.

Copying is an apt model for reference borrowing. If a printed text is photocopied, the words appearing in the copy have the same references as the words in the original. If a lecture is recorded, the words heard on the playback have the same references as the words heard in the lecture hall. The devices that make the copies do not need to be thinking of particular things in order for the copied words to refer to what the originals did. Even if a human being acted as the copier by retyping the text or committing the lecture to memory and reciting it later, she would not have to think of what the words in the originals refer to in order for her utterances to refer to the same things. Likewise, one could view later uses of names in chains of reference borrowing as copies of earlier uses, which preserve the referential features of the earlier uses.

Bianchi (2015) highlights the role for this copying relation in developing an inheritance chain answer to Kripke’s Question. He follows Kripke in noting that not just any causal link between utterances can be a link in an inheritance chain. But, whereas Kripke claimed that the right sort of link has the later speaker intending to use the name with the same reference as the earlier speaker,Footnote 8 Bianchi eschews the involvement of such intentions. What is required according to him is that the later use of the name be a repetitionFootnote 9 of the earlier use. It then refers to a particular thing in virtue of being one in a chain of repetitions stretching back to an utterance that was an introduction of the name for that thing.Footnote 10 Repeating an utterance, for Bianchi, is reasonably understood as copying the utterance.Footnote 11 Thus by repeating an utterance one may refer to a thing by a name, not because one is thinking of the thing (as Cognitive Priority would have it) but because one’s use has inherited the reference of the use of which it is a repetition.

The picture can be made vivid by imagining someone who speaks by splicing together recordings of previous utterances and playing these back to make new utterances. She might collect such recordings of utterances from various sources, assembling a large stock of them. Suppose that in testing or absent-mindedly pressing her playback buttons, this person plays back a recording of a previous utterance of the name ‘Bianchi’ that referred to the bicycle company by that name. Let us allow that the new “linguistic particular” (to use Bianchi’s terminology) produced in this way would also refer to the bicycle company in virtue of the recorded utterance having so referred. This seems similar to the way in which the words in the playback of a recorded lecture would refer to what they referred to when the lecturer spoke them aloud. If the analogy holds, then the case illustrates a difference between referring as a kind of action and referring as a feature of the products of certain actions. Mere copying, repeating, or playing back of a previous utterance is not a referring action, but it can produce an utterance that refers.Footnote 12

Now consider a different situation, in which the speaker uses the recordings to refer to things in her own speech. Suppose she plays back a recording of an utterance of ‘Bianchi’ that referred to the bicycle company as part of referring, for her own part, to the philosopher. She may or may not be aware of what the recorded utterance referred to, but this does not seem to matter. She simply selects a recording with features that serve her purposes. It seems clear that her utterance—the linguistic particular she produces—refers to the philosopher, even though it is made by playing back a recording of a linguistic particular that referred to the bicycle company.

If the speaker’s playing back the recording as part of a referring act were like playing back the recording in testing the system or absent-mindedly pressing buttons, we would expect there to be at least a default sense in which her utterance would refer to the bicycle company. That default reference might then be overridden by some other factor (for instance, that she intends to refer to the philosopher). But I do not think this would be the right way to describe the situation. Rather, the recorded utterance’s having referred to the bicycle company is simply irrelevant to the referring of the speaker’s present utterance. She refers to the philosopher, using a recording that, whatever its source, suits her purposes. The linguistic particular she produces through this referring act likewise refers to the philosopher.

If the speaker instead plays back the recording in order to refer to the bicycle company, there is still no reason to think that her utterance refers to the bicycle company in virtue of the fact that the playback is of an utterance that referred to the bicycle company. Rather, her new utterance refers to the bicycle company in virtue of the same kinds of facts (whatever they are) that made her utterance in the first case refer to the philosopher. Linguistic particulars (utterances) produced by the action of referring with the recordings are different from those produced by testing or by absent minded playback. They sound the same and have some features of their production processes in common, but they are the products of different kinds of actions and thus are different kinds of products. One is a playback of a previous use; the other is a new use produced through a different process, in which a playback of a previous use serves as a kind of raw material. In the latter case, there is not preservation of reference between the original and the copy, by the very nature of the kind of action that produces the copy.Footnote 13

As noted above, there may be ways to produce an utterance that really is a mere repetition of a previous utterance. This could be accomplished by a speaker acting as a repeating or copying device. For instance, in the children’s game of ‘telephone,’ each child in a long line aims to repeat verbatim into the ear of his successor an utterance whispered into his own ear by his predecessor. If the children are successful (it is the point of the game for them not to be, but it could happen), then it might be that the utterance of the last child (suppose it is just the utterance of a name) refers to the same thing as the utterance of the first child, and does so in virtue of the chain of repetitions between them. But again, these linguistic particulars are not the same kind of thing as the similar-sounding particulars that would come into being if the children were instead gossiping about the person referred to by the initial utterance. In the ‘telephone’ game, the children try to act like playback devices, refraining from injecting their thinking of things into their utterances.Footnote 14 If they succeed in this, their utterances of names are of a different kind from the utterances they might have made in gossiping. The latter would be products of their own thinking, not mere reflections of someone else’s.

Thus, it is not clear how appeal to repetition or copying can answer Kripke’s Question. If referring utterances of names are made by repeating or copying previous utterances, this implies nothing about the reference of the present utterances. The previous utterances referred to what they did, and the present utterances are thus repetitions of utterances that referred to those things. But when such repetitions are generated by a process of referring, they are not generated in the right way, and so are not the right sort of linguistic particular, to inherit the reference of the utterances they repeat. The point is not that such utterances inherit reference unless this is overridden by specific features of the situation, as in name introduction or reference change. It is that such utterances lack any such inheritance by virtue of their very nature as products of a referring act.

These considerations might seem to suggest that the Inheritance Chain approach needs to invoke speakers’ intentions after all, as Kripke claimed. An intention-based Inheritance Chain approach could accept that the products of referring acts are not reference-preserving by virtue of being repetitions of previous utterances. Instead, it might be suggested that they are reference-preserving by virtue of the speaker’s intending them to have the same reference as previous utterances from which they are causally derived.Footnote 15

If such an approach is to be an alternative to (rather than a variety of) Cognitive Satisfactionism, the role of the speaker’s intentions cannot be to pick out the referent. The suggestion cannot be that my present use of a name refers to a certain individual in virtue of that individual satisfying a condition given by my intention. (For instance, my intention to refer to the individual referred to in the uses from which my present use derives). This would be a form of Cognitive Satisfactionism.

Instead, the involvement of a reference-preserving intention would have to make the production of an utterance be an act of a kind that preserves the reference of a previous utterance. One kind of intention that might do this would be the intention to merely repeat, as in the game of telephone. It might be that if one’s intention in uttering is merely to repeat, this makes the uttering be the kind of act—a mere repetition—that preserves the reference of a previous utterance. Whatever the merits of this suggestion,Footnote 16 acts of referring are not acts of merely repeating. Thus, if uttering with such an intention made one’s act be an act of merely repeating, it would also make it not be an act of referring. So the intention to merely repeat cannot be what makes acts of referring reference-preserving.

The relevant intention would have to be of some other kind, and it would have to make some acts of referring also be acts of some other reference-preserving kind. The difficulty is that it is not clear what further characterization of this kind of act one should give. Repeating is a kind of act on which we have some independent grip. We also have an independent grip on the way that repeating preserves reference. One could propose that referring in a reference-preserving way is a sui generis kind of intentional act that preserves reference. But to do this is just to insist on an intention-based Inheritance Chain answer to Kripke’s Question, without explaining why such an answer is plausible.

3 Unchained Conventional Determination

It might be thought that the Inheritance Chain approach goes wrong in attempting to explain the reference-determining power of linguistic convention in terms of causal connections between uses of names. An alternative is to advance the general principle that if there is a convention of using a name to refer to a certain thing, and if a speaker participates in that convention in making a referring use of a name, then his use also refers to that thing, in virtue of his participation in the convention. Assuming that his participation in the convention is not itself in virtue of his thinking of the thing, this implies that Cognitive Priority is false. But it does not imply that one participates in a convention in virtue of a causal connection between one’s use of a name and previous uses, as on the Inheritance Chain approach.

If participating in a convention for using a proper name is not some form of copying or repeating previous uses, what is it? One suggestion is that it is a matter of the speaker’s intention to participate in the convention, independent of any causal connection her use has to earlier uses. For instance, Evans (1982) writes: “if a speaker uses a word with the manifest intention to participate in such-and-such a practice, in which the word is used with such-and-such semantic properties, then the word, as used by him, will possess just those semantic properties” (387). The evident causal connections among uses in a practice are then explained by the fact that “it is reasonable to attribute to a speaker the intention to participate, by his use of a name, in the same practice as was being participated in by those speakers from whose use of the name the information he has associated with the name derives”. A similar view is suggested by Martí (2015), who claims that “the job of explaining that [speakers] refer, and what they refer to, is done simply by the fact that they rely on and conform to a practice” (82). The surrounding text suggests that Martí is inclined to view this reliance and conformity as an intention to follow a particular practice.

The danger, however, is that the role for such intentions reintroduces Cognitive Satisfactionism, only now with respect to convention determination rather than reference determination. Instead of associating with her use of the name a condition that picks out a referent, it would seem that the speaker has to associate with her use of the name a condition that picks out a convention.

This requirement should worry theorists who are convinced by the Kripke–Donnellan challenge. For just as many users of the name ‘Feynman’ have no means of uniquely identifying Richard Feynman as the referent of their use of the name, they also have no means of uniquely identifying the particular name-using conventional practice in which they intend to participate. Quite commonly such a speaker “can’t remember from whom he first heard of Feynman or from whom he ever heard of Feynman,” as Kripke noted (1980, p. 91). So the speaker need not be able to identify the convention as the one I am following so-and-so in observing, since the speaker need not be able to identify the person from whom he learned the name.

It might be suggested that the speaker must be able to identify the convention as the one that was followed by whoever it was from whom I learned the name ‘Feynman’. But such a condition also might not uniquely pick out a convention, since the speaker may have learned the name ‘Feynman’ as a name for different people at different times. Even a condition like being the practice followed by whoever it was from whom I learned the name ‘Feynman’ as a name for a famous physicist may not work. For the speaker may have learned the name ‘Feynman’ on different occasions as a name for different famous physicists. (This is quite possible since Richard Feynman’s sister Joan Feynman is also a famous physicist). Without there being any confusion on the speaker’s part he may be referring to one rather than the other.Footnote 17

Furthermore, it is plausible that speakers refer to things using names they have acquired, even when they lack thoughts or intentions about following linguistic conventions. Small children, for example, likely lack the metalinguistic sophistication to formulate the idea of a conventional linguistic practice of using a name to refer to a certain thing.Footnote 18 Nonetheless, they can learn names, forget when and from whom they learned the names, and go on to use them just like the adults in Kripke’s Feynman example. Similar considerations count against views on which conventions apply to uses of names just because the speakers intend to go along with whatever conventions are in force in the relevant linguistic community.Footnote 19 For it is not clear that all who successfully use names to refer are capable of having such intentions.

To sum up this section so far: if the Kripke–Donnellan arguments are correct, then in order to refer to particular things using proper names speakers need not associate their uses with conditions uniquely identifying referents for those names. Parallel arguments suggest that speakers also need not associate their uses with conditions uniquely identifying a linguistic convention that they intend to follow. Thus, theorists convinced by the Kripke–Donnellan arguments should not be attracted to a revival of Cognitive Satisfactionism with respect to convention determination.

One way to avoid this revival of Cognitive Satisfactionism while maintaining the Conventional Stance would be to hold that the ability of conventions to determine the referents of particular uses of names is independent of speakers’ intentions concerning those conventions. Linguistic conventions might apply to and determine reference for a given use of a name in much the same way that the laws of a state apply to the actions of citizens. Conventions might apply to particular uses just in virtue of the speakers being members of a certain linguistic community and being in certain sorts of linguistic situations.Footnote 20

Wettstein (2004) proposes something along these lines as the correct response to Kripke’s Question in light of the Kripke–Donnellan challenge. Wettstein models the actual practice of using names on a hypothetical practice of assigning numbers to people at birth and thereafter using these numbers to refer to the people to whom they have been assigned. In this hypothetical practice, users of the numbers would typically be thinking of the people to whom they refer, would know or believe various things about them, would have various intentions, and so on. Nonetheless, Wettstein claims, “What makes a particular person the referent of an uttered number, …is that it is his number” (87). This determination is independent of the speaker’s intentions. Analogously, an uttered “public name” in our actual practice refers to the individual or thing to which it was given, independently of the speaker’s intentions and independently of the speaker being part of any causal chain of communication reaching back to that thing.

The problem with this view is that proper name conventions are not instituted like laws of the state. They do not come with any specification of the linguistic situations in which they apply. And it is clear that they do not apply across the board. As Wettstein himself acknowledges, “making up a name with no special bearer in mind” does not enable one to refer to a thing that just happens to bear the name (90 n25). Moreover, proper names typically have more than one bearer. If this implied that multiple conventions were reference-determining for a single utterance, then any utterance of a name with more than one bearer would refer to all such bearers (or perhaps fail to refer).Footnote 21

So something needs to be said about the application conditions for name-using conventions. Of course, one thing to say is that a convention applies to those uses in which the speaker intends to follow the convention, or intends to accept it as determining her reference. But to say this would be to give up the independence of reference determination from speakers’ intentions, reviving the difficulties just discussed.

Another option would be to claim that a name-using convention applies to uses that are caused by the convention in some appropriate way. But here there is a dilemma. Either this causal connection merely involves the copying or repeating of linguistic forms used in a conventional practice, or it involves something more. If it merely involves the copying or repeating of certain surface features of previous uses, then it will not suffice to keep a convention applicable from these previous uses. Copying or repeating a linguistic form as part of the process of referring no more preserves the applicability of a convention that applied to the original than it preserves the reference of the original. Consider someone who cuts out the printed name ‘Bianchi’ from a bicycle magazine and uses it in a ransom note about the philosopher. The convention of using the linguistic form ‘Bianchi’ to refer to the bicycle company is not in any sense applicable to the new use.

On the other hand, suppose the appropriate connection involves more than copying. If it involves copying with the intention to refer to the same thing as the original, then we are back to reference determination being dependent on speakers’ intentions. If it involves copying as part of a process that is independently a process of referring to the same thing as the copied use referred to, then Kripke’s Question has not really been answered by appeal to practices. Rather, it has been answered by appeal to whatever independent factors make the use refer to a particular thing. (These might well include what the speaker is thinking of in referring, so Cognitive Priority would no longer be rejected).

4 Two Kinds of Reference?

The foregoing discussion suggests that appeal to linguistic convention will not answer Kripke’s Question in a way that should satisfy those who embrace the lessons of the Kripke–Donnellan challenge. But defenders of the Conventional Stance may reply by citing what they take to be either a key advantage of their approach to Kripke’s Question or, perhaps, a more specific statement of the question. They may argue that in asking Kripke’s Question we need to distinguish between “reference” and “intended reference” (Wettstein), “conventional” reference and intended reference (Martí), or the referring of linguistic particulars versus “what speakers do when they are referring to something” (Bianchi).Footnote 22 Each writer emphasizes that it is the first member of the pair—the unqualified, not merely intended, not action-oriented reference—that is their interest. They are interested in the referring product itself, not what the speaker was doing or thinking in producing it. To illustrate, Martí gives the following example:

Suppose that someone enters a room and overhears a conversation in which the name N is used, becoming convinced that the conversation was about person A, someone she saw leaving the room she was about to enter. She may then join the conversation, start to use N, and continue to use N thinking that she is saying something about A. Yet, if the conversation relies on an established practice by which N refers to B… (2015, p. 80)Footnote 23

The continuation of the thought is that in Martí’s assessment it is clear that the speaker’s real, unqualified, not-merely-intended reference is to B, although she “speaker-refers” to A.Footnote 24 Martí is willing to allow that the speaker does come to think of B through overhearing the conversation, and that her thinking of B plays a role in producing her utterances. But if this is right, Martí asks, then what makes it the case that the speaker’s real, not merely intended, reference is to B rather than to A (to whom she merely speaker-refers)? Martí’s answer is that it must be the fact that the speaker participates in an established conventional practice of referring with N to B, and in no such practice with respect to A. According to Martí, the Conventional Stance allows one to give this answer, while those who uphold Cognitive Priority (Martí calls them “neo-cognition advocates”) have no way to answer the question.

But in cases like this one, it is difficult to apply distinctions between actual and merely intended reference, or between the reference of an utterance and the reference a speaker makes with an utterance. It is in fact not clear what the speaker intends to refer to (both A and B seem like candidates) nor is it clear what she actually refers to. It is not clear whether by her utterance she is referring to A or B, but neither is it clear that the utterance refers to A or B. (There may be facts of the matter. But they are not facts that we can discern from the case as described.Footnote 25).

What about the claim that B is the “conventional referent” of the utterance? Does B not have some clear status as the only referent of the use that is in accord with linguistic convention? Let me allow that B does, and that we need to appeal to established conventional practice to distinguish this clear sense in which B is a privileged referent of the use. Even so, this does not require rejecting Cognitive Priority. For if the privileged reference to B is in virtue of the utterance being generated by the speaker’s thinking of B (as Martí allows that it may be) plus the utterance being appropriately connected to the established conventional practice, then the reference to B is still partially in virtue of the utterance being generated by the speaker’s thinking of B.

Martí considers a response along these lines. She replies by pointing out that there are nonetheless cases in which uses of proper names refer to particular things although it is clear that no thinking of those things generates the uses. She gives the hypothetical example of a speaker who knows that in a certain part of Ireland, it is ensured through some sort of social control that at all times there is exactly one man in each town named ‘Patrick O’Grady’.Footnote 26 This speaker, she claims, may enter a pub in such a town, declare, “I am looking for Patrick O’Grady,” and refer, by her use of ‘Patrick O’Grady,’ to the man in the town who bears that name. The speaker’s use cannot refer to the man in virtue of having been generated by the speaker’s having been told about him in particular and consequently thinking of him in particular, since all she has been told about is the general state of affairs.

In introducing the Patrick O’Grady case, Martí aims to illustrate the possibility that a convention may determine the reference of a use of a name, without the speaker having engaged in any thinking about the thing referred to.Footnote 27 There is some attraction to the judgment that the Patrick O’Grady case is an instance of this. Notice, however, that the speaker’s ability to identify the convention is playing a critical role in motivating this judgment. The speaker is not, to quote Wettstein again, “making up a name with no special bearer in mind” and luckily being taken to participate in the local convention. She clearly associates with her use of the name a condition along the lines of the ‘Patrick O’Grady’ convention in this town, which picks out a unique convention. These features seem essential to the intuition that her use refers to the local man called ‘Patrick O’Grady’.

We might allow that which conventions apply to a given use of a name can be fixed in this manner.Footnote 28 If you stipulate that your use of a name will be subject to a convention which you identify by a certain condition, there is perhaps nothing stopping the convention thus picked out from determining a referent for your use.

But as argued above, this kind of satisfactional determination of applicable conventions should not be part of a general answer to Kripke’s Question. It is not required for speakers’ uses of names to refer. Hence, the Patrick O’Grady case is not the core kind of case for which Kripke’s Question is live. It therefore does not support the view that Cognitive Priority should be rejected in favor of the Conventional Stance in order to answer Kripke’s Question.

I conclude that cases of speaker confusion (or intentional conflationFootnote 29) do not create pressure to abandon Cognitive Priority in favor of a convention-based answer to Kripke’s Question. With respect to “actual reference,” “real reference,” “utterance reference,” “speaker’s reference,” “what the speaker refers to,” “intended reference,” and other similar notions, it is simply not clear in these cases what the desired predictions are. If we want to allow that some candidate referents in these cases are sanctioned by a relevant convention, we can do that without abandoning Cognitive Priority as the basis for answering Kripke’s Question.

5 The Transmission of Singular Thinking

In my discussion of Martí’s A–B example, I noted that she was willing to allow that the speaker does come to think of B by overhearing the conversation, and that this thinking of B is part of what causes her referring use of N. But one might wonder why Martí should allow this. For such cases, as Martí also emphasizes, seem like strong candidates to be instances of referring without thinking of the referent. Wettstein also holds up such cases as support for the Conventional Stance, remarking about someone who overhears a conversation involving the name ‘Cicero’: “…one can then ask, ‘Who was Cicero?’ not having the foggiest idea who he was” (78). The speakers in these cases know next to nothing about the referents of their utterances. They may only have understood that an utterance they heard was a name of something. Now, just by acquiring and using—Bianchi might say, repeating—that name, their utterances refer to the same things as the utterances they heard. They do not identify what is being referred to and start thinking their own thoughts about that thing. They just use the name, letting it refer to whatever it refers to, via the conventional practice they have now joined.

The alternative Cognitive Priority view of these cases is that when the ignorant person hears the utterance of a name—suppose it is ‘Cicero’—this makes her think of Cicero. Her thinking might just be to the effect that she has never heard of Cicero, does not know whether Cicero is a person or a place, and so on. But it is Cicero she is thinking about. Her asking, “Who was Cicero?” is the result of that thinking and the utterance of ‘Cicero’ refers to Cicero because her thinking is of Cicero.

Some support for this alternative view can be found by considering a case that is similar except that the speaker forgets the name, or does not hear it properly. Thus, she is unable to participate in the practice of using ‘Cicero’ to refer to Cicero. Still, she can refer to Cicero, for instance by asking “Who (or what) was that?” or “What were they famous for?” That her use of ‘that’ or ‘they’ refers to Cicero is not in virtue of her participating in the name-using practice she encountered. One candidate for what it is in virtue of would be the fact that the use was generated by her thinking of Cicero, despite not (or no longer) having a name for him.Footnote 30 If this is correct, then it is not clear why her use of ‘Cicero’ to refer to Cicero would not also be in virtue of the use having been generated by her thinking of Cicero, in the case where she does also acquire and recall the name ‘Cicero’.

However, in such cases there is another candidate for what makes the speaker’s use of ‘that’ or ‘they’ refer to Cicero. These uses might refer to Cicero in virtue of the fact that they are anaphoric on the previous speaker’s use of the name ‘Cicero’. That the uses are anaphoric on the previous use could itself be in virtue of the speaker’s participating in linguistic conventions for using pronouns anaphorically. Accordingly, these uses would refer to whatever the previous speaker’s use of the name referred to (i.e. Cicero) as a matter of linguistic convention, not in virtue of being generated by the speaker’s thinking of Cicero. If this is correct, then the case does not give us reason to think that singular thinking is determinative in the original case where the speaker uses the name instead of an anaphoric pronoun.Footnote 31

The question of whether uses of demonstrative pronouns like ‘that’ and ‘they’ have their referents determined by conventional rules is related to but distinct from this paper’s corresponding question about proper names. I endorse a corresponding negative answer, for broadly similar reasons as I have advanced with respect to names. But to make a proper case for this requires more discussion of pronouns than I have space for here.Footnote 32 So, I will observe instead that there are related cases where it is not plausible that a speaker is exploiting any convention of anaphora, yet it is still plausible that her use of a pronoun refers to a particular thing because of an earlier encounter with someone else’s use of a name to refer to that thing.

Consider a case like the Cicero case, except that the speaker does not immediately inquire about Cicero. She lets the conversation go on around her, perhaps pretending that she knows a thing or two about Cicero. Afterwards, she forgets (at a conscious level) the conversation and the name ‘Cicero’. Quite a while later, she is at a pub quiz game and is challenged to name three Roman orators. The phrase ‘Roman orator’ rings a bell and she says, “Oh, him… that guy… who was he? I know I’ve heard of him somewhere but I can’t remember where. Was he even Roman? I’m not sure. He might not have been an orator, either”. We can suppose that she does not consciously remember the conversation and does not know how she is aware of this person. In such a case, it is tenuous to class her uses of ‘him’ and ‘he’ as exploiting conventional rules of anaphora so as to refer to whatever the earlier speakers referred to. But if in fact her associative thinking traces to the Cicero conversation, then it is plausible that her uses of ‘him’ and ‘he’ refer to Cicero. Again, a good candidate for what makes this the case is that these uses are generated by her thinking about Cicero. If this is correct, then it is not clear why the same would not be true in the initial case, where the speaker does also acquire and remember the name ‘Cicero’.

6 Conclusion

The Kripke–Donnellan counterexamples challenge Referent Identification, but they do not directly challenge either Cognitive Priority or Object of Thought Identification. They only suggest that Cognitive Priority and Object of Thought Identification cannot both be true. The choice to reject (at least) Cognitive Priority is motivated by the idea that language has a separate existence—a life of its own—distinct from particular uses of language and the thinking that produces them. When we accordingly abstract a use of a name from its generation by a certain episode of thinking, a question arises as to what makes this purely linguistic entity refer to a particular thing. Proponents of the Conventional Stance think that appreciating the use’s connection to a conventional referential practice answers this question. The investigations in this paper cast doubt on this.

I suggest that Kripke’s Question should instead be answered in a way that embraces Cognitive Priority. In broad outline: a proper name utterance that is the kind of thing produced by a certain kind of act (i.e., an act of referring, as opposed to an act of mere repeating as in the ‘telephone’ game) refers to a particular thing in virtue of externalizing a thinking episode that is of that thing.

So for me, Kripke’s Question makes pressing the following further question: in virtue of what is a thinking episode of a particular thing? If we accept the lesson of the Kripke–Donnellan challenge, so that one or both of Cognitive Priority and Object of Thought Identification must be rejected, then the answer to this further question must not imply Object of Thought Identification.Footnote 33 What I have been concerned to show in this paper is that the development of an answer to this question—an account of singular thought that is free of such epistemic requirements—is the new frontier to which the Kripke–Donnellan arguments point us. The new frontier is not the separation of linguistic reference from thought. For those of us who are convinced by the Kripke–Donnellan arguments, linguistic conventions should not seem up to the task of determining reference. Instead, we should uphold Cognitive Priority and recognize that our inheritance from the Kripke–Donnellan challenge is to free our account of singular thinking from epistemic requirements.