An unjustly neglected theory of semantic reference

Abstract

There is a simple, intuitive theory of the semantic reference of proper names that has been unjustly neglected. This is the view that semantic reference is conventionalized speakers reference, i.e. the view that a name semantically refers to an object if, and only if, there exists a convention to use the name to speaker-refer to that object. The theory can be found in works dealing primarily with other issues (e.g. Stine in Philos Stud 33:319–337, 1977; Schiffer in Erkenntnis 13:171–206, 1978; Sainsbury in Erkenntnis 80:195–214, 2015; Sainsbury, Thinking about things, Oxford University Press, 2018), yet these authors provide no sustained discussion of it. Devitt (Designation, Columbia University Press, 1981) did formulate a view on which semantic reference is conventionalized speaker’s reference, yet his views are assimilated to causalist views. This is a mistake. While the conventionalized speaker’s reference view captures much of what is plausible in descriptivism and causalism, it remains distinct from both.

1 Introduction

There is a simple, intuitive theory of the semantic reference of proper names that has been unjustly neglected. This is the view that semantic reference is conventionalized speaker’s reference, i.e. the view that a name semantically refers to an object if, and only if, there exists a convention to use the name to speaker-refer to that object. This view is so simple that one may well think that it is trivial, and compatible with the main traditional theories of semantic reference. In this paper I will show that it is not, i.e. that if semantic reference is conventionalized speaker’s reference, then both causalism and descriptivism are false. I also diagnose the various confusions that may lead one to think that this view is compatible with descriptivism and causalism.

Section 2 states the conventionalized speaker’s reference view and explains what the conventionalized speaker’s reference view does and does not claim. Section 3 shows that it is distinct from descriptivism and Sect. 4 shows that it is distinct from the causal theory of reference. Section 5 considers descriptivist and causalist theories of speaker’s reference and Sect. 6 defends the view against some obvious objections. In Sect. 7 I speculate about why this theory has suffered such relative neglect.

2 The conventionalized speaker’s reference view of semantic reference

When I speak of the semantic reference of a name I mean the referent that is assigned to the name by linguistic convention. By the same token, when I speak of the semantic content of a name I mean to speak of the conventionally determined contribution that the utterance of a name makes to the proposition expressed by the sentence in which it occurs. In this way both the semantic reference and the semantic content of a name are taken to be properties determined by linguistic convention.¹

The conventionalized speaker’s reference view of semantic reference (hereafter CSR view) claims that the semantic reference of a name can be analyzed as follows:

A name N² semantically refers to an object o, among a coordinating community C, if, and only if, there is a convention among members of C to use N to speaker-refer to o.

The CSR view claims that semantic reference can be fully analyzed in terms of conventions and speaker’s reference, i.e. that ‘Krugman’ semantically refers to Krugman due to the existence of a convention to use ‘Krugman’ to speaker-refer to Krugman, and so on.

The CSR view takes its inspiration from the Gricean view of sentence meaning, according to which sentence meaning can be explained in terms of speaker meaning (Grice, 1957, 1968). It tries to apply this basic Gricean idea to a sub-sentential entity, i.e. names, and claims that semantic reference is no more than conventional speaker’s reference. The key idea behind the CSR view is that people typically use names when they wish to speaker-refer, i.e. make people, places and things, etc. the subject of communicative acts. This leads to the adoption of conventions that pair such people, places and things with names. These claims seem plausible, even bordering on the platitudinous; the CSR view is the view that these claims suffice to analyze the phenomenon of the semantic referent of a name.

The CSR view, as formulated here, makes reference to coordinating communities, instead of speaking of linguistic communities. By a coordinating community I mean any group among which a convention exists. While names are linguistic items, it is not clear whether names form part of specific languages in the same way that other terms do. The formulation employed here allows for the point that linguistic conventions exist among specific communities to be acknowledged, without prejudging the matter of whether such communities should be thought of as coinciding with ordinary linguistic communities.

The remarkable thing about the CSR view is that, despite its prima facie plausibility, it has a very low profile in the literature concerning semantic reference. The basic view is formulated in a paper in Stine (1977), only for her to downplay its novelty and claim that it is merely a way of developing Kripkean causalism. Sainsbury (2015, 2018) has used it in work that primarily deals with other matters, and, although Sainsbury mentioned that he regards it as an alternative to extant theories, his claim has received no subsequent discussion. While Devitt (1981) did formulate a CSR view, his views are typically assimilated to causalist views. This is a mistake. While the conventionalized speaker’s reference view captures much of what is plausible in descriptivism and causalism, it remains distinct from both.

It is not my intention here to argue for the truth of the CSR view. Rather I wish to show that the CSR view is more important than generally realized. To this end, my primary aim is to systematically spell out the argument in favor of taking the CSR view to be an alternative to the main traditional views, and not a mere variant of such views or compatible with such views. I take it that the prima facie plausibility of the CSR view is such that showing it to be an alternative to traditional views thereby shows it to have been unjustly neglected.

Much of the discussion to follow will depend on the distinction between object dependent and non-object dependent conventions.³ Imagine a small island community that has the odd convention of always driving on the side of the road that a prominent person, Bob, drives on. The focal point of the coordinating behavior of the islanders involves a concrete individual, i.e. Bob. Call any convention where the parties to the convention similarly attempt to⁴ coordinate their behavior with reference to a concrete, individual entity an object dependent convention. Most conventions are not object dependent in this way; the convention of driving on the left hand side of the road in the UK is not object dependent as it is not formulated in terms of a real world entity. Rather the convention identifies the correct side of the road to drive on as the side meeting the descriptive condition of being the left hand side of the road.

The CSR view portrays our naming conventions as object dependent, i.e. as formulated in terms of speaker’s reference to individuals. In this way the convention governing ‘Krugman’, i.e. ‘use “Krugman” to speaker-refer to Krugman’ is formulated in terms of a concrete individual, namely Krugman himself.⁵ In construing our naming conventions as object dependent, the CSR view denies that our naming conventions involve non-object dependent criteria like descriptive conditions, etc. as their focal point of coordination. Note that adopting the CSR view of semantic reference does not force one to adopt any particular view concerning the semantic content of names. In this respect it is similar to Kripke’s causal theory. While Kripke’s theory is associated most closely with Millianism about semantic content, and while it may well be argued that the Kripkean should also be a Millian about semantic content,⁶ there is nothing in the theory itself that absolutely forces the theorist to do so. The same goes for the CSR view. The CSR theorist could - and arguably should - adopt Millianism about semantic content. The CSR view itself, however, would also be logically compatible with predicativism about semantic content, i.e. the CSR theorist could hold that names are meta-linguistic predicates so that the content of a name N is ‘bearer of N’.⁷ The adoption of a CSR view of semantic reference underdetermines the theorist’s choice concerning a theory of the semantic content of names, and so the topics are best treated separately.

3 The relation of the CSR view to descriptivism

Descriptivism, traditionally construed,⁸ is the view that an object o is the semantic referent of an utterance of a name N if, and only if, o uniquely satisfies some descriptive condition that the utterer of N attaches to N. One major way in which the CSR view differs from traditional descriptivism is that it rejects the view that semantic reference depends on the mental grasp that some specific, individual utterer of a name has on the object that they wish to say something about. Rather the CSR view claims that semantic reference is determined by interpersonal, social conventions that depend on what members of the relevant coordinating community use the name to speaker-refer to. Construing semantic reference in terms of speaker’s reference implies that semantic reference will, ultimately, depend on what the members of the coordinating community ‘have in mind’ when using the name. This, however, has very little do do with what any individual speaker has in mind; individual deviation from the social standard does not affect semantic reference as the name still semantically refers in virtue of the social standard.

The CSR view can capture much of what is plausible in descriptivism as it need not deny that name-usage can be guided by descriptive criteria. To illustrate this, consider how those who follow an object dependent convention would mentally represent it. Imagine an experiment where subjects’ ability to coordinate their actions is being tested. The experimenter has placed three targets some distance away from the subjects and handed each subject a ball. Every time a whistle goes the subjects have to immediately throw the balls at the targets. If a subject hits the target that is hit by most subjects, then that subject wins a cash prize, and this prize increases in size the more subjects hit the target that is hit by most subjects. Such incentivization renders the context of interaction a game of pure coordination⁹ as each subject should aim to hit the target that everyone else is trying to hit.

We can expect that, either in virtue of natural salience or accident, some target will be hit more than others in one of the first rounds. This target then becomes salient to future rounds of the game, and soon a convention¹⁰ to try and hit this target should emerge. Suppose now that the subjects settle on hitting a specific target, and that it is the target on the left. Stipulate that the targets are now moved between rounds. After some initial confusion, things settle down with everyone trying to, in every round, hit the target that was on the left when the convention arose, irrespective of whether it happens to still be on the left or not. If we call this target K, we can say that the subjects are following the convention ‘when the whistle goes, try and hit target K.’

The convention ‘when the whistle goes, try and hit target K’ is an object dependent convention, i.e. a convention where a real-world object is the focus of the coordinating activity. Subjects, of course, need not have adopted an object dependent convention; they could have adopted a convention which advises them to try and hit whichever object happens to be on the left in each round. Given, however, that their convention is object dependent, this has important implications for how subjects can mentally represent the conventions that they follow. Subjects need to have some conception of target K that allows them to follow the convention, but there is no one mental representation of K that is uniquely preferable. This implies that the de dicto content of the mental representations used by the subjects need not coincide. Stipulate that t is the time the convention stabilized. Now, using Kaplan’s handy dthat¹¹ operator, one subject may mentally represent the convention to be followed as ‘when the whistle goes, try and hit dthat [the target that was most popular at t]’, another may follow the rule ‘when the whistle goes, try and hit dthat [the target that was on the left at t], another may follow the rule ‘when the whistle goes, try and hit dthat [the target I tried to hit at t], and so on. These subjects employ different mental representations of target K, yet it is not the case that these subjects are following different conventions. While they are following rules of action with different Kaplanian characters, these rules are just different strategies used to follow the same convention, namely ‘when the whistle goes, try and hit target K’. Such different strategies function as proxies for the convention, and should not be confused with the convention as such.

It is at least plausible to claim that a subject needs a ‘cognitive fix’ on the object relevant to following an object dependent convention. Once this is allowed, it has to be conceded that different subjects can use different conceptions of the same object in order to follow the same convention. This, however, need not drive us towards the implausible claim that each subject is following a convention of her own. By the same token, the CSR theorist can treat the different descriptive conditions, putatively involved when we communicate by using names, as mental representations that give name-users a cognitive fix on the object conventionally paired with the name, thereby allowing us to all follow the same convention. In other words, descriptive conditions like ‘the richest man in the world in 2022’, ‘the CEO of Tesla’, etc. can, joined to a dthat operator, be used to follow the object dependent convention ‘use “Musk” to speaker-refer to Musk’. In this way speakers who follow rules like ‘use “Musk” to speaker-refer to dthat[the richest man in the world in 2022]’ and ‘use “Musk” to speaker-refer to dthat[the CEO of Tesla]’ can understand themselves - and be portrayed as - following such rules as a way of following the single convention ‘use “Musk” to speaker-refer to Musk’.

In this way the CSR theorist can allow that descriptive conditions play a role in how we use names to communicate, without being tempted to say that such descriptive conditions are the semantic content of the name, and that such semantic contents differ between different speakers.¹² It is simply in the nature of object dependent conventions that they lend themselves to being followed via distinct methods.

4 The relation of the CSR view to causalism

4.1 Trivial and non-trivial causalism

For present purposes I will portray causalism as committed to two distinct claims.

• (1) For all names N, objects o and speakers S, the semantic referent of an utterance of N by S is the o that was baptized N¹³ at the beginning of an appropriately related¹⁴ causal chain from which the utterer of N inherited N.

• (2) For all names N, objects o and speakers S, the fact that an utterance of N by S semantically refers to o is non-trivially explained by the fact that S’s utterance of N is appropriately, causally linked to the event whereby o was baptized N.

The difference between (1) and (2) is crucial to understanding the relation between CSR and causalism. Note that (1) is merely the claim that baptismal uses and downstream uses are co-referential. (2) goes beyond (1) in assigning the relevant causal chain some non-trivial explanatory value in explaining why N semantically refers to o. This explanatory value is normally expressed as the claim that the causal chain determines semantic reference.¹⁵

It will be argued below that the CSR theorist should affirm that (1) will mostly¹⁶ hold true.¹⁷ This is so, as causal chains play an epistemic role in our communicative practices. The matter of the epistemic role of causal chains, however, is trivial, and so the CSR theorist should deny (2). Affirming (1) does not amount to endorse any causalism in any theoretically worthwhile sense.

In addition to acknowledging the epistemic role of causal chains, the CSR theorist may further claim that such chains serve to individuate names. The matter of name-individuation, however, is orthogonal to the matter of theories of semantic reference and so does not commit the CSR theorist to (2). The CSR theorist can still deny (2), the claim that such causal chains non-trivially explain semantic reference, i.e. that such chains determine semantic reference or amount to a ‘mechanism of reference’.

I will discuss the two issues of the epistemic and name-individuating role of causal chains in turn.

4.2 The epistemic role of causal chains

The issue concerning the epistemic role played by such chains can be illustrated by considering the view of Stine (1977), one of the few proponents of the CSR view. She proposed that, “just as Grice goes on to define sentence meaning in terms of concepts he uses in defining speaker meaning, we can go on to define the denotation of a referring expression in terms of concepts he used in defining speaker-referring” (1977, p. 324). Starting from this Gricean foundation,¹⁸ Stine defends the idea that a name semantically refers to an object among a linguistic community if, and only if, the linguistic community have a shared procedure of using the name if they wish the audience of their utterance to identify some object via the recognition of a reflexive intention to so identify the object (1977, pp. 323–325, my italics).

Stine’s view expresses the central insight behind the CSR view, namely that semantic reference can be analyzed in terms of the speaker’s reference of terms being somehow associated with names via a shared social practice. Stine, however, does not take her view to be an alternative to causalism (1977, pp. 331–335). She writes that “most of the time, when we intend an individual by [speaker] - referring to it by a certain name, we intend the same individual as our fellow language users intend by that name, and they in turn do the same, and in this we have a chain, in the history of a people with a common language, going back to the introduction of the name into the language” (1977, p. 334). She understands this point to show that her view is consistent with causalism, though claiming the existence of such chains does not render her view redundant. This can be seen from the fact that the mere existence of a causal link between the downstream use of a name and an original baptism will not do to establish semantic reference; we also need to specify what makes certain causal chains the appropriate ones to successfully transfer reference, and others not (1977, p. 331). She claims that what makes a causal chain one of the appropriate kind is that it satisfies her theory, i.e. that appropriate causal chains are ones which are made up of people who all intended to speaker-refer to the same individual. Based on the above reasoning, Stine claims that her view can be seen as a development of the causal theory. In fact the causal theory “contains much truth” (1977, p. 333); her view and the causal theory support one another (1977, p. 333).

When viewed in this way, Stine’s theory becomes just another in a long line of attempts to give an adequate specification of the causal chains relevant to the causal theory. In fact, her view seems to become very similar to that of Kripke, who requires that “[w]hen the name is ‘passed from link to link’ the receiver of the name must, I think, intend when he learns it to use it with the same reference as the man from whom he heard it” (1981, p. 96).

Stine’s view of the relation between her view and causalism is overly conciliatory. What is correct is that, on her view, as on any CSR view, there will typically exist a causal chain from the use of a name back to some baptism or event whereby the name acquired its semantic reference. If, however, we gloss her reference to a ‘shared social practice’ of using a name to speaker-refer to the same individual as other users of the name in our linguistic community, as a commitment to the idea that there exists a convention to act in this manner, then this radically changes how we should think of the role of causal chains in our communicative practices. The existence of such chains is accounted for by their epistemic import and thereby becomes theoretically uninteresting, as will be explained below. While we should still endorse the fact that (1) will mostly hold true, we should no longer endorse (2).

In defense of the idea that, once we think of semantic reference as conventionalized speaker’s reference, we should no longer endorse (2), consider the fact that the user of a name has to somehow acquire knowledge concerning the convention governing the name. Such knowledge will typically be acquired in virtue of causal contact with those who know the convention already. Acknowledging this fact does not amount to accepting anything worth being called a causal theory of semantic reference. Such chains also exist, for reasons concerning the epistemology of convention-acquisition, for non-referential linguistic elements like connectives, cries like ‘ouch’, and so on. I learned the term ‘and’ from a prior user, who learned it from a prior user, and so on; this means we can trace ‘and’ back to some event whereby it was introduced into English. But this fact does not constitute any deep truth about the semantics of connectives.

In fact, such chains will exist for all conventions. I drive on the left in the UK in virtue of having been told to do so, the people who told me to do so were similarly told to do so, or observed other people doing so, these people, in turn, were similarly told to do so or observed other people doing so, and so on. Information about the UK’s driving convention propagated outward from the event whereby it was established until it reached me, and so there is a causal chain that leads from my act of driving on the left that leads all the way back to the event whereby such a convention was established. Yet it would be absurd to propose a ‘causal theory’ of which side of the road to drive on, or of conventions in general. Independently of what further role one may claim for such causal chains, their epistemic role is trivial; such considerations merely render the fact that such chains exist an empirical fact about the epistemology of convention-acquisition, not a non-trivial explanatory fact about semantic reference.

The above reasoning indicates that the CSR theorist can easily accommodate the fact that (1) will mostly be true. Nothing about doing so amounts to accepting something worth calling a ‘causal theory of semantic reference’, as (2) can still be denied. Classifying the CSR view a causal theory conflates (1) and (2) and so reduces causalism to triviality.¹⁹$^{,}$ ²⁰

4.3 The individuating role of causal chains

It is easy to confuse the CSR view with causalism if one suffers from the related misconception that the CSR theorist has no way of explaining how distinct uses of a phonetic-orthographic type, i.e. ‘Aristotle’ used as a name of the philosopher or as a name of the shipping magnate, can be uses of distinct names.²¹ The CSR theorist, however, can easily discharge the explanatory burden incurred by their adoption of a narrow standard for the individuation of names by treating the epistemically generated causal chains as also providing a standard of individuation for names, i.e. stating that the utterance of tokens of the same name are tokens of the same type (partly) in virtue of being causally related to the same baptism.

Note, crucially, that the adoption of a causal standard of name-individuation does not amount to the adoption of a causal theory of semantic reference. The issue has been clearly stated, and these matters kept strictly separate, in a paper by Sainsbury (2015). In the course of developing his view on name-individuation he formulates as clear an example of a CSR view as one could wish for:

The “semantic reference” of a name, as used in a community, is its conventionalized, stabilized or normalized speaker-reference in the community. “London” refers to London among many speakers who live in England (and elsewhere) because it’s a conventional or stabilized or normal fact about these speakers that they use the specific name “London” (I hope you know which specific name of the genus “London” I have in mind) only if they intend thereby to refer to London (2015, p. 209).

Despite Sainsbury’s formulation of a CSR view, his paper concerns name-individuation, and not semantic reference.²² Sainsbury claims that each name is coupled with a unique act (for example, a baptism) whereby the name originated and is then spread via a historical chain of deferential use (2015, p. 196). This offers the possibility of individuating names in terms of the originating acts to which they are connected by such chains. The basic idea of causal name-individuation is not new,²³ but Sainsbury’s version is the most detailed and plausible development of the idea. Sainsbury also uses this idea to give a plausible account of the cases of unwitting reference change (2015, pp. 210–211).

Note, crucially, that Sainsbury explicitly states that he does not take his causal view of name-individuation to constitute a causal theory of semantic reference (2015, pp. 195, 210). It is easy to run the matters of semantic reference and name-individuation together, but this is a mistake. This can be seen from the fact that we can offer the exact same causal theory of individuation for linguistic items that don’t refer in the same way that names do, e.g. indexicals. Our use of ‘now’ also traces back to an originating use to which we are connected by a causal-historical chain of deferential use. The same goes for our use of non-referential terms; our use of a connective like ‘and’ similarly traces back to an originating use to which we are connected by a causal-historical chain of deferential use.²⁴

Note that Kripke himself treats the matter of individuating names as distinct from adopting a causal theory; he explicitly considers adopting a causal standard for name-individuation and decides to remain agnostic on the matter (1981, p. 9n), even while proposing his causal theory of names. The issues of name-individuation and the nature of semantic reference are distinct; the CSR theorist can relate tokens of a phonetic-orthographic type to a unique name-introducing event without thereby proposing a variant of causalism.

5 Would a CSR view that adopts a descriptivist or causal account of speaker’s reference vindicate descriptivism or causalism about semantic reference?

The CSR view, as formulated here, makes semantic reference dependent on speaker’s reference. This inevitably raises the question of how to account for speaker’s reference. Here the CSR theorist has the option of providing a descriptivist or causalist construal of speaker’s reference - and, relatedly, the option of characterizing our referential intentions as descriptive or singular - and to thereby allow such notions to determine semantic reference indirectly. This could lead to the suspicion that a developed version of the CSR view would be compatible with a descriptivist or causalist view of semantic reference and, in fact, be no more than a variant of traditional causalism or descriptivism.

It has already been shown that a CSR view that includes a descriptivist account of speaker’s reference, i.e. a view on which speaker’s reference is determined by descriptive intentions, would still not be compatible with the descriptivist views about semantic reference held by Russell²⁵ or Frege.²⁶ Both Russell and Frege are standardly interpreted as claiming that semantic reference depends on the descriptive condition that the individual utterer attaches to the name. The CSR rejects this view in favor of viewing the determination of semantic reference as interpersonal. Relatedly, the CSR view naturally leads to Millianism, whereas Russell and Frege defended both a descriptivist theory of reference and a descriptivist semantics.

The descriptivist could, of course, edge ever closer to the CSR theorist. Suppose a descriptivist abandoned the idea that semantic reference depends on the descriptive condition that the utterer of a name attaches to it, abandoned the idea that different names have unique descriptive conditions associated with them and ended up at a social version of the so-called cluster theory proposed by Searle (1958). In other words, consider a descriptivist who stated that the semantic reference of a name is determined by an appropriately weighted average of the descriptive conditions associated with the name by all members of a linguistic community. If a CSR theorist accepted a descriptivist construal of speaker’s reference, then she would have to accept that the main claim made by such a social cluster theory is correct, as far as it goes. The social cluster theory, however, is very different from the CSR view as it omits the mechanism definitive of the CSR view. For it does not claim that such descriptive conditions are only relevant inasmuch as they determine speaker’s reference or that semantic reference is determined by conventions that pair names with such practices of speaker-referring. This is rather a lot to leave out; if such a CSR view turns out to be accurate, then the social cluster theory alluded to above can hardly claim victory, for it offered Hamlet without the prince. If, however, the social cluster descriptivist included such a mechanism in his view, then I fail to see how it can still be taken to be a descriptivist view of semantic reference. On such a view, only speaker’s reference is really determined via descriptions. The real work of determining semantic reference occurs in virtue of speaker’s reference and the nature of conventions.

Alternatively, the CSR theorist may claim that speaker’s reference is only properly understood if we think about it causally. Consider what Gareth Evans called the ‘Photograph model’ of reference (1982, p. 78), which takes its inspiration from the fact that, when we ask which object a photograph is a photo of, we are asking about an appropriate causal antecedent of the photograph. The issue concerns which item causally interacted with the light that left an impression on the photographic plate, film, or was detected by the relevant photo detectors. The CSR theorist may claim that speaker’s reference is analogous; that if I see something and attempt to say something about it by using a name then, even if my conception of that thing is wrong in all sorts of ways, I can be said to have speaker-referred to that thing in virtue of it being the object which caused my attempt at communication.

Nothing stops the CSR theorist from developing a view along these lines. Something like this view has, in fact, been put forward by Devitt. Devitt (2015) expresses his view as follows²⁷:

Speaker-Designation: A designational name token speaker-designates an object if and only if all the designating-chains underlying the token are grounded in the object (2015, p. 125).

Conventional-Designation: A designational name token conventionally designates an object if and only if the speaker, in producing the token, is participating in a convention of speaker-designating that object, and no other object, with name tokens of that type (2015; p. 126).

Matters of phrasing aside, and in terms of the terminology employed here, Devitt analyzes semantic reference in terms of speaker’s reference, and speaker’s reference in terms of causal chains that are grounded in objects. Devitt, however, considers himself a causalist, and is typically glossed as a causalist, about semantic reference in the same way that Kripke is portrayed as a causalist about semantic reference. This is so, despite the fact that Devitt is more accurately portrayed as a causalist about speaker’s reference, with views about naming conventions that serve to render semantic reference causal by proxy.²⁸

Despite the fact that a CSR theorist may adopt a causalist view of speaker’s reference, it cannot be claimed that someone who does so has thereby adopted a causalist theory of semantic reference.²⁹ While this is, to some degree, a matter of terminology, speaking in this way serves to obscure the difference between those, like Kripke, whose view is straightforwardly about semantic reference, and views where the theoretically interesting claims are about a different relation entirely, namely speaker’s reference. The situation is analogous to that of the hypothetical theorist who offers a CSR view that analyzes speaker’s reference in terms of descriptive conditions. While we could classify such a theorist as a descriptivist, it is misleading to do so as, unlike the case of Russell and Frege, whose descriptivism is straightforwardly about semantic reference, they can claim descriptivism only by proxy. On such a view, all the theoretical work is being done by the theory of speaker’s reference; the same holds true of Devitt.

Assimilating descriptivist or causalist views about speaker’s reference to descriptivist or causalist views about semantic reference is pernicious, as the constraints on a theory of speaker’s reference are very different from those on a theory of semantic reference. For a start, the basic data to be explained in a theory of speaker’s reference are attributions of referential intentions, not, as is the case in a theory of semantic reference, intuitions concerning the truth-conditions of utterances. This implies that, if the issue of speaker’s reference were to become the new battleground for causalist and descriptivist views, we can expect the debate to look very different from the traditional one concerning semantic reference.

6 Two objections

Below I will discuss the two most common objections I have heard offered against the CSR view. I will merely argue that the CSR theorist has ample means at her disposal to accommodate the most obvious challenges and/or that the CSR view does not have any difficulty inherent to it that at least one extant, respectable theory does not.

6.1 Objection 1: the CSR view merely passes the buck as semantic reference is explained in terms of speaker’s reference, which is left unexplained

Firstly, it is no objection against a view that analyzes semantic reference in terms of speaker’s reference that it merely pushes the mystery back to another level. The task is to locate the mystery where it actually belongs. This is progress, even if it does not solve all problems at once.

Secondly, some degree of buck-passing is also present in both descriptivist and causalist views. The descriptivist explanation of the semantic reference of proper names presupposes a relation of satisfaction that holds between elements in the world and descriptive conditions. This relation of satisfaction is what ultimately ends up doing the work of relating names to things, and so requires explanation. Kripke, similarly, explicitly admits that his causal picture of the semantic reference of proper names does not serve to fully analyze the notion of reference, but presupposes it.³⁰

There is no denying that the CSR view does depend on it being the case that our notion of speaker’s reference is coherent and that it is more fundamental than semantic reference. It may, of course, be doubted whether speaker’s reference really is more fundamental in the required way. One way of doing so would be to argue that we cannot make sense of mental content independently of making sense of semantic content, or make sense of speaker’s reference independently of making sense of semantic reference. Note, however, that the CSR view’s treatment of speaker’s reference as more basic than semantic reference is analogous to the way in which Gricean views portray the speaker meaning of sentences as more fundamental than their semantic meaning. However, this fact has not been generally thought to constitute a knock-down argument against the Gricean view of sentence meaning. Similarly there is no reason why it should prevent the CSR view from entering the debate concerning names.

6.2 Objection 2: the CSR view cannot accommodate names without bearers

Non-referring names are not, at first glance, a problem for the CSR view. Those who used ‘Vulcan’³¹ to try to speaker-refer to the planet causing the perturbations in the orbit of Mercury failed to speaker-refer to anything; as the CSR view depicts semantic reference as dependent on speaker’s reference, the CSR view correctly portrays ‘Vulcan’ as non-referring. The situation here is similar to that of the causal theory, which also, correctly, portrays ‘Vulcan’ as non-referring.

The question as to the semantic content of ‘Vulcan’ inevitably arises. Here, as discussed at the start of this paper, the CSR view seems to naturally lead to Millianism about the semantic content of names, yet does not force us to adopt Millianism. Given the amount of conceptual independence between the theory of semantic reference and the theory of semantic content I will not pursue the issue here, but do note that the problem concerning non-referring names has not disqualified Millianism from serious discussion. So the fact that the CSR view seems to naturally pair with Millianism cannot be used to disqualify it from the debate concerning names.

There does, however, exist an issue, closely related to the matter of semantic content, that the CSR theorist does need to address. The CSR view portrays name-governing conventions as having the form ‘use “N” to speaker-refer to N’. Can we portray the convention governing a non-referring name like ‘Vulcan’ in the same way?

Here, again, the CSR theorist has different options available. A first option would be to portray the CSR view as a view about ordinary names, and allow for a special class of ‘descriptive names’.³² On such a view we can treat non-referring names as a sub-class of descriptive names and construe the convention governing ‘Vulcan’ as involving the descriptive condition of being the planet causing the perturbations in the orbit of Mercury.

While there is much to be said for proceeding in this way, we should not complicate semantic theory in this way if it can be avoided. The considerations concerning object dependent conventions discussed in this paper suggests a distinct option that preserves parity of form for conventions governing non-referring names. We can treat the descriptive condition of being the planet causing the perturbation in the orbit of Mercury as the type of condition used to formulate a strategy used in order to follow an object dependent³³ convention, but, as is the case with other such conditions, not constitutively involved in the convention itself. While the association between this descriptive condition and ‘Vulcan’ will be ubiquitous in a way that is not the case with a typical name-governing convention, such ubiquity is due to the prosaic fact that the descriptive condition of being the planet causing the perturbation in the orbit of Mercury is the only substantive condition available to formulate a strategy for following the convention governing ‘Vulcan’. This fact, then, also serves to explain how this descriptive condition is readily available for pragmatically facilitated communication when using ‘Vulcan’, yet without involving this condition in the convention governing ‘Vulcan’ itself.

The above reasoning would be compatible with claiming that, while those who introduced the name ‘Vulcan’ attempted to create an object dependent convention, they failed to create a convention at all. It seems preferable, however, to say that there does exist a convention of the form ‘use “Vulcan” to speaker-refer to Vulcan’, and then, if we adopt a general Millianism about the semantic content of names, to then say that the convention is partially empty as ‘Vulcan’ fails to refer.

I am partial to the above construal of the convention governing ‘Vulcan’. However, if the idea of a partially empty convention is considered to be objectionable, do note that it is not unavoidable. Sainsbury has pointed out out that, as speaker-referring is matter of having a certain intention, we can treat the problematic occurrence of the non-referring name as occurring within the scope of the referential intention, and so our statement of the convention need not be read as partially empty (Sainsbury, 2018, p. 176).

Adjudicating between the above options will take us too far from present concerns.³⁴ For now, however, note that the CSR theorist is not without options when it comes to dealing with names without bearers, and so the existence of such names cannot constitute a knock-down argument against the CSR theorist.

7 The relative neglect of CSR views

Given the prominence of the Gricean construal of sentence-meaning in terms of speaker-meaning (Grice, 1968), the relative neglect of CSR views of the semantic reference of names should strike us as puzzling. The relevance of Gricean views to understanding the meaning of sentences arises in virtue of the intuitive idea that speakers typically utter a sentence with the intention to bring some proposition that the speaker has in mind to the attention of their audience. It is similarly intuitive to claim that speakers typically utter a referential, sub-sentential element, for instance a name, with the intention of bringing some individual that the speaker has in mind to the attention of their audience. The relevance of conventions to the understanding of sentence meaning arises in virtue of the plausible claim that the meaning of a sentence is determined by convention. It is similarly plausible to claim that the referent of a name is determined by convention. In fact, the nature of CSR views allow them to avoid the most common complaint against Gricean views of sentence meaning, namely that the existence of an infinite number of sentences commits the Gricean to the existence of an infinite amount of conventions. Given that CSR views of names concern a syntactically simple, sub-sentential entity, as opposed to sentences as such, this concern does not affect such views.

How, then, to explain the relative neglect of the CSR view? One part of the answer must be that it is easy to make the mistake of thinking of the CSR view as a variant of a traditional view, and so relatively uninteresting. This mistake is easily made if one thinks that believing that individual speakers attach descriptive conditions to names amounts to accepting descriptivism, or that accepting the epistemic or individuating role of causal-historical chains amount to accepting a causal theory of semantic reference, or that giving a descriptivist or causal view of speaker’s reference also counts as offering a descriptive or causal view of semantic reference.

A distinct reason for the relative lack of interest in such Gricean ways of thinking about semantic reference could be a background assumption that facts about the nature of conventions are too general to be of help when adjudicating between rival theories of semantic reference. Theorists recognize the conventionality of our use of names when employing formulations like ‘our name-using practices’ and so on, but typically without explicitly considering general matters concerning the nature of conventions. Our degree of reliance on the Russellian ‘method of cases’, whereby we try and construct theories that match our intuitions concerning specific utterances, further serves to keep our glance away from more general matters concerning the nature of conventions.

Another potential source of misunderstanding would be the failure to recognize that a Grice-inspired construal of semantic reference can portray the relevant conventions as object dependent, and so be compatible with Millianism about the semantic content of names.

I suspect, however, that the relative neglect of the CSR view is due not only to substantive confusions, but also to sociological factors. This becomes clear if we compare the history of theorizing about the semantic reference of names with the history of theorizing about sentence meaning. The history of theorizing about semantic reference has mainly taken the form of a dispute between various descriptivist views (that take their basic inspiration from Russell or Frege) and causal views that derive from Naming and Necessity. Given the form that such theories take, it is not obvious how such views are supposed to relate to issues more prominent in theorizing about sentence meaning, i.e. issues concerning speaker meaning and conventions.

Furthermore, theorizing about semantic reference has generally been driven by concerns about how to formalize the logical structure of natural language, whereas theorizing about sentence meaning has been mostly driven by the attempt to understand communication. These projects have mostly been carried out independently from one another with few attempts being made to determine whether the popular views offered on these topics are consistent.³⁵

Even where the same theorist treated both issues they have been treated as being largely independent. Consider, for instance, Lewis’s defense of a causal version of descriptivism (1984, pp. 226–229). One would think that one’s views about sentence meaning would strongly constrain one’s views about a sub-sentential element like names; yet Lewis puts forward causal descriptivism (Lewis, 1984) without making any reference to his work on sentence meaning, or conventions in general. This is done despite the prima facie plausibility of simply adapting his Gricean view about sentence meaning in such a way that it yields some neo-Gricean view about names. The fact that the theory of names and the theory of sentence meaning have been treated as so strongly distinct raises the possibility that, while some theorists may have dismissed Grice-inspired views on philosophical grounds, many may simply have overlooked the possibility of treating names in a neo-Gricean way due to the fact that the theory of names and the theory of sentence meaning sociologically count as independent topics.

The CSR view is, of course, far from bulletproof. One may be uncomfortable with the very idea of analyzing semantic reference in term of intentions to speaker-refer, or worry about how the notion of speaker’s reference is to be analyzed. One may also wonder how the details of the CSR view should be worked out so as to account for fictional names and empty names. Furthermore, the CSR stands in need of a precise theory of the conditions under which a specific convention obtains. But, given that there are powerful objections to just about every theory of reference available, and, more importantly, that the CSR view is simple, intuitive and fits well with how most theorists think of sentence meaning, it is remarkable that this kind of view receives little discussion in the literature. What is more, the CSR view combines the virtues of descriptivism and causalism; it explains how semantic reference can depend on mental states in general, but, in virtue of the fact that the existence and nature of communal conventions are independent of any specific party to a convention, also explains how the semantic reference of a name is independent of the mental state of the utterer. Hence, while it would be hubristic to be confident about the truth of the CSR view as presented here, it seems clear that it presents a ‘picture’, as Kripke would put it, of semantic reference that has been neglected for too long.