Unspeakable names

They call me ‘Bell’, They call me ‘Stacey’, They call me ‘her’, They call me ‘Jane’, That’s not my name, That’s not my name.

—The Ting Tings.

Abstract

There are some names which cannot be spoken and others which cannot be written, at least on certain very natural ways of conceiving of them. Interestingly, this observation proves to be in tension with a wide range of views about what names are. Prima facie, this looks like a problem for predicativists. Ultima facie, it turns out to be equally problematic for Millians. For either sort of theorist, resolving this tension requires embracing a revisionary account of the metaphysics of names. Revisionary Millianism, I argue, offers some important advantages over its predicativist competitor.

1 Introduction

There are names which cannot be spoken, at least on certain very natural ways of conceiving of them. Likewise, there are names which cannot be written down, at least in certain contexts. The reasons for this are not—as either the medieval mystics or the Harry Potter novels would have it—that such names have unexpected and potentially harmful powers. Rather, the reasons for this are perfectly mundane yet still of significant philosophical interest.

Here is the plan for what follows. In Sect. 2, I offer some examples and explain how these raise a very general puzzle about what names are. Since the relevant cases were first introduced by Aidan Gray in his (2015), I will call them ‘Gray Cases’. In Sect. 3, I show how the puzzle raised by Gray Cases poses a serious challenge for predicativists about proper names. Then, in Sect. 4, I develop an analogue of this challenge for Millians.¹ In Sect. 5, I argue that the best prospect for each sort of theorist to overcome this challenge lies in embracing a revisionary theory of what names are. Ultimately, I contend that the revisionary Millian theory is preferable to the revisionary predicativist one. In Sect. 6, I explore what this might tell us about the nature of words in general. Finally, Sect. 7 concludes with some broader methodological reflections.

2 Gray cases

Let’s start with names that cannot be spoken.

Consider a situation in which there are two gentlemen, both of whose names are spelled , drinking champagne. One of these Ralphs is an upper-class Englishman, and so pronounces his name /ref/; the other is from a less rarified background and pronounces his name .² I write:

My inscription of (1) is true, but there is no way of felicitously uttering (1) in this situation. In fact, I would venture so far as to say that there is no way of uttering (1) truly in these circumstances.³ Whichever way one tries to pronounce the name, it will only properly apply to one of these Ralphs.

Now consider the converse: names which are pronounced in the same manner, but which are written differently. My own name can be used to illustrate this phenomenon. Suppose that there are four individuals, each of whose name is pronounced , playing Settlers of Catan. Suppose further that no two of them spell their names alike.⁴ One can say, truly:

⁵

But there is no way of transcribing (2) truly, at least not in ordinary English. (2) can be uttered, and it can be uttered truthfully, but it cannot be written down.⁶

One natural explanation why we can write (1) and utter (2) truthfully is that there is some property that these two or four individuals share with each other, respectively—specifically, the property of having the same name, at least in these contexts.⁷ But what might names be, such that sharing a name in speech needn’t entail sharing it in writing, or vice-versa? In other words, what might names be such that they can sometimes be written, but are unutterable—or vice-versa?⁸

To make the issues here more explicit, consider a simple view according to which names are to be identified with sets of orthographic forms.⁹ In that case, names will be individuated in terms of their orthography. But, if names are to be individuated in terms of their orthography, then there should be no way of uttering (2) truly—since there would be four distinct names in play.

What if we were to treat names as sets of phonological forms instead? In that case, it becomes easy to explain why we can utter (2) truly but hard to see how the written version of (1) could possibly be true. For, if we individuate names phonologically, there will be two names in play.

In fact, things are even worse than this. If we treat names as sets of orthographic forms, then we look to be in no position to explain why (1) cannot be uttered felicitously. Since /ref/ and are transcribed in one and the same manner, this view predicts that there is just one name ‘Ralph’—a name with two acceptable phonological manifestations. On the other hand, if we treat names as sets of phonological forms, then we look to be in no position to explain why (2) cannot be written. For each of these orthographic forms shares a common phonological correlate.

Perhaps then names are best understood neither as sets of phonological forms nor as sets of orthographic forms, but rather of both. Sadly, this won’t do either. The problem is that this view is even more fine-grained than either of those considered above; now any difference in spelling or pronunciation will make for a different name. This would make it impossible to truly write (1) or utter (2). In the former case, the difference in pronunciation would entail that these two individuals differ in their names; in the latter, the differences in spelling would entail that four distinct names are in play.

On reflection, part of the issue here is that the three views just considered, while certainly intuitive, are simply bad views.¹⁰ If names were sets of orthographic forms, then any language that lacked a system of writing—like most in human history—would thereby lack names. If, on the other hand, names were sets of phonological forms, then any purely written language—hardly common, but certainly conceivable—would lack names. And if names were sets of both orthographic and phonological forms, then any language which lacks either a system of writing or of vocalization would turn out to lack names as well. None of these implications looks palatable.

To be clear, our task here won’t be to work out what the best thing to say is about Gray Cases per se, but rather to show how these cases pose a particular challenge for two prominent views of the semantics of names: predicativism and Millianism. Then we’ll turn to how proponents of each view might try and respond to this challenge. This dialectic might initially seem odd, given that neither Millianism nor predicativism aspires, explicitly at least, to be a view about the ontology of names. But, as we will see, each view brings with it substantive commitments regarding the individuation of names. Ultimately, I will argue that the Millian can offer a more natural and theoretically satisfying account of Gray Cases than can the predicativist.

3 Predicativism

First discussed in linguistics by Sloat (1969) and introduced to philosophy shortly thereafter by Burge (1973), predicativism begins with the thesis that names in natural languages like English denote properties, not individuals.¹¹ So, contrary to both appearances and earlier philosophical dogma, a bare use of a name like ‘Rei’ denotes a meta-linguistic property along the lines of being called Rei.¹² Predicativists will then explain the apparent-referentiality of most bare uses of names by positing that a silent, or aphonic, determiner appears in front of the name at the level of syntax or logical form. Some predicativists take this silent determiner to function more like an unpronounced ‘the’, others more like an unpronounced ‘that’. For our purposes, we can leave this matter undecided.

Putting all of this together, predicativists will analyze the meaning of (3) along the lines of (4):

¹³

Alternatively, if the reader prefers to posit a covert ‘that’, they should feel free to insert their preferred analysis of complex demonstratives instead.

Now let us ask: what does it take to satisfy a given instance of the property-schema being called N? A helpful first step, I take it, will be to follow (Graff Fara, 2011, p. 493) in distinguishing between someone’s being called N and this name being used to regularly address this person. As Graff Fara puts it, ‘To be called Willard is to have ‘Willard’ as a name’ (Ibid., p. 493). This, Graff Fara makes clear, does not entail anyone’s actually ever actually calling Willard ‘Willard’ as opposed to ‘Van’.

Still unanswered, however, is the question of what it is to have ‘N’ as a name. Plausibly, the answer to that question is going to hinge, at least in part, on what we think names are. If names are sets of orthographic forms, for instance, then having ‘N’ as a name is presumably going to involve being related in some to-be-specified way to a particular set of orthographic forms. If names are sets of phonological forms, then having ‘N’ as a name is going to involve being related to a set of phonological forms. And so on.

Even without getting into the details of this to-be-specified relation, we can see how Gray Cases are going to pose a challenge for at least some of the natural options. If this to-be-specified relation takes sets of orthographic forms as one of its relata, then it is difficult to see how (1) could be true when written but not when spoken. For each Ralph plausibly bears most, if not all, of the same relations towards the orthographic form ‘Ralph’—meaning that both have the same name, in every context and regardless of whether that name is written or spoken. If this relation takes sets of phonological forms instead, then the inverse problem arises: how to explain why (2) can be uttered truly, but cannot be written down. For orthography, on this view, is stipulated to be irrelevant to whether one is . If we try combined sets, we can bypass this problem. But then we are forced to deny that there is any context where our two Ralphs or four share a name.

Intuitively, what we want is a view on which the extension of having ‘N’ as a name—and, more saliently for our purposes, of having the same name—can vary with context. The question is how to allow for such variation in a principled manner, one which respects some intuitive constrains on the sameness of names. For instance, following (Gray, 2015, pp. 119–120), I take it that one plausible desideratum for any account of what names are is that, if some phonological or orthographic form is in the set corresponding to ‘N’ in a given context, then any sufficiently similar form should also be in that set. This will ensure that dialectical variants are grouped together, in most contexts at least. Likewise with the orthographic forms ‘Eliot’ and ‘${\textbf {Eliot}}$’.

At this point, one might be tempted to try and use the notion of sufficient similarity to offer an analysis of having the same name—then leverage that analysis to offer some insight into having ‘N’ as a name. So, for example, we might try claiming that, in a context, two sets of orthographic and/or phonological forms count as the same name if and only if they are sufficiently similar. Two individuals will then have the same name if and only if they bear the right sorts of relations towards two sufficiently similar sets of forms.

As Gray (2015, p. 120) points out, however, this sort of view faces a problem. Consider a context in which we are discussing my favorite names, which happen to be ‘Helen’ and ‘Alfred’. In such a context, there is a distinct sense in which the corresponding set of phonological forms are very similar to each other: each is one of my favorite sets of phonological forms. Yet this kind of similarity is never sufficient, it would seem, to make it the case that these phonological forms (or any other set derived from them) will count as being the same name. What we need is an account of what constitutes the relevant sorts of similarity when it comes to grouping together sets of orthographic and/or phonological forms into names.

I am skeptical that such an account can be offered. Thankfully, there is a more promising line of response to the question of what it is to have ‘N’ as a name: start by asking ‘In virtue of what do we associate certain sounds and marks with particular individuals, relative to a context?’ Whatever the answer turns out to be, hopefully it will deliver the result that, in most contexts, intuitively similar orthographic and phonological forms are lumped together into names. Only now this surface-similarity is revealed to be a mere byproduct of the real explanation of what it is to have ‘N’ as a name.

Let us call this general approach ‘contextualism’ about names. On this sort of view, the extensions of being N, having the same name, having ‘N’ as a name, and being called N will all vary, in a coordinated manner, relative to the context. The outstanding challenge for the contextualist about names is to fill in the details of the relevant relation in a way that allows them to account for Gray Cases. We’ll return to this challenge in §5. First though, it will help to get a contrasting approach to the semantics of names on the table, one which imposes some rather different constraints on the individuation of names. To that end, I turn now to Millianism.

4 Millianism

Millians start with the claim that names are, fundamentally, referring expressions; bare names denote individuals. In contrast to other sorts of referentialists about names, e.g. Fregeans, Millians also hold that names refer ‘directly’. In other words, according to Millians like Kripke (1980), the meaning of a name is exhausted by its referent.¹⁴ There is a metasemantic story to be told about why any particular name refers to whatever individual it does; for Kripke and many others, this runs in terms of chains of use leading back to some initial baptism. But the semantic story is simple: ‘Ralph’ refers to Ralph, ‘Rei’ refers to Rei, etc.

While perhaps not immediately obvious, the Millian view has substantial implications for the individuation of names. That’s because, according to Millians, it is names themselves that refer, not uses or utterances of names. But since the same phonological and orthographic forms can be used to refer to different individuals on different occasions—there are other Eliots in the world whose names are written and pronounced in the exact same manner as my own—names cannot be individuated according to these forms.

Few Millians have explicitly faced up to this consequence, the notable exception being Kaplan (1990).¹⁵ There, Kaplan proposes to develop the Millian’s favored metasemantic story, a story that runs in terms of chains of use, into a metaphysical story as well: names are to be individuated partly in terms of the chains of use standing behind them.

In more detail: start by considering a particular act of using a name. This, Kaplan claims, will involve (i) drawing on an unambiguous representational vehicle stored in the speaker’s mental lexicon (call this a ‘mental name’), and (ii) externalizing that vehicle by producing particular sound or mark, something which enables both communication and the initiation of others into the practice of using the relevant name. In the first instance, we group these externalizations together not in virtue of their superficial similarities, but in virtue of their having descended from the same ancestor—their being linked back to that ancestor via a series of externalizations, internalizations, and periods of storage in a mental lexicon. The crucial thing is for each of these externalizations to count as a repetition of the relevant name (Ibid., pp. 102–105).¹⁶

So much for uses of names, but what about names themselves? On Kaplan’s view, names are continuants composed of stages of different sorts, i.e. utterances, inscriptions, and mental names (Ibid., p. 98).¹⁷ This stage/continuant thesis has come in for heavy criticism, however.¹⁸ Thankfully, for our purposes, it won’t much matter whether it’s right. That’s because there are alternatives still very much in the Kaplanian spirit, but which jettison that particular aspect of Kaplan’s view: for example, we might accept Kaplan’s claims (i) and (ii) about uses of names but, instead of treating names as continuants, simply identify names and mental name-types. Crucially, on this latter sort of view, we need to be careful to stipulate that, for a mental name to count as such, it must be expressible—that is, it must be associated, by a given speaker at a given time, with some sort of phonological and/or orthographic features, regardless of what exactly they are.

Whichever version of the view one prefers, it invites the following objection: to many, it will seem just plain obvious that multiple individuals can bear the same name.¹⁹ If that’s right, if it’s a basic fact about names that they can be shared, then in adopting anything like the Kaplanian view we would seem to be going seriously off-track.

Unsurprisingly, Kaplan has a story to offer in response to this concern: when we judge that two individuals bear the same name, what we are latching onto is the fact that names can be grouped together into name-genera. Consider the given name ‘David’. Prima facie, this name is shared by David Kaplan and David Lewis. On Kaplan’s view, these two individuals—David and David—bear distinct ‘common currency’ names but share a ‘generic name’. Common currency names always pick out particular individuals; they refer to at most one person or thing. Generic names, in contrast, allow us to talk about classes of common currency names, sorted, initially at least, in terms of common features of their externalizations (Ibid., p. 111).²⁰

With this distinction in hand, we can now say something more concrete about the Gray Cases we began with: while no two individuals can share a name in the common currency sense, they are able to in the generic sense. So the Kaplanian predicts that (1) and (2) will be true when the relevant individuals’ names all belong to the same name-genus and false when they don’t. In other words, on the Kaplanian analysis, (1) and (2) don’t involve unspeakable or unwritable names at all, at least in the common currency sense. What (1) involves is a generic name that cannot be verbalized, (2) a generic name that cannot be written down.²¹

The question we now face is: when exactly do two (common currency) names belong to the same name-genus? Suppose that (1) is true iff there are two people drinking champagne, each of whose name belongs to the name-genus ‘Ralph’. Then, for (1) to be true when written down, it looks like the relevant name-genus will have to be individuated orthographically. But analogous reasoning, in conjunction with the truth of (2) when spoken, pushes us towards the conclusion that name-genera should be individuated phonologically. So we seem to be back to our earlier dilemma, now evinced with respect to the property being a member of the name-genus N rather than with respect to being called N. One could be forgiven for thinking that scant progress has been made.

I don’t think that things are quite so bleak as this, however. For it seems to me that the Kaplanian can take a cue here from recent work in the philosophy of biology and embrace genera pluralism—and that this may offer a way out of this problem. We will consider this strategy in the next section.²²

5 Names as social kinds v. genera pluralism

So far, we have clarified the challenges that Gray Cases pose for two prominent theories of names—predicativism and Millianism—and considered some first steps that each might take to overcome those challenges. In this section, I will expand on those first steps and then compare the resultant views.

5.1 Names as social kinds

Contextualism about the individuation of names was the thesis that what it is to be a name can vary, often quite markedly, from context to context. In other words, in any given context, a particular set of orthographic forms, phonological forms, or combination of the two will count as a name, whereas others will not. Presumably, these forms will bear some sort of socio-historical relation to each other; otherwise, context would seem to be selecting at random. To better evaluate the view, we need therefore to come to grips both with what this socio-historical relation might be and how this contextual selection might work. My suggestion is to try and get a better grasp of both of these by treating names on the model of social kinds.

Recall that predicativists take the basic semantic function of names to be: classifying individuals. This stands in contrast to Millians, who take the basic function of names to be: referring. The question we now face is: how do names classify individuals? What’s clear is that there is no natural property in the vicinity to which we might appeal; on the contrary, names seem to classify objects in a more or less arbitrary manner, a bit like how grammatical gender carves the objects in the world into a relatively small number of classes.

By way of comparison, consider briefly the case of ‘real’, as opposed to mere gramatical, gender. According to Haslanger (2000, p. 39), what it is to be a woman is to be the object of systematic oppression as the result of perceived features of the body presumed to be evidence of a particular role in sexual reproduction. This means that the realizers of womanhood—that is, who counts as a woman—will vary from context to context and culture to culture. In fact, one and the same person might count as a woman in one context but not in another as various background factors shift.

What matters for our purposes is not whether this is the right thing to say about gender, but rather the underlying approach to the metaphysics of social kinds: first, we try and identify some coherent social role (perhaps an obvious one, perhaps a less obvious one) which unifies the kind. Then, we look to see which sorts of things in the world satisfy this social role.

Let’s try to identify a plausible social role for names. One possibility is that names allow us to refer to individuals. So, to be a name, relative to a context, is to be the sort of thing that can be used, in that context, to refer to at least one individual (whether or not they are present in the context). But not only does this proposal run counter to the spirit of predicativism, it also fails to distinguish names from other singular terms like ‘that’ or ‘she’—which pretty clearly aren’t names. Thankfully, there is a more promising option.

This runs as follows: to be a name, relative to a context, is to be the sort of thing that can be used to vocatively address an individual.²³ More specifically, names are sets of phonological and/or orthographic forms marked out by these forms being ways of potentially vocatively addressing a minimal class of individuals—not collectively, but rather as individuals. Importantly, this minimality condition prevents disjuncts like ‘Rei or Ralph’ from counting as names. I assume that such disjuncts are not proper names.

A few clarifications are in order. First, this notion of vocative address is not just a disguised notion of reference. To vocatively address someone or something is, I take it, to put that thing in the addressee position, to make it the target of subsequent uses of the second-person pronoun.²⁴ Perhaps putting something in the addressee position necessarily involves reference, but that would require further argument.²⁵ Second, we can vocatively address inanimate objects. I do this when I name my computer ‘Wall-e’ and then curse Wall-e for crashing. Third, we can vocatively address individuals even when they aren’t present. For example, thinking about my far-away philosophical nemesis, I might exclaim ‘I’ve got you now!’ Fourth, and finally, while the potential for vocative address depends on the socio-historical relations that obtain between orthographic and phonological forms and individuals, this needn’t be so simple as being a form that people are routinely addressed by.²⁶ We are hardly ever addressed by our full legal names, yet these names still retain their vocative potential—as witnessed by legal proceedings, scoldings by disappointed grandparents, etc.²⁷

In its present form, the vocative potential view of names faces two main problems. First, as it stands, the proposal predicts that there is no true reading of (1). That’s because, if we consider the class picked out by the orthographic form ‘Ralph’ alone, there is a clear way of deriving a more minimal class of vocative address: combine the orthographic form ‘Ralph’ with either the phonological form /ref/ or the form . So the view predicts that the orthographic form ‘Ralph’ alone never counts as a name—for it fails to pick out a minimal class of vocative address.

This can be rectified as follows: allow that to be a name, relative to a context, is to be the sort of thing that can be used to address a sufficiently minimal classes of individuals. What counts as sufficiently minimal will itself vary relative to the context, though within limits. In no context will the combination of phonological forms associated with ‘Rei’ and ‘Ralph’ count as a name. In contrast, there will be contexts where the purely orthographic form associated with ‘Ralph’ picks out a sufficiently minimal class of individuals to count as a name. The context in which (1) is written down, for instance, is such a context. And that is why the written version of (1) is true: in this context, both Ralphs are called ‘Ralph’, as the name ‘Ralph’ is individuated purely in terms of its orthography. So this modification to the vocative potential view of names not only allows us to account for Gray Cases, it also helps us locate where and how context figures into contextualism about names.

The second problem is more difficult. There are any number of words which clearly aren’t names, but which are commonly used as means of vocative address: ‘you’, ‘boss’, ‘professor’, ‘dear’, ‘asshole’, etc. Some of these can perhaps be differentiated from names in virtue of their syntactic features. For instance, one cannot put a determiner before ‘you’ while preserving grammaticality.²⁸ It is unclear, however, whether this syntactic strategy will generalize. If not, then we again face the question of whether this social role—or even this social role with some reasonable supplementation—will suffice to allow us to home in on just the names.

Perhaps the contextualist will be willing to accept that, in certain contexts at least, words like ‘boss’ or ‘asshole’ do indeed count as names. But this, I take it, is no small bullet to try and bite. Offering a persuasive response to this second problem is, I take it, a serious outstanding challenge for the contextualist about name individuation—and hence for the predicativist.

5.2 Genera pluralism

According to Kaplan, generic names are ways of clustering together common currency names in terms of certain common features of their outward expression. Kaplan focuses on just phonological and orthographic similarity. But, as we have seen, what we really need is a way of understanding generic names that is more flexible than this, one capable of treating e.g. phonological similarity as sometimes sufficient for two common currency names to be grouped together, but sometimes not.

To achieve such flexibility, I suggested taking a cue from Kaplan’s appeal to genera. Recently, a number of philosophers of biology (and no small number of biologists besides) have urged us to give up on the idea that there is just a single species to which any individual belongs.²⁹ Rather, these philosophers have claimed that there are a number of related species-concepts—phylogenetic as opposed to ecological, for instance—which largely overlap in how they classify things, but which can also come apart in certain circumstances. Likewise for genus-concepts. The suggestion is that we can make progress here by treating name-genera in a similar manner.

Here is how such an account might run. Names, we might posit, can be grouped together via at least three distinct sets of properties: phonological properties, orthographic properties, and socio-historical properties. Combinations of these properties will sometimes form what we might call ‘homeostatic property clusters’, or sets of features such that the presence of one sort of feature favors the presence of the others as well (Boyd, 1999, p. 143).³⁰ /ref/ and will naturally be grouped together under a homeostatic property cluster which combines orthographic and socio-historical properties, but not one which adds phonological properties to the mix. In contrast, the four permutations of naturally fall under a property cluster which combines phonological and socio-historical properties, but not one which is extended to include orthographic properties. Such homeostatic property clusters represent, I would suggest, our best account of name-genera.

What determines which sort of genus is at issue in a given context? That is obviously a difficult question, but we can start with some relatively straightforward rules of thumb: when we are speaking, clusters that include phonological properties will be the default; when we are writing, in contrast, the default will be to include orthographic properties. In instances where there is just one standard transcription of a phonological form or vocalization of an orthographic one, then we readily shift to a richer property cluster, one which includes both sorts of properties. When there is no single standard transcription or vocalization, clusters which include one or the other, but not both, will be more salient.

Put slightly differently: we start with acts of using names to refer. Then we group these names together according to certain features common to their outward manifestations, e.g. their phonological or orthographic forms, their linguistic history, etc. In a given context, these features offer helpful clues as to which act of reference took place, which common currency name was used, so it is natural for us to focus on them initially. When we want to talk about such features, we utter or inscribe a generic name which represents certain of those features iconically—that is, via the very way that it sounds or looks. This iconicity doesn’t fully specify the homeostatic property cluster at issue, but it does help to get us started. Then we use our background knowledge of the context to construct a name-genus on the fly, taking account of spelling conventions, standards for acceptable pronunciation, accommodation of non-standard accents, chains of historical descent, etc.

Returning to our Gray Cases: when (1) is written, the salient name-genus will cluster together names, in the first instance at least, according to their orthographic features. We might try and move from this genus to a richer one that takes account of phonological features as well, but there is no natural way of doing so—for /ref/ and are rival forms, in the sense that /ref/s don’t typically go by and vice-versa. So the written version of (1) is predicted to be true in virtue of the relevant generic name subsuming both Ralphs in this instance. When (2) is spoken, the salient name-genus will start by clustering together names according to their phonological features. But since each transcription of these phonological features is a rival to the others, there will be no natural way of extending this name-genus. So we should expect for the spoken version of (2) to be true, since the relevant name-genus will subsume all four names.

In contrast, if we try and utter (1), we have to pick between one of two rival phonological forms. And there is no way of naturally extending the name-genus that includes one of these forms so as to include the other. Similarly, if we try writing (2) down, then whatever orthographic form we pick will generate a name-genus that subsumes only one of our four . And we cannot extend this name-genus to include any of the rival forms for the same reasons as before: the presence of the one feature makes the presence of the others less likely. So we predict, correctly, that (1) cannot be truly uttered and (2) cannot be written down.

At this point, one might worry that the present view is intolerably ad hoc. This appearance of ad hoc-ness is, however, to be expected in light of the view’s claim that the salience of a name-genus in a context is a socially-contingent matter. According to the Kaplanian, common currency names are the real philosophical goods. These are what allow us to explain how we manage to communicate with each other about particular objects by using names, how we manage to use names to express object-directed thoughts, to transmit knowledge, etc. Generic names are merely ephemeral ways of clustering together common currency names in virtue of the ‘clothing’ they wear. To be sure, some of these ways of clustering common currency names together will be fairly durable; phonology and orthography shift only relatively slowly. But we should not mistake slow change for permanence. Nor should we mistake the fact that we are used to clustering these features together in certain ways for an inability to shift gears when the situation calls for it.

5.3 Comparing the views

On the version of contextualism about names developed above, names are sufficiently-minimal classes of vocative address. Relative to a context, such classes will be realized by particular sets of phonological and/or orthographic forms. On the version of the Kaplanian view on offer, names are univocal, expressible representational vehicles. Generic names, in contrast, allow us to talk about groups of names in virtue of their common orthographic, phonological, and historical features—with the relevant features being determined by the context. In a sense, these views offer rather similar explanations of Gray Cases: the relevant sets of features are determined by the context. Nonetheless, significant differences remain between the two views.

First, although both views cut against common sense, they do so in rather different ways. On the Kaplanian view, strictly speaking, no two people ever share a name. Accepting genera pluralism does nothing to moderate this. Contextualism about names, on the other hand, entails that proper names are highly unstable, shifting from context to context with respect to both their classificatory features and their realizers. For example, what my name is will depend on whether I am in a context with other or not. Perhaps one of these violations of common sense is ultimately less offensive than the other. Since I am unsure how to adjudicate such claims, however, I will set this possibility to the side; both theories look to be highly revisionary.

The second difference hinges on what each view entails regarding communication involving proper names. According to the Kaplanian theory, using a name to communicate requires (i) that the speaker and listener both possess the relevant common currency name, and (ii) that the listener recognizes which common currency name the speaker was trying to express by means of their utterance.³¹ If the listener doesn’t possess the relevant name, then communication will succeed just in case they treat the utterance as an initiation into the practice of using this name to think and communicate.

According to the contextualist, in contrast, successful communication requires that the speaker and listener coordinate on a property being called N. Even if full coordination isn’t required, the speaker and listener will presumably need to coordinate on what counts as being called N relative to the context. Given how much variance the vocative potential view allows regarding what is called N relative to a context, this looks like a rather demanding picture of communication.³² And given further how early proper names are acquired in language-development, I think we should be loathe to accept such a demanding view when a lighter-weight option is available.³³

Third, and finally, the particular kind of contextualism that looked most promising above—the vocative potential view—faced a serious outstanding challenge. It could not differentiate proper names from terms like ‘boss’ or ‘asshole’. The Kaplanian view, in contrast, faced no such worry. Neither ‘boss’ nor ‘asshole’ corresponds to a stable, univocal representational vehicle that can be passed along from one linguistic agent to another. So neither, according to the Kaplanian view, constitutes a name.

These differences, I take it, favor the Kaplanian view of names. While revisionary, that view is compatible with a relatively lightweight view of how names can be used to communicate and is plausibly capable of differentiating proper names from other nearby terms. Thus, we reach a pair of interim conclusions: (i) we have reason to prefer the Kaplanian view to contextualism about names, and hence to prefer Millianism over predicativism; and (ii) if we accept the Kaplanian view, then we should be pluralists about name-genera. For it is only by adopting genera pluralism that we were able to coherently explain the patterns of truth and falsity we observed in Gray Cases. Having established these interim conclusions, I’d now like to step back and ask: what, if anything, do the above arguments have to teach us about the nature of words in general, as opposed to names in particular?

6 Isn’t this all really just about words?

Above, I treated Gray Cases and the issues they raise as peculiar to names. Perhaps that was a mistake, however. Perhaps the issues have to do with words, or ‘lexical items’, more broadly.³⁴

Consider how a similar-looking puzzle can arise with respect to some very different sorts of terms: the Middle English ‘soote’ and the Modern English ‘sweet’, for instance. Are these merely two alternate spellings of what is ultimately the same word? Or are they two distinct words? If the reader is like me, their reaction to this question will run something like: ‘That depends. What are we trying to explain? If it’s how we, as readers of Modern English, are able to muddle our way through the first few lines of The Canterbury Tales, then it’s probably the former. If it’s the number of distinct languages one of our learned colleagues can read, then perhaps it’s the latter.’

Supposing the reader shares my reaction here, they may well note a parallel with the case of names: with names too, how we are inclined to sort them would seem to vary depending on our interests. That is why we are apt to classify both Ralphs as bearing the same name in certain contexts, but not in others.³⁵ If this were the central issue that Gray Cases served to illustrate, then it would indeed be reasonable to claim that our inquiry would be better directed at words in general as opposed to names in particular.

I don’t take this to be an accurate assessment of the state of things, however. Rather, there is plausibly something special about proper names, something which Gray Cases nicely serve to bring out and which we ought not lose sight of in our theorizing. Specifically, we as individuals would appear to have a degree of authority over how our names are written and pronounced for which there is no analogue when we turn to consider other parts of the lexicon. I am an ‘Eliot’, not an ‘Eliott’ or an ‘Elliott’. I am in a position to correct people when they misspell my name. Likewise, our /ref/ can rightly insist that he is a /ref/, not a . In other words, in a great many contexts, proper names are relatively intolerant of variant spellings or vocalizations—in contrast to other parts of the lexicon.³⁶

To see this, consider the case of ‘color’ and ‘colour’: plausibly, these are not rivals, but variant spellings of the same word. So I am not misspelling anything when I write ‘color’, even when I am located in the U.K. and writing for a U.K. audience—and in spite of the fact that I know this to constitute a departure from local standards.³⁷ Likewise, vocalizing ‘tomato’ as isn’t a mispronunciation even when one is in a region where the standard vocalization is . Most items in the lexicon are tolerant in these respects; in contrast to names, it is up to the speaker to choose between variant spellings or pronunciations, and they make no sort of linguistic mistake if they fail to conform to local standards in favor of some other set.

It is the intolerance of proper names which gives rise to Gray Cases: whereas we are never inclined to think of and as constituting different words, in the right contexts we are so inclined with respect to /ref/ and . This is not to say that most of us will not tolerate a range of ways of pronouncing our names; of course we do. Still, when dealing with an unintentional mauling of one’s name, one typically doesn’t respond by saying ‘Yes, that’s right,’ but rather ‘Good enough.’³⁸

In short, there are reasons to think that names are special. They are different from other parts of the lexicon in that we bear a special sort of authority with respect to the spelling and pronunciation of our own names—thereby making them potentially intolerant of alternative spellings and pronunciations, at least in certain contexts. So names, whatever they are, look set to be more fine-grained than other parts of the lexicon. If that’s right, then they will require an ontology which can respect that fineness of grain.³⁹

7 Conclusion

We began by elucidating a puzzle posed by unspeakable and unwritable names: if there are such things, then there looks to be no easy way to individuate names in terms of their phonological forms, orthographic forms, or combinations thereof. This left us facing a question: just what are names? As we saw, both predicativists and Millians need to answer this question if they are to avail themselves of what seemed a natural account of Gray Cases, yet both faced serious difficulties in generating an answer that’s up to the task.

Ultimately, I argued that Millians are better-placed to resolve these difficulties than are predicativists. For predicativists, the best route to an answer looked to be positing that names are a social kind that tracks the potential for vocative address. But even that view couldn’t serve to differentiate names from some other sorts of terms which pretty clearly aren’t names.

In virtue of such difficulties, I suggested that we are better off accepting Kaplan’s development of the Millian view, including his counterintuitive claim that, strictly speaking, no two individuals ever share a name. However counterintuitive it might be, the Kaplanian view—coupled with genera pluralism—at least presents us with a coherent account of what names are, how they are individuated, and what they mean. The predicativist, in contrast, has so far proven unable to clear this bar.

By way of conclusion, I would like to briefly reflect on why it might prove advisable to revise our initial incredulity in the face of the claim that, strictly speaking, no two of us can share a name. Advocates of predicativism often claim as a point in favor of their view that it (purportedly) squares better with the discoveries of contemporary syntax. This provocative claim has given rise to healthy debates regarding whether predicativists are right about this, and, if they are, what exactly follows for the semantics of names. I fear, however, that somewhere in these debates we have lost sight of why we, as philosophers, were interested in proper names in the first place.

Here is one reason that philosophers have long been interested in proper names: we are very clearly able to think about objects with which we have never had any perceptual acquaintance, objects from the distant past for instance. As a matter of social fact, names seem to be standard means by which this capacity to think about specific distant objects gets passed from one agent to another. Once upon a time, most philosophers took such thinking about to run via complex descriptions of these objects; hence, names were associated with complex descriptions. Now most of us, even predicativists, are inclined to reject that thesis. So we need some other explanation of how names can serve this social-psychological function, of how names help us to pass on the ability to think about objects we have never encountered.

The Millian can offer such an explanation precisely in virtue of the fact that, according their theory, names are univocal representational vehicles. It is this very feature of names, the feature which means that they cannot be shared, which makes them such useful tools for our thinking.⁴⁰ So far, the predicativist has offered no equivalent story—nor is it clear how they might. If they claim that names are regularly used to speaker refer to particular individuals, and that, via such speaker’s reference, they can serve to pass on univocal representational vehicles, then we should ask in all seriousness: if meaning depends somehow on use, and if this is how we use names, then why shouldn’t we just treat such speaker referents as the meanings of names?⁴¹

In other words, suppose that the primary use of names in our socio-linguistic communities is to coordinate our thought and action on objects, regardless of whether we have ever perceptually encountered those objects, and to pass on the ability to do so. In that case, it seems far simpler to posit that names just are univocal representational vehicles—that learning a name just is acquiring the ability to manipulate such a vehicle in thought and to externalize it somehow—than to posit that names are complex devices of classification which somehow or other enable the passing on of such representational vehicles. If names are univocal representational vehicles, however, then it follows straightaway that they cannot be shared.⁴²