1. ‘Literal’ versus ‘Non-literal’ Uses

It is a well-known fact about language that words can often be used figuratively. Metaphorical uses of words are a case in point. Consider: 11. My children are still rug rats.

The expression ‘rug rat’ in (11) is used metaphorically for toddlers old enough only to move around on all fours. Such metaphorical use of ‘rug rats’ does not force us into revising our understanding of the meaning of the word ‘rats’. Instead, the meanings of words in metaphorical uses are better thought of as different-but somehow derived-from their basic meanings. (It may be thought that one desideratum for a theory of metaphor is to provide a systematic account of how the metaphorical meanings of words are derived from their basic meanings.)

The same can be true of some other types of uses which are not obviously figurative. Consider the following:

The expression ‘west coast’ in (12) is a metonym for ‘philosophers working in departments located in the vicinity of the west coast of the US’ and (13) demonstrates a creative use of the noun ‘science’ as a verb. The metonymic use of ‘west coast’ does not pose any threat to the usual understanding of the expression ‘west coast’, nor does the use of ‘science’ as a verb in (13) make us doubt its classification as a noun of English.

Instead, the meanings of words in such uses are naturally seen as derived from their basic meanings. The upshot is that certain special uses of words call for a separate explanation using independent linguistic mechanisms pertaining to those uses. And a semantic theory for a class of words cannot be expected to predict or explain the linguistic features of all uses of words belonging to that class.

Without deferring to how these terms are used elsewhere, I now introduce a distinction between ‘literal’ and ‘non-literal’ uses of words. Literal uses of words, by definition, are uses in which the linguistic properties of words are semantically relevant: a semantic theory of a class of expressions must provide a direct explanation of all uses (of expressions belonging to that class) that are considered literal. It would speak against a semantic theory for, say, adjectives, if the literal uses of adjectives do not conform to the account of adjectives provided by the theory. Literal uses of words of a certain category are their basic (or primary, or paradigmatic) uses and constitute the data that any semantic theory of words of that category must explain without the aid of any further semantic or pragmatic mechanisms.

Further, the linguistic properties of words in their non-literal uses, by definition, are not directly relevant in a debate concerning the semantic properties of words belonging to that class. A semantic theory is not under an obligation to explain the uses that are considered nonliteral and such uses may be explained using independent semantic or pragmatic mechanisms that explain figurative, derived, or one-off creative uses more generally. (The mechanisms may rely on the basic meanings provided by a semantic theory, but explaining figurative or derived uses is not a task set out for a semantic theory.)

Thus understood, ‘literal’ and ‘non-literal’ track the relevance of a use of an expression in a semantic debate. This technical use of ‘literal’ must be distinguished from a second, more general sense of ‘literal’, which contrasts with ‘figurative’ uses (e.g., metaphor, synecdoche, hyperbole, irony, etc.). The first sense is relative to a semantic debate about a specific category of expressions, but the second is more permissive. For instance, in the second sense of

‘literal’, the words ‘free’ (adjective) and ‘free’ (verb) are both literally used in the sentences ‘Not everyone likes to receive free food’ and ‘Let us free the birds’. However, with respect to, say a semantic theory of adjectives, it is only the former, i.e., ‘free’ (adjective) which is relevant (and hence literal in the first sense). A semantic theory of adjectives is under an obligation to provide an account of the adjective, but not the verb. (I thank a reviewer for pushing me to clarify this point.)

Although they are distinct, there are connections between the two senses of ‘literal’, which I will frequently use in this article. It is not for a semantic theory of a class of expressions to provide an explanation for all figurative uses of expressions belonging to that class-rather, it is a task set out for a more general theory of figurative uses. In other words: all non-literal uses (in the second, general sense of ‘literal’) are also non-literal uses (in the first, technical sense of ‘literal’). However, not all literal uses (in the second sense of ‘literal’) are also literal (in the first, technical sense of ‘literal’), like in the example of ‘free’ in the last paragraph. Throughout the article, ‘literal’ is used in the first, technical sense.

Names are words. Therefore, like words more generally, names have figurative uses. Consider, for instance:

It is widely acknowledged that the name ‘Napoleon’ in (14) is used metaphorically (^{Burge 1973, p. 434}; ^{Fara 2015a, p. 254}; ^{Jeshion 2015b, p. 230}; ^{Hesse et al. 2023, p. 248}). Metaphorical uses do not fall within the explanatory purview of a semantic theory of names, which makes them non-literal. Similarly, some other figurative uses of names-e.g., cases of metonymy, synecdoche, irony-are also nonliteral.

The judgement concerning literality, however, may not always be so obvious. Note that in labelling the corpus of name-uses that fall outside the explanatory scope of a semantic theory ‘non-literal’, I have not yet said anything about which kinds of uses-e.g., referential, predicative-must be included in it. An important step in the argument of this article is a proposal concerning which uses of names must be considered literal and which others non-literal. This, however, must wait until section 5. For now, I shall go along with an assumption, discussed in the first paragraph of the next section, that most participants in the semantic debate on names have implicitly or explicitly endorsed.

2. The Sceptic’s Challenge

Referential uses are the most prevalent uses of names and are thought to be their quintessential literal use. Every semantic theory of names seeks to explain the properties of names in their referential uses and is judged better or worse based on its success in doing so.⁴ An analogous consensus exists concerning the uses of names in (5)-(10) that have been labelled extended uses. Theorists across the ideological divide treat extended uses as non-literal and hold that the linguistic properties of names in such uses ought to be explained not by a semantic theory of names but by various pragmatic or semantic mechanisms-e.g., deferred interpretation, metonymic generalizations, lexical rules, etc.-that explain the shift in the meaning of an expression from its primary meaning more generally (^{Fara 2015a}; ^{Jeshion 2015a}; ^{Leckie 2013}).

Things are, however, not so straightforward with predicative uses. Defenders of the Predicate View and Variabilism-call them both ‘Predicativists’-treat predicative uses as literal (^{Burge 1973, p. 429}; ^{Schoubye 2020}, Section ‘Neglected Data Points’). However, ^{Jeshion (2015a}; ^2015b; ^2015c) and ^{Leckie (2013)} have highlighted some similarities between predicative and extended-predicative uses that put pressure on the Predicativist’s position.

For one, barring the difference in application conditions, the linguistic properties of names in predicative uses are the same as their properties in extended-predicative uses-e.g., names in both uses occupy the same syntactic position and have the same semantic type (i.e., <e,t>). More importantly, however, these theorists point out that the same semantic/pragmatic mechanisms that explain extended-predicative uses can also be used to explain predicative uses. ^{Jeshion (2015b, pp. 283-289}; ^{2015c, pp. 381-383}) argues that the phenomenon of deferred interpretation-introduced by ^{Geoffrey Nunberg (1995)} to explain metonymic uses of expressions that are not obviously figurative-can be used to explain predicative uses of names just as it can be used to explain extended-predicative uses.⁵

What adds bite to Jeshion’s argument is that a prominent defender of the Predicate view, Delia Graff Fara, invokes the very same phenomenon of deferred interpretation to explain extended-predicative uses of names (^{Fara 2015a, pp. 257-262}). Leckie argues that the same general account that can be used to explain the polysemous nature of words can also be employed to explain how both predicative and extended-predicative uses of names are derived from their primary referential uses.

The essential force of these arguments can be summarized as the following challenge:

Sceptic’s Challenge: Given that the linguistic properties of names in their predicative and extended-predicative uses are remarkably similar, why should predicative uses not be treated as non-literal uses of names?

The significance of the challenge in contemporary semantic debate can hardly be overstated. Kent ^{Bach (2015, p. 777)} counts the Sceptic’s Challenge as one of the two main objections against the Predicate View. If the Sceptic’s Challenge cannot be met, then semantic theorizing can continue as before-the newer semantic theories on the block have mistaken a semantically insignificant phenomenon as significant. But if it can be met, then it would not only bolster the position of the Predicate View and Variabilism as bona-fide contenders against the classical views, but also cement the idea that the corpus of name-uses that any semantic theory must explain include predicative uses.

The next section discusses some ineffective strategies to address the Sceptic’s Challenge. The aim is to highlight some foundational issues pertaining to how the question of literality should be approached and to provide a desideratum for a satisfactory criterion of literality. The positive proposal of section 5 can then be evaluated against this desideratum.

3. Some Ineffective Responses

One may think that the Sceptic’s Challenge can be met by proposing that which uses of names one takes to be literal is relative to the semantic theory of names that one takes to be true. In other words, it is a semantic theory that determines which types of name-uses count as literal. Such proposal, however, would be viciously circular. The decision concerning which semantic account of names is the best depends, at least in part, on which account best explains the literal uses of names. If, however, the corpus of literal uses of names is itself determined by the theory of names that one subscribes to, then such ‘theory-ladenness’ of the corpus would undermine the possibility of evaluating semantic theories based on their relative explanatory potential.

There is a general methodological lesson here for all parties in the debate: the reasons for thinking that a particular use of a name is literal must not already assume the truth of a semantic account of names. The correct account is, of course, expected to provide a direct and satisfactory explanation of all the literal uses of names, but there must be an independent basis for determining which uses it is supposed to explain. Call this the ‘independence principle’. The independence principle mandates that the question of which uses of names must be considered literal must be sharply distinguished from the question of which uses a particular semantic theory of names takes to be literal. In other words, the criterion for distinguishing literal uses of names from non-literal must not already assume the truth of-or beg the question against-a semantic theory of names.⁶

Another strategy to justify the literality of predicative uses has been to give an argument of the following general form:

Premise A: Referential uses are literal uses of names.

Premise B: Uses that have a close affinity with referential uses are literal.

Premise C: Predicative uses have a close affinity with referential uses.⁷

Therefore: Predicative uses are literal uses.

There is something to be said against both the premises A and B of the above argument. If premise A is true, then a semantic theory of names must be under an obligation to explain all uses of a name as a singular term. This, however, is not true. Names can be used figuratively to refer to individuals, and as discussed above, figurative uses of names are non-literal. Consider e.g., the metaphorical use of the name ‘Freud’ in (15) to refer to a contextually salient individual known to psychoanalyse speakers during talks and the metonymic use of ‘Washington’ in (16) to refer to the US government:

It is for a theory of metaphor or metonym (and not for a semantic theory of names) to explain how the names ‘Freud’ and ‘Washington’ can be used to refer to individuals who do not have ‘Freud’ and ‘Washington’ as their names. These examples demonstrate a fact that often gets ignored in the philosophical debate on names-just as there can be literal referential uses of names, there can be non-literal referential uses (for figurative referential uses, cf. ^{Hesse et al. 2023, section 4.2}).

Further, even if premise A were true, there are good reasons to be sceptical of premise B. Even if one has independent grounds for treating referential uses as literal (say, someone who takes referential uses to be literal on the ground that they are the most frequent uses of names), mere affinity with the referential uses cannot be a justification for some other kind of name-use to be considered literal. It is more reasonable to assume that the affinity between the literal name-uses must be explained on the ground(s) of their literality instead of constituting it. A justification for treating some other kind of name-use as literal must include either demonstrating that the other kind of name-use is literal on the same ground that referential uses are, or providing some independent ground to justify the literality of that other kind of name-use.

The above discussion shows first, that it would be a mistake to draw conclusions about the literality of a name-use on the basis of one’s prior commitment to a semantic view, and second, that the basis of the distinction between literal and non-literal uses of names is something other than the semantic role or the syntactic position of a name within a sentence: just as there can be literal referential uses of names, there can be non-literal referential uses. Likewise, when a name appears as a general term within a sentence, its appearance can be either literal or non-literal.

4. The Need for a Criterion of Literality

The need and relevance of a criterion of literality is the most evident for the Predicate View and Variabilism-these views rely on the assumption that predicative uses are literal, and need to justify the assumption in the face of the Sceptic’s Challenge. However, although posed against a handful of views, the Sceptic’s Challenge highlights a general lacuna in the semantic debate about names-the lack of a principled basis to determine which uses of names a semantic theory should aim to explain. Every semantic view would benefit from the justification provided by the criterion for distinguishing semantically relevant from non-relevant uses-a distinction that all parties in the semantic debate rely upon and endorse.

The need for a criterion of literality is thus well motivated. This section makes a case for a further claim-that it is highly desirable to have an across-the-board agreement concerning literality. In other words, it highlights the need for and importance of a criterion of literality to be generally accepted and endorsed by all parties in the semantic debate on names. The argument for this claim relies on some broader methodological considerations governing the semantic debate, and the rest of this section presents some such considerations.

The disagreement among theorists in the semantic debate on names is plausibly a genuine disagreement. Genuine disagreement, however, requires agreement about the object of disagreement. Connecting the debate between different semantic views requires an agreement about which uses should be explained by a semantic theory of names, and which others may be relegated to more general linguistic mechanisms. For instance, there is a sense in which the classical Millian and Descriptive views, which do not consider predicative uses to be semantically relevant, fail to connect with the Predicate view and Variabilism, which consider predicative uses to be not just literal but rather important uses of a name.

An across-the-board agreement concerning which uses are literal is desirable because it will help connect the differing viewpoints within the semantic debate-establishing the common corpus of uses of names that a theory of names must consider relevant data to be explained would allow juxtaposition of distinct explanans concerning the same explanandum, thereby facilitating an assessment of the manner and extent to which the different semantic views of names agree or disagree.

One desirable feature of a semantic theory for a class of expression is its ‘explanatory potential’-i.e., its capacity to provide a satisfactory account of the uses of expression belonging to that class. An agreement on a criterion of literality would provide a common corpus of name uses that every semantic theory is supposed to explain. Competing theories can then be evaluated based on their relative success in explaining the elements of that corpus.⁸ An across-the-board agreement concerning literality is desirable because by providing a desideratum for theory choice it would advance the semantic debate.

Finally, semantic views that investigate names ‘for their own sake’ have a common interest in determining the corpus of paradigmatic, literal uses. A semantic investigation of names may be driven by a number of distinct motivations. One motivation can be to investigate names with an aim to gain insights into some further epistemological, logical, or metaphysical phenomena of interest. (Frege, Russell, and Quine were arguably driven by such motivations, cf. Gray 2012, pp. 3-16 for discussion.) The range of phenomena that such a semantic investigation needs to examine is determined by the purposes and motivations of the investigation. Such an investigation may, with full justification, consider only such uses of names ‘literal’ that help it fulfil its aims. (The scope of this article excludes theories that are driven by such broader philosophical motivations.)

A second motivation can be to investigate names, qua words belonging to a natural language, for their own sake, i.e., to investigate names with an aim of learning more about their linguistic behaviour. The methodology appropriate for any such investigation is closer to one adopted by a scientist studying a natural phenomenon (like tides or pollination) with an intention to know more about the phenomenon itself. It would be methodologically problematic for such an investigation to limit its purview only to some instances of the phenomenon while ignoring others. To be sure, not all instances that are thought to be instances of the phenomenon will be bona-fide instances of the phenomenon-some criterion to identify bona-fide (or paradigmatic) instances would be required. However, given that their goal is to investigate the features of the phenomenon itself, participants in the investigation have a common interest in defining the phenomenon-learning what the phenomenon is-by identifying its paradigmatic instances.

Analogously, the phenomenon of the use of names in natural language-which is the common target of theorists investigating names for their own sake-is constituted by the paradigmatic uses of names. For such theorists then, the task of delineating the literal uses of names is one and the same as the task of delineating their bona-fide, paradigmatic uses. And given their common investigative target, such theorists also have a common interest in determining the corpus of paradigmatic, or literal uses.

But which uses of names should be considered their paradigmatic, literal uses? All paradigmatic uses of names can be thought to have a shared, essential feature, that is not present in derived, figurative uses. This feature of paradigmatic uses of names can also be thought to be distinctive of names, in that it is not displayed by the use of any other kind of expression. In the next section I identify an essential and distinctive feature of our practice of using names, and based on this feature, motivate a ‘criterion for literality’ for distinguishing semantically relevant uses of names from uses that are not relevant.

5. The Criterion of Literality

What Mill says about the use of names in ostensive sentences in the quote above is true of the use of names more generally. When someone (say, Vic) utters truly that ‘Yesterday I met one McTavish’ or ‘I went to Helsinki ten years ago’ to a linguistically competent listener, the listener (say, Mark) comes to learn something about McTavish and Helsinki: Mark learns that McTavish is a person whom Vic met the day before or that Helsinki is the place that Vic went to ten years ago.

This information about McTavish and Helsinki, however, is communicated by the sentence as a whole and must be distinguished from the information conveyed by the names ‘McTavish’ and ‘Helsinki’. By themselves, all that the names convey to Mark is that the things under discussion are called ‘McTavish’ and ‘Helsinki’. If Mark knew nothing about McTavish or Helsinki prior to his interaction with Vic, Mark would learn nothing new about them (beyond the information provided by Vic’s utterances and the fact that they bear the names ‘McTavish’ and ‘Helsinki’).

There can, of course, be some situations in which Mark comes to learn more about McTavish or Helsinki, but they do not speak against the thesis just discussed. First, as a competent user of the English language, Mark may glean that the sentential positions in which ‘McTavish’ and ‘Helsinki’ appear must get filled by noun phrases standing for persons and places; therefore, Mark may infer that McTavish must be a person and Helsinki a place. Such information, however, is not conveyed by the names used in the utterance, but by the semantic requirement that the direct object positions of the verbs ‘meet’ and ‘went’ generally get filled by noun phrases denoting people and places respectively.

Second, given his prior knowledge of the world, Mark may learn more about the people mentioned by Vic-for instance, Mark may already know of McTavish’s desire to meet Vic and upon hearing Vic’s utterance he may learn of McTavish that his desire is fulfilled. Such phenomenon, however, is highly sensitive to the collateral information and worldly knowledge possessed by participants in a linguistic exchange and therefore cannot be considered an essential part of the information conveyed by names (or the sentences within which they appear).

This last point gets further support from the fact that names of individuals can be legitimately and successfully used even when the conversational participants possess no prior knowledge about those individuals. Vic does not need to rely on Mark’s familiarity with McTavish or Helsinki to make her utterances understood.⁹

This feature of names is in stark contrast with other types of expressions like ordinary adjectives or common nouns-it would not be possible for Vic to successfully convey anything through an utterance of ‘I felt dizzy all day yesterday’ or ‘I went to a circus ten years ago’ if Mark were unfamiliar with the meaning/denotation of ‘dizzy’ or ‘circus’. Journalistic reports, stories, or other discourses about unfamiliar individuals can therefore simply begin by introducing their names (‘Elizabeth was a rather popular queen. She had four children and was the head of a large family’).

The use of a name does not require even the issuer of an utterance to possess much knowledge about the bearer. One such example comes from Michael Dummett (although Dummett did not endorse the view that I motivate here using his example):

If, when I come home, one of my children says to me, ‘Mr. Cunningham telephoned and asked if you would ring him back’, the child may no more know the sense or the reference of the name ‘Mr. Cunningham’, which, let us suppose, he has never heard before, than does a piece of paper on which such a message is written’. (^{Dummett 1973, p. 138})

Dummett’s example illustrates that a person’s competence in using a name may involve nothing more than a faithful reproduction of the name-a child can successfully issue an utterance containing the name ‘Mr. Cunningham’ without possessing much information about the individual bearing the name. Although it may be true of many (perhaps most) actual uses of a name that the speaker of an utterance knows a lot about the person(s) whose name she uses in an utterance, such knowledge is not an essential part of the competence required to use a name.

Names are indeed distinctive in this regard-while one’s competence in using an adjective or a verb crucially depends on one’s knowledge of its denotation/meaning, one’s competence in the use of a name-as Dummett’s example shows-does not. One can use the name ‘Gödel’s Theorem’ to talk about one or more of Gödel’s Theorems, all the while having no idea whether these are theorems in logic or in welfare economics (or even that these theorems are named after a person named ‘Gödel’).¹⁰

It is, however, trivially true that every time a speaker sincerely and non-figuratively intends to use a name to talk about an individual (or individuals), she takes the individual(s) to bear the name. What is often meant by a name being a tag is that just like plastic or paper tags (think of valet parking tickets), a name can be transferred from one conversational participant to another without either of them possessing any substantial information about the individual(s) bearing the tag.

Intending to steal a car, a thief may hand over a stolen valet parking ticket to an accomplice without knowing the brand or value of the car attached to the ticket; however, the assumption that the valet ticket is attached to an object is inherent in the very treatment of the valet ticket as a valet ticket. Analogously, any sincere intention to talk about an individual using a name must rest on the assumption that the named individual bears that name. (I return to this point in section 6.)

It is an essential and distinctive feature of our practice of using names that neither the issuer of a name (say, the utterer) nor the recipient (say, the hearer) needs to possess substantial information about the bearer(s) of the name. Substantial information about the bearer may include information about the character traits, nationality, appearance, habits, or stereotypes associated with the bearer(s) of a name. Uses of names which do not require substantial information can thus be thought to be the paradigmatic, literal uses of a name. The following conditional (A) summarizes this result:

A: If a use of a name by participants in a linguistic exchange requires them to possess no more information about the bearer(s) of the name beyond that they bear the name in question, then that use is literal.

Further, uses of names that require the conversational participants to possess substantial information about the bearer(s) of a name must involve harnessing features of a conversational setting that go beyond the essential linguistic features of a name itself. Such features of the conversational setting may include the epistemic state of participants in a linguistic exchange, contextual information, knowledge of linguistic idioms, etc.

If so, then in such uses the linguistic properties of names must be thought to result from a variety of semantic or pragmatic factors and a semantic theory of names alone should not be expected to account for them. Conversely, an investigation into the semantics of names must not assign semantic significance to the linguistic properties of names in uses that substantially depart from the general practice of using names; and such uses must be classified as non-literal uses of names. The following conditional (B) summarizes this result:

B: If a use of a name by participants in a linguistic exchange requires them to possess information about the bearer(s) of the name beyond that they bear the name in question, then that use is not literal.

Conditionals A and B combined yield the Criterion of Literality (COL):

COL: A use of a name is literal iff such use by participants in a linguistic exchange requires them to possess no more informa tion about the bearer(s) of the name beyond the fact that they bear the name in question.¹¹

Note that the Criterion of Literality presented above does not violate the independence principle, a consideration that should count in its favour. The criterion is indifferent both to the syntactic position in which a name appears (it does not assume that a name must appear in a specific syntactic position to be considered literal) as well as to any specific semantic analysis of names (i.e., it does not assume the truth of any semantic theory of names or beg the question against it.)

Instead, the criterion deploys the essential and distinctive features of the practice of using names as the basis for distinguishing literal from non-literal uses. Criticisms of the Criterion of Literality must either question some aspect of the name-using practice that it assumes or provide an alternative (non-semantic) basis for distinguishing literal from non-literal names-an inconsistency with a semantic account of names is not among the possible criticisms of the Criterion of Literality.

In the following two sections, I consider two objections representing two opposite sources of pressure on the Criterion of Literality proposed above. The first objection (‘the Radical Millian’) disputes the conditional A on the ground that it is too strong-it requires the users of a name to possess too much information for the use to be considered literal. The second objection (‘the Minimal Descriptivist’) disputes the conditional B on the ground that it is too weak-in some literal uses of names, participants are required to possess more information about the bearers than the antecedent of B requires.

6. The Radical Millian

Objection: Consider a case-like the one discussed by Saul ^{Kripke (1977, pp. 263-264)}-in which a name is used to refer to an individual who does not bear the name. For instance, two people may hold a coherent conversation about an individual (who bears the name ‘Smith’) all the while using a name that the individual does not bear (say, ‘Jones’). This may be either because they are mistaken about the identity of the referred individual or because they chose to use ‘Jones’ as a one-off word to refer to the individual.

This case shows that our practice of using names allows us to use a name to refer to an individual who does not bear the name. A fortiori, participants in a conversation do not need to possess the information that the individuals they intend to refer to using a name bear the name. Conditional A is therefore too strong-names can be used literally even when participants possess no information about the bearers whatsoever.

Response: Note that this will be a counterexample to Conditional A only on the assumption that through the use of ‘Jones’ to refer to Smith, Smith does not also come to bear the name ‘Jones’.¹²Kripke’s original discussion was intended at arguing that Donnellan cases (i.e., the use of a definite description to refer to an individual who does not satisfy the description, cf. ^{Donnellan 1966}) are best construed as a pragmatic phenomenon-a case of speaker’s reference as opposed to semantic reference. Just as the use of the definite description ‘the man drinking the martini’ to successfully refer to a person not drinking martini is a case of speaker’s reference, a pragmatic phenomenon, the use of ‘Jones’ to successfully refer to Smith is a case of speaker’s reference as well.

The use of a name described in the case above is thus prima facie a pragmatic phenomenon. If so, the use of a name to refer to an individual who does not bear the name cannot be considered a paradigmatic use of a name, and ipso facto falls outside the explanatory purview of a semantic theory of names.

There is little consensus about how to draw the boundary between semantic and pragmatic phenomena, and one may object that Kripke’s classification of the phenomenon as a case of speaker’s reference is not conclusive. However, the phenomenon turns out to be pragmatic on all major proposals for distinguishing semantic from pragmatic phenomena (^{Ezcurdia and Stainton 2013}; ^{Szabo 2006}). The phenomenon is highly context sensitive (‘Jones’ can be used to refer to Smith only if both the speaker and utterer mistake Smith to be Jones), ‘Jones’ is not a linguistic sign assigned to Smith, and although a speaker may use the sentence ‘Jones is raking the leaves’ to convey the information that Smith is raking the leaves, the truth conditional content of the sentence is still false.

Another option would be to treat the use as an erroneous use of a name (upon learning that the person they intend to refer to is not Jones but Smith, language users will shift to using ‘Smith’). If so, then the counterexample fails anyway-erroneous uses, by that very fact, cannot be considered part of the paradigmatic practice of using names. The case described by the Radical Millian therefore does not serve as a counterexample to Conditional A.

7. The Minimal Descriptivist

Objection: Aidan ^{Gray (2016)} has highlighted an interesting connection between the use of names on the one hand and testimonial knowledge on the other, which suggests that the practice of using a name requires possession of much more information about the bearer(s) of the name than is assumed by the Criterion of Literality. Central to our ability to gain testimonial knowledge is the fact that upon hearing a sincere, trustworthy, and knowledgeable interlocutor utter a sentence (say, P) that contains a name, we can come to know that P-e.g., if the child in Dummett’s example knows that Mr. Cunningham telephoned, from the child’s testimony the parent can come to know this too.

If so, then it would be puzzling how the parent can gain such testimonial knowledge unless the parent can also (a) identify the person the child intends to refer to using ‘Mr. Cunningham’ from amongst other people (say, Mr. Buckingham) and (b) disambiguate the intended bearer of ‘Mr. Cunningham’ from other bearers of ‘Mr. Cunningham’. Gray argues that what best explains our ability to gain such testimonial knowledge is the assumption that we associate information with a name that goes beyond the fact that the referent bears the name in question. And although this information may not be enough to determine the referent of a name, it should be enough to allow one to narrow down to the intended referent and disambiguate the relevant Mr. Cunningham amongst other possible bearers of ‘Mr. Cunningham’.

Response: Note that Gray’s argument is limited to the use of names in cases of testimonial knowledge, and it would therefore be reasonable to circumscribe its conclusions to such cases. One theoretical option is to stick to the position defended in this article while also granting that the restricted context of testimonial knowledge requires a special sort of linguistic competence with names-one that goes beyond the minimal, essential competence needed to participate in linguistic exchanges involving names more generally.

But I favour a stronger response. By requiring that an explanation for (a) and (b) must be provided in terms of one’s (linguistic) competence with a name, Gray puts too much burden on the notion of linguistic competence. The task of determining or disambiguating the referent does not have to rely on one’s linguistic competence. A variety of factors-including (but not limited to) the personal histories of the participants, contextual cues, their epistemic state, their knowledge of their mutual epistemic state etc.-play a role in such determination. Indeed, in this regard, names are like other singular terms: what determines or disambiguates the referent of a singular term in a particular instance of its use is not just one’s linguistic competence with the term-i.e., one’s knowledge of its linguistic meaning-but also the sort of factors outlined earlier.

To see this, consider the following modification of Dummett’s example: it is plausible that (at least in certain contexts) the child could have used the expression ‘he telephoned’ or ‘that man telephoned’ to convey the same knowledge that the child conveys using the expression ‘Mr. Cunningham telephoned’. Concerns like (a) and (b) that arise for the use of names in context of testimonial knowledge will also arise for pronouns and complex demonstratives.

However, in the case of pronouns and complex demonstratives, it is obvious that such questions must be answered by appeal to the historical, contextual, and epistemic factors outlined earlier instead of an appeal only to one’s linguistic competence with the pronoun ‘he’ or the complex demonstrative ‘that man’ (which is rather limited and involves knowing things like ‘that man’ can be used to refer to a man in the vicinity of the agent etc.) It is then methodologically sound to assume that it is the general account of transfer of knowledge using singular terms that should be extended to explain transfer of knowledge using names (instead of forcing a modification on the general practice of using names to accommodate cases of testimonial knowledge.)

8. Applying the Criterion

As an initial test of its plausibility, let us apply the criterion to the stock example of a figurative use of a name:

To successfully use the name Napoleon in the intended (metaphorical) manner, both the utterer of (14) as well as the listener must know much more about Napoleon than the fact that he was called ‘Napoleon’-they must know, for instance, that Napoleon was an ambitious and tyrannical dictator. The use of ‘Napoleon’ in (14) is therefore classified as non-literal by the criterion presented above, which is exactly as we would expect it to be.

The criterion also delivers the judgements that we expect it to deliver in other cases. It classifies most referential uses of names as literal-for instance, a lack of familiarity with Alfred Nobel does not render (1) infelicitous:

Therefore, the use of ‘Alfred’ counts as a literal use according to the criterion. Similar considerations apply to the use of ‘Donald Trump’ in (2).

The figurative referential uses of names are, however, a different story. Consider the example discussed earlier:

The use of ‘Freud’ in (15) is figurative because it is used to refer to a contextually salient individual having the same traits as that of the famous Viennese Psychoanalyst. Clearly, a successful use of this name would require participants in a linguistic exchange to know a lot more about both the person referred to in the context as well as the famous Viennese Psychoanalyst. Such competence, however, is not part of our usual competence in using a name and the criterion correctly classifies this use as non-literal. (Similar considerations apply to the metonymic use of ‘Washington’ in (16).)

The criterion also delivers the correct verdict on the non-literality of extended uses of names in (5)-(10). The metonymic uses of ‘Romanov’, ‘Stella’, and ‘Osama bin Laden’ in (5), (6), and (7) crucially rely on the familiarity of conversational participants with the fact that these names refer to a Russian dynasty, a painter, and the mastermind of September 11 attacks (and his looks.)¹³ The mass term, verb, and adverb use of names also rely heavily on familiarity with some relevant bearer of those names. ‘Heidegger’ and ‘Frege’ cannot be used in the intended metonymic manner in (8) unless the conversational participants recognize them as authors of philosophical works; and the use of ‘ChatGPT’ in (9) and ‘Bill Gates’ in (10) both rely on familiarity with ChatGPT as an LLM and with Bill Gates as a billionaire. The criterion therefore classifies them as non-literal uses of names.

Several reasons for classifying extended uses as non-literal have been proposed in the literature, and it is interesting to see how the criterion compares with and complements them. Some reasons include:

All these are strong considerations for taking one or other subtype of extended uses as non-literal-but what speaks for the Criterion of Literality is that it provides an overarching, general, and uniform rationale for taking all extended uses as non-literal. In fact, it is easy to see that common to each of the above points is the condition that the criterion employs to identify non-literal uses-i.e., non-literal uses require collateral information that is not part of the information needed to participate in the paradigmatic practice of using names. The collateral information cited in (i) is the knowledge of the special practices associated with names of certain categories of individuals, in (ii) the knowledge of socio-cultural status of some individuals who bear a name, and (iii) the knowledge of the idioms of a language. The proposal adumbrated in (iv) is very much in line with the Criterion of Literality. However, the discussion in Hesse et.al. 2013 is restricted to the predicative and extended-predicative uses while the Criterion of Literality is general and applicable to any use of names.¹⁴

9. Concluding Remarks

With this much in place, the task of applying the criterion to predicative uses of names will be a quick one. Consider the predicative use of ‘Alfred’ in (3):

The use of Alfred in (3) requires no more information about any (or all) bearers of Alfred beyond the fact that they bear the name ‘Alfred’. (Similar considerations apply for the use of ‘Donald Trump’ in (4).) Further, it follows from the very definition of predicative uses that, in its predicative use, a name ‘N’ truly applies to a ‘bearer of ‘N’’ (uses in which a name appears as a general term with a different application condition are classified as ‘Extended-predicative uses’, which are already shown to be non-literal). Therefore, the Criterion of Literality classifies predicative uses of names as literal.

Recall that the Sceptic’s Challenge would be met if there are independent reasons for treating the extended-predicative uses as nonliteral and for treating predicative uses as literal. The Criterion of Literality, which is based on an essential, distinctive, and non-semantic feature of the practice of using names, provides such reasons. By classifying predicative uses as literal but extended-predicative uses as non-literal, the proposed criterion serves as a response to the Sceptic’s Challenge on behalf of the Predicativists.

The usefulness of the Criterion of Literality goes beyond the contemporary debate surrounding the Sceptic’s Challenge and, before concluding, I will note the generality of the proposed criterion by applying it to an important but ignored class of ‘bound’ uses of names. Consider (18) below, in which the name ‘Ernest’ in the second sentence is bound by the indefinite in the first sentence, and (19) in which (on its intended reading) ‘Quintus’ refers to a different individual depending on the family considered:

There is some debate concerning whether bound uses of names are literal (cf. ^{Rami 2022, chapter 6}). Understanding the intended meaning of (18) does not require one to possess any more information about the relevant Ernest beyond that they bear the name (that they live in Hertfordshire is information that is provided by the sentence.) The use of ‘Ernest’ in (18) is therefore literal. However, the intended reading of (19) can be grasped only if one has the background information that the bearer of ‘Quintus’, as a result of some convention, is the fifth son in a family (before giving the example, Hawthorne and Manley explicitly mention this convention.) The use of ‘Quintus’ is, therefore, non-literal. The criterion of literality thus also bears upon the question of whether bound uses (which are an important motivation behind views that seek to assimilate names into pronouns) can be considered literal.

The semantic properties of names have been a subject of intense debate and scrutiny; however, the question of which types of nameuses are directly relevant in semantic theorizing about names has received little attention. By proposing a principled manner of making the distinction between literal and non-literal uses of a name, the Criterion of Literality proposed in this article remedies this deficiency.^{15 16}