For a semantic theorist, belief and language are two important proving grounds. Belief systems and languages both provide agents with information about our world and influence our activity within it. More importantly, as many have noted, each provides a window on the other: holding belief content and utterances constant, we can solve for linguistic meaning; alternatively, holding linguistic meaning and utterances constant, we can solve for belief content. Given this, semantic theorists must pay heed to the relationship between belief and language, if only to argue that the suggested connections are misleading.
In Belief and Meaning: Essays at the Interface, Wolfram Hinzen and Hans Rott offer eleven essays that address the relationship between belief and linguistic meaning, giving “various answers to how theories of the one notion relate to theories of the other.” The editors go on to say that this collection should help determine “whether the two concepts of belief and of [linguistic] meaning are essential to one another, and how or whether they can be told apart” (vii). The collection is balanced between belief and language, with belief content being the primary focus in five essays and linguistic meaning being the focus in the other six. In all of the essays, though, the authors address implications for theories of the other sort of content. As a consequence, one can learn much about the interface from both directions.
The volume is divided into three sections. Part I, “The Syntax of Meaning,” includes three contributions that defend a syntactic, internalist approach to linguistic meaning and conceptual content. The four essays in Part II, “Belief, Meaning, and Truth,” present various perspectives on the nature and influence of belief content, with considerations of rationality and intentionality framing inquiries into belief change and belief’s role in a theory of linguistic meaning. Part III, “Semantics and Normativity,” is home to four essays that address the normative character of belief content and linguistic meaning, as well as normative elements in and constraints on semantic theory generally. The volume also includes a brief introductory overview by the editors, with Hinzen supplying further introductory material in the first half of his essay in Part I.
In what follows, I sketch the main argumentative conclusions of the essays and their connection to the overriding concern of the volume, viz., the theoretical association of belief content and linguistic meaning. These essays, of course, do not restrict their attention to a single theme, so in a second section I point to a number of related issues and distinctions illuminated by this volume.
I. An Overview
In his opening essay, “Meaning without Belief,” Wolfram Hinzen describes a spectrum along which we can locate theories of belief content and linguistic meaning. On one end stand those theories that regard belief content and linguistic meaning as “different and independent” (16). On the other, one finds theories that take them to be the same thing. In the middle are various theories that take linguistic meaning to depend on belief content, either methodologically or in a more substantive and fundamental way. This spectrum represents the principal theme addressed by essays in this volume, viz., the dependence of linguistic meaning on belief content.
The first group of essays stands at the “independence” end of the spectrum. This group presents a broadly Chomskyan case for a syntactic analysis of belief content and linguistic meaning. According to this brand of empirical internalism, concepts and languages are controlled by different but related cognitive subsystems. While in operation these subsystems can hardly be distinguished, they can be analyzed in isolation for theoretical purposes. This analysis is restricted to the internal grammars of the subsystems. In both places, meaning hypotheses are assessed relative to syntactic evidence, and, in particular, empirical evidence concerning the range of acceptable readings underwritten by grammatically approved structures. Thus, while facts about belief can inform theories of linguistic meaning, and vice versa, these are evidential and not conceptual relations. The semantic theory that results will be a scientifically respectable theory of the internal, syntactic constraints on interpretation and cognition.
Each essay in Part I makes methodological and substantive pronouncements. On the methodological side, the authors express convictions in line with those that set Chomsky apart from other, more traditional semantic theorists. Hinzen argues that language should be studied as a system, with mechanisms posited to account for linguistic meaning that is taken to be antecedent to and independent of belief or use. In “Does Every Sentence Like This Exhibit a Scope Ambiguity?”, Paul Pietroski and Norbert Hornstein contend that anyone who rejects the idea that “meanings of sentences are not individuated more finely than syntactic structures … owes a theory of the alleged nonsyntactic aspects of meaning” (66), where this requires empirical research and not just intuition mining. Turning to the mind, James McGilvray announces early in “MOPs: The Science of Concepts” that his aim is “serious science” of the mind, an internalist brand of empirical science that shuns notions such as intentionality or wide content. On the substantive side, Hinzen crafts a complex argument for the conclusion that rigidity as a feature of name-meaning is a syntactic phenomenon. Pietroski and Hornstein develop a compelling case against the claim that sentences like “Every girl pushed some truck” have a legitimate reading where ’some truck’ takes wide scope. McGilvray defends a Fodorian account of concept structure that is stripped of the externalist baggage of representationalism.
While the essays in Part I introduced us to the end of the spectrum where belief content and linguistic meaning are independent, Part II includes two essays that shift us to the opposite extreme. In “Meaning, Belief, and Truth,” Max Kölbel argues that one can supply a Davidson-style truth-conditional theory of meaning for language and avoid concerns about the limited reach of truth by replacing truth with belief as the central semantic concept. By contrast, Alberto Voltolini spends almost no time discussing language in “Why It Is Hard to Naturalize Attitude Aboutness,” choosing instead to focus on a characteristic feature of representational content, viz., intentionality. But it seems Voltolini would agree that this non-reducible “mental-or-semantic” feature of content is as much a part of linguistic meaning as belief content. Akeel Bilgrami’s essay, “Belief and Meaning,” presents a view that holds to the middle, as he insists that “it is not quite right” to identify belief and meaning (115n2). Even so, his argument for a world-responsive externalism built on an internalist foundation assumes that belief and meaning are closely related. Thus, when he urges us to reject “orthodox externalist doctrine about content” in favor of a single notion that makes an agent rational “by her own lights” without implying skepticism about the external world, he champions a thesis about both belief content and linguistic meaning (107-8).
The wild card essay in this part is Isaac Levi’s “Seeking Truth.” Like the others, it focuses on belief and in particular on “the question of seeking truth in changing beliefs” (125). In arguing that pragmatists can seek truth as a goal of inquiry, he defends the claim that “epistemological infallibilism is consistent with corrigibilism” (134) against challenges from Crispin Wright, Richard Rorty, and Donald Davidson. In the process, Levi suggests that a theorist concerned with inquiry and interpretation need have no “commerce with sense or meaning” (125). In fact, he seems to believe that there is little semantic work for linguistic meaning to do, and this suggests that belief and linguistic meaning should be treated as one topic and not two independent concerns.
In Part III, we shift focus from the relation of belief and meaning to a topic that concerns all who develop a theory along the spectrum, viz., the role of normativity in semantic theory. Theory construction in any domain is normative, if only because one must be sensitive to norms of logic and evidence. In the domain of meaning, though, many regard normativity as part of the problem; in particular, meanings are seen as normative constraints on use and interpretation. For its fans, normativity can figure into the construction of semantic theory in at least two ways: the theory could itself be normative, putting forth laws that govern the economy of contents, or it could be a descriptive account of meanings as normative.
In the first essay, “The Purpose of a Normative Account of the Content of our Beliefs,” Michael Esfeld develops an account of content that is normative, under pressure from rule-following considerations. One difficulty for such a theory is an ontological trilemma between reduction to non-normative considerations, excessive expansion of normativity, or elimination of normativity entirely. Esfeld supplies a first-person account of normative meaning that dissolves the trilemma. The second essay, “Semantic Structure of Belief and Meaning,” by Sebastian Rödl, serves up a social-pragmatist account of the structure of content that is pressed over neuroscientific and psychologistic alternatives. Rödl argues that semantic structure is the structure of the capacity to use and understand sentences, a capacity that is framed by sociocultural norms appropriate to the practice of language use.
By contrast with these normative accounts, the final two essays in Part III approach semantic normativity as a fact to be explained. Diego Marconi, in “The Normative Ingredient in Semantic Theory,” argues that semantic theories of the standard sort do not embed norms but they do obey them. Focusing on formal semantics, Marconi argues that “as a descriptive theory of semantic competence, [it] is informationally incomplete” (219), but informational completeness can only be gained at the cost of normative inadequacy, in particular, provision of the wrong truth conditions. Thus, if one assumes that there is a “norm to the effect that the semantic values of the descriptive constants ought not to be specified in an informative way,” one rationalizes its informational inadequacy by taking formal semantics to be “normatively-inspired”, even if it is not obviously a normative theory (225). In the final essay, “From Within and From Without: Two Perspectives on Analytic Sentences,” Olaf Müller presents a defense of “narrow analyticity” that is intended to capture our intuitive sense of analyticity while withstanding the Quinean critique. With this, he preserves a distinction that poses an important normative question to all language users: “Which sentences am I not permitted to reject—if I want to avoid talking nonsense?” (229).
II. An Evaluation
One measure of a good collection is the number of seminal papers. Most of the essays in this volume are either extensions of projects articulated in greater detail elsewhere, or are contributions to the necessary work of “normal science” within philosophy, i.e., the development of detail in broader theories. Relative to this measure, the collection does not fare very well, but on another it stands out, viz., the number of important themes and distinctions examined.
We have tracked the theme of the theoretical independence of linguistic meaning through the first two parts, and while the essays in Part III have a different emphasis, they can nevertheless be located along its spectrum. Both Esfeld and Rödl defend theses located near the dependence extreme, with Müller not too far away. Marconi, by contrast, views dependence as an obstacle, and so cleaves to the independence extreme. Turning from this theme, we have noted that Part III examines normativity in semantic theory, but again, the rest of the book is not mute on the topic. The essays in Part I contribute to a descriptive, scientific theory of belief and language, but in so doing staunchly support the methodological norm of “serious science.” Part II is a mixed bag. Bilgrami denies “the relevance of norms to word-meaning” (111), but the others are less explicit. Nevertheless, truth for Levi, belief for Kölbel, and intentionality for Voltolini all appear to play the role of normative constraint on semantic theory.
In addition to these prominent themes, several others receive close attention. Two substantive themes concern the nature of content. Well-represented is the debate between externalists who regard content as “outside the head” and internalists who take it to be exhaustively specifiable in narrow, cognitive terms. Voltolini, Esfeld, Rödl, and Marconi supply variations on the externalist theme, while the papers of Part I serve up three statements of internalism. Bilgrami addresses the debate most directly, and in so doing occupies middle ground, building an internalist externalism. Closely related to this is the debate between those who take content to be reducible to the physical, and those who deny that it is so reducible. The reductionist view is given initial expression in the first paragraph of the first essay, where Hinzen takes
the view that the human mind is a natural object that is not in principle investigated in a different way than the immune system, say, or the planetary system. When studying the human language system, we aim at explanatory principles and laws much as in the sciences of the body, hoping for an eventual integration with the core natural sciences. (13)
This view is echoed in Pietroski and Hornstein, and in McGilvray. On the opposite side is Voltolini, who argues that intentionality resists physicalist reduction. Esfeld and Rödl also toe the non-reductionist line, arguing that content is a non-reducible product of social practices.
On the side of methodology, the volume explores two distinctions that frame much of the philosophy of language. Is the proper theoretical approach to belief content and linguistic meaning one that takes them to be present in individual agents or distributed across groups of interrelated agents? The authors of Part I, along with Bilgrami, develop the former answer, while Esfeld and Rödl pursue the latter. A related question concerns the role of empirical considerations—should mind and meaning be treated as proper objects of scientific study, or should they be the subject of conceptual analysis, in the classic philosophical style? McGilvray champions the former view, arguing that “the aim should be a serious science” (73). Hinzen, along with Pietroski and Hornstein, join him in crafting arguments along these lines. A vocal exponent of the latter view is Esfeld, who contends that the theoretical point of this work “is not an empirical or psychological explanation of the capacity of thought, but a conceptual analysis of norms, content, and meaning” (193). The chorus of like-minded includes the remaining authors in the volume, with Rödl standing out as one who explicitly endorses this approach.
This collection presents a number of well-argued essays that cut across several important trends in contemporary philosophy. My chief complaint would be that Part III is rather one-sided. As constructed, it contains no essays that challenge the centrality of normativity to the investigation of meaning and content. The rest of the volume is even-handed, but the view of normativity held by those who believe that theories in this domain should aim at “descriptive and explanatory adequacy” (73) is not in evidence. This weakness, however, is overridden by a wealth of useful insight on a number of other theoretically interesting points.
1 The Predicativist View
According to predicativism, ‘names are predicates in all of their occurrences’ (Fara 2015: 60). The view has prominent philosophical predecessors (Quine 1960; Burge 1973), and in its most elaborated contemporary version holds that ‘When names appear as bare singulars in argument position, they constitute the predicative component of a denuded determiner phrase, a determiner phrase with an unpronounced Determiner’, where ‘the unpronounced determiner is the definite article, so that bare singular names actually constitute the predicative component of a denuded definite description’ (Fara 2015: 60).1 Since many languages do not have a definite article like English ‘the’, and others have more than one kind of definite article (e.g., Catalan or Maori), predicativism can in the first instance only be evaluated as a language-specific claim about English, a point to which we return below. To make the claim about English formally explicit, consider (1), where ‘Russell’ is a name and occurs in an argument position (i.e., that of the grammatical subject):
Since it is linguistically uncontroversial that proper names, viewed as word classes, are nouns (category N), they project noun phrases (NP), which as such form the complements of determiners where these are present. Therefore, predicativism amounts to the claim that as occurring in (1) and despite its intuitively referential reading, the proper name ‘Russell’ in (1) can be represented as in (2a), where it functions grammatically as a common noun, which forms the predicative part of a determiner phrase (DP). Predicativism in its contemporary version,2 adds the twist that the empty determiner position (D) depicted in (2a) specifically is the unpronounced English definite article ‘the’, so that the representation becomes (2b).3 In short, a name is a definite description with an unpronounced definite article4:
- [DP D [NP Russell]]
- [DP Øthe [NP Russell]]
In the remainder, I review nine linguistic reasons against this view, which predicativism needs to overcome to be defensible. It also needs to address evidence from a number of domains of inquiry other than linguistics, which support the conclusion that proper names in their most standard uses function differently from predicative common nouns. This evidence in particular comes from selective proper name anomia or sparing in acquired brain disorders, including double dissociations between proper names and common nouns; from the pattern of acquisition of names and common nouns in normal development; from anomalies specific to proper names in neurodevelopmental disorders such as autism spectrum disorder; and from the distinct neural correlates of proper and common nouns (see e.g., Semenza 2011; Martins and Farrajota 2007; Mottron, Belleville and Stip 1996; Bélanger and Hall 2006; Hall 2009). I will not review this non-linguistic evidence here.
I will also not develop any positive view. A positive view that is not subject to the problems below and has been tried and tested for over three decades in a large range of languages and language families is well documented elsewhere.5 This other view is not the view that proper names are uniformly referential expressions, which is manifestly false in light of their predicative uses. It rather rejects the very meaningfulness of the question of ‘whether proper names are referential or predicative’. An immediate problem with how this question is posed is that, whatever one's specific view, it is agreed that proper names can occur both with an intuitively referential and with an intuitively predicative meaning. The disagreement is over how to account for this and with what grammatical analysis. In short, all names can be used in at least two ways, depending on the grammar of their occurrence. It would therefore be natural to methodologically assume as a null hypothesis that proper names are grammatically predicates only when they are predicates, not when they function referentially. Predicativism, on the other hand, wants to defend the uniformity claim, according to which names are predicates in all of their occurrences. Therefore, it needs to propose a revisionist analysis of the grammar of names on their referential occurrences. The problem here is not only the evidence discussed below that the ‘covert determiner’ posited in (2b) is fictitious in English. It is also that referentiality and predicativity strictly correlate with the (non-revisionist) grammar that we observe in a one-to-one fashion, in such a way that proper names have the predicative type of meaning precisely when they are grammatically used as predicates. A revisionist grammar thus seems also undesired: when the grammar that we see exactly vindicates the readings that we observe, we should not change our analysis of the grammar based on semantically grounded ideas on how proper names should generally function. Since the relevant predictor for the two usages is grammar, positing a uniform, lexically specified predicative semantic value for proper names seems misguided (as much as positing a referential semantic value would be).
2 Evidence against
First, predicativism as stated above makes it immediately surprising why the definite determiner (or any other determiner) can and normally has to be absent in (3a,c) but is obligatorily present in (3b) (stars within a bracket indicate that the determiner is illegitimate, while stars outside of a bracket indicate that it cannot go missing):
- (*The) Russell entered.
- (The) man I met yesterday entered.
- (*The) man is a mammal.
(3a) only makes the determiner overt that is hypothesized to be ‘unpronounced’ in Fara (2015: 60) or ‘null’ in Matushansky (2015). This should not lead to ungrammaticality when (3a) is used in the same context in which (1) is used. Failure to pronounce an element that is otherwise present in a grammatical structure should not lead to problems with grammaticality.6 Compare the case of (4a), where ‘lamb’ arguably occurs as part of a full DP as indicated in (5a), which, in the absence of an overt determiner, receives a default quantificational interpretation (Longobardi 1994; 2005) on the lines of (6a); analogously for (6b) as a quantificational interpretation of (4b), with the structure in (5b). The sentences in (6a,b) thus contain a lexical determiner in the hidden positions posited in (5a,b), making their existential quantificational meaning explicit. The point is that here we crucially do not see a change in either their meaning or grammaticality, contrary to the case above, leading to more confidence that in (5a,b), the empty positions do indeed exist7:
- I had lamb.
- I saw cats.
- I had [DP ∅ [NP lamb]].
- I saw [DP ∅ [NP cats]].
- I had some (amount of) lamb (-meat).
- I saw some cats.
Turning to (3b), it presents evidence exactly complementary to (3a). In English, there is in fact no such thing as a definite description that can lose its determiner while retaining its meaning (for the change of meaning ensuing when it is dropped, see the discussion of (8) below). In short, for any definite interpretation of a nominal, the determiner is obligatorily overt, and there is no such thing, in English, as ‘a definite description with an unpronounced definite article’, i.e., an ‘incomplete’ or ‘denuded’ definite description in the sense of Section 1. Predicativism maintains that ‘proper names are definite descriptions’ (Matushansky 2015: 335). So why should the determiner in the case of (1) and (3a) be obligatorily absent, when in all normal definite descriptions such as (3b) it is obligatorily present?8
Things get even more puzzling when we see in (3c) that all common nouns can occur without an overt determiner, but then precisely cease to function as definite descriptions: as occurring in (3c), ‘man’ has an intuitively referential (i.e., kind-denoting) reading, more similar to that of a proper name (though the referent is now a natural kind). Since the exact same lexical item is present in (3b,c), i.e., ‘man’, this difference in meaning must relate to differences in the grammar of its occurrence. In other words, the obligatory presence or absence of the definite determiner makes a fundamental difference to meaning. It is not a ‘merely syntactic’ fact, which we could decide to simply discard in favor of a ‘uniform’ analysis on which the definite article is present even when we do not see it. Appearances can be deceptive, of course, as a long history in grammatical theory has shown when it has motivated the existence of ‘empty categories’, as in control constructions or null subjects. Empty articles are not grammatically required in the same sense as these empty categories, however, and their obligatory presence or absence is systematically important to what reading we get, whether in the case of common nouns (3c) or proper names (3a), hence demotivating a uniform analysis of both as definite descriptions on all of their occurrences. As a general methodological heuristic, if a hypothesis says that something is there, and in fact it is not there (or even forbidden, as in this instance), then special evidence is required to conclude that it is ‘there, but hidden’. If we can preserve the grammar that we see, everything else being equal, we should do so: the null hypothesis is that when an article is absent, it is absent. In sum, the first problem for predicativism is:
Problem 1. Pronouncing the definite determiner hypothesized to be unpronounced in phrases containing proper names, as well as dropping the determiner in definite descriptions, leads to ungrammaticality.
The second problem has been touched upon above but is worth keeping separate: When the definite determiner hypothesized to be unpronounced in phrases containing proper names ends up being pronounced, we see a systematic change in meaning in the relevant nominal phrase, which correlates with a change that we see when we drop the definite determiner of a definite description containing a common noun. Thus, suppose there are two Johns, one old and one young. I can then utter (7), and I will have said that of the two Johns, the old one came:
The nominal subject in (7) has a descriptive and indeed contrastive reading (note that any contrast presupposes a description).9 Such contrastive–descriptive readings are precisely unavailable, however, when the determiner is absent, as in John came in, and even when the name is preceded by an adjective, as in (8) (from Longobardi 1994), unless the adjective is stressed:
The natural generalization, therefore, is that the presence of the definite determiner in conjunction with a proper name systematically makes readings available that are otherwise unavailable. These readings exploit the (albeit limited) descriptive potential of proper name, usually together with an additional explicit lexical description. The latter can take the form of a modifying adjective, as in (8), but we also get the forms in (9). (9a and b) could for example be uttered after I report from an encounter with a person who I introduced earlier as being called ‘Russell’, while (9c) exploits the fact that Russell is a (single) famous person, different from any other and non-significant Russell:
- The Russell I met had a hat.
- That Russell was a genius.
- THE Russell was a genius.
As occurring (9), ‘Russell’ has a descriptive reading in the following sense: as part of the complex DP in which it occurs, the name serves, together with a modifying relative clause (in (9a)), a deictic element (in (9b)), and a stress (in (9c)), as an identifying predicate for the referent in question. By consequence, the person is referred to and picked out as a Russell, i.e., an x such that x is a Russell. As occurring in (1), however, i.e., without any overt definite determiner, the name lacks this reading: its referent is simply taken to be familiar, without any added definite description. This is crucially not to say that no descriptive information is present in this case: obviously, in particular, the speaker of (1) presupposes the referent to be called ‘Russell’ (which is not to say that he explicitly asserts the referent to be called ‘Russell’, a point to which I return). However, as the facts make clear, in (1), this descriptive information is normally not rich enough to support a descriptive reading (witness the ungrammatical (3a) as contrasted with the grammatical (9a–c)) to allow for descriptive readings as seen in (7) or (9a–c).10 The natural generalization, in short, is that what allows the presence of the determiner is that the reading is rich enough in terms of descriptive content (yes, in the cases of (7) and (9a–c), where more descriptive information than is contained in the lexical proper name is provided to characterize the referent; no, in the cases of (1) or (3a), where we only have the lexical proper name). This suggests that proper names in and of themselves (i.e., lexically) have too little of such information, demonstrating a difference in how they are used to refer, in line with the difference that we see in their grammar. To put this differently, in order to have a genuine definite description, you need descriptive information. If you don't have the latter, you don't have a definite description, exactly as the grammar suggests. In sum,
Problem 2. Adding the hypothesized unpronounced definite determiner (i.e., making it overt) systematically leads to a change in meaning, namely a descriptive one.
Pronouncing unpronounced material in a grammatical structure should not lead to a systematic change in meaning of this kind. All standard examples given in the literature on predicativism face this problem: They all exhibit descriptive and quantificational readings of names, which are systematically unavailable in the absence of the determiner. To illustrate with examples that are standard since Burge (1973):
- 10.There is one Alfred in my class.
- 11.Alfreds are not usually proud of their name.
- 12.Most Alfreds are German.
In all of these cases where the name appears as the NP-complement of an overt determiner, the reading is not referential but quantificational. In (11) and (12), I generically refer to persons called ‘Alfred’. In (10) I could refer to a particular person, but the expression I use only quantifies over people as falling under a certain description, the specified description being that the person in question is called ‘Alfred’. If I am wrong and no one in my class is called Alfred, (10) is false. When saying ‘Alfred is in my class’, by contrast, I directly refer to a particular person, presupposing he has this name and that it is familiar to the hearer. If my presupposition is wrong, I will still have referred to that person, though under a wrong lexical description. In the same way, when using (1) I do not merely refer to whoever is called Russell – nor even to whoever person to whom the description ‘Russell’ applies uniquely. If this was so, (3a) would be predicted to be grammatical, which it is not. It cannot be that a name occurring as a bare singular NP has the structure of a definite description with an unpronounced determiner, when its occurrence with such a determiner systematically deprives it of the way in which it refers when occurring bare, leading to descriptive readings instead.11
Crucially, one can of course always decide to abstract from such differences in meaning and ways of establishing reference – and one could have plenty of reasons in particular to formally eliminate proper names, which are not needed in mathematics, in favor of definite descriptions, in one's logical language, as Quine (1960) proposed. But the present discussion is about natural language and the structure of proper names, which is a matter of fact rather than formal decision. The correctness of one's view should here not depend on having to ignore systematic differences that falsify it. One should not judge the color of cats based on how they look in the dark. Our empirical claims should not be a trivial consequence of our chosen level of theoretical abstraction.
Fara's (2015) response to the problem of distributional differences in proper names and common nouns is to seek to capture these differences in terms of a syntactic generalization. Proper names and common nouns, she concedes, ‘differ syntactically in that names but not common count nouns can appear as bare unmodified singulars in argument position’ (p.79); ‘they are of a different grammatical category: names differ syntactically from common nouns since they distribute differently with respect to the overt definite article’ (p.113). The second claim is technically incorrect as noted, since proper names are lexically precisely of the same ‘grammatical category’ as common nouns, namely, the category N. The difference Fara notes rather relates to the behavior of proper names and common nouns when these occur as parts of phrases, i.e., at a grammatical level. Moreover, the differences obtain only when proper names are used referentially, and we precisely do not see them, if and when proper names are coerced grammatically into behaving like common nouns, as in the standard examples (10–12) above. The crucial point, however, is that these differences in distribution are prima facie counterevidence against predicativism. In order to develop this view empirically, therefore, this counterevidence needs to be unmasked as being only apparent. In direct contradiction to that, Fara proceeds with an attempt to capture the distributional difference through a syntactic generalization, whose virtue is said to lie in providing ‘a simple explanation of the distribution of names with the overt definite article “the”’. Counterevidence is not dealt with by capturing it correctly through a formal generalization, since the generalization, even if correct, captures no more than the counterevidence against the view under scrutiny. Let us now consider the generalization itself,
- Where Øthe:
The definite article must be realized as Øthe when it has a name as its sister, unless it is stressed.
This generalization is technically incorrect12 and also falsified by its prediction that in The Russell I met was a sad memory of his former self, where the definite article occurs ‘with a name as its sister’ and is not stressed, it should ‘be realized as Øthe’, since in fact it must be overt: witness the ungrammatical: Russell I met was…. The deeper problem however is the above: Where Øthe simply restates what we already know: the definite article is obligatorily absent when its complement is a proper name in a referential use. But this is precisely the fact that contradicts the theory. It suggests that the grammar is systematically sensitive to the proper name/common noun distinction, as long as names have their more standard referential uses, which predicativism predicts it should not be.13 There is therefore no point capturing this fact as elegantly as possible through a formal generalization. Moreover, the very fact that the generalization needs to be hedged by the clause ‘unless it is stressed’ illustrates why predicativism is wrong: the generalization that comes out of (9) above is that the determiner can only occur in front of the name if something else happens, too, which increases the descriptive content involved: for example a relative clause, a deictic element, or a stress. But this means no less than that the absence of the determiner goes with different kinds of readings, which are less descriptive and in this sense more referential. To stress the foundational point missed: the grammar of proper names is systematically different from that of ordinary common nouns, different again when they function referentially and when they do not, and in a way that these differences covary with the differences in meaning.
Problem 2 leads to a third: Not only does making the hypothesized definite determiner overt systematically predict the wrong readings for referentially used proper names, but a covert (phonetically null) determiner also predicts the wrong readings. This was illustrated in (4–5) above. There, we arguably have a phonologically null determiner position, but the readings are indefinite-quantificational ones, not the definite-referential one that all parties agree is seen when proper names occur referentially. So there is a double problem for the empty D position that predicativism posits in all occurrences of proper names: an overt version of this hypothesized covert D leads to the wrong results, but leaving it empty also does. Empty determiner positions do not predict the readings that we actually see and need to explain, but other readings that we do not actually see in bare proper names – an observation that was a starting point for Longobardi (1994) when he first raised the issue of how bare proper names can really be. For a definite-referential reading to be obtained, both the lexical NP and the determiner have to be overt, and they are not, in the case of referentially used proper names. In short, Problem 3 is the following:
Problem 3. Covert (unpronounced) determiners predict indefinite-quantificational readings, not definite-referential ones.
The fourth problem is that neither predicativity nor referentiality are lexical notions, i.e., properties that attach to proper names viewed as lexical items. This can be seen from the fact that just as all proper names can (though less standardly) function grammatically as common nouns, all common nouns can (though less standardly) function as names. This is shown in (13–14) for the common noun ‘fish’:
- 13.Fish entered.
- 14.I love fish.
As Borer (2005) noted, far from being ungrammatical, (13) actually has a (single) reading: It coerces the reading that a person called ‘Fish’ entered (e.g., Mr. Fish), in which case the substantive lexical content of the common noun is depleted (in referring to Mr. Fish, one shouldn't be thinking about fish). This means that one problem listed by Fara (2015:77) for her own view, based on a critique of Burge (1973) by King (2006:149), is actually not a problem for her view. The alleged problem is that (13) is ungrammatical, and hence, proper names and common nouns actually behave differently. But (13) is fully grammatical, though crucially a referential reading is triggered when the determiner is absent, exactly as when a proper name occurs without a determiner, functioning referentially.14 As for (14), it is an instance of kind-reference of a common noun: the reading is equally not quantificational, unlike in (15a,b):
- I would love some fish.
- I had fish.
Seen this way, the common noun ‘fish’ becomes a name (for a kind) in (14), while in (13), it becomes the arbitrary name of an individual. The common noun has a referential rather than quantificational reading in both cases. This is the fourth problem: All proper names can grammatically function as common nouns, and all common nouns can function as names (of kinds). This is a systematic fact, which depends on the grammar of the occurrence of the proper name rather than its lexical specifications.15 Thus, for example, in (16), the kind-reading is as unavailable as it is in (15):
It therefore seems as pointless to say that names are predicates as it would be to say that predicates are names. This would mean to be oblivious to the effect that grammar has on how lexical items are used to refer, regulating referentiality and predicativity alike.
Problem 4. Whether proper names function as predicates is a function of the grammar of their occurrence, exactly as in the case of common nouns functioning as names.
This is illustrated throughout above. As noted, on everyone's view, the intuitive difference in meaning between (1), on the one hand, and (10–12), on the other, is to be preserved and to be explained somehow. The intuitive difference is that the use of the name is referential in (1) while it is predicative in (10–12). In all cases reviewed so far, this difference covaries with grammatical factors in a one-to-one fashion. In each case, we can look at the grammar of the occurrence of the name and systematically predict from this whether the meaning will be of the first type or the second. And of course, we can go the other way round, starting from the intuitive difference in meaning and predicting a difference in grammar from that. If grammar and meaning covary in this fashion, we are looking at a difference that is mediated grammatically, not lexically. Consequently, the assimilation of the grammar of names in both types of occurrences leading to the claim that both have a determiner in front of them, whether overt or covert, is the wrong aim from the beginning: there are differences in meaning and these correspond, in a one-to-one fashion, to the grammatically differences that we see. We should not abstract from the grammatical differences that are there in plain sight, when it is precisely the assimilation of the two types of occurrences at some ‘deep’ grammatical level that predicts the absence of the meaning differences we seek to explain, making the wrong distributional predictions as well.
The fifth problem is that predicativism as stated is an empirical, language-specific claim: obviously, it cannot hold for a language like, say, Polish, where there are no definite determiners equivalent to English ‘the’, or Catalan, where there are two possible equivalents of it. Languages differ in their morpho-lexical resources, and no one would be tempted to postulate a ‘covert lexical item’ in a language when there is no overt one corresponding to it. All languages, however, including Polish of course, have nouns functioning as proper names. This does not mean that predicativism could not be adapted as we move from language to language, preserving its basic principles. Moreover, in fairness, any theory of names will begin from the data in some language, and then have to generalize. However, by depending on specific morpho-lexical resources, predicativism is not set up to make its generalizability a promising avenue to pursue. It is useful here to compare the case of the tradition mentioned in footnote 5, which gives a parameterized account of proper name reference based on grammatical principles, which is directly designed to allow a general account. Given that the morpho-lexical resources are a primary dimension of cross-linguistic variation, universals of language are more likely to be found in semantically significant grammatical relations. A general theory of names is clearly what a philosophical approach to language has to aim for. In sum,
Problem 5. Predicativism is not a general claim about proper names.
It is importantly true that in some languages other than English, proper name NPs can systematically appear as the complements of determiners, in such a way that they retain their referential reading. That is, in these languages, an overt definite determiner does not lead to a switch to the descriptive–quantificational reading. However, evidence that in such other languages referential readings of proper names go with determiners is not as such a reason that such determiners exist covertly in English. Moreover, in the often-cited yet rarely explicitly analyzed case of Catalan, there is a specialized article that only goes with proper names and only when the reading of the proper name is referential as opposed to descriptive.16 In any of the above examples with descriptive readings, a switch to a definite determiner with a different morphological form, which is essentially like English ‘the’, has to occur. The facts, with some interesting differences between Balearic and mainland Catalan, are summarized in the Appendix. The grammar of Catalan, in short, systematically distinguishes between referential and predicative occurrences of names, while the prediction of predicativism is that it systematically should not do so, since predicativity is assumed throughout. In the case of Italian, too, definite articles can occur with proper names when they are used referentially, and there is, unlike in Catalan, only a single definite article in this language. Yet even here, a case has long been made that this article, when it goes with proper names under referential readings, is in fact an expletive, despite its morphological identity with the normal, definite article (Longobardi 1994: 26–7). Its meaning is not that of English ‘the’, which as we saw above is only compatible with descriptive readings when occurring with a proper name (see (3) vs. (7–9)). In sum,
Problem 6. Languages in which proper names under referential readings co-occur with determiners falsify predicativism.17
Next, note systematic meaning difference between the following, which are grammatically different in that in (17) and (18), the name ‘Russell’ occurs as part of a naming predicate, which is a sentential predicate in (17) and part of a restrictive relative clause in (18); while in (19), the name appears as the predicative part of a definite description, and in (20), it is in a referential position:
- 17.This guy is called Russell.
- 18.[The guy called Russell] is smart.
- 19.[The Russell I met] is smart.
- 20.[Russell] is smart.
Now, in (17), an assertion is made, which is false if the guy is not called Russell. In (18), the assertion made is that a certain guy called Russell is smart, and it is not asserted that he is called Russell (which would be the case in ‘The guy is called Russell and he is smart’). However, the guy in question is referred to under this description (‘called Russell’), with the speaker clearly presupposing the description to apply, i.e., to be true, as noted in passing earlier. In this case, if the guy is not called Russell, then it can still be true that the person referred to is smart, though there is a presupposition failure related to the use of the name as part of the relative clause. In sum, a naming predicate ‘being called Russell’ is present in both cases, but there is a difference in grammar, which associates with differences related to assertion and presupposition. The same presupposition present in (18) is also present in (19), but it is weaker, since plainly there is now no explicit predicate ‘being called Russell’ anywhere to be seen. However, the name ‘Russell’ is still used, and it occurs in a grammatically predicative position, as part of a definite description. The job of a definite description is to pick out an individual under a description, and if a lexical proper name occurs as part of such a description, it is part of the descriptive information that the speaker provides: he presents the individual referred to as a Russell. But no assertion that the individual is called ‘Russell’ is made, anymore than it is in (18). But equally, nothing is explicitly said about what the person is called, unlike in (18). In line with that, the presupposition in question is weaker in (19). If the speaker of (19) had just overheard the name and the guy in question turned out to be called something else, he might not be much bothered (while in (18), he would more likely have to take back his claim). Still, in both (18) and (19), ‘Russell’ features grammatically as part of a descriptive predicate that is presupposed to apply to the referent:
In (20), finally, there is of course again no assertion present of the kind that we see in (17), and the presupposition just noted to be present in both (18) and (19) (stronger in the former, weaker in the latter) is even weaker. It is present still, since there is no such thing as a use of a proper name without a presupposition that the referent is called by that name. This is not because a special naming relation is attached lexically to proper names, but because that much descriptive information is part and parcel of simply using each and every lexical noun: in exactly the same way, I cannot refer to an object as a bonnet, without presupposing that it is called by that name. Being minimal in this way, however, the presupposition in question is weaker in (20) than in all of (17), (18), and (19). This is why (22) is so easily interpretable in an appropriate context (e.g., two people talking about a common friend, noting that his actual name is not the one they are used to call him by):
- 22.Russell is not actually called Russell.
By contrast, (23)–(25) show an increasing degree of oddity:
- 23.?The Russell I met is not actually called Russell.
- 24.?*The guy called Russell is not actually called Russell.
- 25.*This guy is called Russell and is not actually called Russell.
In other words, while descriptive information is necessarily present in each and every use of a name, it is present in different ways, in line with differences in the grammar of the occurrence of a proper name, which we see systematically in (17) to (20). Predicativism predicts the absence of such differences: rather than stipulating that names are uniformly predicates, or that ‘naming relations’ are associated to them lexically as a ‘semantic primitive’ (Matushansky 2015: 337), we need to look at the specific grammar of their respective occurrences. Depending on whether the name occurs referentially and bare, whether it is part of a definite description, part of an indefinite description, or part of a naming predicate, the dialectic between reference, predication, and presupposition will play itself out in different ways:
Problem 7. Predicativism does not predict differences in how different referential and predicative uses of proper names are sensitive to the satisfaction of the condition of being called by the relevant name, which again systematically depend on the grammar of their occurrence.
Problem 8 builds on Problem 4. There the point was that both lexical proper names and lexical common noun can function grammatically both referentially and predicatively, depending on the grammar of their occurrence. I now formulate Problem 8 to stress a foundational point related to this: The ‘names as predicates’ view makes a conceptual error. We cannot look at a word in isolation and conclude that it is a predicate. ‘Predicate’ is a grammatical notion, and there are predicates only in grammatical configurations. Consider (26), where the common noun ‘man’ is part of a sentential predicate, which as a whole denotes a property, while in (27) it is part of the nominal restriction of a DP, which is the subject and on this occurrence as a whole has an indefinite-specific interpretation. In (28), in turn, it is again the grammatical subject and in a referential position (referring to a kind):
- 26.I am a man.
- 27.A man entered.
- 28.Man comes from Africa.
Clearly, the differences in referentiality between (26) and (27) cannot have to do with either ‘man’, ‘a’, or ‘a man’ (the latter of which, in particular, occurs identically in both); they rather relate to the grammar of the occurrence of the phrase as a whole. Equally, the referential use of ‘man’ in (28) has nothing to do with its specifications as a lexical item, since it identically occurs in both (26) and (27) as well. Hence, claims about a word functioning as referential expressions or as predicates (whether sentential or nominal) are never lexical claims: they are grammatical ones. Therefore, the claim that in (29), ‘Delia’ functions as a predicate (Fara 2015: 67, following Matushansky 2008), if correct, does as such not begin to be evidence that it also functions as a predicate in (30), where as Fara (2015:69) notes, it occurs in an argument position:
- 29.My parents called me Delia.
- 30.Delia is a philosopher.
How a name functions grammatically in one sentence implies nothing for how it functions when it occurs with a different grammar.18 Words never are referential or predicative lexically. Consider (31) vs. (32):
- 31.Trieste is no Vienna.
- 32.Vienna is no Trieste.
(31) means that Trieste is not the kind of place that Vienna is: is not Vienna-like. (32) says about Vienna that it is not Trieste-like. So evidently, ‘Vienna’, as a lexical item, has no problems acting as a predicate in (31) and as a referential expression in (32), and the same applies in reverse for ‘Trieste’. So is ‘Vienna’ a predicate, or a referential expression? The question is meaningless, for being a predicate or being referential are not properties that any word can have lexically. This explains that, as noted, any common noun can also become referential, denoting either an object or a kind. In short,
Problem 8. The claim of the ‘unified view’ that names are predicates uniformly, irrespective of the grammar of their occurrence, confuses grammatical functions of words with their lexical specifications.
As a final point, it is arguably a general fact about natural language that one expression never means the same as another expression with a different grammar. To illustrate the kind of claim intended here, consider that once upon a time, some philosophers thought that minimal pairs as in (33) ‘express the same proposition’:
- Bill kissed Alice.
- Alice was kissed by Bill.
While one could always develop a technical notion of ‘proposition’ that abstracts from various meaning differences, it is now well known that such minimal pairs such as this actually systematically differ in meaning. This not only concerns information structure (such as what is the topic in (33a/b)) but also scope (e.g., The target wasn't hit by many arrows and Many arrows didn't hit the target need not be true in the same circumstances). The pair in (34) illustrates the same point:
- It seems that I am happy.
- I seem to be happy.
Again, one might be tempted to ignore the grammatical difference here and provide a uniform semantic analysis. However, if I don't feel particularly happy, yet a doctor shows me a brain scan from which he concludes that I am happy, I could utter (34a) in preference over (34b). (34b) has a more experiential character, and is preferred as and when the emotion is felt in a first-person way.
Based on such examples, it is a tempting generalization that, whether or not there is such a thing as synonymy at a lexical level, there never is synonymy when systematic grammatical differences are involved, as they plainly are in the present case. This makes it puzzling how it could possibly be that a construction with a proper name in a referential use, on the one hand, and a definite description with its manifestly different grammar, could possibly ever mean the same. And empirically, they don't. A definite description has referential uses, but it also has quantificational ones, without any change in its internal structure as a phrase. A proper name, by contrast, does not have quantificational uses, unless it co-occurs with a determiner or is pluralized, and hence, forms part of a number phrase, as seen above, and it involves descriptive information in a different and more minimal way than a definite description. Therefore, that names are not (‘denuded’) definite descriptions is simply a general instance of what is possibly a general law of language: that two grammatically different expressions never mean the same.
Problem 9. Two grammatically different expressions are likely to never mean the same.
Once upon a time, philosophers of language restricted themselves to the regimentation of natural language – its formalization in a logical idiom. It is a significant change that by now, their claims have become empirical. This makes them testable. If predicativism is to have empirical support, it needs to overcome the evidence above, in addition to the evidence from a number of other domains of inquiry mentioned above, including language development, neurodevelopmental disorders, single case studies, and the processing of language in the brain. Empirically, proper names can function as predicates, but they need not. Moreover, predicativism misses an important foundational insight: that predicativity is a grammatical notion. It is the grammar of a language that regulates whether expressions in it are referential or not on an occasion of their use.
This research was enabled by the Arts and Humanities Research Council UK, grantnr.AH/L004070/1, and the Spanish ministry for Education, Culture, and Sport, grantnr. FFI2013-40526-P.
The data from Balearic Catalan (courtesy Joana Rossello):
In (i.a) below, we see the standard (masculine singular, MSG) definite determiner that goes with common nouns on predicative occurrences of these, while in (i.b), we see the definite determiner specialized for proper names in referential uses. As (i.d) shows, this determiner becomes ungrammatical and has to be switched for the other, when a descriptive is generated by coercing the proper name into a predicative position, similar to that of the common noun in (i.c). Crucially, the determiner for proper names in their referential uses doesn't even have a plural form: in this sense, what the grammar here expresses is that the idea of a ‘plural proper name’, in a referential use, makes no sense. (ii) shows the same facts for the feminine determiner. (iii) shows for both Genders that, where the proper name determiner is used, only a non-restrictive (i.e., referential) reading is available, with the relative clause merely appositive and requiring comma intonation. (iv) shows that when a restrictive reading of a proper name is forced, no good solution is available.
Es ca/*En ca
En Joan/*Es Joan
E(l)s (cinc) cans
Thempl (five) dogs
E(l)s (cinc) Joans
Thempl (five) Joans
Sa cadira/*Na cadira
Na Maria/*Sa Maria
Ses (tres) cadires
Ses (tres) Maries
En Pere, es qui sempre arriba tard, avui no vendrà
Themsg Pere, themsg who always arrives late, today will not come
Pere, the one who is always late, will not come today
Na Maria, sa qui sempre arriba tard, avui no vendrà
Thefsg Maria, thefsg who always arrives late, today will not come
Maria, the one who is always late, will not come today
- ?En Pere/?Na Maria que va suspendre no hi és
*Es Pere/*Sa Maria que va suspendre no hi és
Themsg Pere/Thefsg Maria that failed is not here/that s/he failed is not here
The data from standard mainland Catalan (courtesy Txuss Martin):
(v) illustrates that in mainland (mainly written) Catalan, the standard common noun determiner has to be used when the proper name is pluralized and has a descriptive reading. Example (vi) is from Longobardi (1994