Singular Terms and Ontological Seriousness

Abstract

Linguistic ontologists and antilinguistic, ‘serious’ ontologists both accept the inference from ‘Fido is a dog’ to ‘Fido has the property of being a dog’ but disagree about its ontological consequences. In arguing that we are committed to properties on the basis of these transformations, linguistic ontologists employ a neo-Fregean meta-ontological principle, on which the function of singular terms is to refer. To reject this, serious ontologists must defend an alternative. This paper defends an alternative on which the function of singular terms is not generally to refer and on which they are generally ontologically noncommittal. This is the best way to reject linguistic, ‘easy’ arguments for the existence of properties. The account recommends neutralism about quantification (drawing on Barcan Marcus and Meinongianism), coherently bringing together two important yet uncombined meta-ontological movements. Moreover, it employs Ramseyan insights about the transformations to provide a nonreductionist, non-error-theoretic redundancy approach to explicit talk about properties.

Introduction

Consider the following sentences:

S₁: Fido is a dog.

S₂: Fido has the property of being a dog.

The inference from S₁ to S₂ is generally, though not universally, accepted as valid. Its rejection dauntingly requires (i) a deep revision of our normal linguistic practices and (ii) charging ordinary discourse with underlying errors (see sections 1 and 5.1).

Linguistic ontologists use the acceptability of this inference as an argument for the existence of properties (see Mourelatos 1978: 426; Thomasson 2001; according to Heil [2006: 234, 243] and Lewis [1983: 349ff., 352f.], even D. M. Armstrong, despite his serious, sparse theory of universals and generally ‘appreciat[ing] that linguisticism represents a distraction from serious ontology’, ‘yield[s] to linguisticizing pressures’ [Heil 2006: 243, 234] and employs such ‘semantic argument[s]’ [Lewis 1983: 349–50], in particular against Quine [1953, 1980] and Devitt [1980]; see Armstrong [1978a: 16, 11–17, 58–64; 1978b: 7–59; 1980: 443–45]).

‘Ontologically serious’ philosophers, by contrast, reject the further inference to the existence of properties. So, they can either

1. (a) reject the inference from S₁ to S₂, or

2. (b) accept it and explain, for instance, that the singular term ‘the property of being a dog’ need not refer even if S₂ is true.

Serious ontologists might take route (a) by following Rayo who argues that ‘Ordinary discourse is plagued with unclarities and ambiguities’ (2002: 437), Quine who dismisses ‘daily discourse’ as ‘second grade’ (1969: 24; 1981: 92, himself following J. J. C. Smart 1963: 47; see also Janssen-Lauret 2016: §2.2), or, perhaps, Schipper (2021: §5.2) and Unger (1979a: §5; 1979b: §4) who, on serious ontological grounds, are open to the possibility of widespread, deep errors. However, this paper explores (b): how serious ontologists can best accept normal linguistic practices and thus the validity of the inference (and some other constraints specified in section 1), but reject that this implies that properties exist—thus following Musgrave (2009: 69ff.), Dyke (2008), and Heil (2003: 53–54), who are not error theorists.

The relevant serious ontologist claims, further, that accepting the inference's validity but accepting that this implies that properties exist, as linguistic ontologists do, commits one to the ‘representational fallacy’ (Dyke 2008: ch. 1 & 3; see also Heil 2003: 5–6; Cameron 2010: 8) and to ‘word magic’, which according to Musgrave is: ‘The idea that once we invent or create a word or phrase, we invent or create an entity for that word or phrase to stand for’ (2009: 69; see also 2001: 29f.). Musgrave's target is especially constructivist linguistic ontologists who explicitly claim that many entities are created merely by the use of language (e.g., Goodman 1978; Schiffer 1996, 2003; Thomasson 2001).

Linguistic ontologists would claim, for instance, that the property being a dog is literally created via pleonastic transformations like those from S₁ to S₂, what Schiffer (1994: §5, 2003, 2019: 244ff.) calls ‘something-from-nothing-inferences’. Thomasson calls them ‘trivial inferences from uncontroversial truths’ (2015: 132), ‘to reach a truth that is intuitively redundant with respect to the first, yet leaves us with (apparently new) ontological commitments to the disputed entities—again apparently resolving ontological questions by way of trivial inferences from undisputed truths’ (Thomasson 2015: 134–35). Thomasson (2016: §1.1) explicitly uses such inferences also to argue that ordinary objects, such as dogs and chairs, exist although Schiffer (2019: §2) rejects this. The current argumentation focuses on properties but applies to all analogous linguistic arguments.

To try to diminish the incredulity, Thomasson calls the language-created, pleonastic entities ‘ontologically minimal’ (2001: 323–25). Thus, (a) they are causally inert, ‘come softly into existence, without disturbing the pre-existing causal order in any way’ (Schiffer 2003: 59), and (b) everything about them can be known by ‘study[ing] the language games by means of which they are deposited in our ontology’ (Thomasson 2001: 321; see also Schiffer 1996: 159). Serious ontologists wince at the ontological profligacy: these entities’ ontological dependence on language games and their causal inertness, hence epiphenomenalism, increase, rather than decrease, the sense of illegitimacy. Creation via linguistic transformations alone requires ‘word magic’.

Martin and Heil (1999: 57) describe the ‘linguisticist’ argument style as ‘an evasion or obfuscation of ontological views’ that has led ‘proponents [to] pretend that entailments were possible when they are not’. The pretended entailments are ones to ontological conclusions such as the existence of properties. This reveals an adherence to the ‘widely held’ but ‘ill considered’ and ‘manifestly incorrect’, linguisticist picture of representation (Heil 2003: 5–6), the rejection of which is one of the serious ontologists’ central methodological maxims (adapted from Heil 2003: 6):

the serious maxim (maxim): do not read ontology, the character of reality, off of our linguistic representations of reality and other purely linguistic considerations.

Cameron (2010: 9–11), Dyke (2008: 35, 44, 53, 59, 62, on properties, 67ff., 77), and Heil (2003: 6) employ maxim repeatedly as their methodological touchstone.

The relevant serious ontologists, however, accept the conceptual rules that underlie the transformations; they are not error theorists about ordinary talk or revisionists who recommend paraphrasing into, say, a formal language. This distinguishes serious ontology from approaches like fictionalism (e.g., Field 1989; Yablo 2000), which accepts that literal talk is ontologically committing but then recommends widespread pretence given the widespread error. Some serious ontologists reject the first step as linguisticist. Heil (2003: 53–54) writes:

I am inclined to think that ‘this is a statue’ can be, and often is, literally true. What makes it true is a complex, dynamic arrangement of particles. . . . We cannot hope to paraphrase, translate, or replace talk of statues with talk of such collections.

These transformations, however, are taken not to yield reference to entities over and above, for instance, Fido (in S₁ and S₂), no ontological commitments to extra entities.

However, linguistic ontologists rely only on seemingly standard compositional semantics regarding singular terms and generally accepted inferences from statements employing singular terms to quantified statements. A problem emerges: serious ontology cannot thwart linguisticism without rejecting this underlying and still widely (though not universally) accepted picture of singular terms.

To support maxim, serious ontologists often accept a principle on which truths depend on (fundamental) reality. Some serious ontologists call this a ‘truthmaker principle’ (Heil 2003: 53–54; 2012: 192, 194; Cameron 2010: 17), but they need not (MacBride 2014: §4). (Note that Heil's version of truthmaking is both explicitly antilinguisticist and nonnecessitarian; see especially 2003: 64–65; 2006: 234–42; 2012: §8.3.) Fundamental reality is discovered by nonlinguistic means and need not mirror the linguistic structure of what it makes true. Complex, dynamic arrangements of fundamental particles, for instance, have no structural similarity, let alone structural isomorphism, with truthbearers such as S₁ and S₂. Also, if p depends on some portion of (fundamental) reality π, p need not be reducible to any truth q describing fundamental reality more ontologically perspicuously (see Heil 2003: 53–54; 2012: §1.2, 151, 165, §13; Cameron 2010: 11; compare Austin 1979: 124f., Schipper 2020: 513, §2.3; 2021: §5.3).

However, serious ontologists say nothing about the singular terms in S₁ and S₂. Despite rejecting the linguisticist picture of representation, Heil admits not being ‘in possession of a better, more plausible account of the connection words (or concepts, or thoughts, or representations generally) bear to the world. I have no such account, nor do I know of any’ (2003: 5). Until such an account becomes available, the problem persists.

This paper defends such a serious ontological account. Section 1 unpacks the linguistic argument for properties, identifying which premise serious ontologists should reject. This draws out the ‘neo-Fregean’ (section 1.1) and quantificationally linguisticist (section 2) pictures of language, to which they must present an alternative. Sections 3–5 defend an alternative that draws on quantificational machinery developed by Meinongians and Barcan Marcus (1993) (section 2), and Ramseyan insights (section 3), to persuade us that the function of singular terms is not primarily to refer but to pick out, single out, and focus on objects of thought and talk, in a sense that is ontologically noncommittal. This blocks the linguistic argument and shows us how to talk ontologically seriously.

1. Unpacking the Linguistic Argument for the Existence of Properties

Serious ontology clashes with the following three prima facie plausible claims concerning S₂, which entail that properties exist:

C₁: If S₁ is true, then S₂ is true.

C₂: The expression α, ‘the property of being a dog’, functions in S₂ as a singular term.

C₃: If α functions in S₂ as a singular term, then S₂ is true only if a property exists.

Rejecting C₁ requires a deep revision of our normal linguistic practices, which both linguistic and serious ontologists accept (Heil 2003: 53–54). For instance, Hofweber (2016: §2.3.4), a linguistic ontologist, argues for the obviousness of the equivalence based on linguistic competence. Rejecting C₁ (and the inference's validity) is, however, an option for serious ontologists, especially for those willing to implement such deep revisions. For instance, Quine (1969: 24) recommends replacement by appropriate regimentation into a formal language. Let us call this the ‘regimentation-and-replacement strategy’. Ritchie (2016: §3) and Janssen-Lauret (2016: §§2–4) might accept the ordinary inference but reject the inference after regimentation if no formal inference rules take us from regimented-S₁ to regimented-S₂. But this assumes that appropriate regimentation and paraphrasing is possible. Although not everyone would accept the inference, Martin and Heil's (1999: 34–35, 54, 56–57) ‘ontological turn’ resists such strategies. (Section 5.1 supports C₁ further.) The current paper, recall, explores what would follow from accepting the validity of this inference and other constraints that serious ontologists have previously defended (such as avoiding deep, error-theoretic revisions of our normal linguistic practices and formal paraphrasing).

C₂ is plausible: if anything functions as a singular term, then that expression does. C₃ follows from the following assumptions:

C₄: If an atomic truthbearer χ containing singular terms is true, then the singular terms refer.

C₅: If α refers, then it refers to ψ, the property of being a dog.

C₆: If α refers to ψ, then ψ exists.

C₇: If ψ exists, then a property exists.

Linguistic ontologists might claim that C₆ is an almost analytic, undeniable truth because this distinguishes reference from other similar semantic functions. Some serious ontologists, such as Crane (2001: 23–28), emphasize this, too. There are, of course, philosophers who might legislate and have legislated a different usage of the word ‘refer’, for instance, R. M. Sainsbury (2005: vii), Graham Priest's noneist (2016: 108, 141), Kaplan (1973: 503ff., 505), and Kripke (1971: 146; for discussion, see Hughes 2004: 20–21). However, I shall leave the linguistic ontologists’ stipulation that reference implies existence alone, accepting C₆ for the sake of argument and as a further constraint for the rest of this paper, as we did with C₁ (that S₁ implies S₂).

C₇ is just an instance of existential generalization. Rejecting C₅ entails that α refers to something other than ψ (though clearly not Fido). But then χ (the truthbearer) entails the existence of entities over and above Fido purely because α is a singular term. This is not satisfactory for serious ontology, which must reject C₃ by rejecting C₄. The rest of this paper aims to present and argue for the best way to do this.

1.1 Rejecting C₄ and Neo-Fregeanism

C₄ expresses the semantics standard amongst neo-Fregeans. MacBride is not a linguisticist (2014: §4; 2005: 609f.) but helpfully articulates the relevant principle thus:

Syntactic Decisiveness: if an expression exhibits the characteristic syntactic features of a singular term, then that fact decisively determines that the expression in question has the semantic function of a singular term (reference). (2003: 108)

In short, if it looks, can be used to build sentences, and is built like a singular term, then it functions like one: as a referring term.

I propose that serious ontologists reject this principle by rejecting that the semantic function of singular terms is to refer, that is, accept that syntactic singularity entails semantic singularity but add that this does not entail reference. Others have rejected Syntactic Decisiveness by arguing that syntactic singularity does not determine semantic singularity. Sometimes a syntactically singular term such as ‘Fido’ or ‘the property of being a dog’ can be understood as nonreferring and thus, they conclude, as not semantically singular terms. For instance, Barcan Marcus (1993: 118, 203, 210) distinguishes between (i) ‘genuine proper names’, tags, which are, according to her, directly referential, that is, have the function of just referring directly to their referent without any descriptive content, and (ii) other syntactically singular terms, such as fictional names, that have different kinds of functions. (However, as we will see below in section 2, she must admit nonreferring names as a possible substitution class of variables to keep [substitutional] quantification ontologically neutral.) The serious ontological account I propose takes the semantic function of singular terms generally not to be to refer. This has the benefit of allowing us (a) to retain a closer connection between syntax and semantics and, perhaps more important, (b) not to have to appeal to a more canonical language as ontologically more perspicuous (pace Quine 1969: 24; the proposal is more consistent with the spirit of Martin and Heil's [1999] ontological turn and Heil's [2003: 53–54] non-error-theoretic non-reductionism).

Others have rejected Syntactic Decisiveness in ways that are not congenial to serious ontology. For instance, Hofweber (2016) defends an alternative nonreferential, inferential reading of quantifiers and singular terms, which leads us on the road not just to linguistic ontology but to ‘a strong anthropocentric form of idealism’ (Hofweber 2019: 731). In short, if inferentialism is right and Hofweber's defense of ‘conceptual idealism’ is sound, then maxim is wrong (2019: 719f., 722–23).

Although a defense of maxim against the threat of conceptual idealism must wait, two points are worth making here. Serious ontology must (a) not understand singular terms in a way that vindicates linguisticism; Hofweber, by contrast, argues that we can draw metaphysical conclusions from both readings of the quantifier. Hofweber argues that some singular terms are nonreferential based on empirical evidence about our linguistic practices involving them (2019: 715f., 720). By contrast, serious ontology must (b) reject Syntactic Decisiveness not merely by relying on empirical evidence about linguistic usage. For serious ontology (of Martin and Heil's [1999], though perhaps not the Quinean, sort), metaphysics must come first. The current paper shows how to reject this neo-Fregean linguisticism in a way congenial to serious ontology.

2. Neutral Quantifiers and Serious Ontology: Barcan Marcus and Meinong

One of the main motivations to think that reference is the semantic function of singular terms is an assumption about quantification being existential. What distinguishes reference from other possible functions of singular terms is that reference requires that what we are talking about exists (recall C₆, and pace, e.g., Sainsbury 2005: vii). And, indeed, if (I) quantification requires existence and (II) truths such as S₂ entail that there are properties (that is, quantification over properties), then (III) we can conclude that properties exist. But only (II) is supported by natural and pretheoretical intuitions. (I) is, according to Eklund (2006), characteristic of a linguistic meta-ontological approach to quantification.

Eklund (2006: §4) argues that both Fregean and, what he takes to be, Quinean approaches are ‘fundamentally, linguistic approaches to ontology’ (Eklund 2006: 327):

Both the Fregean and the Quinean assumes that the semantic function of a singular term is to refer, and that so-called existential quantification really is existentially committing (Fregeans tend to focus on singular terms; Quineans on quantification).

Eklund seems largely to follow van Inwagen (1998: 233, 241, theses 2, 3, and especially 4) in calling the relevant view of quantification a central part of the ‘Quinean’ meta-ontology. However, Quine himself believes that ontological commitments only become apparent once a theory has been regimented in first-order logic without individual constants (thus eliminating singular terms). And he rejects the view that properties are created by the use of language. Since Quine's famous rejection of properties is nonlinguistic, he himself is no linguisticist. In fact, he takes what I called ‘the regimentation-and-replacement strategy.’ (See also Janssen-Lauret's [2015: §§1–3; 2016: §§2.1–2.2] articulation of Quine's meta-ontology, which clashes with Eklund's grouping; I thank an anonymous reviewer for discussion.)

Given that Quine's actual views seem to clash with Eklund's grouping, a better way to understand the debate is as follows. If we accept both ordinary discourse and that quantification is existentially committing, then linguistic ontology and its conclusions seem to follow. Quine's serious, nonlinguistic ontological strategy, as discussed earlier, rejects ordinary discourse. An alternative serious ontological strategy that, unlike Quine's, accepts ordinary discourse, can instead reject what Eklund would label linguisticism about quantification by taking quantifiers generally not to be existentially committing.

To do this, serious ontologists can draw on the very important, but unfairly marginalized, twentieth-century, female philosopher Barcan Marcus's substitutional view of quantification or, alternatively, the often misunderstood Meinongian view. Barcan Marcus is explicit that quantification should be understood as ‘ontologically neutral’ (1993: 16–18, 79–82, 85, 122). She writes, further, that ‘Quantification does not itself confer existence’ (1993: 82), at least not on her substitutional view. Meinongian quantification is also best understood as across-the-board not ontologically committing and, as Routley (1966: 255, 275–76) emphasizes, ‘ontologically neutral – at least as far as what exists goes’: ‘Fs exist’ is not equivalent to ‘Some objects are Fs’ or to ‘Some things are F’ (Priest 2016: 198ff., 312ff.; see also 2016: xxix, 11, 13, 152, 331; 2000: 496f.; Eklund 2006: 327–29; Routley 1966: 255f., 275f.; and Berto 2013: §4.3, 70ff.).

According to Meinongians and especially Barcan-Marcusians, quantifiers in natural language are just devices for quantifying, that is, counting or measuring (see also Schipper 2020: §2.2). And one can count and measure nonexistent objects of thought, inquiry, and talk. Infamously for standard, quantificational linguisticism, this sits very well with common usage (Eklund 2006: 327f.; Berto 2013: 210ff.). For instance, we can count sheep while trying to fall asleep even though the sheep we are counting do not exist. One is not pretending to count or doing nothing; one is really counting, most plausibly, imaginary sheep. Also, it may be crucial for solving a case that Sherlock identifies not three but four people in the room (the murderer is in the closet). We can count dormitive sheep, Holmes, merely hypothetical objects such as Vulcan, and many other objects of counting. It is a further question whether they exist.

If this view is correct, then one may accept S₂, draw the inference that something is a property of being a dog and Fido has it, but also accept that this property might not exist. This blocks linguisticists’ ontological conclusion from S₁. According to quantificationally neutralist serious ontologists, S₁ entails S₂, and these each entail that something is a property that Fido has. All accounts of the metaphysics of properties, including tropism, immanent realism, transcendental realism, and even nominalism can accept these claims. The ontological question is whether the properties that we count and quantify over also exist, which we express not with a quantifier ($\exists x\phi$ or, a neutralist alternative, Σxϕ), but with an existence predicate (E!x). However, this further claim, that properties exist, is not entailed by S₂ or even by the quantified statement. It takes further work to establish the existence of the objects we talk about in true, ordinary predications and true, quantified statements.

Thus, replacing (I) above with neutralism about quantification yields an alternative view of the behavior of singular terms that is compatible with serious ontology, bringing together two important yet uncombined meta-ontologies.

Just to be clear, however, one need not import into one's account other Meinongian assumptions, such as the existence-subsistence distinction. Eklund (2006: 328) helpfully distinguishes modes-of-being Meinongianism from noncommitment Meinongianism. When analytic philosophers quickly distance themselves from Meinongianism, they normally cite the existence-subsistence distinction (e.g., Schaffer 2009: 359), hence distancing themselves from modes-of-being Meinongianism. Even van Inwagen (1998: 236) has no arguments against noncommitment Meinongianism, famously only indignation elicited by abuses of the Meinongian approach.

2.1 Barcan Marcus and Serious Ontology

Also helpful here is Barcan Marcus's (1993, e.g., 1972, 1978; see also Belnap and Dunn's [1968: 178] developments) ‘Name-based Meta-Ontology’, as Janssen-Lauret (2015: §5; 2016: §4.1) calls it. Two of Barcan Marcus's key innovations are (a) her substitutional interpretation of quantification, on which an existentially quantified sentence is true iff the open sentence following the quantifier is true on some substitution instance of a name for the variable bound by the quantifier (Barcan Marcus 1993: 80), and (b) her view of names as ‘directly referential tags’ (1993: xiii, 10–12). She writes that her view

frees us to explore generality and existential import in all their subtlety. The satisfaction theory [compare my section 4 below], by contrast, is a kind of logician's monism. But it does not seem to me that the existence of sets or numbers or propositions or attributes or physical objects hinges wholly on the way variables and quantification function in theories. (1993: 82)

Only when the substitution class ‘consists of genuine proper names: where the names link up with objects’ (1993: 118), when they are directly referential tags, do the objectual and substitutional theories ‘overlap’ and can ‘the quantifiers . . . be read with existential import’ (1993: 121). Janssen-Lauret nicely puts it in slogan form: ‘To be is to be the referent of a tag’ (2015: 159).

Importantly, Barcan Marcus must allow that the substitution class can be expanded beyond referring names to keep quantification ‘ontologically neutral’ (1993: 16–18, 79–82). Until we find out that the relevant substitution class contains tags, the true quantified sentences are generally ontologically neutral. Regarding properties, she writes, ‘Substitutional semantics permits quantifiers with predicates as substituends without a prima facie presumption of reference to universals’ (1993: 122). And she allows that the substitution class can contain nonreferring names, ‘non-tag singular terms’ (Janssen-Lauret 2015: 160), when she poignantly writes: ‘If naming may be viewed as the long finger of ostension over time, the case of a syntactical nonreferring name is like pointing at nothing (Barcan Marcus 1993: 210). Her approach is, thus, congenial to serious ontology. The substitution view gives the truth-conditions (though not the meaning [Boër 1975: 266]) of quantifiers, while ensuring that ‘quantification does not itself confer existence’ (Barcan Marcus 1993: 82; for an excellent discussion of Barcan Marcus's meta-ontology, see Janssen-Lauret 2015: §§5–6; 2016: §4.1). Serious ontologists would then emphasize that it takes nonlinguistic work to discover whether the singular terms composing their truth-conferring substitution class are referential or not.

This all reveals that serious ontologists can fruitfully accept generally ontologically neutral accounts of quantification such as Barcan Marcus's (or Meinongians’). But to ensure the ontological neutrality of the quantifiers, especially on Barcan Marcus's view, serious ontologists need singular terms not to be generally referential (see especially sections 3 and 4). Overall, we now have a novel way for serious ontologists to reject linguistic arguments, while accepting (commonsensically) that truths involving singular terms entail quantified truths.

3. Ramseyan Seriousness about Aboutness

Serious ontologists rejecting Syntactic Decisiveness and C₄ need not (only) be Meinongian or Barcan-Marcusian, but can be Ramseyan. Frank Ramsey writes:

It seems to me as clear as anything can be in philosophy that the two sentences ‘Socrates is wise’, ‘Wisdom is a characteristic of Socrates’ assert the same fact and express the same proposition. They are not, of course, the same sentence, but they have the same meaning, just as two sentences in two different languages can have the same meaning. (1925: 21)

Ramsey famously rejects drawing a particulars-universals distinction based on the linguistic fact of the subject-predicate distinction. maxim echoes his overall conclusion:

Nearly all philosophers . . . have been misled by language in a far more far-reaching way than that; that the whole theory of particulars and universals is due to mistaking for a fundamental characteristic of reality what is merely a characteristic of language. (1925: 22)

Ramsey argues that drawing ontological conclusions purely from linguistic distinctions is a mistake. That we normally talk about properties with predicates, rather than subject-position expressions, does not mean that we cannot talk about properties with subject-position expressions. But just because we nominalize the predicate expression ‘is wise’ with the name ‘Wisdom’ does not reveal a special sort of entity called a ‘universal’, distinct from another sort of entity called a ‘particular’ (see MacBride 2005 for further discussion).

A converse point can also be made about other singular terms that seem to refer to particulars, such as ‘the table’. If Unger (1979a, 1979b) is right, no tables exist. According to van Inwagen (1990: 108ff), the only things in the vicinity that exist are simples, most likely subatomic particles arranged table-wise (Unger [1979b: 41] does not endorse this). But according to serious ontology, we can still truly assert that the table is black. We can do this, then, without referring to anything. We are talking about a table and some way the table is, that is, its being black. But ultimately we might not strictly be talking about any entities at all (see Schipper 2018: §4.2.3; 2020: §2.3; Yablo [2014: 75–76] on proportionality; Pietroski's [2005: 276–77] France example). We might only be talking about ways things are with subatomic particles although we are not strictly talking about subatomic particles at all, but about a table.

Similarly, α is about ψ, the property of being a dog, but by using α, we need not be referring to an entity. Reflecting on our use of α reveals only that we can talk about properties with singular terms. We can also talk about properties with predicate-expressions such as ‘is a dog’. S₁ and S₂ are about the same subject matters: Fido and some way that Fido is: being a dog. These subject matters are just spoken of with syntactically different expressions: S₁'s predicate-expression contains the indefinite article ‘a’, while S₂'s says the same thing but contains a singular term. Understood in these ways, C₄ can be rejected: not all singular terms refer; sometimes they are used to talk about ways things are with other subject matters (e.g., with Fido or the table, which in turn may be a way things are with subatomic particles). We now have a different, serious-ontology-compatible way to reject Syntactic Decisiveness and thereby C₄:

Seriousness about Aboutness (SAAB): Some genuinely singular terms, α, used in true truthbearers need not refer to any entities at all, and they can be about ψ without referring to ψ or any entities at all.

The genuinely singular terms in S₁ and S₂, including ‘Fido’, may be about what we think they are about even if there are no such entities as the property of being a dog and even Fido. Given SAAB, the function of singular terms is not generally, and definitely not solely, to refer (compare with Barcan Marcus above). Syntactically singular terms are still decisively semantically singular, but this need not mean that they refer. Thus, Syntactic Decisiveness is false.

4. Two Kinds of Aboutness

There are two relevant, exclusive, and exhaustive views about aboutness that will help us to understand the proper linguistic framework of serious ontology and which I call representationalism and nonrepresentationalism (see also Schipper 2018: §5). According to

Representationalism: an intentional expression α, such as a truthbearer or a name, is about its intentional objects, for instance, ψ, when, and only when, ψ exists and satisfies the relevant satisfaction-conditions Φ.

Searle writes: ‘if nothing satisfies the referential portion of the representative content then the Intentional state does not have an Intentional object’ (1983: 17). Ceusters and Smith write: ‘an [intentional expression] must in every case be about some portion of reality, where the aboutness in question must always be veridical, so that “being about” is a success verb’ (2015: 3). The upshot is that if p's intentional objects do not exist, then p is not about anything.

By contrast, nonrepresentationalists (e.g., Crane 2001: 22ff.; 2013: 39ff.) argue for

Nonrepresentationalism: an intentional expression can be about ψ (or ψ being F) without ψ's existing (or ψ being F really being the case) and, thus, without there existing any ψ to satisfy any relevant satisfaction-conditions Φ.

Nonrepresentationalism is expressed here as the negation of representationalism. These two views are exhaustive and exclusive: one can either (i) think that there is aboutness only when the intentional object ψ exists and when ψ satisfies conditions Φ (representationalism) or (ii) deny this (nonrepresentationalism); one cannot do both or neither if one wants a complete account of aboutness.

4.1 Representationalism Fails to Help Serious Ontologists

If representationalists are right, then there is no aboutness when the putative intentional object, ψ, of a putative intentional expression α does not exist (or does not satisfy the conditions). This gives aboutness the same success-dependent structure as reference: when what α is used to refer to, ψ, does not exist, then α does not refer to ψ. If none of the entities ψ exist, then α fails to refer altogether. Thus, representationalism rejects SAAB. Therefore, serious ontology cannot be understood in this way.

Assume what serious ontology takes to be possible: that we can discover with nonlinguistic, metaphysical work that ψ does not exist and that S₁ is true. Then, the singular term would fail to refer to ψ. Given representationalism, however, ‘the property of being a dog’ would then not be about anything. According to representationalism, if a singular term purports to refer to an entity but fails, then there is no aboutness. If a sentence χ contains singular terms that are not about anything, then its content is gappy and it is not true. Thus, S₂ is not true according to representationalism plus the serious ontological assumption: ‘the property of being a dog’ attempts, but fails, to refer.

By contrast, S₁ does not contain singular terms that do not refer and, thus, contains no gappy content. S₁ is just true. Therefore, S₂ and S₁ are not equivalent. They turn out not to entail each other although they should.

This is problematic. This would mean rejecting the validity of nominalization-inferences (e.g., from S₁ to S₂). Again, this requires a deep revision of our normal linguistic practices. (However, recall that Rayo [2002: 437], Quine [1981: 92], and Schipper [2021: §5.2] are more open to revising ordinary linguistic practices than the main serious ontologists in this paper.) Accepting representationalism means rejecting this equivalence. Thus, serious ontology should reject representationalism and accept nonrepresentationalism.

4.2 Nonrepresentationalism: Success vs Satisfaction Conditions

Now I shall fill out the most plausible nonrepresentationalist picture. Recall, according to representationalists, say, a truthbearer of the form ‘o is F’ is only about ψ if there is a ψ that meets some, perhaps a cluster of, relevant, uniquely identifying, representational satisfaction-conditions such as that there is some ψ that is Φ₁, Φ₂, . . . , or Φ_n (see Searle 1958: 172f.; Strawson 1959: ch. 6; and, for criticism: Kripke 1980: 31ff., 55–67; Barcan Marcus 1993: 211–12). If no such ψ exists, then ‘o is F’ is not about anything.

By contrast, nonrepresentationalists think that ‘o is F’ or α can be about ψ even if there exists no ψ to satisfy any relevant, representational conditions, Φ₁, Φ₂, . . ., or Φ_n. Aboutness has success-conditions rather than satisfaction-conditions. A term's success-conditions vary according to what we are talking about. How α or ‘o is F’ manages to succeed to be about ψ can be understood in various ways, for instance, via a special causal link between α and ψ, via ψ's existence making-true sentences about ψ, via justificatory relations, or via some other way of focusing on ψ.

For instance, if ψ (say, Fido) exists, there are many ways that talking about ψ can succeed (e.g., via causal or noncausal, perceptual means). If Fido never existed, then the success of talking about Fido might require an appropriate causal link with whomever made him up. Or if Fido existed but sadly no longer exists, then success requires the appropriate connections with Fido from when he did exist to the utterance. If Fido is not the kind of thing that could have existed, say if he is just a way things are with subatomic particles and ways do not exist, then success requires other connections, maybe a causal connection with these subatomic particles. Success-conditions depend on the world, especially on Fido and matters relevant to Fido, and not on Fido meeting some linguistically accessible satisfaction-conditions. Furthermore, using ‘is F’, say ‘is a dog’, to say something about ψ, we talk directly and, given the right connections to dogs, successfully about being a dog, some way something could be, rather than about ψ itself, about the success-inducing dogs, or some other way things could be, such as being a wall. Using this predicate or its nominalization successfully, similarly, does not mean that there exists a way things could be.

Serious ontologists codify this as an epistemically plausible, methodological maxim:

serious epistemic maxim: it is never clear purely from what we say whether what we are talking about is an entity or something else, say, a set of entities, an amalgamation of entities, something mistaken such as a hallucinated or dreamt about object, etc.

Without more information about the world and the speaker, we cannot know whether what is talked about exists.

For instance, Sally is talking about a dog she met. After careful questioning, it is revealed that the dog is a figment of her imagination whom she only met in a dream. One can, then, conclude that

Negative Existential (Neg): The dog does not exist

and that she does not talk about the dog with the kind of success required for referring to the dog. But the success-conditions of talking about the dog need not mirror denotational or referential success-conditions. Sally can still successfully talk about the dog, merely by having dreamt of that dog rather than another dog. The dog does not exist, because it is merely dreamt of. Thus, for instance, it is false that the dog exists but true that the dog was merely dreamt of. One is explicitly talking about the dog and saying that he does not exist. One discovered not that anything else does not exist, say Sherlock Holmes or unicorns.

Arguably the original ontological turner, C. B. Martin (1980: 9) writes about ‘the passionfruit is round’ and ‘the passionfruit is purple’ that ‘in each case it is something in particular (and different) about the object [the passionfruit] that makes each statement true. The predicates are built to pick that out’. The semantic function of singular terms (including nominalizations of the predicates) includes introducing particularity and difference, to pick out and talk about some property sometimes, rather than to refer to or denote some entity (e.g., the passionfruit). And any proposed function of singular terms must be consistent with the fact that we pick out the relevant subject matters and meaningfully, coherently, and truly make such negative existential statements as Neg or ‘the property of being a dog doesn't exist’. This indicates that true sentences containing singular terms are not used in a way that entails the existence of what we are talking about. Thus, reference cannot be their only semantic function.

This way of distinguishing between referential and nonreferential functions of singular terms does not vindicate linguisticism as Hofweber's way does for it leaves open to metaphysical, rather than linguistic, discovery what the actual success-conditions are for some intentional expression (that is, whether they are referential or otherwise).

Here is a helpful tool: ψ, Sally's dog, being a dog, and being purple are all talked about intentional objects, that is, ‘just the schematic objects of attention’ (Crane 2001: 15) in the sense that they ‘clearly do not need to have anything in common with each other, except that they are objects of thought’ (Schipper 2018: 3702f.). This lack of commonality, this schematicity, extends not just to their not needing to share the same natures but to some of them not even existing (e.g., Sally's dog). The existence (and nature) of a schematic, intentional object of attention is something that we need further nonlinguistic investigation to discover. But, as we saw, even intentional expressions about nonexistent intentional objects have success-conditions. As we saw in this section, these are similarly external and not epistemically accessible merely by reflection on our language games and practices. This gives us a way of thinking about talking about ψ that undergirds serious ontology's acceptance of maxim.

Thus, competent speakers use ‘the property of being a dog’ to talk in a focused manner about the way that dogs are rather than a way that Tibbles the cat is. Much metaphysical work needs to be done to discover whether, as realists insist, being a dog is a special type of entity, or whether, as nominalists claim, being a dog does not exist but is something we only truly talk about with regard to dogs. For instance, the original serious ontologists Martin and Heil defend tropism via ‘the advancement of a plausible overall scheme, one that minimizes brute facts and meshes with empirical theorizing and common understanding’ (1999: 44, §6). Metaphysics takes much more work than merely accepting conventional, linguistic orthodoxies.

5. Embracing Nonrepresentationalism

5.1 The Argument from Comparison and Sensible Truthseekers

The main argument for antilinguisticism is an argument by comparison. Consider:

S₃: Fido appears to be a dog.

S₃*: Fido has the appearance of being a dog.

S₄: Fido is lucky to be a dog.

S₄*: Fido has the luck to be a dog.

Or less straightforwardly but just as validly:

S₅: Fido's flourishing was stunted by the lack of care by his owner.

S₅*: Fido was left in the lurch by his owner.

These inferences also use pleonastic transformations. But the existence of things called ‘lucks’, ‘lurches’, and ‘appearances’ is accepted as implausible (see also Dyke [2008: 1–6] on ‘Lurch Realism’ and Musgrave [2009: 66ff.], analogously, on creeps and ‘Creeps Realism’). Arguments to their existence from the validity of the transformations are not valid. The inference from S₁ to S₂ has the same structure as that from S₃ to S₃* and from S₄ to S₄*. S₁, like S₃ and S₄, is a normal predication. S₂, like S₃* and S₄*, is a nominalized version of the predication with a singular term. There is no special reason to distinguish property-cases from these cases. Therefore, the argument is invalid also in the case of properties. Even though the transformations are valid, to insist that the singular terms refer is to beg the question against nominalists.

When we use S₂, S₃, S₃*, S₄, and S₄*, we are clearly not just talking about Fido. We also talk about the property being a dog, his appearance, and his luck, respectively. Sensible and competent English speakers might say S₃-S₄*. They may not have thought of this without having read some philosophy, but when a metaphysician asks them whether appearances and lucks exist, it would be perfectly reasonable and coherent for them to say ‘I don't think so’ or ‘That's unlikely’. It follows from representationalism and linguisticism, but not from nonrepresentationalism and SAAB, that our fellow, sensible, and epistemically responsible truth-seekers would for the most part be not just in massive error but confused for taking nonsense to be meaningful. This consequence is implausible or at least unattractive. Normal, sensible people assert S₁ and S₂ as easily as they affirm S₃ to S₄*. This indicates that singular terms in S₁ and S₂ are used in ways similar to S₃* and S₄*: not in a way that commits one to anything more metaphysically and ontologically substantial. Thus, SAAB and serious ontology are preferable.

5.2 The Argument from Pleonasm and Redundancy

When we explicitly talk about Fido's dogness and luck with, for instance, singular terms, we are talking about them pleonastically (rather than referentially), in the sense found in any dictionary. For example, Dictionary.com defines ‘pleonasm’ as ‘the use of more words than are necessary to express an idea; redundancy’. Thus, the two phrases in the pairs S₃ and S₃*, S₁ and S₂, one of which nominalizes what we are talking about and the other of which does not are not just equivalent. In the ways relevant to our discussion, they even mean the same thing and are strictly and fully about the same things and the same ways things are with those objects. When we talk pleonastically about properties of Fido we are talking about properties, about how things are with Fido, merely in a different, though linguistically redundant, and perhaps at best wordy and at worst misleading, but still ontologically noncommittal, way (Schipper 2020: §1.1). In particular, one would be misled if one drew the linguistic ontologists’ conclusion that they are intended thus to be used referentially. Serious ontologists recommend caution. (See also Rayo and Yablo 2001: 79; Wright 2007: 153ff.; and Yablo's [1996: 267] criticism of Lewis's paraphrase argument for modal realism: he explains that we need not quantify entitatively or objectually over hows or ways although we talk about them.)

Thomasson (2015) agrees about the redundancy of property talk. She argues that we cannot explain why Fido is a dog by referring to Fido's having the property dogness because the latter is ‘just a redundant way of restating the former (introducing a new noun term for a property)’ (2015: 157). She compares this to explaining that poppies make us sleepy by referring to their dormitive virtue: ‘saying that something has the dormitive virtue is just a fancy way of saying they make us sleepy . . . and so is redundant[:] it clearly cannot . . . provide any explanation of the fact that poppies make us sleepy’ (2015: 156; for critical discussion, see also Hawley 2019: 232f.). In both cases, there is no explanatory value, because they only utilize transformations via conceptual truths. Unlike serious ontologists, Thomasson uses this to argue that dormitive virtues and properties exist.

Some serious ontologists might already react that if such posits are explanatorily empty, we cannot in fact draw the conclusion that they exist. Indeed, if nothing is added of explanatory or informational value, why think that we can draw a conclusion about the existence of dogness or dormitivity simply from their conceptual equivalences with ordinary predications? Either the existence of dogness was implicit in the predication and it thereby does not add anything, or it has to add something. Thomasson accepts that it is not explanatory, in fact because it does not add anything. Therefore, she must think that it is implicit. But is it plausible that quantification over redness or an existence claim about dogness is implicit in a simple predication?

Thomasson's view is that the redundancy is not explanatory, because x making y sleepy just is y having dormitive virtues and Fido being a dog just is Fido's having the property of dogness. Crucially, Thomasson thinks that these just is facts are conceptual truths. Are they?

Just is statements are understood, for instance, by Rayo (2013: 5) to be no difference statements. They are best understood to be either statements of identity (Hawley 2019: 233) or of grounding (Cameron 2014: 432ff.; Steinberg 2021: §§2&3). (Rayo [2013: §1.1, §2] takes ‘just is’ to be a symmetric, two-place sentential operator.) My point is this: The identity-statements or grounding-statements may be true, that is, an existence-fact may in fact be identical with or grounded in a fact expressed by a simple predication, or the other way around. But these identity-statements and grounding-statements cannot be conceptual truths.

Otherwise, especially nominalists simply do not understand the concept of predication because they think that predication does not involve existential quantification into predicate position. The same holds for anyone who thinks that it is a substantive question whether ‘being red’ and ‘is red’ or ‘being a dog’ and ‘is a dog’ require first-order quantification over an entity signified by the word after the copula.

Put this way, we can turn Thomasson's argument against linguistic ontology. Using ‘is red’ or ‘is a dog’ just does not conceptually involve implicit first-order quantification over redness or dogness. ‘Fido is a dog’ is normally formalized as Fa, where ‘a’ is a constant whose semantic value is Fido and ‘F’ stands for ‘is a dog’. With existential introduction, we can normally only infer $\exists xFx$, not $\exists x\exists y( x$ has y). Unless the standard formalizations and standard inference rules are conceptual falsehoods, Thomasson's view is false. They are legitimate and not conceptual falsehoods. Thus, Thomasson's view is false.

In Thomassonian linguistic ontologists’ favor, implicit first-order quantification, is perhaps the historically most natural way of understanding predication. For instance, Bradley's Regress is pressing because this is the natural metaphysically more detailed understanding of predication.

However, it is clearly not a conceptually required understanding of predication for otherwise there would, plausibly, be no way, on conceptual grounds, even to stop Bradley's Regress. Or, at least, many of the most promising solutions would be ruled out as conceptual falsehoods. For instance, Tractarian Wittgensteinians answer that properties and objects are nonexistent, mere abstractions from states of affairs that are the basic constituents of reality and thus do not require a relation between properties and objects. This might or might not stop the regress. But, it would, problematically, be eliminated on purely conceptual grounds if predication involves conceptually necessary existential commitments.

Further, the following possibility would be ruled out purely on conceptual grounds: that either ‘the house is red’ or ‘the keys are on the table’ does not commit us to the existence of anything answering to any of these sentences’ terms. But this is exactly what truthmaker theorists take to be a live option: that the things that exist do not structurally correspond to any part of a sentence (including the sentence itself), but that what exists is what makes the sentence true. Thomasson (2015: 157), who acknowledges that her ‘simple realism’ is incompatible with truthmaker theory, might respond that, indeed, truthmaker theory can be rejected on conceptual grounds and for conceptually misunderstanding predication.

However, this needs to be defended by the Thomassonian linguistic ontologist. Thomasson's (2015: 123) main engagement with truthmaker theory is just that her simple realism about X does not require her to posit X as an explanatory truthmaker. This obviously is not sufficient to undermine truthmaker theory on conceptual grounds. (Also, Lewis [2001: 611–12], Asay [2020: §2], and Schipper [2022: 3] argue that truthmakers need not be explanatory.)

Overall, it does not seem plausible that normal predications involve hidden, real existential commitments to entities corresponding to noncopulative terms in predicates.

5.3 Talking Pleonastically about ψ with Singular Terms

I propose this. We can talk pleonastically with singular terms about some subject matter ψ without ψ existing, either pleonastically or in a deflationary way as Schiffer and Thomasson claim, or substantially, or in any way at all. Pleonastic entities are explicitly redundant and hence unnecessary: they are prime targets for razors. Pleonastic talk is wordy but innocuous, a device of linguistic convenience and in itself does not reveal anything more substantial (e.g., no ontological commitments to entities). On this, we can learn again from Ramsey. Ramsey (1927) had similar thoughts about the truth-predicate itself, which we can simplify also for convenience. Although

t-equivalence: x is true if, and only if, p (e.g., take x to stand for some singular term for a truthbearer, such as the truthbearer in quotation marks, and p to stand for the truthbearer),

x's being true does not require there to be some property being true to be instantiated by x. The equivalence allows us to assert such things as that everything that Russell said in his 1921 Beijing lectures is true. This is equivalent to reasserting everything that Russell said in those lectures, but it is less time-consuming. Moreover, it allows us to talk about his statements together and say that they are all true or compare them with other statements, saying for instance that Russell said more true things in Beijing than in his 1940 Harvard lectures.

Talking about properties is similarly convenient. One can say that one loves all the properties of Fido, without asserting a long, and potentially infinite, list of sentences of the S₁ form. One can compare properties:

S₆: The properties of a good philosophy essay differ from the properties of a good piece of investigative journalism.

This is more convenient than:

S₇: A good philosophy essay is clear, defends a philosophical position, etc., while, by contrast, a good piece of investigative journalism protects its anonymous sources, reveals something important about current affairs, etc.

S₆ is a more general assertion of difference in properties and more convenient if one wants to state only the contrast and then quickly move on. Explicit talk about properties makes this convenience possible. Such talk is introduced into a language via a comprehension principle such as (see also Schindler 2021: 1, who calls this a ‘schema’):

p-equivalence: an object ψ has the property of being F iff F-ness is a property of ψ iff ψ is F.

p-equivalence is the kind of conceptual truth which Thomassonians have in mind. Indeed, it introduces property-talk into our language and allows for inferences such as from S₁ to S₂. However, such singular terms are plausibly introduced to talk about these properties more efficiently. Their raison d’être in practical terms is to increase our languages’ expressive power, not to expand our ontology. (Cf. Båve's 2015 discussion of the p-equivalence and this raison d’être for introducing property talk. Independently, he defends an error theory of properties [2015: 23]. Schindler [2021: §6] challenges Båve.)

Singular terms are not merely used to refer, and sometimes it would be to jump the gun to think that they do. That is, one cannot conclude this without further nonlinguistic argumentation. Minimally, when we talk about properties, we talk about them pleonastically merely as ways of being without committing ourselves to the existence of ways of being. Maybe properties exist; maybe they are tropes, Platonic universals, or something else. Serious ontologists insist, conservatively, that we cannot decide this based on the equivalence of sentences with and without singular terms for properties or because of the truth of principles like p-equivalence. Instead, serious ontologists, such as Martin and Heil, insist that we seek ‘unified and satisfying answers to persisting questions’ (1999: 49), including: ‘How do we stop Bradley's Regress?’ Practical convenience is sufficient for adopting talk about properties, but such talk would be as convenient if properties did not exist. Accordingly, such convenience provides no serious grounds for ontological expansion.

6. Conclusion

In sum, the antilinguisticist argument has two main steps, one epistemic, one linguistic.

(1) epistemic maxim: We cannot know whether the singular term α refers to ψ, a property, without knowing whether nominalism or realism is true. The answer cannot be read off merely from our ordinary acceptance of the inference from S₁ to S₂. We cannot adjudicate between metaphysical stories (or their unnecessariness) merely by looking at our use of singular terms. Further nonlinguistic argumentation is required.

Serious ontology is ontologically inclusive. It recommends methodological maxims, not ontological positions. Some nominalists reject only abstract objects. Others accept abstract objects but reject universals. Platonic realists accept universals. Russellian realists posit states of affairs as well as universals. Wittgensteinians posit only states of affairs and not universals or particulars. Nihilists reject everything and accept nothing ontologically. All can be serious ontologists if they respect maxim.

(2) linguistic step: Despite the lingering Fregeanism among some philosophers, singular terms are often used nonreferentially. The nonrepresentationalist linguistic picture (i) fits better with natural linguistic practices; (ii) it does not beg the question against nominalism, unlike linguistic arguments. And (iii), it provides a natural Ramseyan way to understand the redundancy of S₂ by analogy with the truth-predicate.

Singular terms are versatile. We can use them to talk about properties pleonastically, for convenience, or to compare, to single out some way of being rather than another, and so on. Their many uses leave open the possibility of substantial discoveries about what they are about.

We can distinguish two uses of singular terms: a relational, referential use where α only refers to β if β exists and a nonrelational or noncommittal use where α can be about β both in meaningful but false truthbearers and also in true truthbearers without β existing. There is then a general sense of aboutness in which α is about β, but where it is not clear, from the perspective of speakers and thinkers who use the singular term α prior to doing metaphysics, whether we are using ‘about’ in the first or the second way. Serious ontologists must accept the second as their general use.

One might wonder: ‘Is there a principled answer to the question whether any particular use of a singular term is referential or merely nonrelationally aboutness-bearing?’ Then one might object: ‘Without a principled answer, we have not moved an inch in the debate about whether the linguistic ontologist or the serious ontologist wins’.

Serious ontology's answer is that metaphysics must dictate which singular terms refer and which allow us merely to single out and focus, nonrelationally, on objects of thought and talk. If nihilism is true, nothing exists, and no singular terms refer. If realism about Fido is true, ‘Fido’ refers. Realists about the property being a dog will insist that S₂ is true and that ‘the property of being a dog’ refers. But neither type of realist, if serious, will accept that we can infer that a property exists from the truth of S₁ and of S₂.

To demand any other principled way of adjudicating between these different uses of singular terms, would mean misunderstanding and rejecting serious ontology, demanding that we read off our ontology from language. There is simply no principled way to decide whether ‘the property of being a dog’ refers prior to doing the hard, diverse metaphysical work of discovering whether there exist properties and in particular the property of being a dog. This, I have argued, is what follows from the serious ontologists’ rejection of linguistic arguments. We now better understand, with a little help from some friends, Barcan Marcus, Meinong (1904), Ramsey, and others, how serious ontologists should understand singular terms. We now also better understand their insistence that metaphysics, not language, comes first.