Existence questions

Abstract

I argue that thinking of existence questions as deep questions to be resolved by a distinctively philosophical discipline of ontology is misguided. I begin by examining how to understand the truth-conditions of existence claims, by way of understanding the rules of use for ‘exists’ and for general noun terms. This yields a straightforward method for resolving existence questions by a combination of conceptual analysis and empirical enquiry. It also provides a blueprint for arguing against most common proposals for uniform substantive ‘criteria of existence’, whether they involve mind-independence, possession of causal powers, observability, etc., and thus for showing that many arguments for denying entities (numbers, ordinary objects, fictional characters, propositions…) on grounds of their failure to meet one or more of these proposed existence criteria are mistaken.

Since Quine (1953/2001, p. 1), the fundamental question addressed in ontology has been: “What exists?” Quine’s simple answer ‘Everything’, while universally agreed upon, did not get us very far, and it might well be said that that’s been the beginning and end of agreement on the subject. And so ontology, the core of metaphysics, has concerned itself with addressing a variety of existence questions: Do numbers exist? Do fictional characters exist? Does consciousness exist? Do social entities, unobservable entities posited by physical science, propositions, or even our familiar tables and chairs exist? Little agreement, of course, has been found on these issues—instead, we have had an increasing variety of surprising answers to the question “What is there”, ranging from those who grant existence even to possibilia or fictional characters, to those who would deny that there are any composite material objects. And then there are those who draw the line in surprising places—accepting animals but not artefacts, apple trees but not apples.¹

Despite disagreement about the answers, there is this much agreement among those I will call ‘serious ontologists’: that existence questions like those above are deep philosophical questions, to be resolved by distinctively philosophical argumentation. The methods used to argue for different answers vary: negative answers are often based on the failure of the alleged entities to meet some substantive ‘criterion of existence’, such as mind-independence, contribution of (distinctive) causal powers, etc. Others ground answers on a Quinean approach that suggests that we accept the existence of all and only those entities over which we must quantify if the statements of our best scientific theories are to be true.

I will argue, however, that thinking of these existence questions as deep questions to be resolved by a distinctively philosophical discipline of ontology is misguided. I begin by examining how to understand the truth-conditions of existence claims. It is a plausible (though not universally held) view that our terms are governed by basic rules of use, such that being governed by those rules is constitutive of using those very (meaningful) terms. In Sect. 1 I will argue for what I take to be a basic rule of use for ‘exists’. In Sect. 2 I will discuss rules of use for general noun terms and suggest how these, combined with the rule of use for ‘exist’, help us assess the truth-values of existence claims. In Sect. 3 I will argue that the prior results provide a blueprint for arguing against most common proposals for uniform substantive ‘criteria of existence’, whether they involve mind-independence, possession of causal powers, observability, or whatnot, and thus for showing that many arguments for denying entities (numbers, ordinary objects, fictional characters, propositions…) on grounds of their failure to meet one or more of these proposed existence criteria are mistaken. Finally, as I will show in Sect. 4, taken together, the results of Sects. 1 and 2 provide a straightforward method for resolving existence questions by a combination of conceptual analysis and empirical enquiry. This leaves little room for a specifically philosophical discipline of ontology, conceived of as the study of what there is. In Sect. 5 I will consider what options remain for the serious ontologist to avoid these conclusions. I will argue that the most common ways of trying to revive a role for serious ontology (in the face of the above arguments) involve either misusing one or more of the terms in the existence questions they address or implicitly changing languages.

1 Existence and reference

There are notorious difficulties for understanding what it is to say that things of a certain general kind K do—or do not—exist. Hume famously held that the idea of existence ‘makes no addition’ to the idea of any object (Treatise of Human Nature, Bk.I, Part II, Sect. vi)—and if we accept that, it is difficult to see what the role is of positive claims of the form ‘Ks exist’. Worse still are problems explaining how negative existential claims can be true, since it seems that for a negative existential claim to be true, the term used in it must refer, in which case the claim would have to be false.

One classic way to avoid this problem is via semantic ascent, noting that—at least for most general nouns K²—the truth-conditions for the object-language claim ‘Ks don’t exist’ may be stated in the metalanguage: ‘Ks don’t exist’ is true just in case the term ‘K’ doesn’t refer—where the latter clearly does not require first referring to some object only to deny its existence. Using the T-schema, we can then get the following schema about existence:

• Ks exist if and only if ‘K’ refers.

This seems to fit well with our use of claims of existence and nonexistence, as nonexistence claims seem to implicitly indict a prior group of speakers as having used a term while (mistakenly) thinking it refers. Positive existence claims are seldom used in standard English, but where they are, it is always with a sort of ‘doch’ function: presupposing that others have doubted or denied the existence of Ks, and rejecting that denial—giving us an idea of what the role of positive existence claims could be.

Despite its attractions, however, certain standard objections have been raised to metalinguistic treatments of existence claims. The first standard counterexample is this:

• ‘K’ could refer and yet Ks not exist, if ‘K’ had a different meaning.

We can avoid this problem by taking a metasemantic rather than metalinguistic approach to existence claims. Following Horwich (1998, p. 118), we can use star quotes to pick out terms individuated not just phonologically or typographically, but on the basis of meaning, and propose instead the following schema as marking out a basic rule of use for ‘exists’:

• E: Ks exist iff *K* refers

The use of star quotes ensures that there is no risk of *K* having a different meaning, for if we change meanings, we no longer have that very term *K* at all.³

The second counterexample commonly raised for metalinguistic views also arises for metasemantic views, however:

• Ks could exist while it is not the case that *K* refers, if the term *K* didn’t exist.

But the existence of the term *K* is presupposed by the use of that term in instances of schema (E). We can, if we like, make that presupposition explicit by rewriting schema (E) as: Supposing the term *K* exists, *K* refers iff Ks exist.⁴ This is not a limiting constraint for our purposes, since the goal was to determine how we can understand and evaluate the truth of different existence claims made in metaphysical debates—and if we are to evaluate the truth of a claim that Ks do or don’t exist, the term *K* must exist as a constituent of the claim to be evaluated. In any case, none of the conclusions to be drawn below are at all threatened if the assumption that the term *K* exists is made explicit, and adding it clearly insulates us from the purported counterexample. Nonetheless, for simplicity I will leave it unstated below.

A third common objection is that this schema misrepresents existence claims as being about language, when they are really about the world and the things in it. But that is not so: on this view, existence claims are the object language correlates of reference claims, but involve a form of semantic descent from claims about reference to claims in the object language, using, rather than mentioning, the disputed terms. Since they use rather than mention the terms, existence claims themselves are of course about the world. The claim is not one of synonymy, but rather that schema (E) demonstrates a connection between the rules of use for our terms ‘refer’ and ‘exist’, which enables us to move up and down the semantic slide, from mentioning terms in discussing whether they refer, to using those terms in talking about whether or not entities of the sort exist.

Indeed the connection between these rules of use for ‘exists’ and ‘refers’ have been noted in another context: in the context of theories of reference. At least for most general nouns, instances of the following schema are platitudinous:

• 1(a) (If *K* refers at all) *K* refers to all and only Ks.

While this is a platitude that has been pointed out by minimalists about reference, it should be noted that one needn’t be a minimalist about reference to accept it—that also requires accepting that there is nothing more to reference than is captured by platitudes like these (Horwich 1998, p. 118).⁵

The schema can be stated more formally as follows (cf. Horwich 1998, p. 119):

• 1(b) ∀x(*K* refers to x iff Kx)

From this we can infer certain conclusions about the connection between quantification and reference, e.g.:

• 1(c) ∃x(*K* refers to x) iff ∃x(Kx)⁶

And if we preserve the Fregean assumption that Ks exist just in case ∃x(Kx),⁷ then we can (beginning from a platitude about reference) reach schema (E) again:

• E: Ks exist iff *K* refers

So we have again some evidence that there is a link between fundamental rules of use for our terms ‘refer’ and ‘exist’, expressed in this schema that enables us to semantically ascend and descend between talk about reference and talk about existence, and ensuring an equivalence in truth-value between claims that a given (meaningful) term *K* refers and claims that Ks exist.

This schema provides some hope of help with evaluating the truth of existence claims since, if an existence claim is made using some general noun *K*, that claim is true just in case the term *K* refers. But of course that only helps evaluate the truth of existence claims if we can say something more about the truth-conditions for claims about reference, leading to the question: when does a meaningful term *K* refer? If and only if Ks exist, of course (as our schema ensures)—but if that’s all that can be said, we cannot hope for much insight about the truth-conditions for existence claims by this route. Fortunately, I think a bit more can be said about when general nouns refer.

2 Reference and application conditions

A natural response to the question when a general noun *K* refers is something along these lines: *K* is associated with some basic rules of use that determine under what conditions it is and is not properly applied—call these ‘application conditions’. Then we can offer another general schema, this time about reference: for most general nouns *K*,

• R: *K* refers if and only if the application conditions for *K* are fulfilled.

I will not have space to argue for this view here, though I have done so elsewhere (2007a); here I will merely try to clarify the idea that our general nouns have application conditions and to respond to some obvious objections. As I conceive of them here, application conditions are among the semantic rules of use established by speakers through their normative practices of applying and refusing terms in various circumstances, and approving or correcting others in their use of them. Mastery of these rules of use is what, in turn, enables competent speakers to evaluate actual and hypothetical situations as ones in which their term *K* would or would not apply, and so also (via semantic descent) evaluate them as situations in which there would or would not be Ks.

Several objections might be raised, first, to the very idea that our terms have application conditions; second, to the thought that appeal to those could help us assess the truth of existence claims. First, causal theories of reference might be thought to threaten the idea that our general nouns have application conditions and refer just in case those application conditions are fulfilled. For according to such theories, the reference of certain terms—at least proper names and natural kind terms—is determined not by any conditions established by speakers, but rather by real causal relations in the world. I have discussed this threat extensively elsewhere (2007a, pp. 38–53) and so will only treat it briefly here.

Two different issues must be distinguished: is a causal relation supposed to determine whether or not a term refers, or rather to what it refers (provided it succeeds in referring)? While causal relations might be thought to aid in determining which thing a term refers to (assuming it refers), the presence of causal relations alone cannot determine whether or not a term refers. For there is always something with which the utterance of a term is connected causally in a (would-be) reference-grounding situation—even if only the speaker’s tongue or the suspended pollen in the surrounding air—and yet we do think our terms (even natural kind terms) may fail to refer.⁸ As causal theorists put it, these are cases in which the chain of reference ends in a ‘block’. So suppose that I attempt to name a new species of fish ‘baubleheads’, but all that is beneath my boat at the grounding is some seaweed and trash on the lake floor. What leads us to think that the grounding in these cases was ‘blocked’ so that the term failed to refer is the fact that (although my utterance is causally related to many things) my term was supposed to refer to a kind of creature.⁹

This suggests that our terms must at least have very basic grounding conditions, establishing what it takes for the term to acquire reference at all. In the case of ‘baublehead’, the grounding condition is plausibly that there is some kind of animal instantiated by all or most of the members of the sample ostended. From the initial grounding conditions we can derive frame-level application conditions: for ‘baublehead’ to refer (in circumstances C), the grounding condition must have been met, and there must (in C) be the same kind of animal as that referred to in the initial sample.¹⁰ (Notice that application conditions like these needn’t be purely ‘descriptive’ conditions, and may be deferential to various facts in the world, enabling us to incorporate insights of causal theories of reference.)

Other objections might be raised to the idea that our general nouns have application conditions—or that if they do they could help us evaluate the truth of existence claims. First, there may be scepticism our terms have informative application conditions. The history of philosophical attempts to discover necessary and sufficient conditions for the application of various central terms is a history of failures, and often the most that can be agreed on is the uninformative disquotational claim, e.g. ‘dog’ applies if and only if there’s a dog. But we clearly cannot expect to get any mileage out of being able to evaluate our existence claims via claims about reference, and those about reference via appeal to application conditions, if those application conditions just say that the term applies (in a situation) just in case a thing of the relevant sort exists (in that situation).

But some terms clearly do have application conditions that can be stated without just saying that the term *K* applies if and only if Ks exist; e.g. we can famously say that ‘bachelor’ is properly applied if and only if there is an unmarried man. More interestingly, application conditions for terms for institutional and legal kinds can generally be given in other terms, as terms for contracts, debts, touchdowns, etc. are explicitly introduced by way of specifying (in other terms) conditions under which a contract is made, a debt incurred, a touchdown scored. Though these may only involve (e.g.) an open list of sufficient conditions for application of the term, rather than a set of necessary and sufficient conditions, they may still be useful in evaluating existence claims. So, e.g., supposing it is sufficient for the term ‘law’ to apply that the members of the legislature vote for a bill which the president signs, we can infer that the term refers and so that there is a law if those conditions are fulfilled (where those conditions don’t appeal to the existence of a law).

Even if application conditions like these for high-level terms are acknowledged, however, the above examples of application conditions appeal to the existence of other sorts of entity (persons, papers, etc.)—requiring that *J* refers if and/or only if (some other sort of entity) Ks exist and…. So while appealing to these high-level application conditions might help resolve high-level existence questions (e.g. about the existence of laws, debts, or touchdowns), it might be thought to simply push back the problem for more basic existence questions (about the existence of Ks).¹¹ Indeed it might be thought that social and legal terms are the exceptions, and that application conditions for the more basic terms of our language still cannot be stated informatively—after all, competent speakers typically can state application conditions for hardly any of the terms they use. But if the application conditions for basic terms can only be stated uninformatively as: *K* refers iff Ks exist, it might seem that we will still have no help addressing basic existence questions.

But for a term to have application conditions is not the same as for those application conditions to be (informatively) stateable—by competent speakers or anyone else. The fact that competent speakers typically cannot state application conditions for most of the terms they use is no evidence at all against the idea that our terms have application conditions. Application conditions should be thought of as semantic rules analogous to grammatical rules; just as competent speakers must be masters of following grammatical rules, but need not be capable of stating them (although it is plausibly those speakers’ normative practices in speaking, correcting the speech of others, etc., that fixes the grammatical rules for a particular language), so must competent speakers be masters at following the semantic rules—but need not be capable of stating them. Moreover, even if the only way application conditions for a given term could be stated in English involves appeal to the existence of certain entities (of another kind or that very kind), this might be visibly avoidable by instead stating the application conditions in what Hawthorne and Cortens (1995) call a ‘feature-placing language’.

Most importantly, however, for a term to have application conditions does not require that those conditions be stateable at all. For it is plausible that any language must include some ‘semantically basic terms’¹²—that is, terms that cannot be learned just by way of learning definitions stated in other terms.¹³ And if that is so, we must allow that for at least some terms, the application conditions (considered as semantic rules of use) needn’t be capable of being (informatively) stated to be learned and to be in force. Instead, speakers may learn to master the rules of use for those terms by other means, e.g., ostensively as we learn that a term is to be applied in situations like this (and not in situations like that), or via judgments of similarity to ostended paradigms—where this does not require saying ‘*K* applies if and only if Ks exist’, or even appealing to the existence of other entities (as in ‘*K* applies if Js exist and…’). And in fact, we teach our children words not by telling them (unhelpfully) that they should apply *dog* wherever a dog exists, but rather by simply demonstratively applying *dog* in some situations and refusing it in others, and applauding or correcting their attempted applications in various situations.

In short, rather than thinking of application conditions as definitions competent speakers (or anyone else) could recite, we should instead think of them as rules for when it is and is not proper to use a term, which speakers master in acquiring competence with applying and refusing a new term in various situations, and that (once mastered) enable competent speakers to evaluate whether or not the term would properly be applied in a range of actual and hypothetical situations.¹⁴ If we think of application conditions this way, we can after all hope to gain help in evaluating existence claims via claims about reference, and those about reference via appeal to application conditions.

3 Consequences for existence conditions

A great many different criteria have been proposed and utilized as conditions for what it takes for entities to exist. Prominent among these is the Eleatic Criterion endorsed by Armstrong (1978, vol. 2, p. 5) and by many others—that only those things that have causal powers exist.¹⁵ Another common proposal is that for things to exist they must be (in some sense) mind-independent—as Lakoff (1987, p. 164) puts it “Existence cannot depend in any way on human cognition”. Azzouni (2004, p. 113), while denying that there are philosophically conclusive arguments for any criterion of existence, nonetheless holds that the criterion at work in our society is ontological independence from “any psychological or linguistic process whatever”. Elder (2004) proposes the distinct but related criterion that objects are real only if they have “real natures”, where this is a matter of having essential properties whose status as essential is mind-independent. Still other criteria for existence are sometimes considered, e.g. trackability, observability,¹⁶ or other forms of epistemic robustness (Elder 1989, p. 440).¹⁷

I will not discuss these proposals separately and in detail, but rather will argue that if we accept instances of the above schemas (E) and (R), we have reason to reject all of these proposals. For (R) ensures that our general nouns refer just in case their basic application conditions are met, where these are semantic rules of use established by speakers in determining when their term is/is not properly applied. And given schema (E), if a general noun refers (has an extension) then entities of the corresponding type exist. This gives us a blueprint for generating counterexamples to any proposed across-the-board criterion of existence: we can show that any proposed criterion is mistaken if we can show that there is some term whose application conditions (established by speakers) are fulfilled although the proposed existence criterion is not met. For if the application conditions are met, then (using an instance of (R) we can conclude that) the term refers, so by using an instance of schema (E) we can infer that the entities in question exist despite failing to meet the proposed existence criterion—showing that meeting the proposed existence criterion cannot be necessary for existence.

Consider, for example, terms for everyday social and cultural entities, including artifactual kind terms like ‘table’ and ‘chair’, terms for works of art like ‘painting’ and ‘sculpture’, and terms for social and institutional entities like ‘baseball’ and ‘money’. Even if we can’t properly state the application conditions for terms like ‘table’ and ‘chair’, it’s clear that most dining rooms provide sufficient conditions for these terms to refer according to the speakers’ ordinary standards, most art galleries meet conditions that ensure ‘painting’ and ‘sculpture’ refer, and so on. And so it seems that (barring radical sceptical hypotheses that would take speakers to be badly deceived in thinking that the relevant conditions are fulfilled in dining rooms and art galleries when they are not) the conditions ordinarily required for these terms to apply are fulfilled.¹⁸ As a result, the terms refer, and (given schema E) we can conclude that tables, chairs, paintings, and sculptures exist. But the corresponding terms require for their application that there be intentional actions or practices of various sorts, and so if they succeed in referring, the entities they refer to must fail the mind-independence condition—at least on the usual construals that take mind-independence to mean that they could exist even if no minds, or psychological events or processes, existed.

Of course it is open to the defender of the mind-independence criterion to define ‘mind-independent’ in some other way, such that these purported entities would not violate it. We could, e.g. (with Elder 2004) take it to require possessing a nature independently of minds (not to require existing independently of minds); or (with Elder 1989) take it to require certain kinds of epistemic independence. I won’t discuss these options in detail here, though I have argued at length elsewhere (2003, 2007a) that artefacts (to name but one case) violate these criteria, although the application conditions standardly associated with the term ‘artefact’ are nonetheless met in spades.

What about the Eleatic criterion? This is interpreted in a number of ways—some take it as the requirement that any entity that exist contribute novel causal powers—i.e. powers ‘over and above’ those already contributed by any other entity (or entities) posited. So interpreted, it is often argued (Merricks 2001, p. 81; van Inwagen 1990, p. 122) that all purported inanimate composite objects (including not just artefacts but also natural objects like sticks and stones) violate this criterion, since all of their alleged work may be better accounted for by the work of the microscopic entities that (allegedly) compose them.¹⁹ But (as I have argued elsewhere (2007a, pp. 155–159)), a term like ‘table’, as ordinarily used, is guaranteed to apply given merely what the eliminativist calls ‘particles arranged tablewise’, that is, intentionally arranged in a characteristic shape and so that they may (collectively) perform certain characteristic functions—without any requirement that a table contribute causal powers ‘over and above’ those of its parts working together. But if the application conditions ordinarily associated with ‘table’ are met, then tables exist even if they do not contribute ‘novel’ causal powers, and so we should abandon this version of the Eleatic criterion as an across-the-board criterion for existence.

On other formulations, the Eleatic criterion is taken to require not that a purported entity contribute novel causal powers, but only that it contribute some causal powers. This is more plausibly thought to be a requirement for terms like ‘table’ and ‘stone’ to apply; it seems that on this formulation, macroscopic inanimate physical objects would pass the Eleatic criterion. But there are nonetheless a great many common sense terms that have application conditions that may be fulfilled without that criterion being met. Consider, for example, terms such as ‘story’ and ‘symphony’.²⁰ Barring radical sceptical hypotheses, it seems that the standard application conditions associated with these terms are regularly fulfilled (given the relevant creative acts, performances, etc.), and so that such things exist. But if such things do exist, as normally understood they cannot be identified with any spatiotemporally located entities such as copies of texts or performances. And so (if we assume that causation requires spatio-temporal location) they cannot be thought to directly have any causal impact on the world at all, although their copies and performances, and our thoughts and beliefs about them obviously may have a great deal of impact. So again if we attend to the established application conditions for common-sense terms like ‘story’ or ‘symphony’, it follows that these terms refer and so (given schema (E)) that stories, symphonies and the like exist, despite their apparent violation of the Eleatic criterion (as well as the mind-independence criterion, and presumably the trackability criterion as well).

Similar troubles arise for those who take a Quinean approach to existence questions, holding that we should accept the existence of all and only things over which we must quantify to make the statements of our best scientific theories true. For again, even if some term, say, ‘hat’, doesn’t appear at all in our best scientific theories, if we accept schema (E) and the view of reference defended in Sect. 2, hats nonetheless exist, provided the standardly accepted application conditions for ‘hat’ are met—and (barring radical sceptical hypotheses or conspiracy theories) it is clear that they are.²¹ We might hope that the Quinean approach will help us to sort out those entities that are scientifically basic from those that are not, but given the above understanding of the rules of use for ‘exist’ and the standard application conditions for our terms, we can’t use it to tell us what does (and does not) exist.

The mistake that lies behind proposing the various uniform criteria for existence seems to come, in each case, from finding an application condition that holds for terms of a certain kind, and a corresponding condition for the existence of things of a certain kind, and illegitimately generalizing it as an across-the-board criterion for the existence of ‘anything whatsoever’. But even if, e.g. mind-independence or contribution of causal powers are genuine criteria for the existence of basic physical entities, it does not follow that these are criteria for the existence of anything whatsoever. Careful consideration of a broader range of examples suggests that there are different application conditions for terms of different kinds. Given the connections between application conditions, reference, and existence, the application conditions for our terms may (by semantic descent) be expressed in the object-language as existence conditions for the objects (if any) referred to by those terms, and the conditions for the existence of objects of different kinds may be as various as the application conditions for our terms. So while each of these proposed criteria might be relevant as conditions for the existence of some sorts of things (e.g. mind-independence might be a relevant criterion for fundamental particles or mountains), they are not legitimately used as criteria for the existence of anything whatsoever (e.g. mind-independence may not be legitimately used as a criterion for the existence of money or works of art).

In fact, the application conditions for our terms seem to be so various (consider the differences in conditions under which terms like ‘electron’, ‘mountain’, ‘debt’, ‘Thursday’, ‘public holiday’, ‘jury’, ‘fictional character’ and ‘symphony’ would properly apply) that the prospects seem very dim indeed for finding a single feature required for the application of each of these terms, and thus required for things of any sort to exist (pace Azzouni pp. 112–113). So if we accept instances of schemas (R) and (E) above, we have reason to doubt that there is any single criterion of existence that is shared across the board for (putative) objects of any kind. We also have reason to call into question many arguments for eliminating entities of various sorts (including mathematical entities, propositions, fictional characters, artefacts, composite macroscopic objects, etc.) on grounds of their failure to meet some across-the-board criterion for existence.²²

Moreover, if the work of Sects. 1 and 2 is correct, arguments for eliminating ordinary objects often rely on misusing language: e.g. those who deny the existence of tables violate a core rule of use either for ‘exist’ or for ‘table’. If they deny that ‘table’ refers, they must be artificially inflating its application conditions beyond the constitutive rules for its use established by speakers (cf. my 2007a, pp. 155–159). If they accept that ‘table’ refers, but deny that tables exist, they violate the link in the rules of use for ‘exists’ and ‘refers’ captured by schema (E). Either way, they are either misusing the standard English terms, or implicitly introducing new terms with different rules of use. I will return to this point in Sect. 5 below.

4 Consequences for ontology

Instead of answering existence questions by determining whether or not some across-the-board substantive existence condition is met, the results of Sects. 1 and 2 above leave us with a very straightforward, pedestrian approach to existence questions—one I have implicitly used in Sect. 3 above. The method is this: We may determine whether entities of a given kind (K) exist by first determining what the application conditions associated with *K* are, and then determining whether they are fulfilled. If they are, then (by application of schema (R)) the term refers and (by application of schema (E)) Ks exist.²³

Thus understood, while there is no shared substantive criterion for existence, there is instead a purely formal criterion for existence—namely that enshrined in schema (E): (provided we have a term *K*), we can say that entities of kind K exist iff *K* refers (has a non-empty extension). Given the variety of different application conditions (conditions under which a term refers) associated with different terms, the substantive conditions on the world that must be fulfilled for terms of different kinds to refer will nonetheless vary considerably—and so (via semantic descent) we can also say that the conditions of existence for things of different kinds will vary considerably.

While this approach will tell us that there are more things on heaven and earth than are acknowledged by those who use across-the-board substantive existence criteria, it does not leave us accepting the existence of such (putative) things as phlogiston and witches. For the application conditions associated with those terms turned out not to be met. Since ‘phlogiston’ was supposed to refer to a kind of chemical that was released during combustion (and no such chemical was released), the application conditions were unfulfilled, ‘phlogiston’ doesn’t refer, and (by schema (E)) phlogiston doesn’t exist. By the same token, if ‘witch’ was supposed to refer to any woman endowed with supernatural powers in virtue of making a pact with the devil, in the absence of such powers and/or devil, ‘witch’ fails to refer, and (given schema (E)) witches don’t exist. So this method of handling existence questions apparently gives us the right results, denying existence just where doing so is the most obviously correct thing to do—where speakers made some mistake in thinking that the application conditions associated with their term were fulfilled, although they were not.

This method makes addressing existence questions so straightforward, however, that it leaves very little room for doing ontology, if we conceive of this as a distinctively philosophical enterprise of figuring out what really exists. For the first step involves conceptual analysis, aiming to determine conceptually relevant conditions for the term to apply, or (correlatively—in the object language) for entities of the sort to exist.²⁴ Many ontological disputes might be understood as implicitly disputes about what it takes for there to be, e.g., fictional characters, numbers, properties, propositions, etc., and the work may be difficult and contentious—especially where the terms used are distinctively philosophical terms that lack rules of use established by ordinary English speakers. But (taken alone) it is not a matter of figuring out what ‘really exists’ by distinctively philosophical means. The second step is straightforward empirical enquiry (cf. my 2007a, pp. 199–201). The empirical work required is often easy (e.g. about whether the situations for ‘chair’ and ‘table’ to apply are met), enabling the existence question to be quickly answered. But even where the empirical work is difficult, it is work for scientists or investigative journalists, not for a distinctively philosophical discipline of ontology.

5 Lines of reply

If the analyses in Sects. 1 and 2 of the rules of use for ‘exists’ and for general noun terms are correct, I have argued, it follows that common proposals for substantive across-the-board criteria of existence are wrong, a great many arguments for eliminating various sorts of entity should be rejected, and existence questions in ontology may be answered straightforwardly by a combination of conceptual and empirical inquiry. It may be useful, however, to close by considering some lines of reply that remain open to those metaphysicians who find these conclusions intolerable.

One option is to accept instances of schemas (E) and (R), but deny that ontological questions are resolvable simply by empirically determining whether or not the associated application conditions are fulfilled. For, it is often said, the application conditions for any of these disputed terms include that there be an object there—and the question of whether some object present is a distinctively ontological (not empirical or conceptual) matter. I have dealt with this response elsewhere (2007a, pp. 157–158; 2008) by distinguishing different uses of the term ‘object’. If ‘object’ is used with its own first-order application conditions, then the question ‘is there an object there?’ is itself straightforwardly resolvable. If (as seems more likely) ‘object’ is used as a covering term guaranteed to apply if some first-order sortal does, then we cannot deny that there is, e.g., a table, symphony, or proposition in a given situation on grounds that there is no object; instead we can only answer the question ‘is there some object’ via answering the various sortal-specific existence questions. Finally, if ‘object’ is used without application conditions, ‘is there an object there?’ is an ill-formed and unanswerable question.

Another option is to hold that while fulfilling application conditions may be sufficient for reference in a thin or loose sense, it is not sufficient for reference in a thick or heavyweight sense, so that we should reject schema (R), at least for the heavyweight sense of REFER, though we may retain schema (E), provided it ties existence to REFERENCE.²⁵ We must then ask what is required for REFERENCE in this stronger sense; if the answer is that it requires that there be some (perhaps mind-independent, discourse-independent) object referred to, the same problems above arise again.²⁶

A more common option for those hoping to avoid the above conclusions is to reject instances of schema (E)—or more commonly, to accept that schema (E) expresses a rule of use for an everyday, ‘thin’ sense of existence, but to deny that it holds for a ‘thick’ sense of EXISTENCE, where the latter is what’s at stake in ontological debates. In fact, it has become commonplace for those defending serious ontology to argue that they are using ‘existence’ or the quantifier in a ‘thick’ or ‘strict’ sense (Dorr 2005; Hofweber 2005; Sider 2008), distinct from the ordinary use that might be captured by schema (E). It is up to the heavyweight ontologist to say what ‘EXISTENCE’ in this ‘thick’ sense means. I have elsewhere (2007a, pp. 110–125, and 2008) attempted to cast doubt on the meaningfulness of various attempts to explicate a thick sense of existence, though new attempts of course must be separately evaluated.

Those serious ontologists seeking higher standards for EXISTENCE typically reject the idea that reference is sufficient for existence, and so reject the ‘if’ direction of schema (E) (Ks exist if *K* refers). Note, however, that if schema (E) really does (as I have argued) lay out a core rule of use for the English term *exists* (where, again, star quotes are used to mark that we individuate terms by their meaning), then as long as we think of meaning as fixed by core rules of use, serious ontologists cannot be properly using the standard English term *exists* in their ontological work. Where they employ the same typography or sounds, they must either be misusing the familiar English term (by intending to use the familiar term and thus subjecting themselves to its standards of use, yet violating a core rule of use) or, more plausibly, implicitly introducing a new homophonic term *EXISTS* governed by different (though perhaps related) rules.²⁷

In the latter case, the serious ontologist’s claims seem best understood as implicit proposals that we adopt a new language, not as claims about what really exists that might surprise or enlighten speakers of English who consider existence questions using that language. Indeed we might ask why those who look to ontology to find out what *exists* should care about the serious ontologist’s claims, or why the serious ontologist should not describe what she is doing in less misleading terms (e.g. as investigating which entities are fundamental).²⁸ In any case, such proposals of new languages are most naturally evaluated in a pragmatic Carnapian vein, as proposals to adopt a new language to better serve certain purpose(s).²⁹

In any case, if serious ontologists are implicitly changing languages, then their work gives no reason for doubting that the above account gives an appropriate and easy method of answering existence questions, whenever these are asked using our familiar English language.