1 Easy Ontology, Regress and Circularity

Easy ontology is a deflationary approach to existence. Central to the theory is the claim that:

(E): Ks exist iff the application conditions actually associated with ‘K’ are fulfilled (Thomasson, 2015: 86).

(E), along with other elements, leads to a deflationary view of existence as ‘we needn’t first settle existence questions of the form ‘‘Do Ks exist?’’ in order to settle questions about whether the application conditions for ‘‘K’’ are fulfilled’ (Thomasson, 2015: 97), where application conditions are rules of use for ‘K’. Application conditions are meaning-constituting, such that ‘a competent speaker has only to make use of her mastery of the rules of use of the term—combined with her access to any relevant empirical information—to determine whether the application conditions are met, and thus to evaluate whether the corresponding entities exist’ (Thomasson, 2015: 113).

For example, rather than requiring ‘deep’ ontological debate, we can conclude that tables exist in virtue of grasping the rules of use of the term, and empirically seeing that those application conditions are fulfilled. This means that answering existence questions requires nothing ‘epistemically metaphysical’ (Thomasson, 2015: 113). All we need is either conceptual or empirical work to find out whether the application conditions a term are fulfilled.

There are a number of further caveats and complications concerning application conditions, but the focus of the discussion in this paper is the restriction that:

“(4) Application conditions must not take the following form: ‘K’ applies iff Ks exist” (Thomasson, 2015: 96)

This restriction is important as it establishes that we cannot appeal to the existence of Ks in order to establish that the application conditions for ‘K’ are fulfilled, and thus that ‘we need not settle existence questions before we can settle questions about whether we may introduce new concepts or terms and indeed about whether those concepts or terms refer’ (Thomasson, 2015: 97).

Brenner (2018) argues that the conjunction of (E) and (4) leads to both an infinite regress and circularity. To see why, take ‘non-trivial application conditions’ to be those application conditions that require more of the world than mere conceptual truths to be fulfilled. That is, the fulfilment of the application conditions makes some demand on the world. For example, if the application conditions for ‘table’ include that there are particles-arranged-tablewise, then those application conditions are only fulfilled if there are particles arranged tablewise. This is in contrast to terms that have ‘trivial’ application conditions where conceptual analysis is sufficient to conclude that those application conditions are fulfilled. For example, ‘there are dogs or there are not dogs’ is a conceptual truth, from which we can infer the property of being a dog is or is not instantiated’, and from this we can infer ‘the property of being a dog exists’ and ‘there is a property’ (Thomasson, 2015: 150–151).Footnote 1

Now say that some term ‘K’ has non-trivial application conditions, where the application conditions for ‘K’ appeal to the existence of something else falling under a different sortal, ‘X’. This means that for the application conditions for term ‘K’ to be fulfilled, Xs must exist. Brenner then asks us to consider the application conditions for the sortal ‘X’. The application conditions for ‘X’ might also be non-trivial, requiring that Ys exist. What about the application conditions for the sortal ‘Y’? Again, the application conditions for ‘Y’ may be non-trivial requiring that Zs exist, and so on and so on. This is a problem as it means that, for some term ‘K’ with non-trivial application conditions, there could be an infinite regress of application conditions, and hence that we can never conclude that Ks exist (Brenner, 2018: 607–608).Footnote 2

Similar considerations give rise to a circularity problem. Considering again the application conditions for the sortal ‘X’, it might be the case that ‘X’’s application conditions require that Ys exist, and ‘Y’’s application conditions require Zs exist, and that application conditions for ‘Z’ require that Xs exist (Brenner, 2018: 608). In such cases, the application conditions for the term ‘X’’s will directly contradict the restriction imposed in (4) as the application conditions for ‘X’ will include the existence of Xs through this chain of non-trivial requirements for the fulfilment of the application conditions.Footnote 3

Brenner argues that both of these are serious problems for the easy ontologist. The infinite regress means we cannot explain how people could come to learn application conditions for terms with non-trivial application conditions given the infinite regress of application conditions (2018: 608), whilst the circularity of application conditions violates a central constraint within the easy ontology that for a sortal term K we need not discover whether Ks exist in order to determine whether the application conditions for ‘K’ are satisfied (2018: 609).

2 An Immediate Response

One possible response to these arguments can be found already in Thomasson’s work. In his argument, Brenner assumes that the application conditions must be stated in other terms of the language. This is required for both the circularity and the infinite regress identified, as both are only possible if it is the case that the application conditions for ‘X’ are stated in terms of ‘Y’, which are then stated in terms of ‘Z’, etc.

However, Thomasson is clear that application conditions for terms need not be statable at all (2015: 92). Rather, Thomasson thinks that for at least some terms:

‘speakers may learn to master the rules of use for those terms by other means, for example, 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’ (2015: 92).

Indeed, Thomasson also suggests that some concepts might be innate, only requiring us to match the concept with the terms of a language (2015: 93). More generally, Thomasson goes on to say that application conditions are more like rules of use, not something that even competent speakers should be required to be able to recite (2015: 93). Brenner acknowledges that if Thomasson is right on these points then there is no threat of either regress or circularity. However, he thinks the response inadequate for two reasons.

First, Brenner thinks it is ‘implausible that an infinite sequence of application conditions could be (or would be, for finite beings like us) innate, or learnable via ostension’ (2018: 608). We might then respond that the entire sequence need not be innate or learned via ostension, but even if only some terms were innate or learned via ostension, as Brenner notes, how could we ever know that a chain of non-trivial application conditions, even one that does end in an ostensively defined term, is met? (2018: 608–609). To know whether it is met would require us to grasp all of the chain of application conditions. Positing innate terms or ones learned by ostension then might solve the learnability problem for creatures capable of grasping infinite sequences, but it is not solvable for creatures like us, and hence cannot serve as the basis for making ontology easy for us.

Second, Brenner argues that this response misses the main target of his argument. The argument is not about the terms we have at our disposal in order to state application conditions. Rather it is about a regress/circularity of the satisfaction of application conditions (2018: 208fn). For example, there could be an infinite regress of application conditions in that for the application conditions of one term to be satisfied relies on the satisfaction of the application conditions another term, which in turn relies on the satisfaction of the application conditions of another, etc. All of this is possible without us having, within our language, the terms to adequately describe that regress.

Both of these responses strike me as, if not knock-down arguments, persuasive. Others will likely disagree, and think that the introduction of innate or ostensively defined terms allows for us to avoid the regress Brenner identifies, and hence solves the learnability problem. It is not my aim to decisively argue one way or another on this matter in this paper. Rather, my aim is to suggest a completely new route of response. Thus, for the purposes of this paper, I will agree with Brenner that both the regress and circularity are major issues for the easy ontologist, but I will argue both problems can be avoided if the easy ontologist accepts a holistic account of application conditions.

3 Atomism and Holism

There are many different forms of semantic holism (see Jackman, 2014 for discussion of some of these). Here, I will follow a definition of holism provided by Dresner in his survey article:

Meaning Holism: the ‘thesis that the relations of a certain kind (or kinds) that obtain among the expressions of a given natural language (all of these expressions, or many of them) are constitutive for linguistic expressions to mean what they do’ (Dresner, 2012: 611).

Putting this another way, the idea is that terms mean what they mean in virtue of the relations that they stand in to other terms in a language. The meaning of any term in a language is constituted by the relations that obtain between that term and some other terms in that language. This means that the meanings of all terms in a language are interdependent; the meaning of all terms can thus be said to be ‘externally constituted’ (Dresner, 2012: 612) through the global relations that any given term bears to other expressions within a language.

As Dresner notes, this definition is more general than some in that it does not specify what kinds of relations hold between the expressions that constitute meaning (cf. Antony, 1993; Devitt, 1996), weaker than Quine’s (1951) suggestion that only language as a whole can be said to be meaningful, but stronger than Fodor and Lepore’s (1992) holism which holds that if one thing has a property then many things must have it (cf. also Pagin, 2006).

This is in contrast to semantic atomism, the view that there is at least some term in the language has a meaning that is independent of the meaning of every other term in the language. Or, more broadly, that (at least some) linguistic expressions have their meaning independently of each other. Note that this conception of atomism is taken to be neutral about how linguistic expressions acquire their meaning, just so long as it is possible that no connection to the meaning of a further linguistic expression is required.

Before we can consider holism and atomism in the context of easy ontology, one small change is needed first. Thomasson’s account of easy ontology does not normally talk about the meaning of terms, but instead their ‘application conditions’, where application conditions ‘are certain basic rules of use that are among those that are meaning-constituting for the term’ (2015: 89, my emphasis).

Adjusting to this, we can distinguish between two views about the nature of application conditions: an atomistic conception of application conditions wherein at least some term in the language has application conditions that are independent of every other term’s application conditions; and a holistic view wherein the application conditions of all terms within a language are interdependent. Depending on which we view we accept, we arrive a subtlety different versions of easy ontology: an atomistic conception where there are terms within the language that has application conditions that are independent of every other term’s applications, and a holistic account where the application conditions of terms are interdependent.

Which version of easy ontology does Thomasson endorse? On the side of holism, Thomasson places a restriction on introducing new sortal terms such that the ‘term(s) must be introduced via a conditional that gives sufficient conditions for its(/their) application, stated using the extant terms of L and/or other minimally introduced terms’, and Thomasson is explicit that we ‘may simultaneously introduce interrelated terms, as with the various terms of baseball and other games, or for corporations, stock, and shareholders, or other social and institutional entities’ (2015: 263). This suggests that at least some terms have application conditions that are interrelated and can only be understood in connection with the application conditions of other terms.

In contrast, on the side of atomism, Thomasson says that ‘it is plausible that any language must include some ‘semantically basic terms’. These are terms that cannot be ‘learned just by way of learning definitions stated in other terms’, meaning that ‘at least for 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’ (2015: 92). These basic terms may even be innate (2015: 90) or arise from basic cognitive functions, in line with some forms of atomism.

However, it would be too quick to read these passages as necessarily ruling out holism. The existence of innate semantically basic terms are not incompatible with holism so long as a semantically basic term only has their meaning through the relations that obtain between it and other semantically basic terms. Holism is compatible with there being more semantically basic terms, only ruling out the possibility of there being only one semantically basic term.

That said, readers may still think that there is too much evidence against a holistic reading of Thomasson. For example, Thomasson also endorses a hierarchical view of sortals, such that the more general sortals depend on the application conditions of more specific sortals (2007: 41). This dependence is one-way—from the general to the specific—making it hard to see how this aspect could be made consistent with holism. It would therefore seem that Thomasson can be read as endorsing a local or partial form of holism where some terms get their application conditions holistically, whilst the application conditions of others are atomistic. Indeed, it seems that Thomasson’s response to Brenner that relies on an appeal to ostensively defined or innate terms is not compatible with a global holism, and thus we should conclude that Thomasson’s own version of easy ontology is at most partially holist.

However, as we will see below, it is a global holism that we need to respond to the arguments of Brenner. If Thomasson does ultimately endorse a partially holist view, then this paper should be best thought of as proposing an alternative version of easy ontology rather than a direct defence of Thomasson’s specific version. It should also be noted that Brenner in his objections also does not specify explicitly what he takes the commitments of easy ontology to be with respect to this holism/atomism distinction. In what remains, I will assume a version of easy ontology that endorses holism about application conditions, and show that within this version of easy ontology, the identified regress and circularity are not vicious.

4 Holism and Regress

Brenner argues that the regress leads to issues about learnability, arguing that we could not come to grasp the ‘meaning or rules of use associated with a term if that would require that we grasp the infinite sequence of application conditions associate with that term’ (2018: 608). That is, Brenner is arguing that due to the regress identified, grasping the application conditions of terms cannot be done gradually, but only by grasping the infinite sequence of application conditions, something that is absurd. Our question here is whether this objection still holds if we accept a holistic account of application conditions.

Brenner’s argument is very similar to objections that have been raised in the semantic holism literature, such as by Dummett (1973, 1991: 221), and is often connected to problems accommodating communication within a holistic account (Fodor, 1987; Fodor & Lepore, 1992). The basic idea is that if we adopt meaning holism then language learning cannot happen in a gradual way, and would instead require an absurd ‘single spasm of seamless cognition’ (Fodor & Lepore, 1992: 9). However, drawing on the semantic holism literature, there are clear options to respond to this problem. First, Jonsson (2014) has argued that semantic holism is not detrimental to language learning, but in fact enhances it. Or, alternatively, Bilgrami (1986) and Dresner (2002) have argued that, within holism, language learning can be taken to be gradual such that a person can have partial knowledge of the meaning of a term and then gradually increase it. The idea is that a speaker need not know all aspects of the meaning of a term, but only some of the restrictions on admissible interpretations of the lexical item, with learning being the process of increasing our knowledge of what restrictions apply for a term.

Analogously, we could hold that easy ontology does not require us to grasp the entire web of holistic application conditions, but only a part of that whole. This would mean that we may, perhaps in part through the process of engaging in easy ontology, subsequently increase our knowledge of the entire system. This would allow for the learning of application conditions without committing to complete knowledge of the entire holistic set of application conditions.

It is important to stress that these approaches commit us to no special ‘metaphysical’ form of knowledge of the sort that easy ontology is designed to avoid. The responses to the learnability problem as applied to semantic holism are intended to be accounts of how ordinary speakers come to learn natural language. Similarly, I propose the analogous account of learning application conditions to be the process through which we come to be competent speakers of a language. This is in line with Thomasson’s claim that to know the application conditions of a term requires nothing more than a competent speaker to make use of her mastery of the rules of use of the term.

Note also that it can still be the case that in a holistic version one can add new expressions and thereby extend it to a larger holistic system, and that this process could continue infinitely. However, this is not a problem for the holist as they only require that the part of the holistic system that we are learning is finite. A holist, mirroring defenders of semantic holism, can provide arguments as to how those application conditions are therefore learnable, and a holist version of easy ontology does not face the regress problem identified.

Perhaps it could be responded here that even if some forms of holism do not require that the speaker grasps an infinite chain of application conditions, they still adopt a view where application conditions would be unlearnable as they require us to grasp the application conditions of all the terms within a language, and all that I have done is define semantic competence so as to not require grasp of the entire holistic ‘web’.Footnote 4 Whether this is an issue for easy ontology depends on whether the view is committed to some particular notion of semantic competence that is stronger than the notion employed by the holist. But there seems to be no such requirement built into easy ontology. For Brenner’s regress argument to work, it would have to be shown that within easy ontology, ‘competence’ cannot be understood in holistic terms. Pending a further argument to that conclusion, the regress is not vicious if we accept holism about application conditions.

This, though, leads to a broader objection.Footnote 5 Within easy ontology, to arrive at the intended deflationary conclusions, we need to be able to infer existential claims from application conditions. This might lead to an objection to the holistic version of easy ontology: given the holistic nature of application conditions, how could we know whether we have grasped the relevant application conditions well enough? Our grasp on holistic application conditions are going to be partial in that no person will be able to fully grasp the full ‘web’ of application conditions within a holistic account. All of the easy ontologist’s existential inferences can thus be resisted on the grounds that the grasp of the application conditions was not sufficiently full.

However, I do not think this is a problem for a holistic version of easy ontology. The reason for this is that Thomasson does not hold that easy ontology will remove all debate about existential claims. Rather, the aim of easy ontology is to deflate debates about the existence of some given entity. The debate is not ‘epistemically metaphysical’, but is instead a conceptual or empirical matter. What this objection highlights is that on the conceptual side, there will still be debates about the precise nature of the application conditions for a given term, and that those debates may be ongoing and difficult to solve. They would not be difficult in the sense invoked by the heavyweight metaphysical realist, but difficult in that they would require complex conceptual work to try to specify as far as possible the application conditions for a given term.

This means that it may remain unclear when it is the case that we have a sufficient grasp on application conditions to be able to infer existential claims. The debate over whether we do have a sufficient grasp though will not be a metaphysical debate—it will be a debate about the application conditions of our terms and that is a form of debate that easy ontology is not meant to rule out.

As a last note on this objection, it is also not clear that this problem is unique to a holistic account of easy ontology. An atomist or partially holist version of easy ontology will also be susceptible to worries about whether we have properly and sufficiently grasped the application conditions of the relevant term to be able to confident in our inferences to existential claims. It is always open for a sceptical objector to argue that someone has not properly grasped the application conditions of a term that is being used within an easy ontology inference and to try to resist the inference on those grounds. We might think that we have sufficient grasp on the application conditions only to find some new use of the term that would suggest that we in fact do not.

Indeed, this is, in part, why the idea of ‘mastery of the rules of use of the term’ is included in the description of application conditions by Thomasson. We can, as a broad guide, be confident that we have sufficiently grasped the application conditions of the term when we have mastered the semantic rules of use for that term. A speaker need not be able to state those rules, but their success in using the term is a (defeasible) guide to them having suitably grasped the application conditions. It is defeasible as Thomasson also allows that we might argue about the application conditions, or defer to experts. The holist about application conditions can accept this same heuristic with no additional sceptical problems than those that could already be posed against the atomist.

This is not to deny the general point that a non-holistic version of easy ontology can maintain that at least in principle, a speaker could come to have a full grasp of the relevant application conditions, while this is not possible on the holistic account. This may be a negative for the holist view for some readers as there could always remain the doubt that we have enough of a partial grasp of the relevant application conditions to warrant deflationary conclusions.

But, if the holist is right about the nature of language, then we can note that existential inferences are already successfully made in natural language communication on the basis of very partial meanings. This observation can be combined with the normal account of easy ontology not requiring that a speaker has full grasp of application conditions to infer the relevant existential claims, suggesting that the deflationary aims of easy ontology are still possible on the holistic account. This may suggest that proposed deflations would have to be handled on a case-by-case basis, examining each to see if sufficient grasp has been achieved to warrant deflation. But, this is consistent with the supporter of a holistic version of easy ontology accepting that grasp of application conditions may remain partial and that this partial knowledge can be enough to warrant deflating metaphysical debates concerning existence.

5 Holism and Circularity

Turning to the circularity objection, accepting holism about application conditions can also help here. The sort of cases that Brenner is worried about are those where the application conditions are of the form: ‘Ks exist iff Ls exist, and Ls exist iff Ks exist’.

In response, it is important to remember that the easy ontologist is not against introducing new terms. If we adopt holism, while it is the case that application conditions might be circular, it is not clear that they are viciously circular. Within a holistic system, it is simply part of the nature of application conditions that, ultimately, there will be circularity of application conditions as no term’s application conditions are independent of all other term’s application conditions. New terms are allowed, but there are conditions and restrictions that ensure that a sortal term is ‘minimally introduced’ (Thomasson, 2015: 263–264). These restrictions include that any new term can be given application conditions using extant terms of the language that the term is being introduced into (Thomasson, 2015: 263). This means that any new term is connected to terms that already exist, and are understood, within a language. This interdependence of new and already existing terms is clearly in line with the proposed holism about application conditions.

Indeed, we can argue that this circularity is actually a positive feature of the view. Katz, inter alia, has argued that when considering theoretical terms, the ‘degree of relatedness exhibited among the concepts in the family is thus a measure, not of circularity, but of the systematizing power of the explanation’ (Katz, 1988: 240–241). Analogously, it might not be a weakness if application conditions are circular. Rather, the circularity actually shows the strength of that particular language as it is one that shows a large degree of systematicity.

The general point is that, granting holism about application conditions, it is again unclear why any circularity is vicious. The holist about application conditions, just like semantic holists, can reject the need for there to be any atomic application conditions, and can accept circularity without issue—even perhaps taking it to be a positive feature of the view.

6 Two Versions of Easy Ontology

This discussion suggests that there are in fact two subtlety distinct versions of easy ontology that we could endorse. The first, as assumed here, takes application conditions to be holistically determined such that the application conditions of all terms of our language are interdependent. The second, an atomistic version, allows that for at least some terms, the application conditions of those terms are independent of the application conditions of other terms in the language.

I have argued that if we accept a holism about application conditions, then both the regress and circularity Brenner identifies are non-vicious. This is at least one, potentially defeasible, reason in favour of embracing a holistic version of easy ontology. Brenner’s arguments implicitly rest on an assumption that the application conditions within easy ontology are taken to be atomistic. Without this assumption, the easy ontologist can render both the regress and circularity non-vicious.

There are also further benefits to embracing a holistic version of easy ontology. A holistic conception of easy ontology further illuminates Thomasson’s response to so-called ‘bad company’ arguments. Such arguments are meant to show that the methodology of easy ontology leads to contradictory ontological commitments. For example, following Thomasson’s adaptation of Eklund (Thomasson, 2015: 260; see also Eklund, 2006: 112), we can introduce the concepts of xheart and xliver such that:

If there is a heart and there are no xlivers, then there is an xheart.

If there is a liver and there are no xhearts, then there is an xliver.

If we accept these conditionals (and therefore assuming the existence of hearts and livers), then xhearts exist only if xlivers do not, and xlivers exist only if xhearts do not. Each of these conditionals taken in isolation seem to commit us to the existence of certain entities (xhearts and xlivers). However, this leads to a claimed contradiction in the ontological commitments of easy ontology, as the conditionals together rule out the existence of both xhearts and xlivers. Thomasson acknowledges that the easy ontologist needs a way to solve such bad company examples, by showing that some existence-entailing conditionals are illegitimate.

Amongst other restrictions, Thomasson suggests as a response to Eklund’s case that we say that new terms in the language must be ‘minimally introduced’ such that:

The term(s) must be introduced via a conditional that gives sufficient conditions for its (/their) application, stated using the extant terms of L and/or other minimally introduced terms (2015: 263).Footnote 6

Under an atomistic conception of application conditions, it is not clear that this restriction is independently motivated, rather than being ad hoc stipulations to avoid puzzles cases raised against the theory. That is, given that under atomism the application conditions of a term can be independent of the application conditions of any other term, what justifies this restriction on the introduction of new terms? Even if this only applies as a restriction on ‘minimally introduced terms’, why this additional restriction? We might also worry that, contra the above discussion, terms cannot be introduced by ostension if we accept this restriction. Above the response to Brenner provided by Thomasson relied on application conditions not needing to be statable, but here the restriction on only ‘minimally introduced’ terms being added to the language seems to require statability.

By taking application conditions to be holistic, we can avoid such an issue in a different way. Holism about application conditions, I suggest, legitimises this restriction imposed on the introduction of new terms. It is not ad hoc as the restriction arises from the holistic nature of the application conditions for terms. Under the holistic account, the application conditions for any given term ‘x’ will make reference to some other term ‘y’, as no term’s application condition can be isolated from the application conditions of other terms. A new term must be stated using the extant terms of L directly due to the holistic nature of application conditions.

Note, that under this account, ostension can still be accepted as a method for speakers to grasp application conditions. Holism does not rule out that we can more easily grasp application conditions via ostension. Ostension can help speakers place a new term within the holistic ‘web’ of terms and their application conditions.

I have indicated some initial reasons to be interested in a holistic version of easy ontology, but is this holistic version of easy ontology even viable? That is, can we endorse easy ontology whilst being a holist? Naturally, a full account cannot be provided in a single paper. However, there are some indications that a holistic version of easy ontology is perfectly possible. The central aspect to consider is Thomasson’s discussion of application conditions. If some aspect of that account is inconsistent with holism, then there could be no way to endorse holism about application conditions and easy ontology. If holism is consistent with these features, then we at least have a prima facie case that holism is consistent with easy ontology more broadly.

Easy ontology requires that we can infer existential claims from an understanding of application conditions. Thomasson lists four features of application conditions. The first three seem immediately non-problematic to a holist treatment; the fourth requires a little more discussion (c.f Thomasson, 2015: 91; my numbering matching the original). Within a holistic account, application conditions:

  1. 1.

    can still be semantic rules of use which speakers master, and there is no need for them to be necessary and sufficient conditions or statable;

  2. 2.

    can specify the conditions under which the term would be correctly applied (entitling us to truly say ‘there is a K’).

  3. 3.

    need not be descriptive, and may involve deference to experts and to the world.

None of these cause problems for a holistic account, and all can be maintained without change from Thomasson’s treatment.

More troublesome, at least initially, is the idea, cited above as (4), that “Application conditions must not take the following form: ‘‘K’ applies iff Ks exist’ (Thomasson, 2015: 96). This is an important principle for Thomasson as it ensures that we do not need to settle existence questions before we can introduce new terms. We might be concerned that this needs to be rejected on a holistic view as, in the response to the circularity problem, the holist has embraced the idea that some sequences of application conditions are of the form: ‘K’ applies iff Ls exist, and ‘L’ applies iff Ks exist.

I think that the holist can maintain this principle, and the reason we can is that the holist can accept this rule, just so long as it is also the case that when a new term is introduced, it is introduced in a way that coheres with the existing holistic semantics of the language.

To see why this is the case, we need to remind ourselves that Thomasson introduces this rule to help govern the introduction of new terms, and that taking application conditions to be holistic does not disallow the introduction of new terms. What is required within the holist version of easy ontology is the introduction of terms that have holistically defined application conditions.

However, it is clear that this is not the case for ‘K’ if the application conditions for ‘K’ are that it applies iff Ks exist. Such an application condition is defined independently of any other term in the language, and thus it is ill-formed within the holist account. This ensures that we can maintain (4).

This is similar in some ways to the response Thomasson gives to the bad company arguments: restrictions are ‘justified given the goal of (merely) extending the language to include new sortal terms that will be usable’ (2015: 266). Disallowing the introduction of a term ‘K’ where the application conditions of ‘K’ are that it applies iff Ks exist is justified because of the holistic account of application conditions we have already assumed. ‘K’ cannot be accepted into the language as terms need to have application conditions that are part of the holistic application conditions of all terms within the language. Any term not so introduced is, assuming holism, ill-formed, and hence will be unusable.

There remains the possibility that application conditions could take the form ‘‘K’ applies iff Ls exist, and ‘L’ applies iff Ks exist’. This is possible in a holistic account, but here we can simply respond, as we did above, that accepting this possibility as part of our holism can be seen as a virtue, and hence the circularity is non-vicious.

Above I suggested that Thomasson’s response in this same vein to the bad company arguments might be thought to be ad hoc. Could my response be thought to be ad hoc too? Perhaps, but I think that it is justified given the prior assumption of accepting holism about application conditions. Granting that assumption, then any attempt to introduce a term that does not cohere with other terms as part of that holism can be legitimately be rejected as having ill-formed application conditions. We could of course reject that assumption, but the entire purpose of this paper was to introduce the possibility of a version of easy ontology that posited holistic application conditions, making any prior rejection of holism guilty of begging the question against the view under discussion.

Thus, accepting holism about application conditions does not seem to be inconsistent with the general rules about application conditions that Thomasson requires. Whether there would be required changes within the rest of the easy ontology framework to combine it with holism is beyond the scope of a single paper. But, that the holist can formulate application conditions within the rules that Thomasson lays out suggests (at least prima facie) that this change would not have further ramifications in the rest of the theory.Footnote 7

7 Conclusion

I have explored the benefits of accepting a holistic version of easy ontology, and argued that for those that wish to endorse easy ontology, it is a viable way to avoid the regress and circularity objections. I have also argued that adopting a holism about application conditions is prima facie consistent with the other stipulations Thomasson makes about the nature of application conditions. Pending further arguments that adopting a holism about application conditions causes problems for some other part of the easy ontology approach, it is not at all clear that there are any reasons that easy ontology cannot be combined with a holistic account of application conditions, and there are benefits from doing so. Future discussions of easy ontology will need to be explicit about whether they are assuming a holistic or atomistic account, and what the possible consequences from this might be.