2. Three Unfriendly Examples

While it may be easy to tell some truths, it is not always easy to say exactly what those truths tell us. We may agree on the truth, but not about its correct analysis. Because true utterances do not always make plain their underlying semantic structure or reality requirements, that structure or those requirements can become the subject of philosophical disagreements.

Some utterances are ornate, prolix ways of expressing truths which misleadingly suggest that the truths that they express are more complicated than they really are. These prolix expressions appear to demand more of reality than is required for their truth. Fortunately, there are other ways of expressing those same truths that convey more transparently what they require of reality. Substituting the plain expression for its prolix counterpart results in a pair of sentences which express the same truth, differing only in how revealing they are of what that truth is.

Two notions play an important role in this article: idiom and paraphrase. We’ll first give some examples of these two notions and how they interact. We’ll then offer some more general comments about how we understand each of these notions.

Consider the following inference, which is unfriendly to easy ontologists:

(4) entails (5) (and vice versa), given the standard idiomatic meaning of the phrase ‘the luck of the devil’. Speakers use (5) as a colorful reformulation of (4). So understood, (5) says no more than does (4) and, since (4) does not say that there is such a thing as the luck of the devil, neither does (5). (5) simply uses a vivid phrase to convey the same information as (4). (5), then, does not imply (6):

To think otherwise is to think that (5) also entails Ned has the same luck as the devil and that there is a unique x such that x = a devil and x has the same luck as Ned. How unlucky for Ned that his extraordinary luck requires an eternal foe of goodness!

This example highlights the distinction between sentences and their semantic content. Any misapprehension that (5) entails (6) should be dispelled by the realization that (5) is only a picturesque formulation of (4) and that (4) does not entail (6). (We do not address here whether (4) entails that there is something Ned has, namely luck; we will dispose of that suggestion in future work.) To the extent that an inference from (4) to (5) is obvious, it is an inference from a commonplace to a prolix reformulation. If the paraphrase were to be indicted (contra our defense in §§7 and 8), however, it is hard to see how the charge of absurdity facing an inference from (5) to (6) is to be resisted.Footnote ¹

Here is another example:

On the reading of (7) whereby it entails (8), (8) does not entail (9):

(7) makes clear what (8) conveys, and what (7) and (8) convey does not entail (9).

Finally, consider (10) and (11):

(10) makes clear what (11) conveys, and what they convey does not entail that there is such a thing as the average family, unlucky as any such family would be. This last example shows that the point is not confined to sentences containing idioms but applies more widely to sentences involving other turns of phrase—to the category of sentences containing the expression ‘average NP’. This is a technical expression defined in terms of ratios. Its definition blocks the inference from ‘The average NP is so and so’ to ‘There is an entity that is the average NP.’

In the next section we introduce the notion of an idiom as part of our criticism of easy ontology.

3. Idioms

As is often the case in philosophy, there is more agreement about the classification of examples than there is about the basis on which that classification is made. One popular view of idioms says that “roughly, [an idiom] is a phrase (or sentence) which is conventionally used with a meaning different from its constructed literal meaning (if it has one)” (Davies Reference Davies1982-3: 68. See also Chomsky Reference Chomsky1980 149-153 and Moran Reference Moran, Hale, Wright and Miller1997 section 1). The constructed literal meaning of a phrase is determined by its semantic structure: the meanings of the phrase’s syntactically individuated constituent words and their modes of combination. By contrast, the idiomatic meaning of a phrase lacks semantic structure.

This popular view has been contested (Nunberg, Sag and Wasow Reference Nunberg, Sag and Wasow1994; Egan Reference Egan2008 section 4 claims that “it can’t be the whole story”). We fix on it because of its popularity, its prima facie appeal, and the fact that it explains at least some of the behavior of idioms. Whatever the correct theory of idioms, we are confident that it could be applied equally well to the terms of particular interest to us.Footnote ² In any case, a corollary of the above view is that a phrase which has a constructed literal meaning and a distinct conventionally used meaning, is ambiguous. There seems to be no restriction on which phrases can have these twin meanings. For example, the category of misnomers is open-ended. Despite what many people might have once thought, shooting stars are not stars, peanuts are not nuts, and lead pencils have never contained lead. If proper names have literal meaning, even they permit this ambiguity. Someone’s full name consists of syntactically individuated constituent words and their modes of combination: Harry S. Truman’s name is not ‘S. Truman Harry’ or some other jumble. But a name might have a conventional use distinct from its constructed literal meaning. Disgruntled workers might routinely call their boss ‘Genghis Khan’ behind his back.

The class of semantically apt idioms is especially noteworthy. This is a class of syntactically complex expressions whose meanings are not determined by the semantic properties of their constituent words and modes of combination, but whose meanings can be seen as somehow felicitous given the semantic properties of their constituents” (Davies Reference Davies1982-3: 70–1). To use an example of Davies which is not an idiom, ‘hydrophobia’ is semantically apt. The semantic properties of ‘hydro’ and ‘phobia’ tell us no more than that ‘hydrophobia’ has something to do with water and a phobia. Semantic aptness admits of degrees and at the other side of the spectrum are, amongst other expressions, idioms. Many idioms are not semantically apt (‘to be driven up the wall’, ‘to draw in its horns’). Moreover, the etymology of some idioms (‘to kick the bucket’, ‘to be sent to Coventry’) is obscure and contested, and so it is far from clear whether they are semantically apt. Lastly, many idioms are circumlocutory and this surface feature veils their idiomatic meaning. Many euphemisms might count as idioms. Rather than call out a lie in a recent political campaign, National Public Radio called it a “disconnect” between the campaign and its messaging. From the use of that innocuous word, you wouldn’t be able to tell that a lie was being reported. Idioms show how a sentence may be true (an Australian might truthfully say ‘Floyd went off like a frog in a sock’) while which truth that sentence conveys may not be immediately apparent. A syntactically different sentence (‘Floyd got very excited’) can then be introduced to make clearer which truth the idiomatic sentence conveys.

Let’s turn to the second key notion that we use in our criticism of easy ontology, the notion of paraphrase. Drawing on a line of thought offered by Szabó (Reference Szabó, Loux and Zimmerman2003: 21), a critic of our claims above might respond as follows. You say there are no such things as gifts of the gab. If you are right, it seems that (8) ‘Nate has the gift of the gab’ would have to be false but (7) ‘Nate is a good talker’ would still be true. But how could semantic paraphrases differ so obviously in their truth-conditions? It is, however, precisely our contention that (7) and (8) do not differ in their truth-conditions. As we noted, it’s an old story: one cannot always read the semantics of a sentence from its syntax. Since a paraphrase need not be obvious, it is no surprise that someone might blithely think (8) to be false but (7) true, if there were no gifts of the gab. What the paraphrase helps show is why once apprised of the paraphrase of ‘Nate has the gift of the gab’ as ‘Ned is a good talker’ one should think otherwise.

We have paired some examples of idioms with paraphrases of those idioms: ‘Ned has the luck of the devil’ and ‘Ned is very lucky’; ‘Nate has the gift of the gab’ and ‘Nate is a good talker.’ As with idioms, there is more agreement in philosophy about certain cases of paraphrase than with any detailed account of what paraphrase is. That said, we hope that the following remarks are relatively uncontroversial and that they locate our views about what paraphrase is.

In the next section we will develop our discussion of paraphrase further.

4. Paraphrase

Broadly speaking, there are two strategies for paraphrasing available in philosophy (see also Keller Reference Keller, Fisher and Maurin2023 section 2.4). There are marked differences between them, but for our purposes a fundamental point of similarity between them matters. One strategy takes paraphrase to be a matter of unmasking and preserving: it takes a paraphrase of a sentence which apparently has one set of truth conditions to have some other set of truth conditions. The unmasking shows that the sentence doesn’t have the truth conditions it appears to have. The preserving shows that, although the sentence has truth conditions other than those it appears to have, our quotidian judgments about the truth values of tokens of that sentence are correct. So, for example, van Inwagen’s paraphrase of folk claims about houses and chairs (Reference Van Inwagen1981: 98–102) denies that those claims are associated with truth conditions requiring the existence of houses and chairs and instead argues that they are associated with truth conditions about simples (i.e., objects without parts) arranged (what he calls) housewise or chairwise. He takes these paraphrases to preserve the truth values that the folk assign to those claims. (For other exponents of this strategy, see Jackson Reference Jackson1998 chapter 2 and Keller Reference Keller2017). The other strategy takes paraphrase to be renouncing and replacing: it takes a paraphrase of a sentence to involve renouncing the use of that sentence (at least outside contexts of loose talk) and replacing it with a sentence with clearly specified and philosophically acceptable truth conditions. So, for instance, Carnap’s project of explication is one of finding replacements for problematic sentences, where these replacements will not have the same truth conditions as the original sentences but will provide more illuminating, fruitful, or simple descriptions (Carnap Reference Carnap1950 chapter 1). Similarly with Quine’s project of regimentation (Quine Reference Quine1960 chapter 5). In Carnap’s well known example, vague and unrefined sentences about the comparative warmth or cold of objects are replaced by more precise, fruitful, and well-informed sentences about the temperature of objects (Carnap Reference Carnap1950 chapter 1 section 4). The recent program of conceptual engineering follows suit (e.g., Burgess, Cappelen and Plunkett Reference Burgess, Cappelen and Plunkett2020).

These strategies agree that often there are better ways of saying what we want to express. What we say may mislead (according to the first strategy) or it may be inapt, erroneous or indeterminate (according to the second) but paraphrase resolves these deficits when it provides a sentence which better expresses what we originally said. The strategies then disagree about the truth conditions of the paraphrased sentence: the first strategy thinks that the target sentence—the sentence to be paraphrased—has the same truth conditions as the paraphrasing sentence, whereas the second strategy denies this. For our purposes this disagreement does not matter. Each strategy of paraphrasing would suit our purposes. For reasons of definiteness, and because it is also a strategy which the easy ontologists explicitly reject in defending their view, we will pursue the first strategy. We now have the requisite machinery in place. The next section will use it to criticize the easy ontology project.

5. What’s Gone Wrong?

Let’s revisit the illustrative inference made by easy ontologists:

Yablo describes inferences of this sort as generating what he calls “hostage crises because they involve a (relatively) thin, innocent claim and a (relatively) weighty, debatable one; the first is hostage to the second in that the second must hold or the first fails” (Yablo Reference Yablo2017: 115). Granting that the threats of paradox and of over-generation can be met, what should we make of such inferences? We will first give our reasons for rejecting certain diagnoses of where the easy ontology project has gone wrong, then we will offer our own diagnosis.

If the easy ontologists’ inferences seem too good to be true, it is tempting to respond to “easy” arguments by echoing, ironically enough, an objection that Schiffer himself raises in another context. The response runs, “If that’s the solution, then what the hell was the problem?” (Schiffer Reference Schiffer1996b: 329). The easy argument strategy makes it a mystery why there was ever a protracted debate about the existence of properties. If all you need to settle the question of whether there are properties is competence with such simple expressions as ‘is brown’ and ‘has the property of brownness’, plus the commonplace knowledge that something is brown, then any fluent and intelligent English speaker would realize that there are properties, as some philosophers countenance them.

A second response to the “easy” argument would be to allow its conclusion to go through, agree that the sentence is ontologically committal, but then to point out that the argument does not settle what properties are. Are they universals? Tropes? Sets? Are they even abstract objects? This response, however, concedes too much because it concedes that if the table is brown at least two entities exist: the table and its color property. We resist this concession. Moreover, we will see similar arguments below where conceding their conclusions would be absurd. This tells us that there is a flaw in arguments of this sort, a flaw that the concession overlooks.

A third suggestion about what has gone wrong with the “easy” argument is that (3)—the sentence ‘There is a property (namely, the property of brownness)’—is not ontologically committing. The truth of that sentence does not require the existence of anything, including any properties (Azzouni Reference Azzouni2004: 63–5; 2007). This apparently radical measure is not without criticism (see Asay Reference Asay2010) but we are neutral on this point because our objection does not entail a sweeping denial of ontological commitments in vernacular expressions. We are content to pool this suggestion with our own diagnosis of what is wrong with the easy argument, thus forming a dilemma. The first horn of the dilemma is Azzouni’s: the “easy” argument establishes the truth of the sentence ‘there is a property’ but such a sentence is not ontologically committing. The second horn is ours: if the sentence ‘there is a property’ is ontologically committing, the “easy” argument fails to establish its truth.

We allow that the inference from (1) ‘the table is brown’ to (2) ‘the table has the property of brownness’ is valid but deny that the inference from (2) to (3) ‘There is a property (namely, the property of brownness)’ is valid. First, consider the inference from (1) to (2). We grant easy ontologists that, to use Thomasson’s terminology, (1) analytically implies (2) and vice versa, and that (2) is a pleonastic reformulation of (1). We take the lesson from this to be that (2) says no more than (1) does and that since (1) does not say that there is a property of brownness which a has, neither does (2). (2) is just a prolix and misleading restatement of (1). (2) is misleading, but it is permissible to utter. This means that it is appropriate to assert (2), but we should be mindful of how it might give the misleading impression that it entails (3). So, on the reading of (2) by which it follows from (1), we say that (2) is true even if misleading.

(2) ‘the table has the property of brownness’ is appropriate for contexts emphasizing the distinction between objects and their attributes, perhaps to the end of helping some grasp the distinction itself. No party disputes that distinction. The concise (1) ‘the table is brown’ is commonly used when predicates like ‘is brown’ are mastered sufficiently. There are good, pragmatic grounds for having these alternatives, even if they are equivalent, ontologically speaking.

The charge against easy ontologists is that they trade on two readings of (2). On the first reading, (1) ‘the table is brown’ entails (2) ‘the table has the property of brownness’ because they have the same truth conditions; (2) is only a wordier reformulation of (1). This reading of (2), however, does not entail (3) ‘there is a property (namely the property of brownness)’ because, we claim, (1) does not entail (3). On the second reading, (3) follows from (2) by existential generalization. This reading of (2), however, does not follow from (1), since (1) does not entail (3). Easy ontologists have devised something of a “duck-rabbit”: a sentence which, read one way, follows from an innocuous premise and so is itself innocuous, but which, when read in another way, entails a conclusion that is far from innocuous. They have not, though, devised a sentence that has a single reading that both follows from (1) and entails (3).

It might be said that sentences containing idioms are not ambiguous and that people who fail to see that a sentence is an idiom simply misread the sentence. We agree that such people will have missed the intended meaning of an utterance of the sentence, but that does not rule out the sentence’s having a secondary meaning. Consider ‘He was shooting the breeze.’ This is naturally read as expressing an idiom. Yet, consider an inebriated farmer enraged at the wind for, as he sees it, ruining his delicate crops. He reaches for his shotgun and literally shoots the breeze.

Finally, to block the easy ontologists’ characteristic inferences, we need only to have shown that there is a rival position which, first, allows that their premises can be true while their conclusion is false and, second, that this is a position which they haven’t eliminated. So, simply for the purpose of undermining the project of easy ontology, we need not establish that our position is correct but only that it remains a live option. We think, however, that our proposal has independent merit. We have noted the perennial philosophical concern that the surface form of sentences may diverge from what those sentences say or from what is required for their truth. When this is combined with the recognition of the pervasiveness of idioms in natural language, a case for our proposal emerges.

It is a commonplace that there are idiomatic expressions. Our principal suggestion is that the domain of such expressions is far wider than is usually supposed. In particular, we suggest that predicates of the form, ‘has the property of being F’ or ‘has the property of F-ness’, have an idiomatic meaning—a meaning which they commonly express when used outside of philosophical contexts. To say ‘the barn has the property of being red’ is an idiomatic way of expressing the claim that the barn is red. Why so?

First, the meanings expressed by their idiomatic use are not determined by the semantic properties of their constituent words and modes of combination. The semantic properties of ‘has,’ ‘the property of’ and ‘being red’ in the sentence ‘the barn has the property of being red’ tell us that the subject of the sentence bears a relation to something else (a property). On our view, that tells us something quite different from what the idiomatic meaning of the sentence tells us, not least the fact that its idiomatic meaning, namely that the barn is red, has nothing to do with the philosophically disputed property of redness.

Second, the etymology of predicates of the forms ‘has the property of being F’ or ‘has the property of F-ness’ is obscure. What purpose in ordinary life do they serve? What benefit do they bring? What do they do that could not be done by the predicate ‘is F’?

Third, those predicate forms are circumlocutory. We suggest that, taken in its idiomatic sense, the sentence ‘the barn has the property of being red’ is a pleonastic way of saying what ‘the barn is red’ says. What strikes us is not the fact that the English language can form the construction ‘the barn has the property of being red’ but that English speakers with only minimal competence can get by without it and use the plain old construction ‘the barn is red.’ This is strong evidence that the former construction is only a cumbersome statement of what the latter says. On those comparatively rare occasions when people utter the former sentence, their primary intention is that the audience should believe that a certain barn is red. Furthermore, they intend that this primary intention should be recognized because the audience will recognize that the sentence is regularly used to “get across” the proposition that the barn is red. This is just what an idiomatic use of the sentence does.

In the next section we will criticize key notions in Thomasson’s easy ontology project.

6. Rules and Presuppositions

Thomasson claims that “easy arguments can settle debates about the existence of concrete, ordinary objects” (Thomasson Reference Thomasson2014: 503). As an illustration, consider argument I:

At least part of what Thomasson means by the ‘constitutive meaning rules’ governing ‘table’ is that it is semantically associated with certain application conditions: non-trivial conditions under which the term ‘table’ is correctly applied. Competent users of ‘table’ have an implicit understanding of these application conditions governing their use of the term.

We find the above argument unpersuasive. There is an unresolved question about how the introduction of ‘table’ takes us to the intended conclusion. To make this vivid, consider a parallel argument which we conjecture easy ontologists would reject:

If easy ontologists do not endorse argument II and its partial vindication of phenomenalism, exactly what account can they give of the supposed difference in logical force between arguments I and II? What non-question-begging account of the semantics of ‘table’ (“the constitutive rules governing the use of our terms”) licenses the conclusion of Argument I but not of Argument II? Moreover, even if easy ontologists can make a case for this discrimination, they need a further argument that the constitutive rules preserve truth.Footnote ³

What should we make of arguments I and II? First, we see no grounds for distinguishing between them in terms of analytic implication. Either both arguments are analytically valid, or neither is. We think neither is, for reasons we explore below.

Second, suppose the validity of these arguments is not to be understood in terms of analytic implication but metaphysical necessity, so that a valid argument is one where the truth of the premises would metaphysically necessitate the truth of the conclusion. Some metaphysical necessities are more apparent than others. If you find Argument I compelling, you think that there is no world in which its premise is true and its conclusion false. The fact that you find the argument compelling does not require that the premise analytically implies its conclusion.

Third, the arguments’ premise sets provide assertion conditions for their respective conclusions. Given that there are particles arranged tablewise, you have reason to assert that there is a table. Given that you are having tablish experiences and are disposed to have tablish experiences, you have reason to assert that there is a table. Accordingly, the premises of the arguments do not concern the application conditions for ‘table’ that the easy ontologist invokes. Instead, they concern widely held (but perhaps defeasible) epistemic reasons for asserting the conclusions (cf. Yablo Reference Yablo2014: 484).

This takes us to our principal reason for rejecting Thomasson’s claim that constitutive semantic rules establish ontological conclusions. The preceding sections expose a dilemma. The first horn: if ‘Socrates is wise’ analytically implies ‘Socrates has the property of wisdom’, that is because the second is simply a cumbersome paraphrase of the first. There is no basis for accounting for the first by way of any entity other than Socrates. The second horn: if ‘Socrates has the property of wisdom’ is an ontological conclusion with anti-nominalistic implications, then the implication is not analytic. It is not analytic because some perfectly competent speakers do not recognize the implication. Philosophical nominalists and quietists resist it.

Constitutive rules, we submit, are obscure. In our view, Thomasson does not provide complete examples of how such rules are given, nor are we aware of following them when we speak or think. Thomasson deflects this objection by claiming that, by the same measure, competent speakers need not be aware of the grammatical rules that govern their languages (Thomasson Reference Thomasson2014: 524). Though correct, the analogy does not suit her purposes. First, stating the grammatical rules of natural languages is a central ongoing and fruitful research programme in linguistics. In contrast, Thomasson tells us that constitutive semantic rules may not even be formulable in language, so there is little prospect of a science of such rules. Second, the basis for the grammarian’s fruitful project is that competent speakers of a language agree a great deal about which strings of symbols form grammatical sentences in their language. By contrast, competent users of sentence pairs of the form ‘a is F’ and ‘a has the property of F-ness’ are not in broad agreement that the first analytically implies the second while uncovering hidden ontological implications. Consequently, the mechanism of constitutive semantic rules lacks support from the intuitions of competent speakers. Lastly, while admitting that competent speakers need not be aware of, nor even follow, constitutive semantic rules, Thomasson claims that speakers are “answerable to” such rules (Thomasson Reference Thomasson2014: 521–523). We are not entirely sure what being “answerable to” amounts to, beyond the claim that speakers’ linguistic behavior can be assessed by reference to these rules. By the same reckoning, speakers’ linguistic behavior is answerable to assertion conditions and principles of evidence, where this means that what people state can be assessed as to whether what they asserted was appropriate and warranted by evidence. We grant that speakers are answerable to any constitutive semantic rules there may be. We have yet to be given an adequate case for any such rules.

Thomasson’s claims about “ontological presupposition” are difficult to reconcile with her main thesis. She writes that:

The above passages seem to be statements of nominalism—the view that there are no numbers or properties or, more generally, abstract objects. This seems to be in tension with Thomasson’s earlier assurance that the introduction of meaning rules for number terms entitles us to talk of numbers. It is difficult to see how these passages could be consistent with the main thesis of easy ontology. If ‘Socrates is wise’ analytically implies that there is a property of wisdom (which Socrates has), it is analytic that: ‘Socrates is wise’ is true only if there is a property of wisdom. In that sense at least, ‘Socrates is wise’ presupposes that there is a property of wisdom. Accordingly, a dilemma arises for easy ontologists. If a form of words has no ontological presuppositions, then competent uses of those words present no grounds for ontological conclusions. But, if ontological conclusions are sometimes warranted by the proper use of some expressions, then that use must sometimes somehow involve relevant presuppositions.

Thomasson denies that there can be ontological errors in addition to empirical errors (Thomasson Reference Thomasson2014: 526–527). Yet it is strange to think that Leibniz made no ontological error when claiming that the world consists of monads, immaterial things with no parts, or that Berkeley didn’t err in claiming that the world consists of ideas and minds alone. If, as Thomasson claims, there are ontological truths—including the alleged ontological truths that there are numbers, properties, and propositions—presumably she would take the nominalist’s denial of such truths to be ontological errors.

Using the example of number talk, Thomasson gives the following reason for her denial:

The challenge to easy ontology, however, is more general, concerning possible factual mistakes, not just possible empirical mistakes. It is a philosophical, not a scientific, project after all. It would be a factual mistake of taking there to be numbers if there are none. On our view, the narrower concern about an empirical mistake—a mistake about the concrete world that is detectable by empirical methods—is a red herring.

Having used the notions of idiom and paraphrase to criticize the easy ontology project, the next two sections will defend paraphrase against some influential objections.

7. Fact and Fabrication

The availability and significance of paraphrase receives rather different appraisals from easy ontologists. Schiffer ignores the notion (as Sainsbury Reference Sainsbury2005: 99 notes) whereas Thomasson is critical. She thinks that paraphrase achieves nothing (Thomasson Reference Thomasson2007: chapter 9). Her criticism is that it matters not whether (2) ‘the table has the property of brownness’ can be paraphrased as (1) ‘the table is brown,’ (1) nevertheless analytically implies (2) and (2) analytically implies (3) ‘there is a property (namely the property of brownness).’ So (1) analytically implies (3). As Thomasson puts it, “[t]he complicated paraphrases do not make way for a sparser ontology; they only hide the commitment to [properties]” (Thomasson 2009: 165).

Yet if paraphrase were to fail to yield a sparser ontology for this reason, it would fail across the board. The would-be inferences from (5) ‘Ned has the luck of the devil’ to (6) ‘there is something Ned has (namely, the luck of the devil)’ and from (8) ‘Nate has the gift of the gab’ to (9) ‘there is an entity (namely, the gift of the gab) which Nate has’ could not then be blocked by paraphrasing (5) as (4) ‘Ned is very lucky’ or (8) as (7) ‘Nate is a good talker.’ If Thomasson is correct, then (4) entails (6) and (7) entails (9). Thomasson seems somewhat insouciant about admitting the existence of counterintuitive entities via analytic implications from platitudes. Yet, if she admits that there is such an entity as the luck of the devil, we would rest our case; something is definitely amiss. Easy ontology is saved from threats from paraphrase at the cost of eliminating any distinction between idiomatic and literal theoretical discourse. That is a distinction which all parties would draw prior to this dispute and one which they would want to preserve, so that, for example, Ned’s luck does not unluckily bring along a lucky devil.

To make the point, let’s return to Thomasson’s criticism of the appeal to paraphrase. We have said that (1) ‘the table is brown’ does not entail (3) ‘there is a property (namely, the property of brownness’, that (2) ‘the table has the property of brownness’ is to be paraphrased as (1), and so that (2) does not entail (3). Thomasson’s claim is that, even if (2) is paraphrased as (1), (2) entails (“analytically implies”) (3), and so (1) also entails (3). The dialectic here is familiar. It echoes Alston’s contention that paraphrase is symmetric. If (2) may be paraphrased as (1), then, following Alston, (1) may equally be paraphrased as 2) and so, if (2) entails (3), (1) entails (3). No ontological reduction is gained by paraphrase (Alston Reference Alston1958, Searle Reference Searle1969: 107f, Melia Reference Melia1995: 224, footnote 1, and Hale and Wright Reference Hale and Wright2001: 156 for similar sentiments).

That this is dubious can be seen by distinguishing between apparently valid inferences and genuinely valid inferences. There is no valid inference from (1) to (3), we have argued. Perhaps there is an apparently valid inference from (2) to (3). But since there is a valid inference from (1) to (2), there is no genuinely valid inference from (2) to (3) (van Inwagen Reference Van Inwagen, Lepore and van Gulick1991: 346–7 and Keller Reference Keller2017 §3). Thus, the relationship is asymmetric. We anticipate that easy ontologists may attempt a matching move and argue as follows: Perhaps there is no apparently valid inference from (1) to (3). But since (2) can be validly inferred from (1), and (3) can be validly inferred from (2), there is a genuinely valid inference from (1) to (3), even if this inference is not apparent. (We read Oliver Reference Oliver1996: 66 as making a rejoinder of this kind).

There is no impasse here for two reasons. First, there is the independent point that certain sentences are open to being interpreted either idiomatically or literally. The inference from (a) ‘Jo didn’t hold Katy responsible’ to (b) ‘Jo let Katy off the hook’ is genuinely valid only if (b) is read idiomatically. Anyone using (b) instead of (a) would think it absurd to ask after the size and type of hook off of which Katy was let. The inference from (b) to (c) ‘There is something that Jo let Katy off’ is genuinely valid only if (c) is read literally and queries about exactly what the something which Katy was let off reveals inattention since that is revealed explicitly in the premise. There is no single reading of (b) on which (c) can be validly inferred from (a). We suggest that property ascriptions (understood as ascriptions of the form ‘a has the property F’ or ‘something has the property F’) can also be read either idiomatically or literally. This suffices to block the ‘easy’ argument. That argument concludes that ‘there is a property F’ from the premise ‘a has the property F.’ But the inference goes through only if the premise is read literally. In that case, however, we take the premise to be question-begging. Alternatively, if the premise is taken to be true, we read it idiomatically and in that case the inference does not go through.

Second, consider how we clarify what we say to one another. When we say something and our audience is puzzled about what we mean, we seek to utter another sentence (or sentences) to clarify what we originally said. Suppose we said during conversation, ‘The table has the property of brownness,’ and our audience was (understandably) puzzled. We could clarify what we said by some plain speaking; we could say ‘The table is brown—that’s all I meant.’ In the normal course of events, our audience would now understand what we had originally said; they might be puzzled only about why we hadn’t said it like that in the first place. By contrast, if we had originally said ‘The table is brown’ and our audience had not understood, they would not have understood us any better if we had sought to clarify what we’d said by adding, ‘That’s to say, the table has the property of brownness.’ Talk of property ascription does nothing to clarify the original sentence. This simple point illustrates an asymmetry between sentences of the form ‘a is F’ and those of the form ‘a has the property of F-ness’ and why the direction of paraphrase between them is one-way. That accounts for the reading of ‘a has the property of F-ness’ that is acceptable (in our eyes, at least) but idiomatic. The easy ontologist’s literal reading of the sentence, however, conveys—analytically implies—the information that there is a property of brownness and that a bears a relation to that property. This extra information would not, though, serve to clarify the sentence ‘a is brown’ for someone who did not already understand it. The information would compound their puzzlement, not relieve it. Yet if this information were analytically implied by ‘the table is brown’, stating what the sentence analytically implied should help clarify the sentence. We conclude that the sentence does not analytically imply this information, and we reject the literal reading of the sentence.

In sum, if we follow Thomasson in talking of analytic implication, we grant that (1) ‘the table is brown’ analytically implies (2) ‘the table has the property of brownness.’ But our explanation of this analytic implication is that (1) is a paraphrase of (2). We have then used this explanation to argue that neither (1) nor (2) analytically imply (3) ‘there is a property (namely a property of brownness).’

8. A Paean for Paraphrase

Paraphrase has poor standing in quarters other than the easy ontology project. Here are two representative passages:

To address Oliver’s charge directly, it is unclear to us why “the general nature” of the argument about the availability and relevance of paraphrase should vitiate that argument. That aside, we do more than produce one or two examples. We have an algorithm for producing paraphrases that block the easy ontologists’ arguments. Given any sentence of the form:

we paraphrase it as:

And given any sentence of the form:

we paraphrase it as:

which in turn we paraphrase as:

Corresponding claims hold in the case of relational sentences. Given any sentence:

we paraphrase it as:

Such algorithms provide the generality required of adequate paraphrases. They also generate the desired result that any sentence of the form ‘a has the property F’ can be paraphrased by a (more) ontologically innocent sentence. (See von Solodkoff Reference Von Solodkoff2014: 577 for these requirements.) Van Inwagen (Reference Van Inwagen1977: 303) says that a paraphrase should provide an account of the logical consequences of paraphrased sentences. We meet this requirement as well: where a sentence of the form of (S1) is paraphrased by a sentence of the form of (S2), the former has all and only the logical consequences of the latter.

Let’s now address a fresh challenge to the project of paraphrase. A paraphrase reveals in a more illuminating or helpful way what a speaker means. But this raises the following questions:

We agree that there must be adequacy criteria for paraphrases, if paraphrases are to be introduced at all. We also agree that one such criterion is that a paraphrase of a sentence should agree with speakers’ judgments about what that sentence means, as we acknowledged earlier. Based on the above paragraph, though, Carrara and Varzi raise a suspicion about paraphrase:

They illustrate their claim with an example. Meinongian objects offended Russell’s “strong sense of reality” (Carrara and Varzi Reference Carrara and Varzi2001: 40),Footnote ⁴ thus motivating his theory of descriptions. We take Carrara and Varzi’s suspicion to be that paraphrase isn’t helpful: paraphrases are introduced only to channel philosophers’ preconceptions about what there is; better, then, to examine directly those preconceptions than to go through the artifice of examining the paraphrases constructed in service of those preconceptions.

We set aside the above oversimplified reading of Russell, who cited the multifarious puzzle-solving nature of his theory (Russell Reference Russell1905). We grant that philosophers’ ontological preconceptions could motivate them to seek paraphrases. Believing in Js but not Ks might well prompt a search for paraphrases of talk of Ks into talk of Js. Motivation is one thing; success quite another. Ontological preconceptions provide little reason to believe that the desired paraphrase can be achieved. If the paraphrase and its successors fail, that is reason to doubt the ontological preconception that guided it. Witness the failures of phenomenalism and logical behaviorism. Since paraphrase is an independent test of prior metaphysical convictions, paraphrase is not an idle exercise. Carrara and Varzi’s suspicion is ill-founded.

Easy ontologists are not well placed to adopt this suspicion in any case. It is precisely their view that sentences of the form (S3) and (S4) are analytically equivalent (although, as we have seen, they eschew describing any of these sentences as paraphrasing the other), so they are in no position to suppose that counterexamples are available to paraphrases made using that template. Yes, many paraphrases offered by philosophers are more haphazard than systematic. Even given an algorithm, it would be still more desirable to have a compositional semantic theory for the fragment of language in question from which the algorithm was derived as a consequence (Lewis Reference Lewis1983: 16–17). That said, such an algorithm and the more informal paraphrases that philosophers devise remain valuable. The algorithm provides data points concerning a language, which a compositional semantic theory for that language would need to generate, insofar as we have reason to believe that those data points are accurate. Szabó (Reference Szabó, Loux and Zimmerman2003: 22) concedes something weaker: that it is possible that semantic theory matches the truth-conditions assigned by the paraphrases. Paraphrases can provide more, though. After all, systematicity and comprehensiveness are matters of degrees. Any account of the semantics of (say) adjectives or modals in a given natural language will itself provide only a fragment of the semantic story for that language. Evidence that the proposed fragment provides an adequate semantics for the aspect of language in question is also evidence that the fragment will be retained as part of any finished and comprehensive semantic account for the language.

Furthermore, unlike many of the paraphrases that philosophers have historically offered, ours do not decrease our ontological commitments at the cost of increasing our ideological commitments. Ours require no increase in our stock of primitive terms.

Even these constraints, however, are not enough to demarcate good from bad paraphrases. Why then think that our favored paraphrases are good ones? Consider the following case of Quine’s (Quine Reference Quine1964). Suppose each closed sentence S of a theory T is paraphrased by a sentence True(n), where n is the Gödel sentence for S and ‘True’ is the truth predicate for T. As we illustrate below, such a paraphrase is ontologically parsimonious, for if T quantifies over numbers, then the paraphrase quantifies over only numbers. And the paraphrase is ideologically parsimonious, given that T already operates with ‘True’ and the numerals. Nevertheless, the paraphrase is bad. Why? The Gödel sentence paraphrases are not logically equivalent to the originals. Whereas, we may suppose, S is a sentence from a biological theory T about the range of prey of foxes—and so has the logical consequence that there are foxes and their prey—the Gödel sentence of S has none of these logical consequences. Those who think that it does are welcome to find S’s analytic entailments by inspection of its Gödel sentence, as well as to reconstruct its use in common argument forms such as modus ponens and modus tollens via Gödel sentences and number-theoretic deductions alone.

Our appeal to paraphrase is curtailed in another important way. We are concerned only with the employment of paraphrase in connection with what we have called “duck-rabbit” cases: cases where an uncontroversial sentence that does not refer to a property is syntactically transformed into another sentence, and it is then concluded that this new sentence does refer to a property and so entails the existence of some property. We do not claim that paraphrase shows that no true sentence refers to a property and so that no true sentence entails the existence of some property. We leave open what to say about (12)–(14):

(12) and (13) involve sentences which apparently refer to properties and which apparently entail the existence of properties. (14) quantifies over magnitudes (a species of properties) and so apparently entails the existence of properties. It is open to repudiate such sentences. Some sentences are only apparently about properties—those can be paraphrased. Others are genuinely about properties—and those we abstain from using. There is nothing untoward in this, since we already repudiate (what is expressed by) ‘there are properties’. About claims about the average family, we do each. ‘The average family has 2.1 children’ can be paraphrased whereas ‘the average family exists’ is repudiated.

The interest of the easy ontologist’s case was that it involved an initial premise, which did not apparently refer to a property and which did not apparently entail the existence of properties, and yet—according to the easy ontologist—it entails that there exists a property. The easy ontologist’s case is that merely the truth of subject-predicate sentences entails that a property exists. We have argued that their case must rest on something else.

9. Conclusion: Prolix Talk

We’ve compared talk of properties to idiomatic talk. Prolixity is an inflated way of speaking. It is characteristic of euphemisms, but it is also found in many idioms: witness ‘He’s all hat and no cattle; truly he has the luck of the devil.’ We further think that talk of properties is prolix talk. Saying this is not special pleading from aspiring nominalists. First, our objection to the easy ontologists’ case for the existence of properties should be welcome to a posteriori realists about properties—those who think that the case for the existence of properties should be made on the basis of total science to explain the causal powers of objects and objective resemblances between them (e.g., Armstrong Reference Armstrong1978: xiii; Campbell Reference Campbell1990: 24–6). Their case is made on the basis of sentences such as (12)–(14), sentences that we did not attempt to paraphrase away.

Second, prolix talk is not confined to talk of properties. Consider the following pairs:

(15) says what (16) says in a more inflated way. Similarly, with (17) vis-à-vis (18). (19) and (20) are a more philosophically interesting pair. It is tempting to think that when we say there are things and they have natures we are talking about two kinds of entity, but talk about the nature of X can be dropped. It is prolix talk; we can just talk about X. Talk of the nature of something is just talk of its defining properties, and if talk of natures is prolix talk, then so too is talk of the defining properties of things. Yet, if that much talk is prolix, it would not be surprising if all talk of properties were prolix, even if devices in addition to paraphrase are needed to show this.

We marplots think that easy ontologists misdescribe their project. Their project is one of easy ideology: they show how we can introduce new terms and constructions into our language to say things which convey no more than could be conveyed by our original language. The added terms and constructions are easily added because they do not increase the expressive resources of our language. The natural contrast is, then, with “hard” ideology, where the introduction of terms is not mere prolixity but increases the expressive resources of our language. Precisely because easy ideology adds nothing to the expressive power of our language, its addition should be recognized as uncontentious and philosophically insignificant.Footnote ⁵