Jump to ContentJump to Main Navigation
Putting Metaphysics First$

Michael Devitt

Print publication date: 2009

Print ISBN-13: 9780199280803

Published to Oxford Scholarship Online: February 2010

DOI: 10.1093/acprof:oso/9780199280803.001.0001

“Ostrich Nominalism” or “Mirage Realism”? *

Chapter:
(p.13) 1 “Ostrich Nominalism” or “Mirage Realism”?*
Source:
Putting Metaphysics First
Author(s):

Michael Devitt (Contributor Webpage)

Publisher:
Oxford University Press

David Armstrong's approach to “the problem of universals” has a contemporary gloss: he leaves it to “total science . . . to determine what universals there are.” Nevertheless his conception of the problem shows him to be a devotee of the “old‐time” metaphysics. The problem is the traditional one allegedly posed by the premise of “Platos One over Many argument”: “Many different particulars can all have what appears to be the same nature” (p. xiii).1 It is a pity that Armstrong takes no serious account of the “new” metaphysics of W. V. Quine and others according to which there is no such problem as Armstrong seeks to solve.2 In my view this Quinean position is a much stronger rival to Armstrong's Realism about universals than the many others he carefully demolishes.

The universals we are concerned with here are properties (what Quine calls “attributes”) and relations. “Realists” believe in them, “Nominalists” don't. After outlining five versions of Nominalism, Armstrong mentions the Quinean position as a possible sixth under the title “Ostrich or Cloak‐and‐dagger Nominalism”:

I have in mind those philosophers who refuse to countenance universals but who at the same time see no need for any reductive analyses of the sorts just outlined. There are no universals but the proposition that a is F is perfectly all right as it is. Quine's refusal to take predicates with any ontological seriousness seems to make him a Nominalist of this kind.(p. 16)

Worse, these philosophers are guilty of trying to have it both ways: denying universals whilst, prima facie, unashamedly making use of them. They commit the (p.14) sin of failing to answer “a compulsory question in the examination paper” (p. 17). In Quinean language, they fail to face up to their ontological commitments.

Ostriches are reputed to ignore problems by putting their heads in the sand. Mirages are another feature of desert life: people see things that aren't there. An “Ostrich Nominalist” is a person who maintains Nominalism whilst ignoring a problem. A “Mirage Realist” is a person who adopts Realism because he sees a problem that isn't there. My major thesis is as follows:

  1. 1. To maintain Nominalism whilst ignoring the One over Many argument is not to be an Ostrich Nominalist; rather to adopt Realism because of that argument is to be a Mirage Realist.

Establishing this thesis would not, of course, show Realism to be unjustified (let alone false): there might be problems independent of the One over Many argument for which Realism is a possible solution. Armstrong thinks there are. I agree. To the extent that he is responding to those problems he is not a Mirage Realist. My thesis about him is as follows:

  1. 2. Armstrong is largely though not entirely a Mirage Realist.

Correspondingly, a Nominalist could be an Ostrich by putting his head in the sand as real problems loom. However correct his stand on the One over Many argument he could otherwise commit the sin that Armstrong complains of. I don't know whether there are any Ostrich Nominalists, but the only philosopher Armstrong alleges (tentatively) to be one, Quine, is not:
  1. 3. Quine is not an Ostrich Nominalist.

Argument for Thesis 1

According to Armstrong, the problem posed by the One over Many argument is that of explaining “how numerically different particulars can nevertheless be identical in nature, all be of the same ‘type’ ” (p. 41). What phenomena are supposed to need explaining here? I take it that what Armstrong is alluding to is the common habit of expressing, assenting to, and believing, statements of the following form:

  1. (1) a and b have the same property (are of the same type), F‐ness.

To settle ontological questions we need a criterion of ontological commitment. Perhaps Quine's criterion has difficulties, but something along that line is mandatory. The key idea is that a person is committed to the existence of (p.15) those things that must exist for the sentences he accepts to be true. What must exist for a given sentence to be true is a semantic question to which our best theory may give no answer in which we have confidence. Furthermore the sentence may, by its use of quantifiers or singular terms, suggest an answer which the person would want to resist. Hence, in my view, the importance of Quine's mention of paraphrase in this context. Suppose the given sentence seems to require for its truth the existence of G's yet the person can offer another sentence, which serves his purposes well enough, and which is known not to have that requirement. This is known because our semantic theory can be applied to this other sentence, in a way that it cannot to the given sentence, to show that the sentence can be true even though G's do not exist. We can then say that the person's apparent commitment to G's in the given sentence arises from “a mere manner of speaking”; he is not really committed to them.

Now in the ordinary course of conversation a Quinean is prepared to express or assent to the likes of (1). (1) seems to require the existence of an F‐ness for it to be true. So he appears committed to that existence. To this extent the One over Many argument does pose a problem to the Quinean Nominalist, but it is a negligible extent. He has a suitable paraphrase readily to hand:

  1. (2) a and b are both F.

When the ontological chips are down, he can drop (1). There is no problem about identities in nature beyond a trivial one of paraphrase.

Armstrong will not be satisfied by this, of course: “You have simply shifted the problem. In virtue of what are a and b both F?” The Quinean sees only a trivial problem here too. It is in virtue of the following:

  1. (3) a is F;

  2. (4) b is F.

Armstrong will still be dissatisfied: “In virtue of what is a (or b) F?” If the One over Many argument poses a problem it is this. That was historically the case and, though Armstrong always states the problem in terms of identities in nature, it is the case for him too.3 If there is no problem for the Nominalist in (3) and (4) as they stand then he has an easy explanation of identities in nature.

The Realist who accepts the One over Many problem attempts to solve it here by claiming the existence of a universal, F‐ness, which both a and b have. The Nominalist who accepts the problem attempts to solve it without that claim. The Quinean rejects the problem.

(p.16) The Quinean sees no problem for Nominalism in the likes of (3) because there is a well‐known semantic theory which shows that (3) can be true without there being any universals:

  1. (3) is true if and only if there exists an x such that ‘a’ designates x and ‘F’ applies to x.

So (3) can be true without the existence of F‐ness. There is no refusal here “to take predicates with any ontological seriousness”. The Quinean thinks that there really must exist something (said as firmly as you like) that the predicate ‘F’ applies to. However that thing is not a universal but simply an object. Further, in denying that this object need have properties, the Quinean is not denying that it really is F (or G, or whatever). He is not claiming that it is “a bare particular”. He sees no need to play that game.

The Realist may reply that this is a mistaken statement of the truth conditions of (3) and that the correct one does require the existence of F‐ness for (3)’s truth. Until a good argument for this reply is produced the Quinean is entitled to go on thinking he has no problem.

All of this is not to say that there is nothing further about (3), or about a being F, that might need explanation. I can think of four possible problems here. None of them pose any special difficulty for the Nominalist: they are irrelevant to “the problem of universals”.

(i) We might need to explain what caused a to be F. (ii) We might need to explain what was the purpose of a being F. Nobody interested in “the problem of universals” is likely to confuse their problem with (i) or (ii) and so I shall set them aside immediately.

It is not so easy to keep the next two problems distinct from “the problem of universals”. (iii) If ‘F’ is not a fundamental predicate then as reductivists we might need to explain what constitutes a being F: perhaps we will want to be told that it is in virtue of being G, where ‘G’ is some physical predicate (a is a gene in virtue of being a DNA molecule). (iv) We might need to explain the semantics of ‘F’: we might want to know what makes it the case that ‘F’ applies to a.

The traditional “problem of universals” has often appeared in a misleading semantic guise: how can ‘F’ “be applied to an indefinite multiplicity of particulars” (p. xiii; Armstrong does not approve of this way of putting the problem)? The strictly semantic problem of multiplicity does not have anything to do with universals. We need to explain the link between ‘F’ and all F things in virtue of which the former applies to the latter. This is not different in principle from explaining the link between ‘a’ and one object, a, in virtue of which the former designates the latter. The explanation of ‘F’’s application (p.17) depends on a theory of one semantic relation, application, the explanation of ‘a’’s designation depends on a theory of another, designation. A feature of the explanations will be that it is F things that are linked to ‘F’, and a that is linked to ‘a’. The F‐ness of F things and the a‐ness of a need not go unexplained in the semantics. Thus I think it is part of a good explanation of the link between ‘tiger’ and the many objects that it applies to that those objects are genetically of a certain sort. So the semantic problem may require some answer to the question: in virtue of what is a F? But the answer required is of type (iii), a reductivist answer.

In denying that there is any problem for the Nominalist about (3) it is important to see that we are not denying the reductivist problem (iii), nor the semanticist problem (iv), nor some combination of (iii) and (iv). What we are denying can be brought out vividly by taking ‘F’ to be a fundamental predicate, say a physical predicate. Then there is no problem (iii): we have nothing to say about what makes a F, it just is F; that is a basic and inexplicable fact4 about the universe. Problem (iv) remains: it is the problem of explaining the link between the predicate ‘F’ and that basic fact. Nothing else remains to be explained.

Why be dissatisfied with this? Explanation must stop somewhere. What better place than with a fundamental physical fact of our world?

Armstrong feels that we need to go further. How can we tell who is right? There is one sure sign that explanation has not gone far enough: an explanation that goes further. Thus if Armstrong's Realist response to the One over Many argument is a genuine explanation then there must be a genuine problem here to be explained. My final remarks in support of thesis I will consider Armstrong's response.

One Realist response, but not Armstrong's, to the One over Many argument runs as follows: a is F in virtue of having the property F‐ness. We explain (3) by

  1. (5) a has F‐ness.

An obvious question arises: how is (5) to be explained? The Realist feels that the one‐place predication (3) left something unexplained, yet all he has done to explain it is offer a two‐place predication (a relational statement). If there is a problem about a being F then there is at least an equal problem about a having F‐ness. Furthermore, the point of this manoeuvre for the Realist is to commit us to universals. In ontology, the less the better. Therefore this (p.18) sort of Realist makes us ontologically worse off without explanatory gain. Any attempt by him to achieve explanatory power by explaining (5) seems doomed before it starts: it will simply raise the same problem as (5); he is in a vicious regress. If there is a problem about (3) this sort of Realist cannot solve it.

Armstrong calls the doctrine we have just considered “relational Immanent Realism”, and rejects it for reasons not unconnected to mine (pp. 104–7). In its place he offers us “non‐relational Immanent Realism.” This doctrine is obscure. Armstrong offers us (5), or the similar, ‘F‐ness is in a,’ and simply declares it to be non‐relational and inexplicable: particulars are not related to universals but bonded to them in a metaphysical unity (pp. 108–11). We have just seen that (5), taken at face value, cannot explain any problem about (3): it is a relational statement and so any problem for (3) is a problem for it. Armstrong avoids this grievous difficulty for Realism by fiat: (5) is not to be taken at face value. How then is it to be taken? Do we have even the remotest idea of what the words ‘in’ and ‘have’ mean here if they are not construed as relational predicates? Armstrong's Realism replaces the explanatory failings of relational Realism with a complete mystery. I suspect that Armstrong views sentences like (5) as attempts to speak the unspeakable: to talk about “the link” between particulars and universals without saying they are related. (Note the scare‐quotes around ‘in’ on p. 108 and the use of a special hyphenating device on p. 111.)

Talk of “particulars” and “universals” clutters the landscape without adding to our understanding. We should rest with the basic fact that a is F. Even the alleged unity of particular and universal can be captured without mystery: a predication must involve both a singular term and a predicate; drop either partner and you say nothing. For the Nominalist the unity of predication is an unexciting linguistic fact. The move to relational Realism loses the unity. Armstrong's non‐relational Realism attempts to bring it back with metaphysical glue. These are “degenerating problem shifts” (Lakatos).

Armstrong sees the One over Many argument as posing a problem for Nominalism and offers a Realist solution. If his solution were real then the problem would be real. The solution is not real. So it throws no doubt on my earlier argument that the problem is not real.

Indeed the Quinean can gain much comfort from Armstrong's book: it is a powerful argument for thesis 1. We have just demonstrated the failings of Armstrong's response to the One over Many argument. Armstrong himself carefully, and convincingly, demolishes every other known response to it. This (p.19) chronicle of two thousand years of failure makes the task seem hopeless. The alternative view that there is no problem to solve becomes very attractive.

I take my major thesis to be established:

  1. 1. To maintain Nominalism whilst ignoring the One over Many argument is not to be an Ostrich Nominalist; rather to adopt Realism because of that argument is to be a Mirage Realist.

Even if there are universals they cannot form part of a solution to the One over Many problem, because that problem is a mirage.

Argument for Thesis 2

The arguments for theses 2 and 3 will be brief.

It follows from thesis 1 that in so far as Armstrong adopts Realism because of the One over Many argument, he is a Mirage Realist. At the beginning of his book he indicates that he sees that argument as the main one for universals (p. xiii). When he talks of “the problem of universals” it is the problem allegedly posed by that argument that he is referring to (e.g. p. 41). Almost the whole book is taken up with the consideration of responses to that argument. Armstrong is largely a Mirage Realist.

In one chapter, drawing on the ideas of Arthur Pap and Frank Jackson, Armstrong offers quite independent reasons for Realism (pp. 58–63).5 We all assent to, express, believe, statements like the following:

  1. (6) Red resembles orange more than it resembles blue;

  2. (7) Red is a colour;

  3. (8) He has the same virtues as his father;

  4. (9) The dresses were of the same colour.

Unlike (3) these seem to require the existence of properties for them to be true. Whether or not they are sufficient for Realism depends on whether or not we can find acceptable paraphrases without that commitment. There is nothing illusory about this problem for a Nominalist. Armstrong is not entirely a Mirage Realist. So,
  1. 2. Armstrong is largely though not entirely a Mirage Realist.

(p.20) Argument for Thesis 3

For Quine to be an Ostrich Nominalist would be for him to ignore the ontological problem posed by his acceptance of statements like (6) to (9). A priori it is unlikely that this would be so. Quine, more than any other philosopher, has pointed out what constitutes an ontological commitment and has preached against ignoring such. Philosophers, like others, can fail to practise what they preach, but I suggest that it is unlikely that Quine would fail here, about as unlikely as that he would confuse use and mention.

A quick glance through Word and Object6 shows that he does not fail. In a section on abstract terms he considers, for example, the sentence,

  1. (10) Humility is a virtue.

a sentence that raises much the same problem as Armstrong's (8), and sees it as committing him to the existence of “an abstract object” (p. 119), in fact to “an attribute”, what Armstrong would call “a property”. He goes on to “deplore that facile line of thought” that allows us to ignore this (pp. 119–20). He considers ways to paraphrase away this apparent commitment to attributes and admits the difficulties (pp. 121–3). The issues are postponed until chapter 7. He does not there discuss sentences like (6) to (10) directly, so far as I can see, but his strategy for them is clear enough: all talk of attributes is to be dispensed with in favour of talk of eternal open sentences or talk of classes (p. 209). Whatever the merits of this approach it is not the behaviour of an Ostrich. So,
  1. 3. Quine is not an Ostrich Nominalist.

Postscript to “ ‘Ostrich Nominalism’ or ‘Mirage Realism’?”

1. Criteria of Ontological Commitment

1.1 Semantic and Quinean criteria

My presentation of the Quinean criterion of ontological commitment begins with the claim that “a person is committed to the existence of those things that (p.21) must exist for the sentences he accepts to be true”. I go on: “What must exist for a given sentence to be true is a semantic question” requiring a “semantic theory” to answer it. I propose part of a theory for that purpose: ‘a is F’ is true iff there exists an x such that ‘a’ designates x and ‘F’ applies to x. This was a mistake7 and is at odds with “Putting Metaphysics First”, the title and guiding idea of the present volume.

The real Quinean criterion of ontological commitment is simpler than the semantic criterion I presented. The Quinean criterion is that a person is committed to the existence of a or Fs if he says that a or Fs exist. As Quine points out:

We commit ourselves to an ontology containing numbers when we say there are prime numbers larger than a million; we commit ourselves to an ontology containing centaurs when we say there are centaurs; and we commit ourselves to an ontology containing Pegasus when we say Pegasus is. (1961: 8)

Having emphasized that mere talk of red houses, roses, and sunsets does not commit us to the existence of redness, Quine goes on:

We can very easily involve ourselves in ontological commitments by saying, for example, that there is something (bound variable) which red houses and sunsets have in common. (p. 12)8

If we said this we would commit ourselves to the property redness, a universal.9 But no semantic theory is needed to show this commitment. For someone to show this, the person must of course understand what we say but this understanding does not require the application of a semantic theory. Our ordinary understanding of a language is a skill quite unlike any theoretical knowledge of language. Or so I have argued at painful length (1981: 95–110; 1991b: 270–5; 2006d).10

(p.22) 1.2 The Priority of the Quinean Criterion

This is not to say that there is anything wrong with using the semantic criterion to establish commitment. However, it has the disadvantage of requiring a semantic theory that may be controversial. Furthermore, it is important to note that the Quinean criterion is explanatorily prior. Consider Tom's assertion:

  • (1) Lulu is a cat.

Applying the semantic criterion, and using our semantic theory, we see this assertion as ontologically committing Tom to Lulu and cats, because, first,
  • (a) (1) is true iff there exists something such that ‘Lulu’ designates it and ‘cat’ applies to it;

and second,
  • (b) If ‘Lulu’ designates, it designates Lulu;

  • (c) If ‘cat’ applies, it applies to cats.

What is it about (a), (b), and (c) that shows that Tom is ontologically committed to Lulu and cats in asserting (1)? Clearly it is our talk of existence. Tom is ontologically committed to cats if we are right in saying, as a consequence of (a) and (c), that cats must exist for (1) to be true. So the semantic criterion presupposes that saying that cats exist establishes an ontological commitment to cats; i.e. it presupposes the Quinean criterion.11

1.3 Consequences of Using the Semantic Criterion

In using the semantic criterion I had to apply a semantic theory. This invited an objection made by David Armstrong in his response to my paper: “there may be alternative, and perhaps more satisfying, ways of giving the semantics (p.23) for ‘Fa’. Devitt offers no argument against this possibility” (1980: 445). Hugh Mellor and Alex Oliver object similarly in their helpful introduction to a collection of papers that includes mine and Armstrong's: “Since [Devitt] and Quine link ontological commitment to a semantic theory, we need some way of telling when such a theory is correct, and an argument for its supposed link with ontological commitment” (1997b: 14–15).

Now, whatever the merits of these objections they do nothing to support Armstrong's traditional claim that “the main argument for the existence of universals is Plato's ‘One over Many’ ” (1980: 440). And that claim is the main target of my paper and, of course, of Quine's “On What There Is”. To support the traditional claim, we need an argument to show that the semantic theory I use is not correct. For, if that theory is correct, the One over Many poses no problem for the Nominalist. To sustain the traditional claim it is not sufficient to suggest, rightly, that the semantic theory might not be correct. The onus is on those who make the traditional claim to show that the theory is not correct.12

As noted, my semantic proposal for the truth conditions of ‘a is F’ was:

  • a is F’ is true iff there exists an x such that ‘a’ designates x and ‘F’ applies to x.

Mellor and Oliver continue their objection as follows:

suppose the following clause appears in the correct semantic theory:

  • a is F’ is true iff there is a ϕ such that ‘is F’ designates ϕ and ‘a’ falls under ϕ.

This suggests that ‘a is F’ is only committed to F‐ness, not to an entity designated by ‘a’. (1997b: 15)

According to my Nominalistic semantic proposal, the semantic criterion takes ‘a is F’ to be ontologically committed to an F object; according to the present Realist proposal, the criterion takes ‘a is F’ to be committed to the universal F‐ness. Which semantic proposal should we prefer? That is a good question. But whatever the answer, it will do nothing to sustain the traditional claim about the importance of the One over Many. Clearly, if my Nominalistic proposal should be preferred, the semantic criterion shows that the One over Many is no problem for Nominalism, just as I argued. If, on the other hand, (p.24) the Realist proposal should be preferred, then the One over Many is irrelevant: the argument that justified the preference for a Realist semantics would have already established Realism.13

So why should we prefer one semantic proposal over the other? There can be no purely semantic answer. Rather, we to look to our best scientific theories of the world and apply the Quinean criterion. According to that criterion, our best theories clearly and unequivocally commit us to objects but the jury is still out on whether they commit us to universals. So we should prefer a Nominalistic semantics that is committed only to objects. In this way we put metaphysics first.14

1.4 Revisiting the One over Many

Putting all the weight on the semantic criterion was a mistake. Set that criterion aside and apply the Quinean one. Then the basic case against Armstrong and the One over Many is that the “identity in nature” of red houses and sunsets need not commit us to the universal redness. We accept that the houses and sunsets exist and explain away their “identity in nature” as simply a matter of their all being red. No more need be said. The One over Many is a pseudo problem and to adopt Realism about universals because of it is, indeed, to be a Mirage Realist.

2. Paraphrasing

Mellor and Oliver have another problem with the Quinean approach to ontological commitment, arising from the role that the approach gives to paraphrasing. They describe this role as follows:

Suppose we have a sentence S, apparently committed to some entity e, and an equivalent sentence S’ which is said to be uncommitted to e. This, it is said, shows that S is only apparently committed to e.

Calling on Alston (1958), they object:

Why should we think it is S and not S’ that deceives us? Why not say the S’ is really committed to e because its equivalent S is? What we need here is what we do not have, namely a test for when a sentence is only apparently, and when really, committed or uncommitted to some entity. (1997b: 15)

(p.25) This objection fails because it rests on several misunderstandings.15 First, we need to distinguish the ontological commitments of sentences and of people. Quine's criterion “applies in the first instance to discourse and not to men” (1961: 103). When it is applied to S and S’, their commitments will be real not simply apparent: either they do or do not “quantify over” e by saying that e exists; neither “deceives us”. And there is no role for paraphrasing here. Second, paraphrasing has its role in determining the commitments of people. Suppose that S is committed to e but that S’ is not. Suppose that a person asserts S thereby apparently committing himself to e. Yet suppose, further, that he is prepared to withdraw that assertion in favor of S’ because S’ lacks the commitment and yet will serve his purposes well enough. He thus “frees himself from ontological commitments of his discourse” (ibid.). His commitment to e is only apparent not real: it arose from “an avoidable manner of speaking” (p. 13). Third, in these circumstances, we might say that S and S’ are “equivalent” in some sense—they serve the same purpose—but we should not say that they are equivalent in the sense of being synonymous. For, if they were synonymous, it is hard to see how they could differ in their ontological commitments.16

3. Response to Armstrong

3.1 The Genuine Problem of Universals

In my paper I followed Quine in allowing that there is a genuine problem of universals arising from our apparent quantification over properties; see Arguments for Thesis 2 and for Thesis 3.17 Armstrong has quite a bit more to say about this problem in his response (1980: sec. II). In assessing this response, it is important to note that I made no attempt to solve this genuine problem in my paper. In particular, I presented no argument against Realism about universals. And I shall not do so here.

3.2 The One Over Many

I wrote my paper largely because I felt that Armstrong, in his book (1978a), had taken “no serious account” of Quine's argument that the One over Many (p.26) is not a genuine problem for Nominalism. This Quinean point is the concern of my Argument for Thesis 1, occupying almost the entire paper. It seems to me that Armstrong, in his response (1980), still takes no serious account of the Quinean argument. Rather, his attention is elsewhere, looking for other possible reasons for Realism. We have already seen evidence of this in section 1.3 above. Here is more.

Armstrong again insists that the One over Many “shows that there is a strong preliminary case for accepting universals” (p. 440). But the Quinean point is, of course, that the One over Many provides no case at all. Armstrong draws attention to ubiquitous talk of sameness of type. He claims that this sameness is “a Moorean fact” that needs an account. He then simply repeats the charge that Quine is an ostrich for not giving such an account, for “refusing to answer a compulsory question” (p. 441). But, of course, my Argument for Thesis 1 is precisely that Quine has answered this question. To illustrate, a and b might be “of the same type” in virtue of their both being red; no talk of the universal redness is called for. Armstrong has largely ignored the Quinean argument.

Armstrong does add a new wrinkle: he raises the sameness‐of‐type problem at the level of predicates. He claims that the Quinean answer I have just illustrated rests on a rule along the following lines:

Suppose that we are given sentences that of the form ‘a is—and b is—.’ If but only if the two blanks are filled by the same predicate, it is permitted to rewrite the sentence as ‘a and b are both—,’ with the same predicate in the new blank. But ‘same predicate’ here is a type‐notion. . . . Some account must then be given, reductive or otherwise, of what sameness of type is. (p. 442)

But the Quinean treats talk of the sameness of type of predicates in just the same way that he treats talk of the sameness of type of stones, trees, cats, or whatever. Just as we paraphrase away talk of objects sharing the property redness by saying that they are all red, we paraphrase away talk of objects sharing the property of being of the word type ‘red’ by saying that they are all ‘red’ tokens. The Quinean treatment is quite general. The One over Many is a pseudo problem at all levels.

3.3 “Solutions” to the Pseudo Problem

In my paper I emphasized that the Quinean does not deny that an object really is red or whatever. Armstrong is unmoved, insisting that, on the Quinean view, “particulars are a sort of structureless blob . . . they lack real internal structure” (p. 446). This is a caricature. It foists on the Quinean an ontological framework that is motivated by the One over Many problem, (p.27) just the problem that the Quinean rejects. So the problem does not lead the Quinean to traffic in “bare particulars”, “mere thisnesses”, and the like; as I remarked, “he sees no need to play that game”. Suppose that, according to the Realist, an object has an internal structure F‐ness. Then, according to the Quinean, it really is F, said as firmly as you like. Nothing more need be said.

The Quinean gains comfort from the dismal failure of attempted solutions to the One over Many problem through more than two millennia. Armstrong poses the issue vividly: Realists hold that “a particular involves a factor of particularity (haeccitas, thisness) together with properties which are universals. The question is then this: how are the two components of a particular to be put together?” (p. 446) Relational answers always generate a vicious regress, as Armstrong has demonstrated convincingly. Armstrong's non‐relational answer is that “the thisness and the nature are . . . not related”. Armstrong admits, with admirable frankness, that this is “profoundly puzzling” and that my claim that it is an “inexplicable mystery” is not implausible (p. 447). Indeed, as I say, non‐relational answers seem to involve “attempts to speak the unspeakable: to talk about the ‘the link’ between particulars and universals without saying they are related”.

3.4 Armstrong's ad Hominem

I agree that there is a genuine problem of universals but have made no attempt to solve it. I think, as Armstrong points out, “that it may be necessary to postulate universals”. Suppose that it is. Armstrong asks how I will “solve the problem of how universals stand to particulars?” He thinks that I will “end up saying something similar” to what he and other Realists have had to say (p. 448). I might well end up saying something similar to what the Relational Realists have had to say. But, if I did, I would not be open to the vicious‐regress objection that Armstrong has leveled at these Realists. For, that objection is effective only against those who are Realists as a solution to the One over Many problem. Those Realists must accept a further universal every time they relate a universal to a particular. My Realism would not be a solution to that problem and so I would not have to accept the further universal.18 A Relational Realism that is a response to the pseudo problem of universals must indeed generate a vicious regress; a Relational Realism that is a response to the genuine problem need not. If the One over Many were a genuine problem, the vicious‐regress objection would show that it (p.28) was insoluble. So, we have very good reason to think that it is not a genuine problem.

4. Lewis's One over Many

David Lewis also finds Armstrong's discussion of the One over Many problem “unconvincing” (1983: 351) and agrees with much of what I say about it (p. 354). However, he thinks that this problem, concerned with

  • a and b have the same property (are of the same type), F‐ness,

is “too easy”. There is another One over Many problem concerned with the less definite
  • a and b have some common property (are somehow of the same type).

that is not so easy to deal with (p. 355). I agree, but I think that it is still not that difficult.

First, what are we to make of the less definite claim? Lewis has remarked earlier: “Because properties are so abundant, they are undiscriminating. Any two things share infinitely many properties, and fail to share infinitely many others” (p. 346). So, if we are prepared to talk of properties it seems trivial to say that two objects share one. In any case, it is easy for the Quinean to paraphrase such an apparently trivial claim with the similarly apparently trivial:

  • a and b resemble each other.

But doubtless someone making the less definite claim has in mind that a and b share a significant property, “one of an elite minority of special properties” which Lewis calls “natural properties” (p. 346). If so, we can paraphrase the less definite claim well enough with:
  • a and b significantly resemble each other.

In sum, in the paper we paraphrased away an apparent commitment to a and b sharing a particular property F‐ness by talking of their both being F. We now paraphrase away an apparent commitment to a and b sharing some property or other by talking of their resembling, or significantly resembling, each other.

(p.29) 5. A Diagnosis

The One over Many is a pseudo problem. Why, then, are philosophers so beguiled by it?19 I suspect that the reason is an implicit commitment to the “ ‘Fido’‐Fido” theory of meaning. This theory starts from the appealing idea that the meaning of a proper name like ‘Fido’ is the object it names, Fido.20 The theory generalizes this view of meaning to all terms. The theory has had a persistent hold over the minds of philosophers and many others.

Consider, for example, what Armstrong has to say about Quine's “extraordinary doctrine” of ontological commitment. He claims that, according to this doctrine, “predicates involve no ontological commitment”; a predicate “need not be taken with ontological seriousness” it has “what has been said to be the privilege of the harlot: power without responsibility” (1980: 443). Yet, as we have seen, the Quinean takes predicates with great ontological seriousness: ‘There are Fs’ is committed to the existence of Fs because of the role of the predicate ‘F’. Why is that not serious enough for Armstrong? It looks as if he thinks that to be really serious we must take ‘F’ to name something: we must adopt the ‘Fido’‐Fido theory.

It is easy to see how this theory leads us to universals. Consider ‘That rose is red’. This sentence, like all others, has a certain complexity. It has two terms, the singular term ‘that rose’ and the general term ‘red’, of different grammatical categories and playing quite different roles. How can the ‘Fido’‐Fido theory cope with this complexity? It has to see the two types of term naming two types of entities: the different roles of the terms require different types of entities. The entity named by ‘that rose’ is a particular rose; that named by ‘red’ is the universal, redness, which can be shared by many particulars. The One over Many begins to look like a real problem.

In my view, the ‘Fido’‐Fido theory is false. The present chapter illustrates a quite different way of coping with the complexity of sentences. It is not that each term stands in the one semantic relation of naming to different kinds of entities. Rather, the terms stand in different semantic relations to the same kind of entities, neither “particulars” nor “universals” but just plain objects. Thus ‘that rose’ designates a certain object, a rose, while ‘red’ applies to many objects, including many roses. Where the ‘Fido’‐Fido theory catches the complexity (p.30) with different sorts of entity, we catch it with different sorts of relation. The only entities we need are objects of the familiar sort.

If this is the right diagnosis, the One over Many problem shows the cost of not putting metaphysics first: our semantics should be driven by our metaphysics not vice versa. Perhaps that metaphysics cannot be Nominalist but we have seen that the One over Many provides no reason for thinking that it cannot.21

Notes:

(*) First published in Pacific Philosophical Quarterly, 61 (Devitt 1980). Reprinted with kind permission from Wiley‐Blackwell. I am indebted to Elizabeth Prior for help with the first draft of this paper and to David Armstrong and Frank Jackson for helpful comments on that draft.

(1) Such references are to Armstrong 1978a: i.

(2) See particularly Quine's discussion in “On What There Is” (1961: 9–14), which Quine's discussion is largely aimed at a position like Armstrong's (“For ‘McX’ read ‘McArmstrong’ ”: Elizabeth Prior).

(3) See e.g. his remarks on Ostrich Nominalism (quoted above) and his discussion of the varieties of Nominalism, pp. 12–16.

(4) Lest an uncharitable reader should take this talk as committing me to the existence of facts, let me hasten to add that such talk is a mere manner of speaking, eliminable at the cost of style and emphasis.

(5) Given the importance Armstrong attaches to the One over Many argument for Realism, this chapter's title, “Arguments for Realism”, is misleading.

(6) Quine 1960.

(7) As I soon came to realize (1984: 40–3; 1991b: 50–3). This subsection and the next draws on that discussion.

(8) So, note, the Quinean criterion is not “biased against universals”. The criterion makes it easy for us to commit ourselves to absolutely anything.

(9) Jody Azzouni (2007) has offered an ingenious argument against the Quinean criterion. I am not convinced. Azzouni is on strongest ground in denying the ontological commitment of ‘there is . . . ’ in ordinary language. If he were right about that (which I don't think he is), the Quinean would simply fall back on ‘there exists . . . ’, or even ‘there really exists . . . ’, as the locus of commitment. But Azzouni denies that these expressions, indeed any expression in the vernacular, have a meaning that entails ontological commitment. How then do we convey that commitment? For Azzouni this is a pragmatic rather than semantic matter: we use rhetorical indicators, stage‐whisperings, and the like. This is prima facie very implausible. Given our ordinary and scientific interest in ontological matters it would be strange indeed if we had no expression in our language that entailed ontological commitment. It would be extraordinary e.g. that there could be no words in a theory that committed it to a certain ontology. And ‘exists’ seems perfectly designed for the job.

(10) Quine sometimes uses ‘true’ and ‘refer’ in stating his criterion: “a theory is committed to those and only those entities to which the bound variables of the theory must be capable of referring in order that the affirmations made in the theory be true” (pp. 13–14). But this use of ‘true’ simply exploits the “denominalizing” (“disquotational”) property of ‘true’ and does not make the criterion semantic; similarly, ‘refer’. So these uses do not make Quine's criterion dependent on the application of a semantic theory. Indeed, Quine has a deflationary view of truth. For more on the deflationary view, and the denominalizing property of the truth term, see ch. 2, part I, and ch. 8, particularly sec. 3, in the present volume.

(11) And just the same goes for the “truthmaker” criterion proposed recently by David Armstrong (2004: 23–4). For, on this criterion, a theory is ontologically committed to certain entities only because when we apply the criterion to the theory we conclude that those entities must exist in order to make the theory true. The criterion is parasitic on the Quinean criterion, as Jonathan Schaffer points out (2008: 16). In 7.3, of the present volume, I argue against the view that we need a metalinguistic statement of the truth conditions of a statement in order to establish the commitments of the statement.

(12) A similar response is appropriate to Armstrong's claim about the technical term ‘applies’ that features in my semantic theory: “The Realist may well argue, correctly I believe, that a convincing account of the semantics of ‘applies’ cannot be given without appeal to the properties and/or relations of the object a” (p. 445). Of course the Realist might argue this but he hasn't. So the problematic nature of the One over Many has still not been established. In any case, ‘applies’ is surely no more problematic for the Nominalist than any other two‐place predicate.

(13) The One over Many would be similarly irrelevant if Armstrong (2004) were right in supposing that the entities that must exist for our theories to be true are “states‐of‐affairs”. Whatever justified that supposition would establish Realism about states‐of‐affairs without appeal to the One over Many argument.

(14) I draw here on Devitt 1984: 45–6; 1991b: 57–8.

(15) My own discussion of paraphrasing in the present chapter is not as clear as it should have been.

(16) Despite all this I do accept that the role of paraphrasing raises a difficulty for Quine's criterion (1984: 44; 1991b: 54).

(17) I cited Word and Object (1960) as evidence of how seriously Quine takes the problem of universals. I might also have cited “Logic and the Reification of Universals” (1961: 102–38).

(18) As Quine says in his response to Armstrong on the regress problem, “the use of a two‐place predicate is not itself a reference to the relation” (1980: 451).

(19) My discussion in this subsection draws on Devitt and Sterelny 1999: 279–80.

(20) Although the idea is appealing I think it is not right (1981, 1996a, 2009 d).

(21) My thanks to David Armstrong and Jonathan Schaffer for comments on this postscript.