Paraphrase and the Symmetry Objection

ABSTRACT

There is a puzzle about the use of paraphrase in philosophy, presented most famously in Alston's [1958] ‘Ontological Commitments’, but found throughout the literature. The puzzle arises from the fact that a symmetry required for a paraphrase to be successful seems to necessitate a symmetry sufficient for a paraphrase to fail, since any two expressions that stand in the means the same as relation must also stand in the has the same (unwanted) commitments as relation. I show that, while this problem does undermine some conceptions of paraphrase, on a proper understanding of paraphrase the puzzle gets no purchase. Since paraphrase is an important component of Quinean approaches to meta-ontology, this paper constitutes a partial defence of Quinean meta-ontology. Since paraphrase is an important component of traditional methods of philosophical inquiry, this paper constitutes a partial defence of traditional modes of philosophizing as well.

KEYWORDS:

• paraphrase
• ontological commitment
• meta-ontology
• Quine, Alston

Notes

¹ See Correia and Schnieder [2012] for a nice introduction to the grounding literature.

² Alternatively, we might realize that a certain understanding of a sentence is inadequate and so aim to replace that way of understanding it with a better one. Thanks to an anonymous referee for pushing me to say more about the forms that revisionary paraphrase can take. See also note 13 and note 28.

³ In Keller [2015b], I argue that paraphrases must (typically) preserve truth-conditions—an important aspect of meaning—but need not preserve semantic content, in so far as truth-conditionally equivalent sentences can have different semantic contents. See note 11 for related discussion.

⁴ See, e.g., Quine [1948], Lewis and Lewis [1970], Lewis [1986], and van Inwagen [2009]. I call this approach ‘neo-Quinean’, since typical ‘Quinean’ meta-ontologists like David Lewis and Peter van Inwagen embrace an ontology and ideology of meaning (propositions, synonymy, etc.) that Quine himself repudiated. See Manley [2009] for a discussion of why this matters.

⁵ Most importantly, for avoiding unwanted ideological commitments: e.g. providing a reductive modal semantics that gives truth-conditions for the modal facts without making use of ‘possible’ or ‘necessary’.

⁶ Other materialists allow for the existence of sets and other abstracta, restricting their materialism to the concrete realm.

⁷ Versions of this argument are discussed in many places, including Lewis and Lewis [1970], Casati and Varzi [1995], and Varzi [2007].

⁸ Would non-fundamental immaterial souls be acceptable to such materialists?

⁹ At least, the Crack Argument provides no reason for privileging (1) over (1*). But what about the fact that (1), but not (1*), is formally entailed by ‘There is a thin crack in my favourite vase’? (Thanks to an anonymous referee for pressing me to address the parallel with Davidson [1967] on adverbial modification.) Well, first, there are strategies for handling inferences like these that don't involve quantification over cracks (see, e.g., Thomason and Stalnaker [1973]). Second, and more importantly, my ultimate purpose here is not to defuse the Crack Argument or to argue that (1*) is ultimately more perspicuous than (1). It is to show that paraphrases can be ‘non-symmetric’ in philosophically important ways. And if, contrary to what I suggest below, we should privilege the apparent logical form of (1) over that of (1*), this shows that paraphrases can be non-symmetric in philosophically important ways.

¹⁰ Here and below, I have substituted ‘crack’ (etc.) for ‘possibility’ (etc.), for the sake of continuity with our example.

¹¹ They clearly express the same proposition on a possible worlds conception of propositions (as defended in, e.g., Lewis [1986] and Stalnaker [1999]), where propositions are individuated by their truth-conditions. And, as I argue in Keller [2015b], sameness of truth-conditions is the aspect of meaning that a successful paraphrase must preserve: if (1) and (1*) are true in all of the same worlds, and (1*) is true in some worlds where materialism is true (as seems evident), then (1) is consistent with materialism. So, even if (1) and (1*) express different ‘fine-grained’ propositions, their truth-conditional equivalence would suffice to undermine the Crack Argument.

¹² (1*) appears to express a claim of the form is a vase & is cracked)—a claim that does not even suggest that is a crack).

¹³ Alternatively, we might realize that, although (1) and (1*) express different claims, we were, all along, intending to express the content of (1*) when we asserted (1). If so, there is no ‘cost’ to abandoning (1) in favour of (1*), since it requires no revision in what we believe. (Well, no revision other than our beliefs about the content of (1).) And if we abandon (1) in favour of (1*), the Crack Argument cannot get off the ground. In this case, (1*) would not be a (non-revisionary) paraphrase of (1) itself, but it would be a (non-revisionary) paraphrase of the belief that we were (trying to) express with (1).

¹⁴ For a nice discussion logical consequence in general and the view I favour in particular, see Blanchette [2012].

¹⁵ Similarly, a theory is ontologically committed to the existence of just in case it entails , and a person is ontologically committed to just in case she accepts a theory that is. The norms for attributing commitment are subtle: if I accept a theory that entails , and I don't accept any theory that entails , and if all s are s, it may appropriate in some contexts to say that I am committed to s.

¹⁶ While the central argument of this paper is compatible with many views about commitment, it is incompatible with views according to which the commitments of a theory cannot be hidden or non-obvious. The view defended in Brogaard [2008], according to which the commitments of a theory are the things that it would be irrational for someone who accepted the theory to deny, might seem to be such a view. However, the idealized rational commitments of a theory —the things that it would be irrational for an ideal agent who accepted to deny—will often be non-transparent. And the things that an actual agent who accepts would be criticizable for denying are in fact revisable (via paraphrase), for the reasons outlined in section 4.

¹⁷ A minimal constraint on successful non-revisionary paraphrase would seem to be material equivalence with the original. Hence, those who paraphrase (1) with (1*) and (2) with (2*) must accept (P1) and (P2). One could avoid this result if one held that while, e.g., (1) and (1*) have the same content, they (can) have different truth values because they belong to different ‘linguistic levels’. (Thanks to an anonymous referee for suggesting this.) It is, however, difficult to see how two sentences with the same content could have different truth values, so I prefer the solution offered in the main text.

¹⁸ More generally, a perspicuous paraphrase wears its commitments on its sleeve.

¹⁹ At least if we take ‘successful paraphrases’ to be those that express the target claim without any objectionable formal consequences. As John Searle has argued [1970: sec. 5.3, 1993], sentences making grammatically primitive predications of reality appear to be formally consistent with each other and almost anything else. So, for example, ‘Reality is such-that-my-favourite-vase-is-cracked’ is formally consistent with materialism, and seems to express the fact expressed by (1). The problem is that its formal consistency with materialism comes at a high cost: ‘Reality is such-that-my-favourite-vase-is-cracked’ is, formally, a logical island: its form gives us no clue about the logical properties of the claim it expresses. Indeed, if one blindly applied the standard heuristic rules for translating English sentences into the regimented idiom of logic, ‘Reality is such-that-my-favourite-vase-is-cracked’ would be translated as a sentence letter (or perhaps , where reality), and one would have to conclude that ‘Reality is such-that-my-favourite-vase-is-cracked’ is consistent with ‘There are no cracks in my favourite vase’, ‘Nothing I own is cracked’, etc. Since other ways of expressing (1) do not have this unwelcome lack of logical perspicuity, ‘Reality is such-that-my-favourite-vase-is-cracked’ is an inadequate paraphrase of (1), and cannot be used to reconcile materialism with the fact that my favourite vase is cracked. For a more detailed critique of such paraphrases, see Turner [2010]. Related discussions can be found in Lewis [1983] and Sider [2012: sec. 9.2]. Note that, according to the argument discussed in note 9, (1) is a perspicuous paraphrase of (1*), rather than vice versa. If that's true, perspicuity still functions as a symmetry breaker.

²⁰ Another response to Burgess and Rosen's argument would be to reject (P1) and (P2), retreating to a revisionary conception of paraphrase. On this view, (1) and (2) do not in fact have the same truth values as (1*) and (2*): strictly speaking, (1) and (2) are falsehoods, to be replaced with (1*) and (2*). We are sometimes reluctant to give such revisionary paraphrases, since it is not clear that we have more evidence for philosophical theories like materialism than we do for pre-theoretic facts such as that my favourite vase is cracked. But in this case the revision appears to be costless: (1*) and (2*) are, nearly enough, stylistic variants of (1) and (2). If there is a difference between the claims they express, it is a difference that falls below the threshold of cognitive significance. Hence, if (1) and (2) have undesirable implications, they can be replaced with (1*) and (2*) without altering our conception of reality in any noticeable way. In other words, such ‘revisionary’ paraphrases would entail no change in what we think, but only in what we say. On this view, our paraphrases would be semantically revisionary while remaining cognitively equivalent. Since our pre-theoretic evidence doesn't support (1) over (1*), it is hard to see what if any costs we would incur by replacing (1) with (1*). Likewise with (2) and (2*). Cf. note 13. Thanks to an anonymous referee for pressing me to address this point.

²¹ Of course, there will be a ‘special glow’ around the paraphrase that seems perspicuous (to you). But seeming perspicuous and being perspicuous are, sadly, not co-extensive. Recall again that we are focusing on non-revisionary paraphrases.

²² Assuming that (2) is interpreted in a heavyweight way. Cf. section 3.2.

²³ At least if we assume that (1) is true. Of course, as noted in note 9, other apparently true sentences about cracks, such as ‘There is a thin crack in my favourite vase’, might appear to undermine the efficacy of paraphrases like (1*). But, again, this doesn't undermine my general point: we can use differences in perspicuity to ‘break’ other symmetries. If we have reason to think that (1) is more perspicuous than (1*), we have reason to accept an ontology of immaterial objects. And that's important news!

²⁴ Strictly, every minimally perspicuous way of expressing or regimenting the truth, where a minimally perspicuous expression of a claim is a sentence that has formal entailments corresponding to the claim's uncontroversial or known entailments. See section 3.2.

²⁵ If this doesn't seem obvious, see Keller [2015a] for a defence.

²⁶ At least if we assume that I have no reason to doubt that this seeming is veridical.

²⁷ Of course, if I come to believe that () on the basis of its seeming that (assuming that there are such beliefs—see Edgington [1986]), then will be an objective commitment of my ‘total theory’—of and this additional belief together. This doesn't undermine the distinction between the objective and subjective commitments that I have in virtue of accepting itself, however, and there are reasons to think that it won't undermine the distinction even as applied to total theories.

²⁸ Thanks to Patricia Blanchette, Jeffrey C. Goodman, Lorraine Juliano Keller, Daniel Z. Korman, Michael C. Rea, Alexander Skiles, Jeff Speaks, Peter van Inwagen, and two anonymous referees for helpful comments on earlier drafts of this paper.