Pleonastic propositions and de re belief

Abstract

In The Things We Mean, Stephen Schiffer defends a novel account of the entities to which belief reports relate us and to which their that-clauses refer. For Schiffer, the referred-to entities—propositions—exist in virtue of contingencies of our linguistic practices, deriving from “pleonastic restatements” of ontologically neutral discourse. Schiffer’s account of the individuation of propositions derives from his treatment of that -clause reference. While that -clauses are referential singular terms, their reference is not determined by the speaker’s referential intentions. Rather, their reference is determined in a top-down manner—in Schiffer’s words, “by what the speaker and audience mutually take to be essential to the truth-value of the belief report.” While this accounts for a deep disanalogy between belief reports and other relational propositions—a disanalogy emphasized by Schiffer—I argue that the proposal runs into trouble when we consider the case of de re belief. I close by showing how a modification of Schiffer’s approach—one differing in essential ways from the theory developed in Things—is capable of handling these difficulties.

1 Introduction

In Chapter 2 of The Things We Mean,¹ Stephen Schiffer notes a curiously unremarked disanalogy between the following cases:

1a.	Henri admires Picasso.
1b.	Michelle believes that Barack jogs.

In interpreting (1a), we first identify the references of the terms flanking the main verb, and then check whether or not they instantiate the relation it expresses, thereby determining whether or not what (1a) says is true. But when interpreting (1b) we seem to work in the opposite direction. We don’t first identify the referent of its that-clause and then check to see if the referent of ‘Michelle’ bears believes—the relation expressed by the main verb—to it. Rather, says Schiffer, we first attend to “the [contextually-determined] criteria for truth-evaluating the belief report” (81). It is only with these criteria in hand that we are able to fix the reference of the contained that-clause.²

Of course, the introspective data that Schiffer appeals to might be misleading about how reference is actually determined. It is, after all, entirely possible that (1b) says what it does in virtue of the referential properties of its that-clause, in much the same way that (1a) says what it does in virtue of the referential properties of ‘Picasso’. But an alternative way of stating the contrast is available. What intuitively grounds the reference of (1b)’s that-clause is a sentence-level (as opposed to a constituent-level) property—in particular, the property, determined by the context, of having such-and-such truth-conditions. It would be absurd to hold, however, that the reference of ‘Picasso’, in (1a), is similarly grounded in a corresponding sentence-level property—of (1a)’s being true just when Henri admires Picasso. Clearly, what grounds the relevant sentence-level property is, among other things, the fact that ‘Picasso’ refers to Picasso. This intuitive contrast is solid and does not rely on judgments about relative priority in the order of interpretation.³

While the contrast is striking, it is in tension with something that Schiffer, in Chapter 1, takes as a default assumption regarding the logical form of (1b). According to the “face-value theory” he endorses, (1a) and (1b) are identical in point of logical form: both involve binary relations, differing only in that (1b), unlike (1a), involves a relation between an individual and a proposition. But if both (1a) and (1b) are made true by the fact that, in each case, two things stand in a given relation to one another, how can they differ so markedly in how we assign them truth conditions?

This contrast motivates the treatment of propositions that completes the face-value theory—the theory of pleonastic propositions.⁴ The latter theory individuates propositions in a manner that comports with our intuitive judgments about the truth conditions of belief reports while also promising to make sense of the above-cited contrast. That is, the theory preserves the idea that both (1a) and (1b) are relational while at the same time explaining why it is that we assign them truth conditions in wildly divergent ways.

I will argue, however, that the pleonastic theory fails to provide an adequate account of de re belief reports—the very sort of belief ascribed to Michelle in (1b).⁵ Section 2 provides a brief overview of Schiffer’s approach to belief reports; Sects. 3 to 6 develop a problem for his account. In a final section I argue that a modification of the pleonastic theory can avoid these criticisms.

2 The top-down approach

It is a well-advertised fact that the face-value theory of belief reports provides the most straightforward analysis of the following familiar inference patterns:

A.	Malia believes everything Michelle believes.
	Michelle believes that Barack jogs.
	So, Malia believes that Barack jogs.
B.	Adele believes that arithmetic is incomplete.
	That arithmetic is incomplete = Gödel’s Theorem.
	So, Adele believes Gödel’s Theorem.

(A) involves apparent quantification into that-clause position; (B) involves what appears to be an application of Leibniz’s Law involving that-clauses. Obviously, appearances can be deceiving, but the binary relational analysis clearly earns its face-value status—as being “the default theory that’s to be defeated if it’s not to be accepted” (11).⁶

As we have seen, the reference of ‘that Barack jogs’ in (1b) is grounded in sentence-level properties; this is in contrast to the reference of ‘Picasso’ in (1a), which is not. A related contrast concerns the referential intentions of a speaker who utters (1a) and a speaker who utters (1b). On any approach that conforms to the face-value theory, (1a) and (1b) share their logical forms. And this naturally suggests the following: Just as the speaker who utters (1a) uses a referential singular term to tell us who it is that Henri admires, so the speaker who utters (1b) uses a referential singular term—a that-clause—to tell us what it is—which proposition—that Michelle believes.

Schiffer rejects this familiar idea. He denies that an ordinary speaker, in uttering (1b), intends to refer to the proposition that Barack jogs in anything like the way the speaker who utters (1a) intends to refer to Picasso:

The crucial disanalogy [in the case of (1b)] is that, imagining oneself in [the speaker’s] position, it is difficult to find in oneself referential intentions that would determine a particular one of the [that-Barack-jogs] propositions. This is because the particular referent is not determined by the speaker’s referential intentions. It is determined by what the speaker and audience mutually take to be essential to the truth-value of the belief report. (81)

This account of reference determination dovetails with Schiffer’s account of proposition individuation. On Schiffer’s view, propositions exist in virtue of the existence of certain practices, practices that license “something from nothing transformations”—inferences that take us from ‘Barack jogs’ to ‘that Barack jogs is true.’ The former is ontologically neutral with respect to propositions; the latter is not. The individuation of pleonastic propositions is similarly deflationary. Just as there is no more to the existence of the proposition that Barack jogs than the existence of a practice of going from ‘Barack jogs’ to ‘that Barack jogs is true,’ there is no more to the nature of an individual proposition than is revealed by our practices of reporting beliefs:

[T]he propositions we believe enjoy no more intrinsic conditions of individuation than those provided by their truth conditions and the requirements for believing those propositions that are determined by the criteria for truth-evaluating belief reports in which reference is made to them. (86)

To clarify matters, it will help to consider an utterance of a belief report:

2.	Ada believes that George Eliot wrote Middlemarch.

For Schiffer, “what fixes the referent of the that-clause are the criteria for truth-evaluating the belief report” (81). For example, the that-clause “may refer to a proposition that, intuitively speaking, requires thinking of George Eliot as a famous author, along with various other George Eliot related things not so easily articulated …” (83).

Schiffer thus adopts what I will call a “top-down” approach to that-clause interpretation, deriving certain constituent-level properties from sentence-level properties, and not conversely.⁷ Not only does this reflect the fact that the referent of (2)’s that-clause is not determined by an associated referential intention on the speaker’s part. It also accords with the above-noted asymmetry between ordinary relational predications and belief reports.

As I will show, Schiffer’s treatment of de re belief reports follows naturally from his principle of proposition individuation. This principle reflects neo-Fregean intuitions concerning truth conditions. Indeed, for current purposes, we can conceive of pleonastic propositions as neo-Fregean in the following sense: while they are object-dependent they are also sufficiently fine-grained to explain standard substitution failures—bearing in mind that they differ from neo-Fregean propositions in being unstructured.⁸ However, the attempt to incorporate these intuitions into a formal theory reveals difficulties with Schiffer’s proposal and raises serious questions about its viability.

3 Pleonastic propositions and object-dependency

For the direct reference theorist, the variable x in (3) is no different from the variable that occurs in ‘x flies’—in both cases we have an intelligible logical form:⁹

3.	Lois believes that x flies.

But then the same holds true for the variable in its negation, ‘Lois doesn’t believe that x flies.’ And this causes trouble if we take these open sentences to regiment (4) and (5), respectively:

4.	Lois believes that he flies.
5.	Lois doesn’t believe that he flies.

Imagine (4) uttered pointing to Superman as he is leaping from the top of a building and (5) uttered pointing to Kent, i.e., Superman, as he is entering his cubicle at the Daily Planet. While these utterances appear thoroughly consistent, the first-order regimentation fails to capture this. (5), relative to an assignment of Superman to ‘he,’ contradicts (4) relative to the same assignment.

In view of this, Schiffer denies that (3), understood as an open sentence, can be used to regiment (4); at best, we take the variable to be a mere pseudo-variable. As he writes:

I do not… claim that the occurrence of ‘he’ in (4) refers to Superman in any sense that allows us to say that it follows from the truth of (4) that Superman has the property expressed by the “open sentence”

3.	Lois believes that x flies

[For] my view is that (3)… is not an open sentence, since it expresses no property, and this because the occurrence of ‘he’ in (4) does more than refer to Superman in the sense in which I allow; it also plays its contextual role in determining the proposition to which [its] that-clause refers, even though the referent of the that-clause isn’t compositionally determined by its structure and the referents of its parts (i.e., ‘he’ in (4) refers to nothing other than Superman). (85; emphasis added, numbering changed)

This shows clearly that Schiffer’s denial that (3) expresses a property is intimately related to his endorsement of the top-down strategy for that-clause reference determination. The pronoun in (4) does more than simply introduce a referent: in addition to referring to Superman, it also helps determine the reference of the containing that-clause. (Although this is perhaps easier to see when the that-clause contains a proper name).¹⁰

But there is a question. The direct reference theorist assigns truth conditions to belief reports in a straightforward manner, even if the assignment fails to meet a desideratum that Schiffer takes to be non-negotiable—showing (4) and (5) to be consistent. This is achieved by treating the pronouns in these sentences as variables. The pleonastic theorist, as we have seen, denies that the relevant pronouns are variables. But then how does he propose to assign truth conditions to (4) and (5)? In particular, how does he propose to assign truth conditions in a manner that follows the direct reference theory in capturing their de re aspect—their Superman-dependency—but improves on that approach in preserving their mutual consistency?

Given the above quotation, one option is clearly off the table. Although (4) expresses a Superman-dependent proposition, one that is true at a world w just in case Lois believes that Superman flies at w, it cannot, for Schiffer, be construed as an open sentence with Superman assigned to the free variable. Similarly, mutatis mutandis, for (5). Of course, the most straightforward way of capturing the Superman-dependency of the proposition expressed by the imagined utterance of (4) is with its pronoun construed as a variable under a contextually-given assignment. But, as indicated, this leads to well-known problems. If we construe the imagined utterance of (4) as equivalent to (3) (relative to an assignment of Superman to x), then, although we thereby explain the Superman-dependency of (4), we are unable to account for the consistency of (4) with (5). On the other hand, if we deny that these forms are to be construed as open sentences, relative to an assignment to their free variables, we seem incapable of explaining their Superman-dependency—or, more specifically, the fact that they have the particular possible-worlds truth conditions that they have.

Schiffer addresses this worry by considering the inferences that we are permitted to draw from sentences such as (4). The default strategy, endorsed by the direct-reference theorist is simple: we explain what it is for (4) to express an object-dependent proposition by construing it as an open sentence relative to an assignment. Moreover, this proposal receives confirmation from the fact that (4) entails (6), which, on the view in question, can only be construed as an existential quantification, with the pronoun functioning as a bound variable, as depicted in (7):

6.	There is someone such that Lois believes that he flies.
7.	∃x (Believes(Lois, that x flies))

Again, if we are to reject the idea that (4) can be analyzed into an open sentence and a variable assignment, then we must also reject the idea that (4) entails (6), understood as (7). But then how can we make sense of (4)’s object-dependency?

In response to this, Schiffer adopts a contextual-definition strategy. In particular, he denies that (6) has the logical form it appears to have, namely, that given in (7). On his view, (6)’s surface form, which involves a quantifier binding a position in its that-clause, is misleading. (6) is, as he describes it, a “masquerader” (2016: 435), its logical form captured, not by (7), but by (8):

8.	∃x∃p (p is an x-dependent proposition that is true iff x flies & Lois believes p)

The contextual definition invoked here can be stated as follows:

[CD]	There is someone such that A believes that … they … = def ∃x∃p (p is an x-dependent proposition that is true iff … x … & A believes p)

As Schiffer writes:

[W]e don’t explain the truth of (6) by saying that someone has the property expressed by the ostensibly quantified open sentence in (7), ‘Believes(Lois, that x flies)’, for there is no such open sentence and not only can’t (7) be true, it’s not even well formed, since the occurrence of ‘x’ in it can’t be that of a variable bound by the initial quantifier. Instead, we explain the truth of (6) in terms of Lois’s believing, for some x, an x-dependent proposition that is true iff x flies. (Schiffer 2016: 435; numbering changed, emphasis added)¹¹

Let’s review the strategy: claim that (4) has the expected modal profile (true at a world w just in case Lois believes that Superman flies at w) and is thus Superman-dependent, but deny that this involves Superman’s having the property expressed by ‘Lois believes that x flies’ (indeed, deny that there is any such property).

It is standardly thought that for (4) to have the modal profile it has is for it to entail (6), where (6) is understood as having the form represented by (7). While Schiffer accepts the entailment, he claims that (7) misrepresents (6). Instead, (6)’s true logical form is revealed by (8).

4 The validity of the crucial inference

Schiffer claims that, (9), although “valid in that there is no possible world in which its premise, but not its conclusion, is true, [is] not logically valid” (2016: 435; emphasis in text).

9	a.	Lois believes that Superman flies.12
	b.	∴ There is someone such that Lois believes that he flies.

He continues:

Nevertheless, like the inference ‘Jones is a bachelor; therefore, Jones is unmarried’, it’s valid because it’s necessarily equivalent to [an] inference … which is logically valid:

• 10 a. Believes(Lois, S*)

  b. S* is a Superman-dependent proposition that is true iff Superman flies.

  c. ∴ ∃x, p(p is an x-dependent proposition that is true iff x flies & Lois believes p) (435; emphasis and numbering added)

(9), that is to say, is necessarily equivalent to (10).¹³ Moreover, this necessary equivalence explains (“it’s valid because”) the intuitive validity of (9)—which is made explicit in (11)—and, in so doing, the Superman-dependency of ‘Lois believes that Superman flies.’

11	(a)	Believes(Lois, S*)
	(b)	∃p (p is a Superman-dependent proposition that is true iff Superman flies & S* = p)
	(c)	∴∃x∃p (p is an x-dependent that is proposition that is true iff x flies & Lois believes p)

Note that while the position occupied by the dummy letter S* is an expression whose reference is stipulated to be whatever it is to which the that-clause in (9a) refers, it is not itself a that-clause. On the proposed analysis, (9) is an enthymeme, containing a suppressed premise, (10b), stating, among other things, the Superman-dependence of what Lois believes, thus linking the minor premise, (10a), to the conclusion, (10c). If (10) accurately captures (9), the proper inference to draw from (an utterance of) (9a) is not what we traditionally thought it was—that someone is such that Lois believes that he flies (where ‘he’ is treated as a bound variable)—but rather that someone, x, is such that, for some x-dependent proposition p that’s true iff x flies, Lois believes p.

Thus, that, e.g., (9a) is de re is not captured by the fact that it logically entails ‘there is someone such that Lois believes that he flies.’ It does not logically entail this. Rather, it is captured by the fact that its counterpart in a necessarily equivalent argument, (10a), entails (with the help of a supplementary premise) a conclusion that reports, not a de re belief on the part of Lois but—what comes to the same thing—her believing-true an object-dependent proposition.

A parallel story can be told about (12a):

12a.	Lois doesn’t believe Clark Kent flies.
12b.	There is someone such that it’s not the case that Lois believes that he flies.
12c.	∃x, p (p is an x-dependent proposition that is true iff x flies & ¬(Lois believes p))

Just as (9a) entails (9b), whose form is given by (10c), (12a) entails (12b), whose form is given by (12c). Since (12c) is fully consistent with (10c), contradiction is avoided.¹⁴

Rendering (9) as (10) thus allows Schiffer to meet the relevant desiderata: we capture (9a)’s Superman-dependency while also maintaining its consistency with ‘Lois doesn’t believe that Kent flies.’

While the analysis meets the required conditions, one might reasonably worry whether it is independently justified. The next two sections addresses this and related concerns.

5 Evaluating the strategy

I will assume that it is a requirement on any theory of the semantics of de re belief that the theory provide an account of the sort of inference on display in (9). That is, every such theory must meet (what I will call) the Exportation Requirement:

Exportation Requirement A condition of adequacy on any theory of de re belief reports is that it explain the intuitive validity of the inference from ‘A believes that b is F’ to ‘there is someone such that A believes that they are F,’ or some variant thereof (where ‘b’ is a referential singular term).¹⁵

Whatever its drawbacks, the theory of direct reference does give a clear and intuitive account of this sort of inference. What about Schiffer’s pleonastic theory?

We can concede, at least for argument’s sake, that (10)’s conclusion gives us the required information—that Lois has an object-dependent belief of a certain sort—even if it is delivered in unfamiliar packaging. That is to say, we need not insist that an account of the inference in (9) adhere to the standard logical form assigned to (9b). So there’s no principled objection there. But, as we shall see, to derive it Schiffer has helped himself to a premise, (10b), that raises more questions than it answers.

If anything is clear, it is that the step from minor premise to conclusion does not flow from Schiffer’s account of pleonastic propositions. It has to be supplied by a seemingly ad hoc major premise. Only then can the relevant information, verifying the minor premise’s object-dependency, be derived. The problem with the pleonastic theory is thus not merely that it changes the form that the relevant information takes (regarding the object-dependency of the ascribed belief), it is also that it can only account for how that information is forthcoming from the report by pulling an extra premise out of thin air. For all that has been said, the fact that Lois’s belief involves Superman is an addendum to the analysis.

It might be objected that the extra premise, (11b), follows analytically from (11a) and thus hardly “falls out of thin air.” But this seems false. If the entailment is analytic, then to grasp (11a) is inter alia to grasp that it cannot be true unless (11b) is true. Yet it is unclear how grasping either the linguistic meaning of (11a)—which seems equivalent to that of ‘Lois believes that’—or the proposition (11a) expresses could put one in a position to know that the truth of (11a) guarantees the truth of (11b). (Note that this is independent of whether or not the argument from (11a) to (11b) is modally valid).¹⁶

There is a related worry to the “thin air” objection—perhaps the same worry under a different guise. Schiffer in effect argues that we accept (9) as intuitively valid because it is necessarily equivalent to an argument, (10), which is formally valid. But there is a problem with this methodology. We can, let us assume, grasp that (9) is modally valid; but appeal to such grasping is epistemologically suspect. What is needed is a representation of the inference that displays its validity—an assignment of logical forms to the premises and conclusion on which the conclusion is shown to be a logical consequence of the premises. True, one standard account on which the validity of (9) can be demonstrated—direct reference—has problematic consequences. But that approach at least provides an account of the validity of the inference. What we are interested in here is whether the pleonastic theorist can provide such an account. What seems suspect is to be told that, while (9) is not, and cannot be shown to be, formally valid, there is an argument that is modally equivalent to it, one that is formally valid.

Indeed, the problem with this proposal is now staring us in the face: If an appeal to the modal validity of (9) is suspect—after all, the inference still needs to be explained¹⁷—an appeal to the modal equivalence of (9) and (formally valid) (10) is hardly in better shape. Indeed, that is just an appeal to the modal validity of the claim that (9) and (10) are logically equivalent!

The following argument may show the problem more clearly:

‘(C) is an analytic entailment: it is modally valid, but not logically valid.

C.	Javier is a bachelor.
	So, Javier is unmarried.

But how, then, do we explain its validity? What formal features of the argument allow us to demonstrate that the conclusion follows from the premise? Here is how: (C) is necessarily equivalent to an argument that is logically valid, to wit:

D.	Javier is a bachelor.
	All bachelors are unmarried.
	So, Javier is unmarried.

It is this necessary equivalence to (D) that explains (C)’s intuitive validity.’

There are two problems with this form of argument. First, the second premise of (D) functions as a Carnapian “meaning postulate.” Quine rightly exposed appeals to such postulates as unprincipled—we have been given no way of determining when such an added premise is or is not necessarily true. In addition, the argument is suspect because the appeal to necessary equivalence is on no firmer ground than the original claim that (C) is modally valid.¹⁸

To summarize: Schiffer denies that the following entailment of ‘Lois believes that Superman flies’ should be taken at face value:

6.	There is someone such that Lois believes that he flies.

Rather, the entailment that captures the de re nature of the report is more accurately captured by (11c):

11c.	∃p∃x (p is an x-dependent proposition that is true iff x flies & Lois believes p)

To generate the entailment, Schiffer is forced to supply the tacit premise, (11b):

11b.	∃p (p is a Superman-dependent proposition that is true iff Superman flies & S* = p)

But the addition of (11b) seems a blatantly ad hoc maneuver. Worse, the modal equivalence of enthymemic (9) to (11) has to be taken on faith. But then we might just as well take the modal validity of the original argument on faith.

It is essential to Schiffer’s view that pleonastic propositions are unstructured: the content of (9a) cannot be parsed into the individual, Superman, and the property expressed by the open sentence (3) (‘Lois believes that x flies’) on pain of rendering (9a) and (12a) inconsistent. This necessitates changing how we analyze the characteristic entailments of these propositions—the entailments that certify, so to speak, their object-dependent status. But, in arguing that there is nothing in (9a)’s logical form that, properly understood, permits the derivation of (7), Schiffer seems to have rendered himself incapable of showing how (9a)’s logical form permits (without supplementary premises) the derivation of (11c) as well. The claim that some other argument, distinct from (9), is valid, is hardly to the point—unless we can be assured that (9) is just this other argument in disguise.

6 The inference as analytic

We have overlooked perhaps the most straightforward interpretation of Schiffer’s argument.¹⁹ On this reading, Schiffer simply denies that the inference, (9), admits of explanation. He thus rejects the demand for an account of its validity. Yes, (9) is modally valid, on this reading, but not in virtue of its logical form, nor in virtue of its necessary equivalence to a formally valid argument. While (9) is equivalent to (formally valid) (10), this fact cannot be invoked to explain (on pain of circularity) (9)’s modal validity.

At first pass this seems reasonable and certainly immune to any principled objection. After all, no one denies that there are countless inferences like (C)—valid, but not by dint of logical form or of equivalence to an argument that is formally valid. Valid, that is, but with no explanation of their validity. What prevents us from placing (9) among such arguments?

In fact, there are reasons to be skeptical of the proposal. First, the analogy with (C) and other standard analytic entailments is imperfect. (9)’s conclusion, (9b) (‘There is someone such that Lois believes that he flies’) involves the introduction of a logical constant—indeed, it has the surface form of an existential generalization. But, unless the meaning of the logical constant is not its usual meaning, or is simply idle (as in the case of vacuous quantification), its presence should signal that the argument is valid (if indeed valid) in virtue of its logical form and thus is not properly analytic.²⁰

But even if we set this aside and take the inference to be analytic,²¹ Schiffer still owes us an explanation of how the conclusion can be an existential quantification—which it obviously is—without his account’s incurring the very inconsistencies it was designed to avoid. How, for example, do we construe the open sentence within the quantifier’s scope? If we have no answer to this question, we have no guarantee that (9b)’s form is not after all given by (7) (repeated here for convenience)

7.	∃x (Believes(Lois, that x flies))

And it cannot be given by (7). If (9b)’s form is (7), this means (9b) analytically entails that ‘Lois believes that x flies’ is true under an assignment of Superman to x. But then, by parity of reason, (12a) (‘Lois doesn’t believe Clark Kent flies’) will presumably analytically entail (12b) (‘There is someone such that Lois doesn’t believe that he flies’). And this means that it entails that ‘Lois does not believe that x flies’ will also be true under an assignment of Superman to x. But then we have a contradiction.

As we have seen, Schiffer avoids the contradiction: (9b)’s form is given by (10c). And this suggests a parallel strategy for (12a). While (12a) entails (12b), the latter gets unpacked as (12c). On the current reading, Schiffer’s supplying (9b) with logical form (10c) (and (12b) with logical form (12c)) is not part of an explanation of (9)’s modal validity, as was argued in Sect. 5. Rather, Schiffer provides this logical form to complete the analytic entailment story. Again, without such a story, we have no guarantee that the proposal avoids contradiction.

But providing a story doesn’t dispense with the worry that (9) can be construed as an analytic entailment only if a rather revisionary claim is made about its logical form—one whose positing seems to have no motivation aside from the fact that it avoids the above-described contradiction. (Similarly for (12).) But then the concerns expressed in Sect. 5 still stand. There, I objected that an implicit premise had to be “pulled out of thin air” in order to explain (9)’s validity. A version of the same objection still applies. Here a claim about (9b)’s logical form (unremarked in the previous section) is also seemingly pulled out of thin air—not, here, to explain (9)’s validity, but to show how it avoids contradiction relative to a standard test case.

A final concern is that the proposal jettisons the Exportation Requirement. This means not only that we have to accept, without argument, that the inference is valid relative to the claims that Schiffer makes about the logical form of (9a) and the nature of the entity Lois is said to believe. We also have to accept—again, without argument—that (9a) is object-dependent relative to these same claims. Schiffer, that is, has provided us with a theory of a de re belief reports that does not even attempt to explain what makes them de re. As we have seen, the direct reference theorist, in contrast, does provide such an explanation.

To sum up: the analyticity strategy maintains that (9)’s validity admits of no further explanation. But, setting aside the fact that (9) does not sit comfortably with arguments like (C)—arguments that are canonical cases of analytic entailments—this strategy requires that we accept specific claims about the logical form of (9b). Otherwise the analyticity strategy is no better than the direct reference theory it is supposed to supplant. And these claims about logical form seem arbitrary. In addition, without any indication as to how the conclusion follows from the stated premise, the proposal fails to meet an intuitive condition of adequacy—the Exportation Requirement.

7 A structured alternative to pleonastic propositions?

In Sects. 2 to 6 I argued that Schiffer’s proposal cannot meet the following conditions: accommodate the object-dependency of (9a) (‘Lois believes that Superman flies’) while also maintaining its consistency (relative to the described scenario) with (12a) (‘Lois doesn’t believe that Kent flies’). In this section, I will propose an alternative to Schiffer’s account: one that meets these dual desiderata and, in addition, preserves the top-down strategy for that-clause determination. (This latter aspect of the theory also enables it to explain the contrast with which we began—namely, between binary relational assertions generally and the special case of belief reports.) The theory is not pleonastic, however. On Schiffer’s view, propositions are individuated in terms of their truth conditions “and the requirements for believing those propositions that are determined by the criteria for truth-evaluating belief reports in which reference is made to them.” On the current view, propositions are not individuated in terms of the combined criteria Schiffer cites—truth conditions and (context-sensitive) reference conditions—but rather in terms of their structure and constituency: if p and q are composed in the same way (this, of course, needs fleshing out) from identical constituents, then p and q are the same proposition; if not, they are distinct.

The pleonastic theory requires that we adopt a top-down procedure for determining that-clause reference (recall that “what fixes the referent of the that-clause are the criteria for truth-evaluating the belief report”). However, it is worth noting that there is no necessary connection between the top-down procedure for reference determination—which explains the contrast between (1a) and (1b)—and the thesis that propositions are unstructured.²² For example, we might take the object of belief to be an ordered pair of a Russellian proposition 〈x, F〉 and a guise presenting it. Thus, to say that Lois believes 〈Superman, flying〉 under m (as the Russellian is wont to say) is to say that she straight out believes 〈〈Superman, flying〉, m〉. 〈Superman, flying〉 is supplied by the constituents of the that-clause (relative to their syntactic roles in the clause) while m is supplied either by a hidden-indexical or through a process of free enrichment. The details don’t matter here.²³

Now for some stipulations. 〈〈Superman, flying〉, m〉 is true at a world w just in case Superman exists at w and flies at w. The constituent m is thus, following Recanati (1993), “truth conditionally irrelevant”: 〈〈Superman, flying〉, m〉 is true exactly where 〈Superman, flying〉 is true. Then how does the Russellian proposition differ from its enriched counterpart? As follows: Lois might be in a belief state whose content is captured by 〈Superman, flying〉 but not by m and thus not by 〈〈Superman, flying〉, m〉. Why not? Because the content 〈Superman, flying〉 although grasped by Lois, is either not grasped in way m or is grasped by Lois in way m, but not believed by Lois when thus grasped.

The elements here are all familiar. I’ve just inserted the guise or mode of presentation of 〈Superman, flying〉 into the object of belief.²⁴ This might be flagged as an odd move: Russellian propositions are already contents. Wouldn’t pairing them with guises just create a monstrosity? I’m sympathetic to this idea. But note that the advocate of pleonastic propositions is in no position to object. On his view, propositions are entities that play a certain role: they function as the contents of our utterances and as the referents of that-clauses. (As such, they are abstract, mind- and language-independent entities possessing truth conditions, and possessing them essentially and absolutely.) And, in playing these two roles, they have the modal profiles we expect them to have. In a word (or two), a proposition is whatever plays the proposition role. Now, clearly, certain entities are immediately disqualified—Julius Caesar, to mention one. But nothing about our pre-theoretical intuitions regarding what precisely that-clauses refer to disqualifies an ordered pair of a Russellian proposition and a guise. Whatever these expressions refer to, they are not things about which ordinary speakers will have intuitions.

How do these hybrid entities satisfy the relevant desiderata? First, note that (9a) relates Lois to a proposition that, in virtue of containing 〈Superman, flying〉 has the desired possible worlds truth conditions. How does (9a) entail (9b)? Well, first, although the position occupied by ‘Superman’ in (9a) is not open to substitution salva veritate with respect to a co-referential term, this does not mean that the position is not bindable by an antecedent quantifier. A term occupying this position does two things: it refers to an individual and it partially indicates the way that the individual (and thus the referred-to proposition) is to be grasped by the subject to whom belief in the proposition is being ascribed. Replacing the term with a co-referential alternative will in all likelihood induce a change in how the referred-to individual is to be grasped. If so, then, even though the position occupied by ‘Superman’ in (9a) is bindable by an external quantifier, it does not follow that we can freely substitute a co-referential singular term for that name and wind up with a sentence expressing a truth-conditionally equivalent proposition.²⁵ This is because the choice of a term may suggest thinking Superman of as a superhero, “along with various other [Superman] related things not so easily articulated.”

Quine argued that failure of substitution of co-referential singular terms with respect to a position (such as that occupied by ‘Superman’ in (9a)) showed the position to be inaccessible to quantification: quantifying in would be “improper.” But this is a fallacy.²⁶ As we have seen, substitution salva veritate is not a reliable diagnostic for determining whether a position is bindable.

This means that the inference from (9a) to (9b) is valid, on the current proposal. But doesn’t this get us into hot water with respect to (9a) and (12a)? After all, (9a) is true just in case ‘Lois believes that x flies’ is true with respect to an assignment of Superman to x, and (12a) is true just in case ‘Lois doesn’t believe that x flies’ is true with respect to the same assignment. But then we get:

13.	There is someone—namely Superman/Clark Kent—such that Lois believes and does not believe that he flies.

The first thing to note is that this consequence would seem to be a reductio of Schiffer’s own approach. But Schiffer denies this. It is, he claims, true:

But wouldn’t common sense regard (13) as a contradiction? I don’t think so. Uttered out of the blue, it would no doubt provoke a ‘Come again?’ response. But first, when we get to this level of complexity it is not clear what philosophers should make of the reactions of ordinary speakers, and, secondly, I don’t see why the non-philosopher in whom the response was provoked shouldn’t be assuaged by the explanation that (13) is true because, after all, Lois believes that Superman flies but doesn’t believe that Clark Kent flies. (86; numbering changed)

While (13) might appear contradictory, when properly understood, the air of absurdity dissipates.

If this response is satisfactory as a defense of the pleonastic theory, it should equally work as a defense of the structured proposal I am advocating. Once the incredulous non-philosopher is provided with the aforementioned explanation, he or she should relent since, “after all, Lois believes that Superman flies but doesn’t believe that Clark Kent flies.”

There are two worries here, which I will just note. First, it is not clear that the non-philosopher will be satisfied with such a story. It remains a fact that, after the dust has settled, (13) is a contradiction. Second, a variant of this strategy now becomes available to the direct reference theorist. Consider the charge that (9a) and (12a) are inconsistent. If Schiffer’s just-quoted response is acceptable, the following, made on behalf of the direct reference theorist, should be equally acceptable (although here we are explaining why something apparently consistent is in reality contradictory, whereas above we were explaining why something apparently contradictory is in reality consistent—indeed, true): “I don’t see why the non-philosopher in whom the response was provoked shouldn’t be assuaged by the explanation that the utterances are contradictory because, after all, Lois believes that he, Superman, flies but doesn’t believe that he, Superman, flies.”

In sum, if we accept Schiffer’s rejoinder, it appears that we are in no position to object to the direct-reference theorist’s explanation of our anti-substitutivity intuitions.²⁷

My concern, however, has been to show that the current alternative to the pleonastic theory—on which propositions are structured—is preferable to Schiffer’s version. I will have to leave consideration of the challenge posed by (13) for another occasion.

8 Conclusion

An advantage of the pleonastic theory over its competitors is that it enables us to make sense of the contrast between how the interpretations of (1a) and (1b) are determined.

1a.	Henri admires him.
1b.	Michelle believes that Barack jogs.

The explanation involves the claim that, while singular term reference is generally determined bottom-up, so to speak, that-clause reference is determined in a top-down manner. I’ve argued that the particular top-down approach Schiffer endorses—the pleonastic theory of propositions—leads to apparently insurmountable complications when we try to make sense of the object-dependency of certain attitude ascriptions. While Schiffer claims that the object dependency of (1b) consists in its entailing that Michelle believes an object-dependent proposition—one true at a world just in case Barack jogs at that world—he does not give a clear indication of how that entailment is generated. As I’ve suggested, the fact that Michelle’s belief involves Barack appears an addendum to the analysis. And this means that the object-involvingness of such reports must be taken on faith. One might argue that there is no principled objection to this idea—perhaps both the object-involvingness of the report and the fact that it entails ‘There is someone such that Michelle believes that he jogs’ must be taken as basic. While it is true that certain valid inferences must be acknowledged as basic, recognizing their validity doesn’t require us to revise their logical form. But taking, e.g., (9) as basic requires that we accept a revisionary claim about the logical form of the entailment—something that has no independent motivation.

If, however, we change course and simply drop a defining feature of Schiffer’s approach—that propositions are individuated in terms of their truth conditions and the requirements to make reference to them—then there is no barrier to making sense of the fact that the proposition expressed by (9a) is both object-dependent and consistent with the proposition expressed by (12a). Yes, reference determination goes exactly the way that Schiffer says it does, but this does not entail that the reference determination story doubles as a theory of proposition individuation. Of course, we do have to grapple with the fact that the revised theory entails (13), given the facts about (9a) and (12a). But this is a consequence the original theory faces as well.²⁸