Abstract
Russellianism is characterized as the view that ‘that’-clauses refer to Russellian propositions, familiar set-theoretic pairings of objects and properties. Two belief-reporting sentences, S and S*, possessing the same Russellian content, but differing in their intuitive truthvalue, are provided. It is argued that no Russellian explanation of the difference in apparent truthvalue is available, with the upshot that the Russellian fails to explain how a speaker who asserts S but rejects S* can be innocent of inconsistency, either in what she says or, at least, in what she implicates. Yet, while there is no semantic or pragmatic explanation of the substitution failure consistent with Russellianism, there remains the possibility of a purely psychological explanation that is, nonetheless, Russellian. This is an attractive option. It comes at a cost, however, since, in abandoning the project of providing a semantic or pragmatic explanation of anti-substitutivity intuitions, the Russellian is no longer in the business of explaining how a rational, well-informed speaker, with no incentive to mislead, can avoid inconsistency in reporting the facts as they appear.
This is a preview of subscription content, log in via an institution to check access.
Similar content being viewed by others
Notes
For discussion see Schiffer (2003, pp. 12–15). Bach (1997), McKinsey (1999) and Moltmann (2003) have argued, however, that substitution of a ‘that’-clause by an appropriate propositional description (or “PD”) is not in general truth preserving. While ‘Lois knows that Superman flies’ is true, (C1) ‘Lois knows the proposition that Superman flies’ is not. But then we must re-examine the claim that ‘that Superman flies’ and ‘the proposition that Superman flies’ co-denote. One must proceed with caution here, however. One suggestion, following King (2002, pp. 345–346), is that when a PD occurs as the complement of ‘knows’ the non-propositional sense of that verb is given preference over the propositional sense; similar remarks apply to ‘fears’, ‘remembers’, etc. This would be due to the fact that a sentence of the form ‘A knows the NP’ invariably (barring rather exceptional circumstances) invokes the non-propositional sense. If so, it should be difficult to hear C1 as invoking the propositional sense. But then we have an explanation of the substitution failure that does not require us to give up the idea that the relevant PD has that Superman flies as its denotation.
Similar considerations apply to a related argument, according to which the aforementioned substitution does not in general preserve grammaticality. While ‘Lois hopes that Superman flies’ is grammatical, (C2) ‘Lois hopes the proposition that Superman flies’ is not. Again, this provides no immediate reason to question Russellianism. The occurrence of a PD as the complement of ‘knows’ automatically selects the sense appropriate (in almost every linguistic context) for such occurrences—namely, the non-propositional sense—and thus delivers an odd but not wholly unintelligible reading. The occurrence of a PD as the complement of ‘hopes’ (and similar verbs, such as ‘wishes’), on the other hand, cannot select an appropriate sense, since there is no such sense. While the result is unintelligible to the hearer, this need not be because the denotation of ‘the proposition that Superman flies’ differs from that of ‘that Superman flies’; it may be due to the fact that ‘hopes’ never takes a (non-proposition-denoting) definite description as a complement. If so, it should be difficult to hear C2 as invoking the propositional sense. (In fact, since ‘hopes’ never takes a non-propositional quantifier as a complement (e.g., ‘Mary hopes everything’ is ungrammatical) it is hardly surprising—pace (Rosefeldt 2008)—that it never takes a propositional quantifier as complement either (e.g., *‘Mary hopes every proposition’)). Once again, we have an explanation of the substitution failure that does not require us to give up the idea that the relevant PD has that Superman flies as its denotation.
I adopt the convention of using set-theoretic notation to represent Russellian propositions. The idea that belief reports (with singular terms in subject position) make implicit reference to modes of presentation is developed in Crimmins and Perry (1989) and Crimmins (1992). The parenthetical restriction is crucial: neither of these approaches takes belief reports with quantified subjects to refer to particular modes of presentation. In such cases, context will constrain the mode of presentation parameter, restricting it to modes of presentation that satisfy a given contextually relevant property.
As indicated, m* is a mode of presentation of a proposition, not of any one propositional constituent.
The example also affects the neo-Fregean analysis. In fact, Schiffer originally directed it against neo-Fregean approaches to attitude reports. This style of counterexample first appears in Kripke (1979, p. 109), where it is directed against classical Fregean approaches.
As Schiffer shows, a parallel argument levels the Fregean analysis of (3a). The Fregean holds the following: (i) A proposition p is a structured entity containing modes of presentation as constituents. In the simplest case, it involves a mode of presentation of an individual paired with a mode of presentation of a property. (ii) ‘that’-clauses refer to mode-of-presentation-containing propositions; (iii) ‘believes’ is a binary relation holding between believers and mode-of-presentation-containing propositions. On its Fregean reading, (3a) asserts (3c):
$$ \left( { 3 {\text{c}}} \right)\quad \forall y \left( {y\,{\text{has met Madonna}} \supset {\text{Believes}}\left( {y, < m,m^\ast>} \right)} \right), $$where m and m* are modes of presentation of Madonna and having musical talent, respectively. Note that (3c) implies (3d):
$$ \left( {3{\text{d}}} \right)\quad \exists p\forall y\left( {y{\text{ has met Madonna}} \supset {\text{Believes}}\left( {y, p} \right)} \right) $$Given that the Fregean takes the object of ‘believes’ to be a Fregean proposition, (3d) and (3d*) are analytically equivalent:
$$ \left( {3{\text{d}}^\ast} \right)\quad \exists p\left( {p\,{\text{is a Fregean proposition}}\,\,\&\,\, \forall y\left( {y{\text{ has met Madonna }} \supset {\text{Believes}}\left( {y, p} \right)} \right)} \right) $$But (3d*) is false in a circumstance of evaluation E at which there is no single mode of presentation m such that all those who have met Madonna think of her under m. Yet, (3a) may be true at E. Thus, the analysis fails. There are other Fregean analyses, ones according to which ‘that’-clauses are not singular terms. But there is no reason to think them any more likely to provide a truth-conditionally adequate treatment of (3a).
Crimmins and Perry (1989) and Crimmins (1992) hold a combination of the implicit-reference approach and the implicit-constraint approach (however on Crimmins’ view, the implicit-constraint approach applies exclusively in cases of belief reports with quantified subjects). Although I will focus on Schiffer’s hidden-indexical theory in what follows, my arguments apply with equal force to the proposals of Crimmins and Perry and of Crimmins. If, as I argue, neither the implicit-reference nor the implicit-constraint approach ultimately succeeds, there is no reason to consider the combined approach, which must fail if both of the aforementioned approaches fail.
On normal notions, see Crimmins (1992, pp. 158–161).
I use the restricted quantifier notation here for perspicuity.
I’m assuming that a necessary condition for the truth of (4a) (and (4b)) holds, namely, that everyone who has read Middlemarch believes, under some mode of presentation or other, that George Eliot (Mary Ann Evans) was intellectually gifted. I’m also assuming that the context at which we evaluate the inference from (4a) to (4b) is (in the terminology of Moore 1999) an ‘enlightened’ one—that is, speaker and audience are mutually aware that ‘George Eliot’ and ‘Mary Ann Evans’ co-refer. Otherwise the substitution will be rejected for reasons that are unrelated to the validity of the inference. (Similarly, the fact that someone confused about the meaning of ‘attorney’ may reject the inference from ‘Smith is an attorney’ to ‘Smith is a lawyer’ tells us nothing about the validity of the inference).
Of course, trivially, all the individual modes under which Middlemarch-readers conceive of the proposition in question do determine a disjunctive condition—that of being identical to m 1* or to m 2* or … (where the m i *’s exhaust the modes in question). I do not take such a disjunctive condition to constitute a style.
These considerations apply with equal force to proposals, such as Récanati (2004b), according to which ‘that’-clauses effectively quantify over propositional modes of presentation (or, as Récanati prefers, belief states). Since it allows context to restrict the quantifier to conversationally relevant modes of presentation, the proposal counts as a version of the hidden-indexical theory, one that is similarly vulnerable to the considerations presented in the text.
As we have seen (§2), it is not an option to hold that a normal utterance of (4a) will restrict the domain to ‘normal’ readers of Middlemarch—namely, those who remember who wrote it. Surely a speaker can use (4a) to quantify over all readers of Middlemarch, even the forgetful ones. In any case, even if we adopt this suggestion, it fails to explain the following substitution failure:
-
(4a′) Everyone who has read Middlemarch, including Jones, believes that George Eliot was intellectually gifted.
-
(4b′) Everyone who has read Middlemarch, including Jones, believes that Mary Ann Evans was intellectually gifted.
Surely (4a′) can be uttered truly even given our assumptions regarding Jones; if so, (4a′′) must provide a trivially satisfiable style. Yet the substitution appears to fail nonetheless.
-
The argument assumes, against the hidden-indexical theory, that the sentences uttered contain no unarticulated elements or parameters.
This information is conveyed by the choice of content sentence. In uttering (2a) the speaker will convey (inter alia) that Lois accepts the content sentence ‘Kent flies’ and, in general, is disposed to behave in ways that can be predicted from this fact (for example, not screaming when Kent leaves her apartment by jumping from a window). More generally, the view is that: In uttering ‘A believes that S’ a speaker implicates that A accepts the proposition expressed by S via a guise similar to S, where a guise m is similar to a sentence S just in case having m in one’s belief box disposes one to assent to S, to behave in ways that can be predicated from one’s assent to S, etc. Although there are problems with this view, (cf. Green 1998; Saul 1998) I will stick to it for the sake of concreteness.
Récanati (2004a) uses the label “syncretism” for any doctrine on which speakers’ intuitions about semantic content are at variance with the actual facts regarding semantic content. This label thus applies both to the naïve theory as well as the view, to be considered below (§5.2), according to which the semantic content of a belief report is something essentially incomplete—something not assertible and thus something not recognizable, to ordinary speakers, as a semantic content.
The diehard naïve theorist might argue that the universal quantifier in the implicated proposition is restricted to people with normal guises of the proposition that Eliot was intellectually gifted. But the best way of putting this would be to claim that the literal meaning of (4a) involves an implicit restriction of the quantifier to individuals with normal guises of the proposition that Eliot was intellectually gifted and that the implicated proposition (4a″) incorporates this restriction. However, as we have seen above, this view about the literal meaning of quantified belief reports such as (4a/b) cannot be sustained.
We are here using the naïve theorist’s strategy, according to which speakers systematically mistake the proposition implicated by a token belief report S for S’s semantic content.
Soames’s view is thus similar to the Contextualism advocated by Récanati (2004a) and others, which holds that what a speaker asserts (or says) in uttering a sentence S is often an enrichment of S’s semantic content. (In fact, Soames goes beyond most Contextualists (as well as the view defended in his 2002) in holding that a speaker, in asserting or saying an enrichment of S’s semantic content, does not necessarily also assert or say S’s semantic content; see Soames 2005, pp. 376–379).
Soames leaves it open whether the literal content of (7a) is a proposition such as (7b) or a propositional template such as (7c) (Soames 2005, p. 365, n. 14). It bears noting, however, that the claim that that the literal content of (7a) is a template doesn’t sit well with the idea that enrichment is a non-semantic notion.
It should be noted, however, that Soames has recently gone on record as endorsing a combination of the pragmatic strategy and (what I’ll call) the enrichment strategy (see Soames 2006a, p. 649; b, p. 727). If, as I intend to show, both strategies fail to explain the data we have been considering, then a combined strategy, whatever its other advantages, seems misguided.
The strategy is obviously counterintuitive. Worse, as Braun (1998) has pointed out, it fails to explain our intuitions in certain straightforward cases (see note 24, below).
I follow Braun in assuming that [NT*] quantifies over sentences in an internal code.
Braun claims that his account “does not require that speakers grasp or believe propositions about ways of believing or the ternary BEL [i.e., assent under a guise] relation” (Braun 1998, p. 577; emphasis in original). Note that [NT*] does not involve Salmon’s concept of a guise. Nor does it suggest, as [NT] does, that speakers can consciously assent to a proposition under a guise.
Braun also has independent worries about the naïve theory’s pragmatic strategy (579n). As he points out, recognition failure is not restricted to belief contexts, but can arise with simple sentences. Indeed, it can arise with propositions of the form ‘A assents to p under S.’ If so, this spells trouble for the pragmatic strategy. To see this, consider an utterance u of (i):
-
(i)
Clark Kent believes that he can fly.
Lois would not accept u, even though she believes the proposition it expresses. According to the naïve theory’s pragmatic strategy, her failure to accept u is explained by the fact that u implicates a proposition that she takes to be false, namely (ii):
-
(ii)
Clark Kent would assent to [Clark Kent can fly] under a guise similar to ‘I can fly.’
The problem, however, is that Lois believes the implicated proposition—which is identical to the proposition expressed by (iii) (although she would not recognize it when put this way):
-
(iii)
Superman would assent to [Clark Kent can fly] under a guise similar to ‘I can fly.’
But then the pragmatic strategy fails to provide an explanation of our intuition that u is false: the proposition it is held to implicate is true.
-
(i)
Salmon notes that a speaker who is trained in distinguishing semantic from pragmatic phenomena may still accept (8) and reject (9).
All that needs to be said about the correlation between S and the relevant mental sentence is that the latter is similar to the former. (On “similarity”, see note 14).
This is true on the naïve theory as well. However, Braun’s view has the advantage of being able to explain how an enlightened speaker (i.e., one aware that ‘George Eliot’ and ‘Mary Ann Evans’ co-refer) can adopt conflicting attitudes to the pair <(8), (9)>, accepting one but not the other (see Braun 1998, pp. 583–590). This is especially notable when we consider pairs, like the following, towards which an enlightened speaker may quite plausibly adopt conflicting attitudes.
-
Superman was more successful with women than Clark Kent.
-
Clark Kent was more successful with women than Clark Kent.
The examples are due to Saul (1997).
-
The naïve theory preserves the speaker’s rationality by appealing either to the implicated content of (4a/b) (this is the pragmatic strategy, discussed in §5.1) or to the assertoric content of (4a/b) (this Soames’s recent proposal, discussed in §5.2).
Another option is that our practices are indeed consistent, but that speakers don’t fully grasp the rules underlying them. See (e.g.) Sider and Braun 2006, who claim that speakers “fundamentally misunderstand the rules that govern language use” (p. 678).
The central argument of this paper dates back to a presentation I gave some years ago at the Pacific Division Meetings of the APA. I’m grateful to my commentator on that occasion, Krista Lawlor, for very helpful remarks and to Ray Buchanan, David Chalmers, Russell Dale, Stephen Neale, Frank Pupa, Gurpreet Rattan, Jennifer Saul, Stephen Schiffer, Peter Smith, Zsofia Zvolensky and an anonymous referee for their comments on subsequent drafts. I owe special thanks to Ray and the anonymous referee for last-minute structural recommendations that greatly improved the flow of the paper.
References
Bach, K. (1997). Do belief reports report beliefs? Pacific Philosophical Quarterly, 78, 215–241. doi:10.1111/1468-0114.00036.
Braun, D. (1998). Understanding belief reports. The Philosophical Review, 107, 555–595. doi:10.2307/2998375.
Crimmins, M. (1992). Talk About Beliefs. Cambridge: The MIT Press.
Crimmins, M., & Perry, J. (1989). The prince and the phone both: Reporting puzzling beliefs. The Journal of Philosophy, 86, 685–711. doi:10.2307/2027014.
Green, M. (1998). Direct reference and implicature. Philosophical Studies, 91, 61–90. doi:10.1023/A:1004212614842.
King, J. (2002). Designating propositions. The Philosophical Review, 111, 317–341.
Kripke, S. (1979). A puzzle about belief. In A. Margalit (Ed.), Meaning and use (pp. 239–284). Dordrecht: Reidel. Reprinted in Propositions and attitudes, N. Salmon & S. Soames (Eds.), New York: Oxford University Press, 1989.
McKinsey, M. (1999). The semantics of belief ascriptions. Nous, 33, 519–557. doi:10.1111/0029-4624.00192.
Moltmann, F. (2003). Propositional attitudes without propositions. Synthese, 135(1), 77–118. doi:10.1023/A:1022945009188.
Moore, J. (1999). Saving substitutivity in simple sentences. Analysis, 59, 91–105. doi:10.1111/1467-8284.00155.
Récanati, F. (2004a). Literal meaning. Cambridge: Cambridge University Press.
Récanati, F. (2004b). ‘That’-clauses as existential quantifiers. Analysis, 64, 229–235. doi:10.1111/j.0003-2638.2004.00490.x.
Rosefeldt, T. (2008). ‘That’-clauses and non-nominal quantification. Philosophical Studies, 137, 301–333. doi:10.1007/s11098-006-0002-8.
Salmon, N. (1986). Frege’s puzzle. Cambridge: MIT.
Saul, J. (1997). Substitutivity and simple sentences. Analysis, 57, 102–108. doi:10.1111/1467-8284.00060.
Saul, J. (1998). The pragmatics of attitude ascription. Philosophical Studies, 92, 363–389. doi:10.1023/A:1004290112630.
Schiffer, S. (1992). Belief ascription. The Journal of Philosophy, 89, 499–521. doi:10.2307/2941005.
Schiffer, S. (2003). The things we mean. Oxford: Oxford University Press.
Sider, T., & Braun, D. (2006). Kripke’s revenge. Philosophical Studies, 128, 669–682. doi:10.1007/s11098-005-2983-0.
Soames, S. (2002). Beyond rigidity. New York: Oxford University Press.
Soames, S. (2005). Naming and asserting. In Z. Szabó (Ed.), Semantics vs. pragmatics. Oxford: Oxford University Press.
Soames, S. (2006a). Précis of Beyond Rigidity. Philosophical Studies, 128, 645–654. doi:10.1007/s11098-005-2981-2.
Soames, S. (2006b). Reply to critics. Philosophical Studies, 128, 711–738. doi:10.1007/s11098-005-2985-y.
Author information
Authors and Affiliations
Corresponding author
Rights and permissions
About this article
Cite this article
Ostertag, G. A problem for Russellian theories of belief. Philos Stud 146, 249–267 (2009). https://doi.org/10.1007/s11098-008-9254-9
Received:
Accepted:
Published:
Issue Date:
DOI: https://doi.org/10.1007/s11098-008-9254-9