Salmón, Schiffer and Frege’s Constraint

Abstract

In his (Philosophical Perspectives 1:455–480, 1987) and (Noûs, 40:361–368, 2006), Schiffer devised a puzzle about Salmón’s (in: Frege’s puzzle, MIT Press, 1986a) Millian-Russellian theory of belief reports, which Salmón resolved in his (Philosophical Perspectives 3:243–285, 1989) and (Noûs, 40:369–375, 2006). My paper has three objectives. First, I will argue that the strategy employed by Salmón (in: Noûs 40:369–375, 2006) to solve Schiffer’s puzzle and his argument for such a strategy are disputable. Second, I will raise a new puzzle, inspired by ideas from Saul (in: Analysis 57:102–108, 1997) and Braun and Saul (in: Philos Stud 111:1–41, 2002), which achieves similar results to Schiffer’s puzzle but to which Salmón’s overall strategy for resolving the latter does not apply. Third, I will contend that the import of both puzzles is neither what Salmón maintains nor the alleged inadequacy of the Millian-Russellian semantics of belief reports as Schiffer suggests, but is the failure of Frege’s Constraint—a constraint to which several conceptions of modes of presentation, including Salmón’s (in: Frege’s puzzle, MIT Press, 1986a) in terms of guises and Schiffer’s (in: The things we mean, Oxford University Press, 2003) in terms of unstructured and fine-grained concepts/propositions, are committed.

1 Preamble: Millian Russellianism and Salmón’s theory of belief reports

Millian Russellianism is a contemporary theory of meaning, which traces its roots to ideas of John Stuart Mill (1843), Bertrand Russell (1903) and David Kaplan (1989a). According to this theory,

• proper names and singular demonstrative/personal pronouns in contexts of use are Millian terms, i.e. terms whose semantic content (i.e. what they express) is solely their referent (in those contexts);

• the semantic content of a predicate is an attribute (i.e. a property or a relation);

• the semantic content of a declarative sentence in a context of use or a statement (i.e. an utterance or inscription of a declarative sentence) is a Russellian proposition, i.e. a structured proposition whose basic constituents are individuals and attributes.

In his seminal book Frege’s Puzzle (1986a), Nathan Salmón advocated a Millian-Russellian theory that assigns to belief reports the semantic content and truth conditions specified in the claims (A) and (B) respectively.

• (A) The semantic content of an instance of the (sentential) schema ^⌜α believes that Φ_β^⌝ (where the substituends for α and β are Millian singular terms and the substituend for Φ is a predicate) in a context of use is a Russellian proposition of the form <a, believing, <b, F>> , where believing is a two-place relation between the subject or agent a (i.e. the referent of the substituend for α) and the Russellian proposition formed by the object b (i.e. the referent of the substituend for β) and the property F (expressed by the substituend for Φ).

• (B) ^⌜α believes that Φ_β^⌝ is true if and only if <a, believing, <b, F>> is true; in turn, the latter (and consequently the former) is true if and only if there is at least one guise g such that a is disposed to inwardly or mentally assent (BELs) to the Russellian proposition <b, F> grasped by means of g.

On Salmón’s view, a guise is a mode of presentation that has no impact to the semantic content of sentences, including belief sentences. A mode of presentation is whatever satisfies Frege’s Constraint below, with respect to which Salmón has affirmed: “I am prepared to grant … that something along the lines of … [this c]onstraint is indeed correct” (1989, p. 267).¹

Frege’s Constraint: A subject a rationally (and simultaneously) believes and disbelieves (i.e. believes the negation of) b to be such that it has (the property) F only if in (simultaneously) believing and disbelieving so (i.e. b to be such that it has F), (i) a thinks of the object b under different modes of presentation which a takes to present distinct objects or (ii) a thinks of the property of being such that one has F under different modes of presentation which a takes to present distinct properties.²

It is important to highlight that Frege’s Constraint involves belief reports of the form (D) which contains the phrase “believes … to”, instead of the phrase “believes that” appearing in (C).

• (C) α believes that Φ_β.

• (D) α believes β to be (something/someone) such that Φ_it.

Salmón (2006, p. 371) draws a semantic distinction between the de dicto schema (C) and the de re schema (D): as established in claim (A) above, instances of the schema (C) express Russellian propositions of the form <a, believing, <b, F>> ; by contrast, instances of the schema (D), where the pronoun “it” does not occur free, express Russellian propositions of the form <a, believing*, b, being (one) such that one has F> or identically <a, believing*, b, λx[Fx]> ,³ where believing* is a three-place relation among the subject a, the object b, and the property of being (one) such that one has F or identically the propositional function λx[Fx].⁴ For example, the de dicto report (1) expresses the Russellian proposition (1p), whereas the de re report (2) expresses the Russellian proposition (2p) or identically the Russellian proposition (2p*).

• (1) Lucy believes that Aristotle is a philosopher.

• (1p) <Lucy, believing, <Aristotle, being a philosopher>> 

• (2) Lucy believes Aristotle to be such that he is a philosopher.

• (2p) <Lucy, believing*, Aristotle, being such that one is a philosopher> 

• (2p*) <Lucy, believing*, Aristotle, λx[x is a philosopher]> 

Based on the semantic contents assigned by Salmón to (C) and (D), it turns out that in the de dicto schema (C) the Millian singular term instantiating β is substitutable salva veritate with any coreferring Millian singular term but not always with codesignative definite descriptions,⁵ whereas in the de re schema (D) the Millian singular term instantiating β is also substitutable salva veritate (though not salva significatione) with any codesignative definite description. So, to reconsider the example above, even in case Lucy has the disposition to utter sincerely, on reflection and competently “I believe that Aristotle is a philosopher, but I don’t believe that the teacher of Alexander the Great is a philosopher” failing to realize that Aristotle is the teacher of Alexander the Great, “Aristotle” is substitutable salva veritate with “the teacher of Alexander the Great” in (2), not in (1) though.⁶

Despite the highlighted differences between (C) and (D), it seems appropriate, from a Millian-Russellian viewpoint, to uphold thesis (T1) below, called by Schiffer (2006, p. 362) the special-case consequence.

• (T1) Every instance of de dicto schema (C), i.e. ^⌜α believes that Φ_β^⌝, entails the corresponding instance of the de re schema (D), i.e. ^⌜α believes β to be such that Φ_it^⌝, under the twofold assumption that β is instantiated by a Millian singular term and “it” in (D) does not occur free.⁷

2 Subject matter and goals

In his article “The ‘Fido’-Fido Theory of Belief” (1987), Schiffer devised a puzzle about Salmón’s (1986a) Millian-Russellian theory of belief reports; a more sophisticated version of the same puzzle, involving reports of the form (D), was later presented by Schiffer in “A Problem for a Direct-Reference Theory of Belief Reports” (2006), while solutions to both versions of the puzzle can be found in Salmón’s articles “Illogical Belief” (1989) and “The Resilience of Illogical Belief” (2006). More recently, Salmón has slightly modified his (2006) article and republished it under the title “Constraint with Restraint” (2016), while Schiffer has replied to them in “Why Believing isn’t a Relation to Russellian Propositions” (2008) and in “De Re Subtleties” (2016). This thirty-year debate between Salmón and Schiffer is of great interest and, in my opinion, would deserve more attention from philosophers of language, logicians and linguists working on the topics of belief reports, de dicto/de re distinction, propositions and modes of presentation.

My paper focuses on the most compelling version of Schiffer’s (2006, 2008) puzzle, involving reports of the form (D), which will henceforth be exemplified by the “George Eliot”/“Mary Ann Evans” case (Sect. 3). A first goal of my paper is to show that Salmón’s (2006) strategy to solve this puzzling case, consisting in the rejection of thesis (T1), and his (2006) argument for such a strategy, viz. a specific counterexample to (T1), are disputable (Sect. 4). Furthermore, based on well-known discoveries about simple sentences by Jennifer Saul (1997) and related considerations by David Braun and Saul (2002), I will raise a new puzzle, the “Superman”/“Clark Kent” case, which achieves similar results to Schiffer’s puzzle but to which Salmón’s overall strategy for solving the latter—consisting in the replacement of the rejected thesis (T1) with an amended version of it—does not apply (Sect. 5). I will finally argue that the import of both puzzles is neither what Salmón maintains nor the alleged inadequacy of the Millian-Russellian semantics of belief reports as Schiffer suggests, but is the failure of Frege’s Constraint—a constraint to which several conceptions of modes of presentation, including Salmón’s (1986a) in terms of guises and Schiffer’s (2003) in terms of unstructured and fine-grained concepts/propositions, are committed (Sect. 6).

3 Schiffer’s “George Eliot”/“Mary Ann Evans” puzzling case

3.1 The puzzle

Consider two rational English speakers, Jane and Ralph. Jane is aware that “George Eliot” is the pen name of Mary Ann Evans, whereas Ralph isn’t. Having heard Ralph affirm “I believe that George Eliot was a man, but I don’t believe that Mary Ann Evans was a man”, Jane acquires the disposition to utter sincerely, on reflection and competently (5) and (6) below. Hence, it is correct to report (7) and (8)—where the latter sentence is a convenient rephrasing of “Jane believes that Ralph does not believe that Mary Ann Evans was a man”.

• (5) Ralph believes that George Eliot was a man.

• (6) Ralph does not believe that Mary Ann Evans was a man.

• (7) Jane believes that Ralph believes that George Eliot was a man.

• (8) Jane disbelieves that Ralph believes that Mary Ann Evans was a man.

Within Salmón’s theory, the truth of (7) and (8) is justified by claim (B) (beginning of Sect. 1): given Jane’s aforementioned linguistic dispositions towards (5) and (6), there are two guises corresponding to these sentences under which Jane BELs respectively the Russellian proposition—(5p) below—expressed by (5) and the Russellian proposition expressed by (6). Alternatively, the truth of (7) and (8) can be justified by simply appealing to the Disquotational Principle below.

• (5p) <Ralph, believing, <Eliot, having been a man>>

Disquotational Principle: If a normal speaker a of a language L, under normal circumstances, has the disposition to sincerely, on reflection and competently utter or assent to a sentence S of L in a context where S expresses a proposition p that a understands, then a believes that p.⁸

Now, in accordance with thesis (T1) (end of Sect. 1) (7) and (8) respectively entail (9) and (10) below—the last sentence being a convenient rephrasing of “Jane believes Mary Ann Evans not to be such that Ralph believes that she was a man”.

• (9) Jane believes George Eliot to be such that Ralph believes that she (i.e. Eliot) was a man.

• (10) Jane disbelieves Mary Ann Evans to be such that Ralph believes that she was a man.

In turn, the de re report (10), where the position occupied by the name “Mary Ann Evans” is open to substitution salva veritate with any codesignative singular term, entails (11) below, according to Salmón’s semantics (second part of Sect. 1) and more generally according to any semantics involving de re belief reports.⁹

• (11) Jane disbelieves George Eliot to be such that Ralph believes that she was a man.

Finally, the truth of (9) and (11) leads to the following falsification of Frege’s Constraint (first part of Sect. 1): Jane rationally believes and disbelieves Eliot to be such that Ralph believes that she (i.e. Eliot) was a man (~ii) thinking of the property of being such that Ralph believes that one was a man under only one mode of presentation and (~i) thinking of Eliot without having different modes of presentation of her which Jane takes to present distinct individuals.

3.2 Three strategies to solve the puzzle: Schiffer’s, Salmón’s and an alternative strategy

Based on the exposition proposed in Sect. 3.1, which draws from Schiffer (2006, pp. 363–364, 2008, §1–§2), the “George Eliot”/“Mary Ann Evans” case appears to be generated by the following four elements (all of which have been introduced in Sect. 1 and explicitly mentioned in Sect. 3.1):

• (a) Salmón’s claim (B) about the truth conditions of de dicto belief reports;

• (b) thesis (T1);

• (c) Salmón’s semantics of de re belief reports;

• (d) Frege’s Constraint.

Since, on Schiffer’s view, Salmón’s theory commits to all elements (a)–(d), Schiffer concludes that the “George Eliot”/“Mary Ann Evans” case poses a problem for Salmón’s theory.¹⁰ Schiffer’s (1987, 2006) solution to this puzzle consists in rejecting (a) and, ultimately, the Millian-Russellian semantics of belief reports, which he (2003) replaces with a remarkably different semantics involving unstructured and context-dependent (as regards their granularity) propositions (Sect. 6.4).

As we will see in Sect. 4.1, on the other hand, Salmón denies that his theory is committed to the conjunction of (a)–(d). Specifically, he (~b) rejects thesis (T1), in this way resolving the “George Eliot”/“Mary Ann Evans” case.

While acknowledging the extraordinary interest and value of Salmón’s and Schiffer's solutions, I dissent from both. After critically scrutinizing Salmón’s solution in Sect. 4 and objecting to Schiffer’s in Sect. 6, I will conclude in Sects. 6 and 7 that the “George Eliot”/“Mary Ann Evans” case should be solved by (~d) rejecting Frege’s Constraint. Hence, the Millian-Russellian semantics of belief reports can be retained pace Schiffer, whereas Frege’s Constraint can be revised along the lines indicated in Frege’s Constraint* below, which several theorists of mental files/nodes including Crimmins and Perry (1989), Forbes (1990), Saul (2007), and Récanati (2012, 2016) seem inclined to endorse.¹¹ Alternatively—that’s my (2020, 2021) proposal—Frege’s Constraint could be replaced by the Relationist Constraint below, a cognate of Frege’s Constraint* that involves the relation of cognitive coordination (characterized in fn. 12 due to reasons of space) instead of modes of presentation.¹²

Frege’s Constraint*: A subject a rationally believes and disbelieves b to be such that it has F only if in believing and disbelieving so, (i*) a thinks of the object b under different modes of presentation or (ii*) a thinks of the property of being such that one has F under different modes of presentation.¹³

Relationist Constraint: A subject a rationally believes b to be such that it has F and believes b not to be such that it has F only if in believing so, (i_RC) a negatively coordinates (in the sense specified in fn. 12) the two occurrences of the object b or (ii_RC) a negatively coordinates the two occurrences of the property F.¹⁴

4 Salmón and the “George Eliot”/“Mary Ann Evans” case

4.1 Salmón’s solution to the puzzle

Pace Schiffer (1987, pp. 464–465), Salmón thinks that the truth of the de dicto reports (7) and (8) (Sect. 3.1) is compatible with the assumption that Jane is rational and with her knowledge that the names “George Eliot” and “Mary Ann Evans” corefer. He (1989, pp. 267–268, 1995, p. 219, 2006, p. 373) offers the following explanation for such a compatibility: Jane mistakes the Russellian proposition <Eliot, having been a man> for two Fregean propositions or the like as if she were a “proto- or closet neo-Fregean” (2006, p. 373), with the result of thinking of the more complex Russellian proposition (5p) (Sect. 3.1), believed and disbelieved by Jane and containing <Eliot, having been a man> as a constituent, under different modes of presentation, viz. guises corresponding to the sentences (5) (Sect. 3.1) and “Ralph believes that Mary Ann Evans was a man”, which she mistakenly takes to present distinct things, in accordance with Frege’s Constraint.¹⁵

So far so good. But then Salmón also reckons that the truth of the de re reports (9) and (10) (Sect. 3.1) is incompatible with Jane’s rationality, given her knowledge that “George Eliot” and “Mary Ann Evans” corefer.¹⁶ In order to avoid the allegedly problematic conclusion that both (9) and (10) are true, he (1995, p. 218, 2006, pp. 370–371) denies the implications and, consequently, the entailments from the de dicto reports (7) and (8) to the de re reports (9) and (10) respectively, thereby rejecting thesis (T1), of which the entailments in question are instances. In this way, Salmón resolves Schiffer’s puzzle.

4.2 Some objections to Salmón’s solution

Although I personally endorse Millian Russellianism and the Millian-Russellian semantics of belief reports,¹⁷ I am not convinced by Salmón’s proposal for resolving the “George Eliot”/“Mary Ann Evans” case by denying the implication and therefore the entailment from (7)/(8) to (9)/(10), with the result of rejecting thesis (T1). In fact, unlike Salmón (2006), I do not think that the truth of (9) and (10)/(11)  (Sect. 3.1) along with Jane’s knowledge that “George Eliot” and “Mary Ann Evans” corefer imply that she is irrational. Jane would be irrational if (12) below were true; however—and this is where, in my opinion, Salmón (2006) and I disagree—the conjunction of (9) and (10)/(11) does not imply (12) in the specific case under discussion, namely in the “George Eliot”/“Mary Ann Evans” case.

• (12) Jane believes George Eliot both to be and not to be such that Ralph believes that she was a man.

By utilizing the Russellian propositions (9p), (10p) and (12p) below, which are respectively expressed by the sentences (9), (10)/(11) and (12) above, the strategy of solution I am proposing can be restated more perspicuously as follows: in the “George Eliot”/“Mary Ann Evans” case, (9p) and (10p) are true and so is their conjunction, hereafter labeled “(9p)&(10p)”; however, (9p)&(10p) does not imply (12p) in such a case.¹⁸

• (9p) <Jane, believing*, Eliot, λx[Ralph believes that x was a man]>

• (10p) <Jane, believing*, Eliot, λx[Ralph does not believe that x was a man]> 

• (12p) <Jane, believing*, Eliot, λx[Ralph believes that x was a man & Ralph does not believe that x was a man]>

My own Millian-Russellian explanation for the falsity of the implication from (9p)&(10p) to (12p) in the case under discussion invokes the relation of cognitive coordination (and is spelled out in fn. 19, due to space limitations).¹⁹ However, resorting to this relation is not indispensable in order to adopt the strategy I have suggested: a Millian-Russellian advocate of modes of presentation could deny the implication from (9p)&(10p) to (12p) by maintaining that two modes of presentation are associated to the two occurrences of Eliot in (9p)&(10p) whereas neither of them is associated to Eliot’s sole occurrence in (12p). It is important to add that due to her knowledge that “George Eliot” and “Mary Ann Evans” corefer, Jane realizes that these modes of presentation present the same individual, Eliot, in accordance with Frege’s Constraint* (end of Sect. 3.2) but in contrast with Frege’s Constraint.²⁰

So, my strategy for resolving the “George Eliot”/“Mary Ann Evans” case ultimately violates Frege’s Constraint. For this reason, Salmón, who advocates it, does not endorse my strategy.²¹ Instead, coherently with Frege’s Constraint, he maintains that Jane (unlike Ralph) has only one mode of presentation (viz. individual guise) of Eliot, with the inevitable outcome that the conjunction (9p)&(10p) implies (12p);²² and since the consequent of this implication, (12p), is false (otherwise Jane would be irrational, against our initial supposition), he concludes that its antecedent, (9p)&(10p), must be false as well. What is unclear to me, nevertheless, is why Salmón does not consider the alternative option of abandoning Frege’s Constraint for Frege’s Constraint*, thus safeguarding thesis (T1) and resolving the “George Eliot”/“Mary Ann Evans” case by denying, as I propose, the implication from (9p)&(10p) to (12p).²³

Salmón’s solution also presents another reason of concern. As pointed out above, according to him, Jane has only one guise of Eliot. On the other hand, she also has only one guise of the property of having been a man (end of Sect. 3.1). But, then, how is it possible that she has two guises of the Russellian proposition <Eliot, having been a man> and, consequently, of the Russellian proposition (5p), which she mistakenly takes to present distinct things (Sect. 4.1)? Salmón answers this question by rejecting (e).

• (e) Propositional guises are (standardly) structured entities,

  [namely, they are] entities whose basic components are [guises] of the basic components of the Russellian propositions of which the propositional [guises] are [guises.] (Schiffer, 2006, p. 365)

Indeed, once (e) is rejected, Salmón can maintain that in the “George Eliot”/“Mary Ann Evans” case, Jane’s two propositional guises corresponding to “George Eliot was a man” and “Mary Ann Evans was a man” are entities not constituted or not solely constituted by her guise of Eliot corresponding to both “George Eliot” and “Mary Ann Evans” and her guise of being a man corresponding to “was a man”. Some perplexities arise at this point, though: what are the additional constituents, assuming that there are any, of Jane’s propositional guises?²⁴ And why should a non-philosopher and ordinary thinker like Jane believe/BEL Russellian propositions under sui generis guises, lacking the structure of the believed/BELed propositions of which they are guises?

To recap, thanks to a couple of undoubtedly clever maneuvers, viz. (~b) rejecting thesis (T1) and (~e) resorting to sui generis (viz. unstructured or non-standardly structured, fn. 24) guises, Salmón is able to solve the “George Eliot”/“Mary Ann Evans” case. However, manoeuver (~e) appears somewhat artificial. Concerning maneuver (~b), it seems like he arrives at defending it with a reasoning by exclusion, viz. after excluding that the rejection of Frege’s Constraint and its replacement with Frege’s Constraint* is a viable option to solve the “George Eliot”/“Mary Ann Evans” case; nevertheless, no justification for such exclusion is provided.²⁵

4.3 Salmón’s argument for the strategy underlying his solution: the “Ortcutt” case

The argument that Salmón puts forward in support of the falsity of (T1) (the thesis he rejects in order to solve Schiffer’s “George Eliot”/“Mary Ann Evans” case) is an alleged counterexample presented in his article “The Resilience of Illogical Belief” (2006, p. 371). Suppose that a rational subject, Ralph, sincerely, on reflection and competently utters “He is taller than he is”, pointing first to a picture portraying a man in a brown hat and then to another picture portraying a man at the beach. Failing to realize that the two pictures portray the same man, Bernard J. Ortcutt, Ralph comes to believe the Russellian proposition <Ortcutt, being taller than, Ortcutt> . Based on claim (B) (beginning of Sect. 1) and utilizing the guise corresponding to Ralph’s aforementioned utterance, report (13) below turns out to be true.

• (13) Ralph believes that Ortcutt is taller than Ortcutt.

Of course, (13) does not imply (14) below: otherwise, Ralph would also believe the Russellian proposition <Ortcutt, being taller than oneself> , to the effect of being irrational or extremely confuse. All the more reason and more interestingly for our discussion, (13) (which is an instance of the de dicto schema (C)) does not imply and, therefore, does not entail (15) below (which Salmón (2006, p. 371) identifies as an instance of the de re schema (D)). Since “the resulting instance of [(T1), i.e. the thesis that allows the inference from (C) to (D), is] false” (2006, p. 371), Salmón concludes that such a thesis is violated.

• (14) Ralph believes that Ortcutt is taller than himself.

• (15) Ralph believes Ortcutt to be such that he is taller than himself.

4.4 An objection to Salmón’s argument

Salmón is certainly right in maintaining that (13) does not imply and, therefore, does not entail (15). However, it is debatable whether the falsity of this entailment provides a counterexample to (T1): whereas (13) unquestionably has the de dicto form (C), it is far from obvious that (15) has the corresponding de re form (D) required to falsify (T1). An instance of (D) is (16) below, where the two occurrences of the pronoun “he” are anaphorically linked to the name “Ortcutt” (fn. 3); yet, as we will see shortly, it can be argued—based on Kaplan (1986: Appendix B)—that (17) below, and not (15), is an appropriate paraphrase of (16). Now, if (16) (whose form is (D)) is paraphrased as (17), it turns out that (13) (whose form is (C)) implies (16), with the result that thesis (T1) is no longer falsified in the “Ortcutt” case.

• (16) Ralph believes Ortcutt to be such that he is taller than he.

• (17) Ralph believes Ortcutt and Ortcutt to be such that the former is taller than the latter.

The aforementioned implications can be restated more perspicuously using the propositions (13p), (15p) and (17p) below, instead of the sentences expressing them, (13), (15) and (17): in the “Ortcutt” case, proposition (13p) does not imply proposition (15p), whereas it does imply proposition (17p).

• (13p) <Ralph, believing, <Ortcutt, being taller than, Ortcutt>> 

• (15p) <Ralph, believing*, Ortcutt, λx[x is taller than x]> 

• (17p) <Ralph, believing*, <Ortcutt, Ortcutt>, λxλy[x is taller than y]>

Let’s now focus on the aforementioned claim that (17), and not (15), is the appropriate paraphrase of (16), which is crucial for the rejection of Salmón’s counterexample to (T1). Notably, in his article “Relational Belief” (1995, pp. 213–214), Salmón himself observes that a procedure from Kaplan (1986: Appendix B) called “articulation” yields precisely such a claim as an outcome:

[A] serious flaw in Quine’s [(1956)] proposal was uncovered by Kaplan in “Opacity” ([1986: Appendix B]). … Kaplan pointed out … that Quine’s method … is insensitive to subtle distinctions in content involving the phenomenon that I call “reflexivity”. … [S]uppose Ralph is under the illusion that the man in the brown hat is taller than the man at the beach. It would seem then that the following sentence is true: [(16)]. Quine’s procedure translates this sentence into [something like (15p)] which may be read: [(15)]. Unless Ralph is insane this is false. Kaplan improved upon Quine’s scheme by employing a procedure that Kaplan calls “articulation”. Kaplan translates the problem sentence[, i.e. (16),] instead into something along the lines of: [(17p)]. This may be read: [(17)].” (Salmón, 1995, pp. 213-214—underlining mine)

So, by paraphrasing (16) as (15) in “The Resilience of Illogical Belief” (2006), Salmón appears to make the same “mistake” that a decade earlier, in “Relational Belief” (1995), he, following Kaplan (1986), ascribed to Quine (1956). Salmón’s shift is surprising and challenging to justify, leaving room only for speculative explanations.²⁶

Regardless, having ascertained a sharp discrepancy between Salmón’s (2006) position and Kaplan’s (1986)/Salmón’s (1995) position, the question arises as to which of the two is more plausible. In my opinion, there are good reasons to counter the former position and hence favor the latter. For, first, if (15) were the correct paraphrase of (16), then in (16) the second occurrence of the pronoun “he” would be anaphorically linked to its first occurrence, since in (15) the position of that second occurrence is occupied by the reflexive pronoun “himself”. But nothing in (16) compels us to interpret the second occurrence of “he” in that way: (16) only mandates the interpretation of both anaphoric occurrences of “he” as linked to the common antecedent “Ortcutt”, with no obligation to consider anaphora as a transitive relation, given that “Ortcutt” in (16) is placed outside the linguistic context “to be such that …”. We should also bear in mind that (16) is the de re sentence corresponding to the de dicto sentence (13), and no feature of (13) suggests that in (16) the second occurrence of “he” is anaphorically linked to the first; quite the opposite, the “Ortcutt” case supports the conclusion that no anaphoric link holds between those two occurrences, as a consequence of the fact that in such a case, Ralph fails to realize that the man in a brown hat is the man at the beach.²⁷

As regards the alternative (in my opinion, preferable) paraphrase (17) of (16), one could object that it is also unsuitable: (17) contains two occurrences of “Ortcutt”, whereas (16) contains one occurrence of this name. In reply, I should note that if we aim, using the lambda calculus, to render the idea that in (16) both occurrences of “he” are anaphorically linked to “Ortcutt” while kee** open the possibility that the second occurrence is not anaphorically linked to the first (a possibility that, as I have argued above, is a fact in the “Ortcutt” case), the only available option seems to be Kaplan’s (1986) proposal (17p); and proposition (17p) is inevitably expressed by something like (17), where two occurrences of “Ortcutt” appear.²⁸

I must then conclude that no (clear) counterexample to thesis (T1) stems from the “Ortcutt” case: a counterexample emerges if, following Salmón (2006, p. 371), (16) (undoubtedly an instance of (D)) is paraphrased as (15); on the other hand, the counterexample vanishes if, following Kaplan (1986) and Salmón (1995), (16) is paraphrased as (17); and, as I have argued in this sub-section, Sect. 4.4, there are good reasons to favor the Kaplanian paraphrase (17) over the Quinean paraphrase (15).

4.5 More on Salmón’s strategy and further objections to it

Let’s return to Salmón’s strategy for resolving the “George Eliot”/“Mary Ann Evans” case, which, as explained in Sect. 4.1, consists in the rejection of thesis (T1). It is worth noting that Salmón considers valid (rightly, from a Millian-Russellian viewpoint) the inference from the de dicto report (5) (Sect. 3.1) to the de re report (19) below, which is also an instance of the (rejected) thesis (T1). In light of this, he (2006, p. 371) finds it appropriate to replace (T1) with (T3) below. Thesis (T3), which is a variant of (T1), allows him to claim that in the “George Eliot”/“Mary Ann Evans” case, the first-order de dicto report (5) entails the corresponding de re report (19), since the predicate “was a man” is monadic; at the same time, (T3) is taken by Salmón to be compatible with his solution to Schiffer’s puzzle—namely, blocking the move (viz. denying the implication and thus the entailment) from the second-order de dicto report (7)/(8) to the corresponding de re report (9)/(10) (Sect. 3.1)—due to the fact that the predicate “believes” is not monadic.

• (19) Ralph believes George Eliot to be such that she was a man.

• (T3) Every instance of the de dicto schema (C), i.e. ^⌜α believes that Φ_β^⌝, entails the corresponding instance of the de re schema (D), i.e. ^⌜α believes β to be such that Φ_it^⌝, if Φ_it has monadic-predicational form, “It”+VP, and under the threefold assumption that β is instantiated by a Millian singular term, VP is a monadic predicate, and “it” in (D) as well as in VP does not occur free.

I am not persuaded that the replacement of (T1) with (T3) allows Salmón to achieve his goals: the denial of the implication from (7)/(8) to (9)/(10) falsifies the former as well as the latter thesis. In fact, the exact instance of Φ_it in the report (9)/(10) (instantiating (D)) is not “believes” (as I conceded above) but is “Ralph believes that she was a man” or rather: “She” + “is believed by Ralph to have been a man” (concerning this harmless shift from de dicto to de re, see fn. 16). Now, since “is believed by Ralph to have been a man” is a complex monadic predicate, (T3) successfully applies to (7)/(8), licensing the move from (7)/(8) to (9)/(10), in contrast with Salmón’s intended solution to the “George Eliot”/“Mary Ann Evans” case.

In order to avoid this drawback, Salmón could replace thesis (T1)/(T3) with the even more sophisticated thesis (T4) below, where by simple-monadic predicate I mean a monadic predicate from which no polyadic predicate can be carved out (e.g. “was a man” is simple monadic, whereas “loved Evans” isn’t because from the latter a dyadic predicate, “loved”, can be carved out).

• (T4) Every instance of the de dicto schema (C), i.e. ^⌜α believes that Φ_β^⌝, entails the corresponding instance of the de re schema (D), i.e. ^⌜α believes β to be such that Φ_it^⌝, if Φ_it has simple-monadic predicational form, “It”+VP, and under the threefold assumption that β is instantiated by a Millian singular term, VP is a simple-monadic predicate, and “it” in (D) does not occur free.

Unfortunately, thesis (T4) does not provide justification for a case like the following, signaled by Schiffer (2008, §3), where an instance of (C), viz. (5*) below, seems to entail (from a Millian-Russellian viewpoint) the corresponding instance of (D), viz. (19*) below, despite the fact that the relevant instance of Φ_it, viz. “loved Evans”, has complex-monadic predicational form:

Consider … Dr. Fritz Lauben, an acquaintance of George Henry Lewes who knew Mary Ann Evans and knew about the love affair she and Lewes were having. I should think that if

• (19) Ralph believes George Eliot to [be such that she was] a man

counts as true by virtue of its being the case that

• (5) Ralph believes that George Eliot was a man,

then

• (19*) Fritz believed Lewes to [be such that he] loved [Evans]

… should count as true by virtue of its being the case that

• (5*) Fritz believed that Lewes loved Evans.

Finally, and more problematically, as we will see in the upcoming Sect. 5, the strategy of replacing thesis (T1) with thesis (T4), and a fortiori the strategy of replacing (T1) with (T3), does not resolve the new puzzle I am going to present in that section.

5 The “Superman”/“Clark Kent” puzzling case

5.1 The puzzle

The “Superman”/“Clark Kent” case aims to achieve results similar to Schiffer’s “George Eliot”/“Mary Ann Evans” case, with the advantage that the overall strategy proposed in the second part of Sect. 4.5 for resolving the latter case, i.e. rejecting thesis (T1) and replacing it with thesis (T4), does not apply to the former case. The key difference between the two puzzles lies in the fact that the “Superman”/“Clark Kent” case, unlike the “George Eliot”/“Mary Ann Evans” case, involves first-order (instead of second-order) belief reports where the instances of Φ_it have simple-monadic predicational form.

The new puzzling case originates in ideas, presented in Saul’s article “Substitution and Simple Sentences” (1997) and in Braun’s and Saul’s later article “Simple Sentences, Substitutions, and Mistaken Evaluations” (2002), about simple sentences that provoke anti-substitution intuitions.

Many competent speakers initially judge that [(20) below] is true and [(21)] is false, though they know that (22) is true.

… But perhaps you [do] not have these intuitions. No matter: it’s undeniable that many competent, rational, relevantly well-informed speakers who understand these sentences do have these intuitions, at least initially—the reactions of readers to Saul (1997) are sufficient to establish this. (Braun & Saul, 2002, pp. 1–2—underlining mine)

• (20) Superman can fly, while Clark Kent cannot fly.

• (21) Superman can fly, while Superman cannot fly.²⁹

• (22) Superman is identical with Clark Kent.

The data provided in the quoted passage above are grounded on experience; as such, they can be taken at face value (unless stronger empirical data that disprove the former are put forward). Now, let Emily be one of the many rational English speakers, mentioned by Braun and Saul, who, although they know that (22) is true, have the intuition that (20) is true and (21) is false. In light of the provided data, we expect Emily to have the disposition to utter sentence (20) sincerely (due to her judgment that this sentence is true), on reflection (since she is “relevantly well informed”),³⁰ and competently; as confirmed in the quoted passage, she also understands such a sentence. Based on these considerations and by applying the Disquotational Principle (Sect. 3.1),³¹ we conclude that Emily believes the proposition expressed by (20).

It is worth observing that such a conclusion can also be achieved without applying the Disquotational Principle but by employing Scott Soames’ characterization of belief as a disposition to judge:

To believe that B is red is … to be disposed to judge that it is. (Soames, 2015, p. 18)

In fact, starting from Braun’s and Saul’s (2002, p. 1) datum that Emily judges that (20) is true and using Soames’ characterization of belief, we derive that Emily believes that (20) is true. Since Emily understands sentence (20) (2002, p. 1) and therefore entertains the proposition it expresses, we conclude that Emily believes such a proposition.

Finally, note that this conclusion is also immediately derivable from the following further passages by Braun and Saul:

… the proposition that [(20)] … expresse[s is] false, but [Emily] c[o]me[s] to believe it [is] true. (2002, pp. 16–17—underlining mine)

[Emily] c[o]me[s] to believe that [(20)] is true and [(21)] is false. (2002, p. 20—underlining mine)

Now, having established—in three different ways, viz. (j) by applying the Disquotational Principle, (jj) by employing Soames’ (2015) characterization of belief, and (jjj) by simply exploiting some quotations by Braun and Saul (2002, pp. 16–17, 20)—that Emily believes the proposition expressed by (20), we can infer that she also believes (simultaneously) the propositions expressed by (23) and (24) below. Consequently, reports (25) and (26) below are true.³²

• (23) Superman can fly.

• (24) Clark Kent cannot fly.

• (25) Emily believes that Superman can fly.

• (26) Emily disbelieves that Clark Kent can fly.

Crucially for our discussion, since “can fly” is a simple monadic predicate, thesis (T4) (Sect. 4.5) applies to (25) and (26), licensing the moves from these de dicto reports to the de re reports (27) and (28) below respectively. At this point, the substitution in the de re report (27) of the name “Superman” with the coreferring—according to Saul (1997, pp. 103–105, 2007, Chap. 2) and Braun and Saul (2002, §3.1)—name “Clark Kent” yields (29) below.

• (27) Emily believes Superman to be such that he can fly.

• (28) Emily disbelieves Clark Kent to be such that he can fly.

• (29) Emily believes Clark Kent to be such that he can fly.

Finally, from (28) and (29) the following falsification of Frege’s Constraint can be drawn: Emily rationally believes and disbelieves Clark to be such that he can fly (~ii) thinking of the property of being such that one can fly under only one mode of presentation and (~i) thinking of Clark without having different modes of presentation of him which Emily takes to present distinct individuals.³³

5.2 Two related objections to the “Superman”/“Clark Kent” case and a reply

There is a couple of prima facie easy ways one could explore in order to dismiss the “Superman”/“Clark Kent” case: arguing that the names “Superman” and “Clark Kent” refer to different entities or personae, say Clark-in-his-superhero-role and Clark-in-his-ordinary-life, to the effect that the substitution salva veritate of “Superman” with “Clark Kent” in (27) is not permitted (semantic objection); or arguing that in Emily’s statement (20), the token names “Superman” and “Clark Kent”, though (semantically) coreferential, are pragmatically/cognitively taken by Emily as referring to the aforementioned personae,³⁴ with the result that Emily ends up thinking of Clark under different modes of presentation which she mistakenly takes to present distinct entities, viz. the two personae in question, in accordance with Frege’s Constraint (pragmatic/cognitive objection).

In reply, it could be pointed out, first of all, that both objections do not match with Emily’s identity judgment that (22) is true (beginning of Sect. 5.1).³⁵ Secondly, I would like to present a variant of the “Superman”/“Clark Kent” case—echoing analogous cases from Braun and Saul (2002, p. 11) and Saul (2007, §2.2.3.2, §2.2.3.3) but with a wider impact than them—where the token names “Superman” and “Clark Kent” in a particular utterance of (20) (semantically) corefer and are pragmatically/cognitively taken by Emily as coreferring, viz. as referring to Clark rather than to two distinct personae. Suppose that Emily sincerely, on reflection and competently assents to an utterance of (20) made by Lois Lane, who, unlike Emily, does not know that the identity sentence (22) is true. Assuming that Emily is aware of Lois’ ignorance, it seems intuitively correct to claim that Emily fully (i.e. semantically, pragmatically and cognitively) understands Lois’ utterance and thus grasps the proposition communicated (i.e. semantically expressed and/or pragmatically conveyed) by such an utterance. If ab absurdo two personae (instead of the individual Clark) were pragmatically/cognitively associated by Emily to Lois’ token names “Superman” and “Clark Kent”, then Emily, contrary to all evidence, would fail to understand Lois’ utterance pragmatically/cognitively: surely, by uttering (20), Lois does not intend to communicate a proposition about the entities Clark-in-his-superhero-role and Clark-in-his-ordinary-life; nor does Lois intend to cognitively take “Superman” and “Clark Kent” to refer to such entities, for the simple reason that she is unaware of Clark’s dual role in his life (in fact, she wrongly judges that Clark Kent and Superman are distinct individuals). Lois’ aforementioned intentions—both communicative and referential—combined with her competence in uttering (20) also rules out that her tokens of “Superman” and “Clark Kent” (semantically) refer to Clark-in-his-superhero-role and Clark-in-his-ordinary-life, and therefore that her utterance expresses a proposition about such entities: if ab absurdo those token names referred to the entities in question, then Lois, as a competent user of (20), would notice it, which instead does not happen. In sum, in order to account for the intuitive fact that Emily fully understands Lois’ utterance of (20) and for Lois’ communicative/referential intentions combined with her competence in uttering (20), we must conclude that at least in the present variant of the “Superman”/“Clark Kent” case, the token names “Superman” and “Clark Kent” (semantically) corefer and Emily follows Lois in pragmatically/cognitively associating the individual Clark to them as their referent.³⁶

5.3 A third doubt about the “Superman”/“Clark Kent” case and a couple of responses

Let’s return to the “Superman”/“Clark Kent” case presented in Sect. 5.1. As we assumed in that sub-section, Emily, a rational speaker, has the disposition to utter sincerely, on reflection and competently conjunction (20). She is then expected to have the same sort of disposition towards its two conjuncts, (23) and (24). Let’s suppose that she in fact utters, under the aforementioned conditions, (23) and (24). One might wonder how, despite uttering (23) and judging that the identity sentence (22) is true, Emily refrains from drawing the conclusion (31) below, being instead disposed to utter sincerely and on reflection the negation of (31), namely (24).³⁷

(31) Clark Kent can fly.

Braun and Saul (2002, §4) and Saul (2007, Chap. 6, Appendix B.2.1) have offered various empirical/psychological explanations for such phenomenon, the most basic of which (basic explanation) relies on the hypothesis that when Emily utters (23) and (24), her identity judgment that (22) is true is somehow set aside—I would say, it is stored into a separate compartment of her mind.³⁸ As a result,

[Emily does] not consider [(22)] during [the] evaluation procedure; or if [she does], [she fails] to go through the sort of reasoning that would [allow her] to detect [her] mistake. (Braun & Saul, 2002, p. 17)

I am inclined to think that there is an even more basic explanation for the phenomenon in question, which does not presuppose the isolation of Emily’s identity judgment within her mind and which is written between the lines of Braun’s and Saul’s (2002) article:

[Emily] treat[s] “Superman” and “Clark[ Kent]” … as if they were about different people.

… Clark’s double life gives [Emily] good reason to [do so]. (Braun & Saul, 2002, p. 19−underlining mine)

On my own interpretation of these passages, treating as (which I also call “cognitive coordination”) is a non-semantic and non-pragmatic but cognitive relation between a subject and two token terms; such a relation is grounded (not in personae but) in the relations of taking as and doing as if, the latter being a sort of simulation of taking as (more details in fn. 12). Now, it happens that although Emily (unlike Lois Lane) correctly takes “Superman” in (23) and “Clark Kent” in (24) as coreferring, she does as if they were non-coreferring, in doing so simulating someone (like Lois) who wrongly takes them as non-coreferring. As a consequence of it, Emily (like Lois but for a different reason than her) does not treat “Superman” in (23) and “Clark Kent” in (24) as coreferring, i.e. she negatively coordinates them. This fact allows Emily to utter the quasi-contradictory sentences (23) and (24) sincerely, on reflection and rationally while knowing that (22) is true, and thus justifies her unwillingness to draw at least in some occasions, for instance the one under discussion, the conclusion (31) above (which plainly contradicts (24)) from the premises (22) and (23).

6 From Salmón and Schiffer to an alternative strategy for solving the two puzzles

6.1 Objections to Salmón: a summary

Let’s sum up the criticisms I have addressed to Salmón in the previous sections:

• (I) Salmón’s solution to Schiffer’s “George Eliot”/“Mary Ann Evans” case appears not entirely justified with respect to its endorsement of Frege’s Constraint, and artificial with respect to its appeal to sui generis (viz. unstructured or non-standardly structured) guises (Sect. 4.2);

• (II) Salmón’s argument for the strategy underlying his own solution, i.e. rejecting thesis (T1), is disputable (Sect. 4.4);

• (III) Salmón’s overall strategy of replacing the rejected thesis (T1) with thesis (T3) does not allow him to achieve the intended solution to the “George Eliot”/“Mary Ann Evans” case (Sect. 4.5);

• (IV) Such a strategy and even the more sophisticated one that replaces thesis (T1) with thesis (T4) fail to solve the “Superman”/“Clark Kent” case (Sect. 5.1).

Among (I)–(IV), the objections of greatest impact are, I think, (III) and (IV). In response to these two objections, Salmón could, in principle, maintain that it is possible to formulate a thesis, even more sophisticated than (T4), which resolves the “George Eliot”/“Mary Ann Evans” case by blocking the move from (7)/(8) to (9)/(10) (Sect. 3.1), and the “Superman”/“Clark Kent” case by blocking the move from (25)/(26) to (27)/(28) (Sect. 5.1). On the other hand, the desired thesis is also expected to offer a justification for the valid (from a Millian-Russellian viewpoint) inferences from (5) to (19) and from (5*) to (19*) (Sect. 4.5). I wonder how all these desiderata could be fulfilled without sharpening the criticism (I), viz. without rendering Salmón’s solution to the “George Eliot”/“Mary Ann Evans” case and the “Superman”/“Clark Kent” case more artificial.

6.2 Some objections to Schiffer

So, the strategy of resolving the two puzzles by replacing (T1) with some more sophisticated thesis does not seem very promising. In order to devise a better strategy, it is useful to reconsider the list of elements that, based on Schiffer (2006, pp. 363–364, 2008, §1–§2) and based on what we discovered in Sect. 3 and in the second part of Sect. 4.2, underpin his “George Eliot”/“Mary Ann Evans” case, i.e.:

• (a) Salmón’s claim (B) about the truth conditions of de dicto belief reports;

• (b) thesis (T1);

• (c) Salmón’s semantics of de re belief reports;

• (d) Frege’s Constraint;

• (e) Propositional guises are (standardly) structured entities.

As anticipated in Sect. 3.2, Schiffer’s (1987, 2006) strategy to solve his own puzzle consists in rejecting (a) and, ultimately, the Millian-Russellian semantics of belief reports.

Nevertheless, various criticisms may be raised against this strategy. First, as the reader can verify by going over my exposition of the “George Eliot”/“Mary Ann Evans” case in Sect. 3.1 (and considering the additional point made in the second part of Sect. 4.2), elements (a) (enriched with (e)) and (c) are not indispensable in order to generate this case: the move from Jane’s linguistic dispositions towards (5) and (6) to the de dicto reports (7) and (8) can be performed using the Disquotational Principle instead of (a); and the move from the de re report (10) to (11) is authorized by any semantics involving de re belief reports (fn. 9), not exclusively by (c). Given these ascertainments, it can be alternatively and perhaps more plausibly argued that the “George Eliot”/“Mary Ann Evans” case is generated by the following four elements: (a*), (b), (c*) and (d). If so, pace Schiffer, the import of his puzzle would not lie in the failure of the dispensable element (a).

• (a*) Disquotational Principle.

• (c*) any semantics involving de re belief reports.

Regarding Schiffer’s rejection of the Millian-Russellian semantics of belief reports in the light of his puzzle, it could be critically pointed out that none of the elements (a*), (b), (c*) and (d) is exclusive of this semantics, but they are also involved in other semantics. This is obvious for (a*). Concerning (b) and (c*), remember that, as famously shown by Kaplan (1968), Fregeanism can also encompass de re constructions and thus the de dicto/de re distinction. As for (d), Schiffer (1990, pp. 249–254) himself has maintained that most (Millian-Russellian as well as) non-Millian-Russellian theories of belief reports involving modes of presentation (more or less explicitly) commit to Frege’s Constraint. In addition, as argued in Sects. 3.2 and 4.2, it is a viable option for a Millian-Russellian theorist to dismiss modes of presentation, thus completely renouncing (d). Based on all these considerations, which can be extended mutatis mutandis to the “Superman”/“Clark Kent” case, the following conclusion can be drawn:

• (V) Pace Schiffer, the “George Eliot”/“Mary Ann Evans” case poses a problem not specifically to the Millian-Russellian semantics of belief reports, and so does the “Superman”/“Clark Kent” case. Therefore, the import of both puzzles is not the failure of this semantics.

6.3 An alternative strategy

What is, then, a good strategy to solve the “George Eliot”/“Mary Ann Evans” case and the “Superman”/“Clark Kent” case? Taking for granted that (a*) and (c*) should not be rejected,³⁹ two options remain: either rejecting (b); or rejecting (d). In Sect. 6.1, I turned down Salmón’s proposal of rejecting (b), in light of (I)–(IV). By exclusion, (d) should be rejected. So,

• (VI) the import of the “George Eliot”/“Mary Ann Evans” case and the “Superman”/“Clark Kent” case is the failure of Frege’s Constraint.

On my own view, Frege’s Constraint is replaced by the Relationist Constraint (end of Sect. 3.2), involving cognitive coordination (fn. 12); accordingly, the two puzzles are solved as indicated in footnotes 19 and 33.

6.4 A further doubt about Schiffer

In his book The Things We Mean (2003), Schiffer has proposed a very original semantics for simple sentences and belief reports, according to which:

• type linguistic expressions lack semantic content but have character*, which serves to constrain the semantic content of their tokens (2003, pp. 132–133, 156);

• statements (or token declarative sentences) express Schifferian propositions, i.e. propositions which are pleonastic (2003, p. 71), unstructured (2003, pp. 82, 84), and more or less fine-grained (2003, pp. 82, 84) depending on the conversational contexts where the statements are made (2003, pp. 80, 81);

• token singular terms and token predicates in belief statements are semantically innocent, with the former retaining their customary referent (2003, p. 84) and the latter preserving their customary content (Schiffer’s Replies in Ostertag, 2016, pp. 387, 408);

• believing is a two-place relation between a subject and a Schifferian proposition (2003, p. 84).

Nothing seems to prevent us from introducing a de dicto/de re distinction within Schiffer’s semantics: token sentences of de dicto form (C) could be said to express Schifferian propositions with a granularity comparable to that of propositions of the form <a, believing, that Fb> , where that Fb ranges over fine-grained Schifferian propositions; and token sentences of de re form (D) could be said to express Schifferian propositions with a granularity comparable to that of propositions of the form <a, believing*, b, being such that one has F> (second part of Sect. 1). If so, in the “George Eliot”/“Mary Ann Evans” case, from Jane’s rationally believing the fine-grained Schifferian propositions that Ralph believes that George Eliot was a man and that Ralph does not believe that Mary Ann Evans was a man we deduce, in accordance with Schiffer’s theory and using thesis (T1), Schifferian propositions having a granularity comparable to that of the propositions (9p) and (10p) in Sect. 4.2, despite the fact that Jane does not have different ways of thinking (Schiffer’s Replies in Ostertag, 2016, pp. 411, 433) of Eliot which she takes to present distinct individuals,⁴⁰ in contrast with Frege’s Constraint.⁴¹ It is therefore evident that:

• (VII) Schiffer’s “George Eliot”/“Mary Ann Evans” case, and mutatis mutandis the “Superman”/“Clark Kent” case, pose a problem to Schiffer’s theory of belief reports as well.

Given this outcome, which aligns with the previous conclusion (V) (Sect. 6.2), one may wonder how Schiffer would solve within his theory the two puzzles—generated by (a*), (b), (c*), (d) (Sect. 6.2)—and in particular his “George Eliot”/“Mary Ann Evans” case: it is highly probable that Schiffer (1987, p. 465, 2006, p. 363) endorses some version of the (a*) Disquotational Principle; and he would certainly have no objection to incorporating (c*) de re belief reports into his theory, as proposed above; nor, in light of his (2006, pp. 365–367), would he abandon (d) Frege’s Constraint in favor of Frege’s Constraint*. Perhaps, in line with his (2016, p. 452), he is inclined to resolve his puzzle by (~b) rejecting thesis (T1) like Salmón; on the other hand, in this paper, I have extensively criticized such a strategy (see Sect. 6.1 for a summary of my criticisms) and I have also defended thesis (T1) (see fn. 7).

7 Conclusions

Although I (unlike Schiffer) endorse the Millian-Russellian semantics of belief reports (fn. 17), I (pace Salmón) think that the “George Eliot”/“Mary Ann Evans” case should not be solved by denying the implication from (7)/(8) to (9)/(10) (Sect. 3.1), nor should the “Superman”/“Clark Kent” case be solved by denying the implication from (25)/(26) to (27)/(28) (Sect. 5.1), leading to the rejection of thesis (T1). Arguably, this thesis is true within a Millian-Russellian framework—considering that potential counterexamples to (T1) like the “Shorty” case can be discharged as proposed in fn. 7.

My solution (Sect. 4.2 and footnotes 19, 33) to the two puzzling cases consists in denying the implication from the conjunction of (9) and (10)/(11) to (12) (Sects. 3.1, 4.2) and the implication from the conjunction of (27)/(29) and (28) to (30) (Sect. 5.1 and fn. 33). The truth of (9)–(11) and (27)–(29) is ensured by the rejection of Frege’s Constraint (Sect. 6.3) and its replacement with the Rationalist Constraint (end of Sect. 3.2), the latter involving cognitive coordination (fn. 12) instead of modes of presentation.

As became evident in Sect. 6, the rejection of Frege’s Constraint poses a challenge not only to Salmón (1986a) but also to Schiffer (2003) and to all mode-of-presentation theorists of belief reports who (more or less explicitly) assume that modes of presentation fulfill this constraint.⁴²