Many philosophers continue to be attracted to a Fregean (1892) approach to indirect discourse and attitude ascription (e.g. recently Yalcin 2015; Lederman 2022). But there is a hanger-on accompanying that theory, the debate over which is at something of a standoff. It is the “hierarchy view”: that for each word there is an infinite hierarchy of semantic values that it contributes in different kinds of occurrence. Frege does not mention this idea when explaining his theory, but many readers have taken his theory to entail it. Saul Kripke (2008, 183) even called it a “familiar consequence” of the theory. Others deny the entailment, but nonetheless maintain that the hierarchy view is a natural extension of Frege’s stated doctrine, to cover cases that he didn’t originally have in mind. Those are cases of so-called “multiple embeddings,” as in

Susan thinks that John believes that Paris is beautiful.

The idea is that since “Paris” is doubly embedded by indirect discourse verbs, its semantic value in that occurrence is not its customary sense—which would be its value if it were only singly embedded—but its “indirect” sense, a distinct sense that “presents” its customary sense. And so on, for higher degrees of embedding: when embedded to degree n, a word’s semantic value is its level n sense.

Almost all of the discussion of this view has concerned its bare tenability rather than its plausibility. My concern here is with the latter. I will examine what seem to me to be the two strongest considerations given in support of the view.

One consideration appeals to cases of multiple embeddings in which intersubstitutability salva veritate seems to be more stringent than it is within single embeddings (Mates 1950). The claim is that the hierarchy explains this apparent phenomenon. For it allows that two words can share their level n senses while having different level n + 1 senses. In such a case, the words are guaranteed to be intersubstitutable when embedded to degree n, but not so guaranteed when embedded to degree n + 1. Many philosophers (e.g. Fine 2007; Horwich 2014) have endorsed this idea. But there is a methodological problem: this is not the only possible explanation of the Matesian phenomenon. There is another sort of explanation which involves no hierarchy. So the hierarchy view does not come close to being supported by an inference to the best explanation.

There are also arguments meant to show that the hierarchy view follows from Frege’s semantic theory. The best argument of this sort is due to Tyler Burge (1979; 2005). Burge argues for the hierarchy on the basis of claims about how Frege’s core principle plays out in the formalization of a natural language. His argument is careful and detailed—dauntingly so. It is widely endorsed (e.g. Salmon 2005). Its crucial step rests on a claim about the well-formedness of expressions in a “logically perfect” (Frege 1892, 58) language. On examination, it turns out that that claim is plausible only if the hierarchy view is true. The argument, then, begs the question, and gives the hierarchy view no support.

So the hierarchy view has even less going for it than has been thought. What then explains its enduring appeal—such as it is, mostly among Fregeans? How is it that so many readers, even while differing in their alliegance to Frege’s overall approach, have thought that his view at least strongly suggests the hierarchy? That common reaction is not due to some complex argument such as Burge’s. What then is behind it?

I diagnose it as resting on an error in thinking of the base case of Fregean reference shift. The error is mistaking what a word’s semantic value does, for something the word itself does. When embedded, an attitude or discourse verb does not induce embedding, precisely because its semantic value in such an occurrence is not what it is when it occurs unembedded (and does induce embedding). On the Fregean approach, understood correctly, there is no such thing as “multiple embedding”—let alone a theory of it. This illustrates again the power of Frege’s dictum “never to ask for the meaning of a word in isolation, but only in the context of a sentence” (1884, xxii).

References

Burge, Tyler. 1979. Frege and the hierarchy. Synthese 40:265–81.

———. 2005. Postscript to “Frege and the hierarchy”. In Truth, thought, reason: essays on Frege, 167–210. Oxford: Clarendon Press.

Fine, Kit. 2007. Semantic relationism. Oxford: Blackwell.

Frege, Gottlob. 1884. The foundations of arithmetic. Translated by J. L. Austin. Evanston, IL: Northwestern University Press.

———.1892. Über sinn und bedeutung. Zeitschrift für Philosophie und philosophische Kritik100: 25–50. Translated as “On sense and meaning” in Translations from the philosophical writings of Gottlob Frege, 2d edition (1960), ed. and trans. Peter Geach and Max Black, 56–78. Oxford: Blackwell.

Horwich, Paul. 2014. Critical notice of Seven Puzzles of Thought and How to Solve Them: an Originalist Theory of Concepts, by R. M. Sainsbury and Michael Tye. Mind 123:1123–39.

Kripke, Saul. 2008. Frege’s theory of sense and reference: Some exegetical notes. Theoria 74: 181–218.

Lederman, Harvey. 2022. Fregeanism, sententialism, and scope. Linguistics and Philosophy 45: 1235–75.

Mates, Benson. 1950. Synonymity. University of California Publications in Philosophy 25:201– 26.

Salmon, Nathan. 2005. On designating. Mind 114(456):1069–1133.

Yalcin, Seth. 2015. Quantifying in from a Fregean perspective. Philosophical Review 124:207–53.

The functional composition of thoughts.

Frege scholars have long wondered how the thought expressed by a sentence relates to the senses of that sentence’s constituent expressions. According to the orthodox answer, a thought is a whole which contains, as its constituent parts, the senses of the relevant expressions (Dummett 1973; Heck and May, 2011). Others claim that thoughts compose functionally: senses of functional expressions, such as predicates, are themselves functions. On this heterodox approach, there is no need to hold that a thought contains the relevant senses as its parts (Church, 1951; Pickel, 2021).

In this paper, I will present a novel argument for the functional approach. However, I do not want to engage in Frege interpretation. I want to engage instead with the many contemporary philosophers of language and mind who help themselves to the Fregean notion of a thought. Indeed, my argument is, in one respect, crucially different from the arguments put forward so far by Frege-interpreters favourable to the functional approach. These arguments are presented as turning on an idiosyncratic commitment, one which Frege just so happened to hold (reference to senses). By contrast, the argument I produce in this paper begins upstream from there: it is not dependent on any commitment that one who goes in for the notion of a thought can reasonably opt out of.

I first argue that anyone who takes the notion of a thought seriously must accept reference to thoughts. This is (with respect to English) the claim that one can refer to a thought by prefixing “that” to a sentence which ordinarily expresses the relevant thought.[1] One who goes in for the notion of a thought must thus accept that a that-clause can be used to refer to the thought which the embedded sentence ordinarily expresses. The paper then contends that reference to thoughts, together with three additional but innocuous claims, forces the functional approach to thought composition.

The first additional claim, functional compositionality of reference, is this: the referents (semantic values) of the names into which a complex expression can be logically analysed yield the referent of the entire complex expression through functional application. The semantic value of a complex whole expression is thus the result of applying a function which is the value of some component expression to the value of some other component expression. In formal semantics, the principle has obtained the status of a virtually unchallenged orthodoxy (Heim and Kratzer, 1998). The second claim is this: a sentence as it appears inside a that-clause has the same (syntactic and semantic) structure as when it occurs alone. One who has mastered the semantic and syntactic structure of “Snow is white” exercises her capacity to discern that very same structure in understanding “S believes that snow is white”. Following Pickel, I call this embedded structure (2021). Finally, the third claim: transparency. Transparency states that a that-clause which refers to a thought cannot be understood unless one thinks the very thought it names. This has been noticed before by a variety of authors (i.e. Peacocke 2008). One cannot understand “It has not yet been proven that every even integer greater than two is the sum of two prime numbers” but fail to understand “Every even integer greater than two is the sum of two prime numbers”. The sense that every even integer greater than two is the sum of two prime numbers is thus transparent, for grasping this sense involves grasping the very thought it determines, namely every even integer greater than two is the sum of two prime numbers.

These are all the premises needed for the argument. I first show that reference to thoughts and transparency together yield reference to senses. (Reference to senses says that an expression inside a that-clause refers to the sense it ordinarily expresses. Where reference to thoughts claims that an entire that-clause refers to a thought, reference to senses governs the constituent expressions of a that-clause.) Then, relying on Pickel's 2021, I argue that reference to senses, functional compositionality of reference, and embedded structure together yield the functional approach to thought composition.

The dialectical significance of this new argument is as follows. In the context of Frege interpretation, arguments for the functional approach to thought composition have standardly begun with reference to senses. But reference to senses has the look of an optional commitment, one which Frege could have perhaps done without. This has made it seem as if one could take the notion of a thought on board without accepting that thoughts compose functionally, so long as one does not follow Frege in accepting reference to senses. Indeed, this is precisely Dummett’s reaction: he “charitably” amends Frege by rejecting reference to senses. My claim that reference to senses in fact follows from reference to thoughts modifies the dialectical situation. For no one who goes in for the notion of a thought should deny either reference to thoughts or transparency. (Indeed, Dummett does not). This shows that one cannot in fact go in for the notion of a thought without committing oneself to reference to senses. And this, in turn, forces the functional approach.

Bibliography

• Church, Alonzo (1951). “A Formulation of the Logic of Sense and Denotation” in P. Henle, H.M. Kallen & S. K. Langer (eds.), Structure, Method and Meaning (New York: Liberal Arts Press), pp. 3 – 24.
• Dummett, Michael (1973). Frege: Philosophy of Language (London: Duckworth).
• Evans, Gareth (1982). Varieties of Reference (Oxford: Oxford University Press).
• Heck, Richard, & May, Robert (2011). “The Composition of Thoughts”, Noûs 45: 126–66.
• Heim, Irene, & Kratzer, Angelika (1998). Semantics in Generative Grammar (Malden, MA: Blackwell Publishers).
• Künne, Wolfgang (2015). “Frege on That-Clauses” in B. Weiss (ed.), Dummett on Analytic Philosophy (London: Palgrave Macmillan), pp. 135 – 173.
• McDowell, John (1996). Mind and World (Cambridge: Harvard University Press).
• Peacocke, Christopher (2008). Truly Understood (New York: Oxford University Press).
• Pickel, Bryan (2021). “The Functional Composition of Sense”, Synthese 199: 6917-6942.

[1] Different languages will of course admit of different versions of the principles.