by David J. Lobina
A new post, a novel series, and back to all things cognitive science about language. In this chapter of this column, which hopefully won’t feature any politics but perhaps a little bit of ideology, I will focus on the very thorny issue of how language and thought relate. Or said otherwise, the issue of whether we think in a natural language such as English or Italian, or instead in something else together – in The Language of Thought (LoT), as I shall in fact argue!
The question of how language and thought relate is possibly time immemorial, and the extant literature too large to review satisfactorily in a series of posts. Such a description of the general state of affairs won’t surprise anyone; what may raise a few eyebrows, perhaps, is the claim that large tracts of the modern literature are afflicted by two rather important problems. Firstly, scholars focused on this relationship are hardly ever clear as to what they take language and thought to be exactly, as definitions are rarely put forward; and complementing this, the very same scholars are usually rather vague regarding how language and thought are supposed to relate at all.
Myriad possibilities can be imagined in either respect, but identifying the different prospects necessitates significant exegetical effort. Natural language, to begin here, may be defined in many different ways: as a (computational) system of the mind, along with its connection to other mental systems, like Noam Chomsky has long argued; as a set of mental representations, perhaps in terms of propositional attitudes, as exemplified by a sentence like I believe/fear that this post will go unread, a position the philosopher Jerry Fodor once defended; as a purely communicative system, thereby closely connected to mind reading (the ability to ascribe mental states like beliefs and desires to others) and the like, as the psychologist Michael Tomasello proposes; as an exclusively real-time processing phenomenon (language as a ); etc
Whatever perspective one adopts, the particular choice will surely have an effect on what sort of issues are considered in the study of how language and thought are actually related. In fact, it is not an uncommon feature of the language-and-thought literature that the eventual terms of comparison employed in many a study are in fact the result of the conclusion to an argument on how to relate the two phenomena rather than a principled way to approach the problem at hand.
Among the different possibilities, we find the following, prima facie incompatible options: linguistic representations may be the actual vehicles of the main medium of thought humans employ; thought may be in fact impossible without language, this capacity necessitating the representation of the propositional attitudes, an ability some scholars have argued to be only achievable by employing language (or so have philosophers Donald Davidson and Ruth Millikan argued); language may connect different conceptual systems of the mind via specific linguistic representations (most probably, syntactic representations), these systems otherwise unconnected in the absence of a fully formed language faculty, the favourite take of the philosopher Peter Carruthers; the employment of language in speech (either inner or outer) may augment one’s computational powers during explicit mental processes, language therefore at least enhancing (but probably not replacing) thought; natural language may be in fact inadequate as a system of thought, its main function being the communication of thoughts rather than its representation, thought therefore constituting a (perhaps slightly) different mental phenomenon altogether, Fodor’s take; etc.
The result is that scholars are often seemingly talking about very different phenomena when discussing the reputed relationship, from representational and architectural issues to processing phenomena. What’s more, these scholars also differ in terms of how central they take language to be for thought, from those who defend that the most sophisticated forms of thought are conducted in natural language (Davidson, Carruthers), in some cases concluding that there is no medium of thought other than language, thus cutting the distinction very loose indeed, to those who believe that thought is instead conducted in a medium other than natural language, language and thought thus (again, perhaps slightly) different domains of the mind (Fodor).
So how does language and thought relate, then? And what is “thought”, anyway? Perhaps one way of answering both questions is to determine whether any of the representations language provides – syntactic, phonological, semantic, etc. – are suitable for the “fixation of belief”, which is what many philosophers will take thinking to be primarily, with problem solving, planning, etc. constituting particular instances of belief fixation. I should stress that by focusing on belief fixation one does not necessarily have to focus on the process of thinking per se, even if various psychological data will be discussed in this series. Rather, I am more likely to concentrate on what thought representations must be like so that belief fixation is at all possible, which is a slightly different issue.
Thought will here be characterised by paying close attention to two properties of human cognition have engaged philosophers such as Fodor a great deal: a) what philosophers usually call content, or a proposition, the kind of objects over which propositional attitudes range, and the objects of beliefs themselves (e.g., the proposition that this post will go unread); and b) the representational vehicle, or format, of a thought representation, that is, the structural properties of a thought representation – as an analogy, consider the syntactic properties of a sentence and the way in which sentences are divided into clauses, phrases, etc. “To have a thought”, according to this view, is to entertain a proposition along with the relevant representational vehicle, while “to think” is to combine propositions and representational vehicles in various ways, from embeddings and combinations of various kinds (via logical connectives such as and and or, for instance) to the premises-and-conclusion scheme typical of reasoned thought.
To this end, a thought representation must be fully explicit and complete, bear truth values (a proposition should be evaluable as a True or False statement of the state of affairs in the world), and exhibit a constituent structure (viz., a predicate with its arguments, as it is central to human thought to ascribe properties such as go unread to objects such as this post). Consequently, I take the question of “what thought representations must be like” to be primarily about the format properties of such representations, and not so much a question about its content properties.
That out of the way (more or less), what shall I understand “language” to be? Readers of previous posts will know the answer to this. By natural language I will here understand, following the line defended by the generative grammar enterprise in the last 60 or so years, the faculty of language (FoL, in short), an architectural state of the mind composed of the following components: a combinatorial operation that combines linguistic pieces (words, phrases, etc.); a set of lexical items (bundles of syntactic, phonological, and perhaps semantic, features; lexical items are realised as words in speech); and two interfaces that connect the faculty with other systems of the mind, namely the sensorimotor (SM; roughly, the sound/sign systems) and the conceptual-intentional (C/I; even more roughly, the thought/meaning systems).
According to this architecture, the language faculty generates a number of
prima facie very different types of representations:
1: one for the SM systems, termed a PHONetic representation (a flat and linear object ready for externalisation, be in speech or in any other modality);
2: one for the C/I systems, this one called SEMantic and the central concern of many linguists (a hierarchical object encompassing the containment relations inherent in a linguistic object; despite its name, SEMs are syntactic structures);
3: an actual, linguistic semantic representation put together, it must be assumed, at the C/I interface and meant to establish/compute the interpretation of a sentence;
4: a phonological representation mediating between the combinatorial operation and the SM interface before the linearisation of a SEM into a PHON (the linearisation process perhaps operating over this phonological representation).
In simpler terms, 1 is a phonetic representation, 2 is syntactic, 3 is semantic, and 4 is phonological. Representations 1, 2, and 4 have been more thoroughly studied than 3, but the general point I shall make in the series applies to all four. To wit, these four domains manipulate atomic elements that, combined with principles specific to each domain, yield representations that appear to be extraneous to thought, since these properties would appear to be specific to language. This is perhaps most obvious in the case of phonological representations, given its primitives (phonemes, etc.) and structural tiers (segmental, syllabic, metrical), and as such I won’t have much to say about them. The other three representations (1, 2, and 3), however, have all been argued to play some sort of role in thought and thinking (sometimes a very central role). Be that as it may, I should perhaps make clear that SEMs are the central representations within the design of the language faculty, for the other representations (PHONs, the phonological, and the semantic) are supposed to be derived from SEMs, either right after the syntactic engine has finished operating or in tandem with it (perhaps even strictly pari passu).
By analogy to the FoL and SEMs, the internal structure of propositions will be one a key issue here; propositional constituents, in this sense, must allow for the flexibility and creativity involved in belief fixation, as evidenced in the very common phenomenon in which different types of perceptual inputs (modalities) can be combined with each other and with many other beliefs during the construction of thought, something that needs to be explained. The usual way to do this is to take thought constituents to be what some philosophers and psychologists call concepts, the mental particulars that underlie propositions – mental particulars in the sense that the concept CAT is a different particular from the concept DOG.
Concepts are abstract, therefore amodal, stable and thus re-usable, and must be embedded in a conceptual repertoire of not insignificant structure, a conglomerate of properties that would allow for the combination of mental representations into ever more complex representations. As Fodor has argued since at least his 1975 book, The Language of Thought (shown in the photo above), conceptual representations are a bit like natural language: thought has a syntactic structure and a compositional semantics; this, in a nutshell, is the LoT story, and more details will be provided in 4 weeks.