What’s The Language of Thought, That A Person May Grasp It?

by David J. Lobina

Pete Wishart MP, inadvertent expert on The Language of Thought.

What I really mean, of course, is what The Language of Thought (LoT, postea) is like; after all, in previous entries of this series on the relationship between language and thought, I have stated what the LoT is supposed to be, and thus, that can hardly still be an issue this late in the day – if anything, the question now is not what the LoT is, but what’s in it. In order to approach the latter question, we can do with a reminder on the former.

Starting from two reasonable, but by no account, universally accepted, assumptions about human cognition – namely, that much of cognition involves mental representations (possibly symbolic representations) as well as computational operations over these representations (in philosophy, these two assumptions are known as the representational and computational theories of mind, respectively) – one of the most striking features of the kind of thinking we conduct on a daily basis is how flexible it is. What I have in mind by this is the ability to combine information from different modalities (aural, visual, etc.) into a single representation, the result a thought, a decision, or what have you.

Such cognitive flexibility has often been taken as suggesting that the merging of different kinds of information must take place in a representational system that is in fact amodal – that is, a common code composed of, not words, pictures, or sounds, but abstract concepts, the mental particulars I have claimed in this series subsume most of cognition. And just like the words of a natural language, mental concepts can combine with each other into ever more complex representations (sometimes called conceptions), thereby explaining the richness of human thought – and, in turn, accounting for the “language” in the language of thought.

Despite the choice of words, the correspondence between the structure of the LoT and the structure of natural language should not be taken as being isomorphic, or even commensurable. This is a point I have already stressed in previous entries of the series, but which will become even more significant here, as I intend to show how the structure of thought is best ascertained through an analysis of the structure of natural language.

Though the contemporary locus classicus on the LoT, as mentioned before, is Jerry Fodor’s 1975 book, The Language of Thought, the idea of a “mental language”, a verbum mentis we use to do most of our thinking, is a rather old one. A relevant figure in this respect is William of Ockham, the 14th century Franciscan friar whose ideas on the mental language once made Fodor feel that his own take on the matter was a bit blasé.[i]

Among other things, and contrary to what is the case with natural languages, Ockham believed that there could be no ambiguity in the mental language, either at the level of its syntax or in terms of the content of whatever “units” the mental language is composed of (so no synonymous terms, and possibly no polysemy, either). This is clearly not a million miles away from modern characterisations of the LoT.

The point I want to make in this entry is that there is a general kind of structural misalignment, or mismatch, between language and thought, and that has a number of repercussions for the study of thought itself. Consider this curious sentence, reported in a newspaper article by a journalist during a debate in the British parliament:

Pete Wishart says he wants to put the question that the question not be put be now be put.

Let’s put aside the question of the sentence’s grammaticality along with such issues as a possible dialectic origin and the licence the written medium usually confers to linguistic production. Suffice to say that we can often ascribe meaning, and thoughts, to many ungrammatical sentences and that this sentence was once the product of someone’s linguistic knowledge. What thought did the speaker have in mind in this case?

As mentioned, the sentence was once employed by a journalist covering a debate in the British Parliament in order to report a phrase apparently used by one Pete Wishart MP. The sentence is a convoluted but somewhat typical example of British political discourse, its repetitive nature certainly too much for the uninitiated: too many questions, bes and puts. The strange phrasing is the result of the often idiomatic or formulaic nature of British parliamentary discourse, and ignorance of the relevant idioms can only confound.

This is because we need to know that the phrase that the question be now put is used in Parliament to request a “closure motion”, that is, a closing of the debate at hand so that the question under consideration can be put to a vote. A British parliamentarian has to literally say something along the lines of ‘I move that the question now be put’ so that a vote can be taken on whatever issue is being discussed. In order to know what the sentence means, then, one would have to be privy of this knowledge; one would have to be acquainted with this particular idiom.

In the case at hand, we have a situation in which there was a question under discussion that was not currently being considered for a vote (i.e., the question not put, or in parliamentary terms, the question not be put) and Pete Wishart MP wanted to put this question to a vote (viz., that the question now be put).[ii]

That’s a fairly informal way of describing what the sentence means, but what would the corresponding thought look like in an amodal representational system? Consider the following sentence:

Coffee grows in Africa.

What kind of thought underlies this sentence? We can follow a three-stage process to work this out. We start with the syntactic analysis of the sentence in order to compute its meaning. Once the semantic representation has been put together in the way this is usually done in the study of semantics (this is the second stage), we would then be able to appraise the structure of the conceptual representation, which can be carried out in terms of the old-style Aristotelian distinction between subjects (or arguments) and predicates.

Thus, the proposition or thought underlying coffee grows in Africa would be composed of the property being in Africa, the predicate, and the argument the growing of coffee, in clear contrast to the syntactic and semantic structures, where coffee would be the subject/agent and grows in Africa the verbal phrase/topic (very roughly, subject and predicate are syntactic phenomena, agent and topic semantic ones).

If so, predicates and arguments (or subjects) manifest themselves rather differently in the domains of language and thought. The general situation can be regarded as a case in which sentences such as coffee grows in Africa encompass different layers of structure. The analysis of the “linguistic” layer would outline the syntactic structure of a sentence along with its semantic representation, whilst the analysis of the “conceptual” layer would instead identify the logic-like subject-predicate structure of the underlying proposition/thought.

The first analysis would be linguistic in nature (syntactic and then semantic) and best left for the linguist, whilst the second is more philosophical and logical. In the case of the latter, it is noteworthy that,, according to this take predication is all-important to an analysis of the conceptual representations subsuming thought. In such an approach there is a sort of progression from a sentence’s syntactic/semantic structure to the corresponding (logical) structure in thought. Therefore, the overall strategy I am describing is based on the idea that differences in meaning point to specific differences in conceptual representations, even if each domain exhibits its own set of structural relations.

To come back to the Wishart sentence, and to, moreover, keep it all simple, let us focus on the meaning I have identified for the sentence – namely, that there is a question that has not been put to a vote, which Wishart now wants to put to a vote.[iii]

In terms of the internal structure of propositions, the thought-bearing segment of the sentence – i.e., that the question not be put (be) now be put – reveals two specific predications in the conceptual representation. The first relates to the question not being put to a vote – an unconsidered question, that is. This is the first internal structure the conceptual system would have to put together: the predication of <not being considered> to the argument <the question> (I use angled brackets to represent the internal structure of propositions, a standard way to do so). That is, the proposition/thought <not being considered <<the question>>.

After the initial predication, the conceptual system would have to construct a further predication of the so-far constructed proposition; namely, it would now have to assign the predicate <being put to a vote> to the proposition <not being considered <<the question>>, resulting in the proposition <being put to a vote <<unconsidered question>>.

Thus, the derivation of the conceptual representation would be mostly a two-stage affair, predicating <not being considered> of <the question>, and then, predicating <being put to a vote> of <not being considered <<the question>>.

So, what does all this mean for the match-up between the linguistic and conceptual representations underlying the original sentence? Even though I am using a similar vocabulary to the sentence’s actual words to describe the conceptual representations (being put to a vote, the question, etc.), it is very doubtful that the right concepts are all that similar to the corresponding lexical items. That is, the relevant concepts are likely to be very different indeed, and it is worth noting that I have had no need to employ so many bes and be puts in the conceptual representations. After all, the various be puts in the Wishart sentence refer to different concepts: an idiom in one case, a subjunctive verbal form in another, etc. Instead, I have simply used the copula to be plus a predicate as a way to account for the conceptual facts in terms of predicate-argument structures – the connection to the various bes and be puts is coincidental.[iv]

The overall point should be clear enough, though: there is some distance between how you put things in language – parliamentary or not, in Wishartian terms or not – and how the requisite thoughts are in fact represented in conceptual terms.[v] There is, of course, a fair amount of working out to do when it comes to many other details of the internal structure of the LoT, but the proposal that the representation of thought involves predication in one way or another will have to be, I think, a central feature of any account. But if you want to know more about it, do send some research funds my way and I will rewrite The Language of Thought all over again, though not quite, I promise, à la Pierre Menard.

This brings an end to the series on Language and Thought, and to series on series (at least temporarily). Starting next month, I shall write on more topical issues related to linguistics, philosophy, and psychology, each entry its own island, as it were.


[i] Claude Panaccio (2017), Mental Language, New York, NY: Fordham University Press, chronicles the history of the ‘old idea’, including Ockham’s contribution, though for the latter Paul Vincent Spade’s Cambridge Companion is a better source.

[ii] The standard phrase along the procedure employed to close a motion in Parliament can be found here: http://www.parliament.uk/site-information/glossary/closure-motion. Worth adding that, interestingly, Pete Wishart seems to have gone beyond the parliamentary formula, creating his own idiom – not so much that the question now be put but that the question be now be put. Indeed, Wishart is playing with the original wording of the idiom a little bit and appears to have taken the phrase be put to be sort of detachable from the whole, that the question be now put expression. That is, he is closing the motion in a slightly different way; namely, in putting the question to a vote, Wishart takes it that one might want to request that the question be put to a vote. Accordingly, the be put within the that the question be now put would be part of Wishart’s own idiom in addition to being the subjunctive verbal form within the more conventional parliamentary idiom. This would explain the presence of what seems to be an extra be, the second last. As a matter of fact, the sentence sounds much better without this extra be, as shown in (1) below, which I gather most people would consider unquestionably grammatical (even if nonsensical).

(1) Pete Wishart says he wants to put the question [that the question not be put [now [be put]]].

(2) ? Pete Wishart says he wants to put the question [that the question not be put [now [be [be put]]]].

In the original example, then, Wishart seems to have thought that, in addition to the subjunctive be, which the sentence must obviously have, the sentence can also exhibit a be put as a unit (itself a subjunctive form), given that, for Wishart at least, in order to close a motion in parliament one must be putting the question to do so. If so, the word be appears in two guises at the end of the original sentence: as part of a common parliamentary expression – the original idiom – and as an independent subjunctive. Technically, therefore, Wishart could have uttered (2) instead, considering that the adverb now can appear on either side of the verb. Needless to say, it would be very hard to convince anyone that (2) ought to be regarded as grammatical too (hence, the question mark in front of the sentence, one way linguists have of marking the dubious status of a given sentence).

[iii] I’m simplifying a great deal indeed, as I’m glossing over all the linguistic details of the Wishart sentence (i.e., the nitty-gritty of how the sentence is syntactically put together and how the semantic meaning is read off the syntax), and it is by comparing the linguistic details with the conceptual details that the mismatch is better exposed.

[iv] The difference in vocabulary does point to the fact that the atoms at play in conceptual representations are different to the atoms of language (concepts and lexical items are different things, after all), above and beyond the fact that we are dealing with an idiom. The linguistic and conceptual representations must connect at some point if what is meant (or thought) is to be said, but the details need not match up very closely.

[v] Another issue, related to note (ii) above, is the fact that the parliamentary idiom is manifestly more transparent as to its meaning – and, thus, as to how it relates to the underlying conceptual representation – than Wishart’s own idiom. That is, given the relevant knowledge of parliamentary procedure, the sentence in (1) in note (ii), which does not have the extra be of the original, is readily accepted as grammatical and its meaning easily computed in a way that the Wishart sentence clearly is not. Careful reflection may well help work out the actual meaning of Wishart’s idiom, this being an interesting issue in itself, but the sentence could very easily be rejected as ungrammatical. The extra be in the Wishart sentence is having an inordinate effect on comprehension, and in turn on how the overall linguistic representation connects to the corresponding conceptual representation. There really aren’t any extra bes in the conceptual representation, for the thought itself is something else altogether – these are additions at the “what is said” side of the equation that underlies the relationship between language and thought. It is this very fact that highlights the actual mismatch I have been focusing on, which generalises to many other cases, the non-idiomatic included, but that’s a story for a different blog.