When I planned to write this, I wanted to start by saying that language—our ability to speak and write in a form that others can understand—is backed by a non-linguistic form of thinking. The idea is that language, the vocabularies and grammars that comprise a particular speech, are not the basis of thinking but rather a form of signaling invented specifically to communicate with others. After some additional reflection, it seems the situation is more complicated than that. It seems that some kinds of thinking require a linguistic basis.
I think most of us have had the experience of trying to say something but the exact word we want to use won’t come to mind, so we pause looking for the word. Colloquially, we may even excuse this pause by explaining that the word “is on the tip of my tongue.” Several frustrating moments pass while looking for that word that we know so well but can’t recall. Maybe we think of it. Maybe we give up and use some other word or phrase as a crutch that doesn’t quite do the job we intended. But this phenomenon is very revealing. If language were the basis of thought, how could you have a thought to express without already having the word in mind that you want to use? Clearly the situation indicates that expressing a thought, and the thought itself, are two different matters. The only way to be looking for a word to fit a need is if the need is already clearly in mind. It also suggests that there is some sort of search process intervening between a thought and the steps involved in producing the sentence that expresses it.
There are other hints that we have an inner mode of thinking that is pre-linguistic. Some people are bilingual or multi-lingual: they know and can speak more than one language. A beginner learning a new language starts by thinking in his native language and then translating to the new language. He hasn’t fully internalized the new language yet. But people who are fluent in two languages can report that they can think in either language. There is a conscious decision to speak in one or another language, and then from that point forward, they think normally but gather words and structure from the language they’ve chosen, and it all happens rather effortlessly. Speaking in Spanish, for example, is just as easy and normal as speaking in English for someone who is fluent in both languages, and there is no translation step. You just produce sentences in the desired language. This type of process is easy to understand if you suppose that thinking is the same regardless of the language and there is some final step of generating sentences that intervenes between thinking and speaking.
Looking deeper, it seems that a thought is a highly structured kind of object. I think of it like an old-style wagon wheel with spokes. There is a hub that represents the entire thought. It has no content itself; it’s merely an anchor for connecting concepts. Individual concepts are associated with the thought by creating a link, a spoke, to connect it to the hub. One of the spokes of the wheel connects a subject of the thought. Another spoke associates an action with it. Other spokes might associate an object receiving the action. Additional spokes can associate modifiers with the thought that changes the weight or sense of another spoke. When this thought occurs in the mind, it appears all at once, instantaneously. There is no process of putting the thought together that we can sense; rather, when you have the thought, all of its parts are already there.
I’m not suggesting that thoughts are physically represented in the mind as wheels of hubs and spokes. This is a symbolic way of speaking. Thoughts occur in the mind in a way that is as if it consists of a unit handle with subordinate thoughts attached. It’s a flexible structure that allows a very complicated idea to be assembled and worked with as if it were a single unit.
To speak the thought to an audience, it is necessary to take this fully formed, structured idea and translate it into a linear series of verbal elements. First the sentence must be planned using a basic grammar. The speaker decides whether it will be subject first, or a passive voice with the action first. The subject is then assembled by looking up the word in the language you’re using that names that subject, adding any necessary modifiers, and then speaking it. The process continues, assembling subsequent parts of the sentence until sometime later it has all been pushed out. In actual structure, a subject can be a secondary hub-and-spoke thing that anchors a complex subject, for example “a big cat with stripes and a black face.” In your mind you have an image of the cat, but you can’t say that all at once. Instead, you have to take that idea, the picture, and describe it to the audience using words. The process is complicated, time consuming, and requiring numerous references to memory, and this process is not perfect. If you speak too fast, you can run out of planned sentence parts before you’ve found all the words you need, so you have to pause. People throw out filler sounds like “um” and “ah” to bridge the gaps until sentence assembly can catch up. Sometimes a speaker will change the plan of the sentence midway through it, get confused, and say something that makes no sense.
The architecture of thoughts can be quite complex. The subject could be a secondary hub with spokes that associates a number of different ideas modifying the meaning, like the way a house can be a red house, with a porch, and a garage, and a second story with dormers. Similarly the action can be a hub and spoke arrangement that pulls together a number of different ideas to represent a detailed, complicated action. The more elements in this hierarchically structured thought, the more words it will take, and time it will take, to produce a sentence to communicate it.
This general scheme supports the experience we have of thoughts that pop into the mind whole, already full of references and meanings. In general, we can say the mind thinks in “meanings,” not in words. Pictures would be a type of meaning but other types of meanings can be more abstract. In any case, by thinking in meanings, the complicated business of assembling sentence structures can delay all that work which has nothing to do with the thought itself, and it also delays the choice of which language to use to say it. Such a model reflects the difference between the intent of the thought itself and the complicated linguistic details of creating sentences to express it; by keeping them separate, thinking is faster, simpler, and more economical.
There is one aspect of this puzzle that I haven’t worked out yet, and it is this: Why do people talk to themselves silently, in the mind? If it’s true that thinking is mostly pre-linguistic, which is supported by the kinds of mistakes we make in speaking—the defects in the process illuminate some of the nature of the process—then it would seem that thinking in words would be unnecessary, just adding extra work to the business of thinking. And yet there are problems where some sort of verbalization process in the mind seems essential to working it out. I would like to explain why such a verbalization process helps, but I don’t quite see the answer yet.