We've all caught ourselves projecting phrases into our mind's ear, such as phone numbers you've just been read - "072255...", snatches of dusty arithmetic -"7 times 8 is 56", rhymes - "30 days hath September...", and the like. It is little auditory loops such as these, together with the absence of an alternate hypothesis, that convince most of us that we think "in English" (or whatever). But do we? If we believe that "earth is a planet", is there a bit of the brain dedicated to "earth", connected sequentially to another one for "is", followed by one for "a" and finally one for "earth", all the while keeping track of which part is subject vs object vs indefinite article, and so on? This sounds like a stretch, but then what else could be going on?
Well, it may surprise you, but we don't think in any earthly language. All we need to show to refute the hypothesis is that some English sentences, while perfectly intelligible, don't contain enough information to cut it as a language of thought. Let me explain:
- Ambiguity - Real newspaper headlines such as "Chinese Apeman Dated", or "Stolen painting found by tree" strike us as funny because they unintentionally have a second (bizarre) meaning. But we realise that the mistake was inadvertent; the writers knew which meaning of "dated" or "by" they meant. But if an English phrase can have more than one meaning, then English can't be the medium of thought.
- Co-reference - In your history essay, you may begin by referring to "Adolf Hitler", but then later make reference to him with phrases like "Hitler", "the Führer", or "the leader of the Third Reich". As long as you write reasonably clearly, it is perfectly clear that you are referring to the same person - yet the language has changed! Once again, this means that there is something in the brain that is treating the phrases as the same thing. And that thing can't have been the language, which was as fickle as last season's fashions.
- Deixis - Deixis is what linguists name those parts of language that are intelligible only in context. Take an old joke like "Every four seconds, a woman gives birth. She must be found and stopped." This only works because the first part of the sentence leads you down one interpretative alley ... and then the second corrects it. Again, this shows that the phrases themselves don't inexorably tie themselves to one - and only one - meaning, as would have to be the case if English really were the language of thought. Rather, there is some underlying (non-English!) interpretative function beneath the language that the joke is able to temporarily thwart.
There are several other proofs too, but I think these will suffice. So if we don't think in English, in which language do we think? The answer from cognitive neuroscientists and others is that we think in an fairly abstract, subconscious "language" often called Mentalese. Calling it a language is a little confusing though, since we've just proved that it doesn't use "words" in the everyday sense of the term. Nonetheless, there must be some sort of information processing going on inside my cranium (using concepts and objects instead of words); one that obeys a specific logic ("grammar") that is robust enough to grant us "intelligence". For want of a better term then, we all think in Mentalese.
The above ideas were largely taken from Steven Pinker's masterly "The Language Instinct". Read it.