Over the past week, I have looked at quite a lot of research for this blog and, figuratively speaking, put quite a lot of it back on the shelf. Not because it was bad research. But because I often found myself thinking: we largely knew this already, and I have written about studies confirming this idea before. But this study by Hope Kean and colleagues immediately caught my attention. Although I quickly realised that it may disappoint some education die-hards. The practical consequences are very limited. But please, do read on, because the story is more than fascinating. It touches on a very old question: do you need language to think? For logical reasoning, at least, the answer seems to be no. And the researchers tested this in an extremely interesting way.

The debate?

Part of my personal fascination with this study comes from the fact that the question of whether we need language to think is about as old as thinking about thinking itself. Aristotle already distinguished between words and the mental processes to which those words refer. Many centuries later, the idea of a language of thought emerged. This is the idea that our thinking has a kind of language of its own, consisting of elements that are combined in a structured way. Quite logically, this raises the question of whether this supposed language of thought is also our ordinary, natural language.

And that is less obvious than you might think. We can, after all, express complex ideas using mathematical symbols, diagrams or, as I like to do myself, even music. And here is a sentence to chew on: the fact that a thought can be put into words does not necessarily mean that the thought itself came into being in words.

There have also been indications for some time that language and thought can, at least partly, be separated. This is where people with aphasia become particularly interesting. These are people who have suffered brain damage and, as a result, developed a language disorder that can make speaking, understanding, reading or writing difficult. But… despite this language disorder, they can sometimes still reason mathematically or causally and understand what other people are thinking. Previous brain research has also suggested that mathematical thinking, programming and social reasoning do not simply rely on the brain areas we use for language.

But for what is perhaps a crucial form of thinking, the evidence was much less clear: abstract logical reasoning. And that is precisely what Kean and colleagues wanted to investigate.

How did they study this?

The researchers chose a combination that I find ingenious in its simplicity. First, they had healthy adults perform different reasoning tasks while their brain activity was measured using fMRI. For each participant, the researchers first identified individually which brain areas belonged to the language network. The participants were then given various logical problems to solve.

For inductive reasoning, for example, they had to infer the rule being applied to a series of numbers, somewhat like the problems you may know from intelligence tests. For deductive reasoning, the researchers gave the participants syllogisms. They had to determine whether a conclusion logically followed from the premises. The tests also included matrix problems involving geometric shapes, similar to those you might encounter on intelligence tests.

The reasoning behind the experiment is simple. If we use natural language to reason logically, one would expect the language areas in our brains to become more active as the logical reasoning becomes more difficult.

That barely happened. The language areas responded strongly, as expected, when people processed language, but much less or not at all when the logical demands of the tasks increased. There was a small effect for inductive reasoning, but the effect for language itself was more than four times as large. For deductive and matrix reasoning, the researchers found no significant effect in the language network.

Step 2?

But the next step provided even more evidence. The researchers found two people with very severe aphasia resulting from extensive brain damage. Their language abilities were seriously impaired. On several grammatical tests, they even performed around chance level, meaning that on some tasks they did barely better than they would have done by simply choosing an answer at random.

If language is necessary for logical thinking, you would expect these two people to have enormous difficulties with the reasoning tasks as well. But no. The opposite was true.

On the inductive task, one solved 19 out of 25 rules and the other 39 out of 40. Their performance did not differ significantly from that of people without aphasia. On the matrix problems, they scored 25/30 and 26/30. That was actually considerably better than the average for their age.

Something else interesting happened in the brain scanner. Inductive reasoning and matrix reasoning mainly activated the so-called multiple demand network, a network we use for all kinds of cognitively demanding tasks. But deductive reasoning using syllogisms did not even do that and appeared to activate yet another set of brain regions. So there may not even be one general reasoning system sitting alongside our language system. Different forms of reasoning seem to rely partly on different systems.

So what does this actually mean?

First, to make sure we don’t accidentally create a new myth. This does not mean that language is unimportant for thinking. What the researchers do show quite convincingly is that natural language does not seem to be necessary for adults to engage in formal logical reasoning. What this language of thought actually consists of, we still don’t know.

And here comes the possible disappointment for some people in education. You cannot conclude from this that language is unimportant for learning. Language can support our thinking and may also be important when learning abstract concepts. This study was conducted with adults and therefore tells us nothing about whether children need language in order to learn to reason logically. The researchers explicitly identify this as an open question.
There are limitations too. The brain-imaging study involved 29 adults, while the spectacular part involving people with severe aphasia involved exactly two people. Fortunately, people with such profound aphasia are relatively rare. But this limited pool of potential participants is also less problematic than it may seem. If language really were necessary for logical reasoning, two people who can barely process grammatical language but can still reason extremely well already pose quite a difficult problem for that claim.