For thousands of years, philosophers have debated the purpose of language. Plato believed that language was essential to thought. Thought, he wrote, is “a silent, inner dialogue between the soul and itself.”
Many contemporary scholars hold similar views. Beginning in the 1960s, MIT linguist Noam Chomsky argued that humans use language for reasoning and other forms of thought. “If language is seriously flawed, thought will be seriously flawed,” he wrote.
As an undergraduate, Evelina Fedorenko took Chomsky's classes and heard him explain his theories. “I really liked the ideas,” she recalls, but was troubled by the lack of evidence. “A lot of what he was saying was stated as if it were fact, or truth,” she says.
Dr. Fedorenko went on to become a cognitive neuroscientist at MIT, using brain scans to study how the brain produces language, and 15 years later, her research has led to a surprising conclusion: we don't need language to think.
“When we start to evaluate, we find no evidence to support the role of language in thinking,” she said.
When Dr Fedorenko began his work in 2009, research had shown that the same brain regions needed for language are also active when people reason and make calculations.
But Dr. Fedorenko and other researchers found that this overlap was an illusion. Part of the problem with the early results was that the scanners were relatively crude. The scientists made the most of the fuzzy scans by combining the results from all the volunteers and creating an overall average of brain activity.
In her study, Dr. Fedorenko used more powerful scanners and performed more tests on each volunteer. These procedures allowed her and her colleagues to collect enough data from each person to create a detailed picture of their individual brains.
Scientists then went on to work to identify the brain circuits involved in language tasks, such as recalling words from memory and following grammatical rules. In a typical experiment, subjects read nonsense sentences and then read real sentences. The scientists found that specific brain regions were active only when subjects processed real language.
Each volunteer had a language network — a collection of regions that become active during language tasks. “That's very stable,” Dr. Fedorenko said. “Whether you scan them today, or 10 years from now, or 15 years from now, it's going to be in the same place.”
The researchers then scanned the same subjects while they engaged in different types of thinking, such as solving puzzles. “While they were doing these thoughts, other areas of the brain were highly active,” she says. But the language networks were silent. “It was clear that none of these thoughts were engaging the language circuits,” she says.
In a paper published in the journal Nature on Wednesday, Dr Fedorenko and her colleagues argued that studies of people with brain injuries point to the same conclusion.
Strokes and other brain injuries can destroy language networks, leaving people struggling to process words and grammar — a condition known as aphasia. But scientists have found that people with aphasia can still do algebra and play chess. In experiments, people with aphasia can see two numbers (for example, 123 and 321) and recognize that 456 is followed by 654, using the same pattern.
If language is not essential for thought, then what is it for? Dr. Fedorenko and his colleagues argue that language is for communication. Dr. Chomsky and others reject this idea, pointing to the ambiguity of language and the difficulty of vocalizing intuitions. “The system is poorly designed in many functional respects,” Dr. Chomsky once said.
However, a large body of research suggests that language is optimized to communicate information clearly and efficiently.
In one study, researchers found that shorter, frequently used words make a language easier to learn and allow information to flow faster. In another study, researchers looking at 37 languages found that grammatical rules place words close together, making their combinations easier to understand.
Kyle Mahowald, a linguist at the University of Texas at Austin who was not involved in the study, said separating thought and language could explain why artificial intelligence systems like ChatGPT are so good at some tasks but so poor at others.
Computer scientists train these programs on huge volumes of text to uncover the rules about how words are connected, and Dr Mahowald believes they are beginning to mimic the language networks of the human brain, but that they fall short when it comes to reasoning.
“It's possible for a sentence to be very grammatically fluent, but the underlying ideas may or may not be coherent,” Dr Mahowald says.
But Guy Dove, a philosopher at the University of Louisville, thought Dr. Fedorenko and his colleagues had taken language too far away from thinking, especially complex thinking. “When we think about democracy, we might rehearse a conversation about democracy,” he said. “We don't need language to think, but language can enrich thinking.”
