TedEd: How do our brains process speech?

A TedEd by Gareth Gaskell

We learn a lot of words over the course of our lives. By age 20, we know as much as 52,000. By age 60, we average between 35 and 56 thousand. Most words take less than a second to say out loud, and when we hear them, our brains have to make a split second decision on which word, out of the many thousands, it is. We are correct about 98% of the time. So how do we do it? One important part is that our brains can do many activities at the same time. Many scientists believe that each word in our brain is connected to a different neural network that identifies the likelihood of heard speech matching that word. These neurons are in the brain’s cortex. When we hear the beginning of a word, many of these units become active because many words can have similar beginnings. As the word goes on, the words that are missing crucial parts begin to go inactive, eventually narrowing down to one word. This often occurs before the end of the word, and the point at which one word is selected is called the recognition point. Most people can recognize up to 8 syllables per second. We also need to know the meaning of the word, and our brain accesses many different words/meanings during the word before settling on the correct one. Recognition happens more quickly when we are given context, especially with words that have multiple meanings or homophones. For multilingual people, they must also match the language they are hearing in order to eliminate incorrect words or meanings. But what happens when we learn new words? What prevents new neural patterns from erasing old ones? Scientists believe that to avoid this problem, new words go first to the hippocampus, the memory center of the brain, instead of the cortex so that words can’t overlap. Then when we sleep, the new words gradually transfer over with the old words, becoming part of our neural network.