One of the first steps to learning a language is figuring out where one word ends and the next one begins. Since fluent speakers don’t generally pause between words, it can be a daunting task. We’ve discussed one of the ways people do it in this post — they focus in on consonant sounds. Other researchers have found that we also focus on the statistical properties of language.
Certain syllables are likely to follow each other within individual words, but unlikely to follow each other between words. Take the phrase “between words.” In English, within a single word we’re much more likely to hear bet followed by ween than ween followed by wor.
Researchers have found that if you make up nonsense words like gimysi and mimosi and play a constant stream of these words to listeners, the listeners will eventually figure out the boundaries of the words based solely on the statistical properties of the words.
But still, it can take a long time to pick up the word boundaries. A team led by Daniele Schön invented just six words: gimysi, mimosi, pogysi, pymiso, sipygy, and sysipi, and after seven minutes of listening to these words repeated in random order, student volunteers couldn’t distinguish between them. It took over 20 minutes for listeners to learn where one word started and the next one ended.
Schön’s team suspected that singing the words might improve listeners’ ability to parse them. After all, mothers often sing to their infants. Perhaps one purpose of singing is to help children learn language faster. In a second experiment, the researchers assigned a unique pitch to each of the syllables used in their six words (gi was C5, my was D5, and sy was F5, and so on). A speech synthesizer played back the words in a sing-song fashion, with a musical note assigned to each syllable.
After listening to the words for seven minutes, the volunteers were tested. They heard three-syllable “words” from the original list and partial words composed of fragments of real words (for example, mysimi, made from gimysi and mimosi). How accurate were listeners at identifying the original words? Here are the results:
The dotted line in each graph represents the average score for all listeners, and each square is the average score for an individual listener. As you can see, in the speech-only experiment, listeners did no better than chance. But in the second experiment, nearly everyone did better than chance, and the average score was 64 percent correct — significantly better than chance performance. Simply associating each syllable with a musical note improved performance.
But in real songs, syllables aren’t always matched with the same notes. Sometimes different syllables get the same note, and sometimes the same syllable is sung with a different note. In a third experiment, Schön’s team allowed the notes to vary with each syllable. Again, listeners could identify words at a rate better than chance (though they weren’t as good as in the second experiment).
Schön and her colleagues don’t go so far as to argue that music is a requirement for learning language, but they do make the case that the extra information provided in music can facilitate language learning. They also suggest that other information, like gestures, might be equally helpful for learning a language.
But there is additional evidence suggesting that music plays an important role in language. Similar areas of the brain are activated when listening to or playing music and speaking or processing language. Language and music are both associated with emotions. And of course, we know that children — especially small children — really like music. This study offers another bit of evidence that the link between language and music may be a fundamental one.
D Schon, I Peretz, M Besson, M Boyer, R Kolinsky, S Moreno (2008). Songs as an aid for language acquisition Cognition, 106 (2), 975-983 DOI: 10.1016/j.cognition.2007.03.005