Babies and Language

Babies are ravenous language learning machines. Much of this is hardwired, not in the sense that a language is hardwired, but that language in general is built-in.

The sounds we make in speaking are phones. There is a wide variety of sounds the human can make with with many variations in between. Up to a few months old, babies are very good at distinguishing different phones. I wonder if they can beat linguistics students.

In language we consider some distinct phones to be the same. The are different sounds, but we often cannot tell them apart. These are allophones, and the collection is called a phoneme, designated by the anchor allophone. Some allophones occur in certain speech context, others are free. In a sense, the aggregation of phones into a phoneme is a form of error correction.

And here is where it gets weird. Babies shift to hearing phonemes, not phones. That is, many phones are the same. The performance in distinguishing phones goes down in the 5-month to 14-month period. However, the ability to recognize sounds as being in the same phoneme in the context of a particular language goes up. In a sense, some of the bits of information in hearing phones are converted to error correcting bits.

With this, language acquisition skyrockets. Between 1 year and 3 years, the child is learning language at a high rate. At 30 months a child is learning at a rate that is very hard to duplicate in adults.

What about learning two languages? Will learning one, handicap learning the other? Surprisingly, this is not the case. The context is part of the language rules. When speaking to grandfather, one might speak one language, but when speaking to strangers one might speak the other. There is a blip of a slowdown at around 20 months, but by year 3, there is no difference.