Thursday, December 20, 2007

Lack of data vs. Poverty of the stimulus

NewScientist recently published an article about the subconscious that included the following claim:
"Infants do not need tutoring to acquire their native language; they pick it up subconsciously. What's more they do this with remarkably little linguistic data - what the Harvard University linguist Noam Chomsky has called the "poverty of stimulus" - suggesting that this subconscious learning allows youngsters to use information very efficiently."

The nativist language acquisition argument from the "poverty of the stimulus" seems as though it may have been misconstrued. It is not a quantitative argument but a qualitative one. Miller and Chomsky claimed in 1963 that children's input included many grammatical errors and disfluencies. How were children to discover which input to attend to and which input to ignore?

More recent arguments suggest that, errors aside, given the lack of negative evidence, it is hard to see how children could learn what is ungrammatical. We have evidence for children overapplying rules such as regular past tense endings. Given that, it would be logical to assume they would overapply other patterns such as the following:

  • She sent a letter to him. / she sent him a letter.
  • She explained the letter to him. / *She explained him the letter.

Without evidence that this last construction is not possible, the 'poverty of the stimulus' argument goes, how are children to know that they merely have not yet heard it?

The idea that children simply don't get enough input, however, is odd. Research suggests that average North American children get somewhere in the range of 26 million (plus or minus many million) words of input in their first four years of life. It's hard to see how this can be called "remarkably little linguistic data."

No comments: