Re: Skinner on Language Learning

From: HARNAD Stevan (
Date: Mon Jun 03 1996 - 21:01:32 BST

> From: "Laud, Emma" <>
> Date: Thu, 23 May 1996 13:12:32 +0100 (BST)
> Skinner suggests that we learn language through being shaped. This is
> done by trial and error and reward feedback. An example we can look at
> is when a child babbles and imitates gramatically correct words it
> hears. When the child repeats this word correctly Skinner says they are
> rewarded through positive feedback.
> However, a child hears and sees too little to learn syntax (the rules
> based on arbitary shape manipulation) through reward and punishment. A
> child can not possibly learn every possible combination and permutation
> of gramatically correct sentences - there are just too many.

The problem is not to memorise every possible combination, it is to
learnt the RULE that generates only the correct ones, and none of the
incorrect ones. To learn rules like that, one needs positive and
negative instances, cases that are right and cases that are wrong.
The child does not produce or hear cases that are wrong, so it cannot
learn the rules. Yet it can soon speak correctly. Hence it must have
been born with the rules.

> A child
> does know how to manipulate words and symbols in a sentence in order to
> make it gramatically correct even though it may never have seen the
> sentence before. So if this ability is not learnt through reward and
> non-reward as Skinner suggests, then it must be a process we are born
> with, which then leads us to look at Universal Grammar, (the ability to
> recognise correct/incorrect symbols). If we are born with this ability
> it does indeed put hold to Skinner's theory that we learn all that we
> know in language through shaping by trial and error and rewards and
> non- rewards. It may be so that we learn new words/symbols through
> Skinner's explanation of language learning but we are born with the
> ability to manipulate symbols and not as Skinner suggests.

For an A, you'd need to discuss Universal Grammar and negative
instances, and make sense of it all together.

This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:43 GMT