Re: Chomsky vs. Skinner on Language

From: Stevan Harnad (
Date: Fri Feb 23 1996 - 19:16:55 GMT

> Date: Wed, 21 Feb 1996 12:12:39 +0000 (GMT)
> From: "S.J.Petrie" <>
> Skinner claimed that behaviour was based on a reward and punishment
> basis; that is that people only do things when there is something in it
> for them.

Not quite. He said that what they DO do has been (largely) shaped by
reward much the way a rat is shaped to press a lever for food.

(This does not rule out doing something for no particular reason, or
because you were born that way. It's just that most behaviour is the
result of a prior history of rewards.

> Language is also a behaviour so it can be based on the same
> principles. Children learn a language by reinforcement from their
> parents so eventually their use of it is pretty skilled by the age of
> four.

Let's be more specific: We were talking more about grammar than about
vocabulary. Chomsky found out that the REAL grammatical rules that give
us the capacity to recognise what it and is not grammatical, and to
produce all and only grammatical strings of speech are very abstract
rules that we do not know consciously (indeed, linguists are still
trying to work them out). Yet the child of 4 years old already "knows"
them, because he can speak.

Skinner would have said that, whatever those rules might be, the child
must have learned them by having been rewarded for saying things
grammatically and not rewarded (say, ignored) for saying things
ungrammatically. Then gradually, like the rat learning to press the
lever, the child learns to produce only the grammatical strings of words
and not the ungrammatical ones, according to Skinner.

> Language acquisition is shaped by its consequences. For example,
> a child asks for things by trial and error until he or she is
> successful in getting what he or she wanted in the first place.

Right, except we are now supposing that the child says a lot more than
just "may I have that" -- all the things a child can say, in fact, and
is rewarded for saying them correctly by being understood and replied
to, and by failing to be understood or replied to when he says them
incorrectly. By trial and error and feedback from the consequences, the
child learns the right grammar.

> However, evidence seems to suggest that knowledge of a language is
> innate; everyone already comes born with it. This was proposed by
> Chomsky. Somehow children are equipped with a knowledge of syntax or
> grammar, universal to every language;

True, but this was Chomsky's CONCLUSION form the evidence, not just an
assumption he made. It turns out that grammar HAS to be innate, because
the trial and error evidence the child gets -- i.e., everything he says
and hears till age 4 -- is insufficient for the child (or anyone) to
derive the rules from it by trial and error.

> there are certain things that
> every language has like a subject, object and a verb.

Yes, but that could be learned. That is not the kind of abstract rule
that Chomsky found was true of all languages and yet not learnable from
the data any child encounters. The example was that you can say "Who did he
hit that went out?" but you can't say "Who did he think that went out?"
gives you the flavour of the kind of abstract rule we all "know," yet
we could never have learned from trial and error (because no one ever
says or hears "who did he think that went out" -- except in a
linguistics seminar on Universal Grammar!)

(By the way "Who did he hit that went out?" is also ungrammatical in the
old, non-Chomsky way. It should be "Whom did he hit that went out?" But
that's not the kind of grammatical rule at issue here, because that rule
CAN be (and is) learned by trial and error and instruction and

> He argued the
> "Poverty of stimulus" view concerning only grammar. At a young age,
> grammar is impoverished so the child begins to pick up more with
> practice; grammar has evolved until it sounds better.

A bit garbled here: The "stimulus" is what the child HEARS (and also
what he says, actually). There are the things that are either the trial
and errors attempts to say something grammatically (like the trial and
error in the rat, gradually being shaped to press the lever) or the
feedback that indicates whether the trial was successful or unsuccessful
(the child should either not be understood if he says something
ungrammatical, or should be corrected, like the rat: but the child never
makes the mistake, and hence never gets corrected; so he must already
know right from wrong for rules like that).

It is the stimulus that is impoverished, not the grammar.

> There is no
> particular rule with grammar. If someone tried to work it out from lots
> of examples there wouldn't be one; there doesn't seem to be even one
> constant rule.

No. There are rules. It's just that you can't LEARN them from examples,
because the only examples you ever hear are the "positive" ones -- i.e.,
grammatically correct sentences. You never either produce or hear any
violations of universal grammar (UG) that would allow you to learn UG
from them. So you must already know UG.

> Everyone already has grammar rules implanted into their
> brain so where does it come from? An example of it already being there
> is when a child can say a new sentence out of the blue, not hearing it
> from its parents. He or she doesn't just repeat sentences parrot
> fashion but produces new unheard of expressions. That is what's so
> strange; they already know how to string a sentence together.

No, the fact that a child can make new sentences it has never heard
before is not what is so strange. If the child learns, by trial and
error, that "Please, may I have a X" will get him an X, then he can do
this with any X, even in sentences that he has never heard. If a child
hears that you can use "and" to combine two things, he can make new and
long sentences he has never heard stringing things together with "ands".
MANY rules of grammar are like that; they are learnable, and will allow
you to say completely new things.

The strange case is the one where the rule itself cannot be learned from
examples (because the examples are too impoverished, they are missing the
"negative evidence" from getting the rule wrong, and having it

> Can you define language for me again, I didn't quite catch it? Is it
> something like a language is a device which can say anything with
> words?

Remember it wasn't a definition but a criterion (sort of a test for
whether something is or is not a language): There seem to be no
"primitive" languages. Language seems to be all-or-none: If you can say
something in it, you can say anything in it. So the criterion is that
if you have a symbol system that can be translated into and out of any
natural language, then it is a language. The criterion is
"intertranslatability." A related criterion is "effability"
(expressibility in words): All language can express any proposition
(any statement about anything). If you think otherwise, tell me
something that cannot be expressed in words...

This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:24:15 GMT