Re: Language

From: Stevan Harnad (
Date: Tue Apr 29 1997 - 16:01:48 BST

> From: "Liz Lee" <>
> Date: Mon, 28 Apr 1997 09:59:52 GMT
> I'm having difficulty in persuading myself that Pinker is correct (and
> therefore Chomsky).

You need to distinguish what Pinker and Chomsky say about human
SYNTACTIC capacity in particular (a part of our language capacity) and
what they say about LANGUAGE in general:

The property of language that is very likely to be innate is our
syntactic capacity: our ability to recognise and produce all and only
the utterances that conform to Universal Grammar (UG). The reason this
capacity is very likely to be innate is that it is "underdetermined" by
what the child says and hears during the short time in which it
"learns" it (the "poverty of the stimulus"): The child hears and
produces only POSITIVE instances -- utterances that conform to UG. It
doesn't hear or utter NEGATIVE instances -- utterances that violate
UG. Since it only hears and produces utterances that conform to UG. it
must already know UG.

It's as simple as that. But UG is not simple; and that's the point.
Chomskian grammarians are still trying to "learn" the details of UG,
more than 30 years since it was discovered, yet every child already
"knows" UG without having to learn it (the child learns only the where
to set the "dials" on UG: the machinery itself, in all its complexity
is already there).

If the child cannot learn it, so must already have it, where could UG
have come from? This is an unanswered question. We are tempted to say
it evolved, just as the wings or fins or eyes did: UG is an inborn
organ like those. The trouble is that it is much easier to envision the
evolutionary process that would have "shaped" fins, wings or eyes,
driven by their adaptive advantages, than it is to envision the
evolutionary process that shaped UG.

Pinker insists (and no one would object) that language must have
evolved, like everything else. But in treating the details, he talks
about properties of language other than UG. UG itself is a tough case,
for the reasons I mentioned above.

> Why can't language be an evolutionary capacity (as Pinker and Bloom
> argued) that is also shared with other primates? Why must there be this
> innate "organ"- a LAD - which only we have? The evidence of apes'
> ability to understand language from people like Sue Savage- Rumbaugh
> seems overwhelming.

Because language involves many capacities, some of which certainly
evolved, some of which are shared with other primates and others of
which (UG) are not shared with other species, are not learned, and
are not readily explicable as having been shaped by Darwinian

The problem is with speaking about "language" in general. That's like
talking about "behaviour" or "cognition" in general: We need to specify
the bit we have in mind. "Modularity" is probably too fashionable these
days, but UG seems to qualify as a module which is part of language,
but a part no other species shares.

But even forgetting about grammar, the chimps that do the symbolic
communication don't seem to be be doing the same things we do, if for
no other reason than that we can't carry on a conversation with them:
You can talk to any child, whether hearing (in speech) or nonhearing
(in sign language). Through language, they can say and understand any
sentence in, say, English. "Yerkish," one of the artificial "languages"
that the Chimps use is just a finite set of symbols associated with
certain objects, acts, or wishes. The ability to use them that way is
certainly something interesting, and may have been one of the prior
cognitive capacities on which language drew, but it isn't language.

One of the interesting features of natural language is that
it is all-or-none: There is no language in which you can say only some
things, but not everything that you can say in (for example) English.
(A language may lack the vocabulary, of course, but then it just needs
to be coined: You can't increase the scope of "Yerkish" that way.)

> Pinker suggests that apes don't 'get' the point of language, whilst it
> may be that they are unable to converse for its own sake, couldn't the
> human ability to do so just be because we have a more evolved language
> capacity, in the same way that some mammals have a more evolved
> swimming ability than we do?

Like Humpty Dumpty, we can declare that a word means whatever we mean
it to mean. So, for example, we could say that what we mean by flying
is not only what birds and planes do, but also what people do when they
jump -- it's "flying," only less evolved than in birds or planes!

Or, to pick a more cognitive example, we can say that animals can do
arithmetic, only a less evolved kind of arithmetic, because they can
only "count" to 4 or 5.

(Now this is becoming a bit controversial, because there are forms of
brain damage in which humans are reduced to being able to count only to
4. See
There are also forms of brain damage in which language is disturbed
and it's not clear exactly what capacity they have left.)

What Pinker means by chimps not "getting" it is that we can't
understand, given what they CAN do (understand a finite number of
symbol/object pairings, and use them to communicate), why apes can't
go on to do all the rest of what we can do.

(My own cliche for this is: If they can pick up the stick, how come they
can't run with it?)

> Premack showed that apes who had been taught to use a basic language
> (plastic) were better able to complete tasks than apes with no such
> training, so our use of language over the last (disputed) 30,000 years
> or so has increased our cognitive abilities and (possibly) allowed us
> to expand our vocabulary and use syntax. Surely there is no evidence
> that the earliest language speakers used syntax?

If you train apes on symbolic tasks, they'll do better on other symbolic
tasks. But the symbolic tasks were not language. (Or, to put it another
way, connecting a symbol with an object or state is not language, though
it is drawn upon by language.)

Those cognitive capacities that Premack tapped are probably
among the precursors of language (just as inborn animal signals
are probably also precursors). But by this token even hearing
and seeing are precursors of language!

Syntax, as I said earlier, is a special case. A trivial,
arbitrary syntax (say, rules about the order of use of symbols) could
probably be trained. Because there are so few symbols, the rules about
order could be memorised. And you might want to call this a precursor of
UG (in which case walking -- left-right-left... -- is also a precursor
of UG). Animals have certainly never shown signs of knowing UG;
nor are they likely to be able to learn it, given that we can't learn it

There could, by the way, be a perfectly good artificial language, one
that could say everything you could say in English -- and could do it
without UG! The prediction then would be that children could never
learn this language in place of a natural one. That means that UG
is necessary for human language. (Once you have UG, you can learn other
symbol systems with non-UG syntax as a second "unnatural" language;
you could even go on to "learn" UG explicitly, which is something that
the child does not and cannot do, but teams of MIT linguists

The evolutionary story about the origin of UG is a problematic one.
It is easier to make some plausible guesses as to what might
have shaped us into language-users; why it had UG syntax is another
story. No plausible account has come so far.

> We know, from studies of deaf children, that those not given access to
> sign language do not just develop it by themselves, they need another
> person as a facilitator, so the capacity to use language is not the
> same as the instinct to walk after one has mastered crawling.

Right, it's an interactive, communicative capacity. I'm not sure
it's correct that deaf children don't develop a sign language between
themselves, though. They just wouldn't develop it if they were on their
own and spoken to by others only orally: they would then learn to

> The more I read, the less sure I am about this "unique" human
> ability.

I think you are mixing up the continuities and precursors that there
will always be, for any ability, with the ability itself. Sometimes the
boundary is arbitrary (whether we consider knuckle-walking apes to be
bipedal -- walking on two rather than for limbs) but sometimes it's not:
with language as a whole, and with UG and a part, the boundary is
not arbitrary: for each you either have it all, or not at all.

This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:24:17 GMT