Re: Evolution of Universal Grammar (UG)?

From: S.Harnad (
Date: Fri Feb 13 1998 - 17:16:15 GMT

> From: "Jon Wright" <>
> The question says "why is UG a problem for evoluionary theory and
> why can it not be treated like any other evolved trait?" It is a problem
> for learning theory because of poverty of the stimulus and the rapid
> development of language after 2years old. Not a problem.

Yes, learning theory can simply back off and grant that -- apart from
the parameter setting -- grammar is simply not learned; it's inborn.
So this is not a problem for learning theory.

> Evolutionary theory would say that 5% of an eye is better than
> no eye at all. No doubt, so within the variation possible in the design
> of an eye (or light sensitive patch) whatever is going to be more
> sensitive will have a differential effect on the reproductive ability of
> the vehicle and the eye evloves further. The eye has a complex design
> and could not have occurred by a chance recombination of atoms/genes.

Correct. So the complex structure of the contemporary eye can be
understood as having been shaped, mostly gradually, perhaps in spurts,
by evolution in the usual way.

> UG is more complex than it would need to be if it were designed
> from scratch, but it is not. It builds on the adaptations of the last
> generation of speakers and is refined.

Here I can't follow: Are you suggesting that UG was shaped gradually by
evolution the way the eye was? But what advantage did all those rules
convey? Why was it not much simpler? There are many other forms a
useable grammar could take: what was the advantage (whether overall or
in the small steps you are imagining occurring from one generation of
speakers to the next) of UG over a much simpler grammar, perhaps even a
learnable one?

> Language is flexible, which is
> why there would be no point in programming in a full vocabulary since
> the child might not be able to talk with it's mother and community.

Yes, but there's no problem with vocabulary: It's learnable and it's
learnt. UG is not learnable (based on what the child hears and says) and
it is not learned. It is already there, a complex detailed structure.
How and why did it get there? What was its selective advantage? What
were the alternatives? And how did it get shaped? No understandable
story like the story about the eye suggests itself. THAT's the problem.

> Animals do not have language, in the sense that they cannot
> communicate about things which are not there, referring to them with
> arbitrary symbols (detached representations - words/motions).

Probably not, though that story is not over yet. But that has nothing to
do with UG.

> The start
> of meaningful communication involves more than the present cues of the
> environment the use of symbols. As soon as this starts, the evolution of
> 2-word sentences might begin: simple subject/object and verb. This is
> the beginnings of grammar. ("protolanguage" - creole, chimpanzees
> signing, Genie and wild children, two year olds)

So far that's fine. Word order conventions are learnable. (By the way,
there are no "protolanguages." All natural languages can say anything
any other language can say. And animal communication and symbol systems
-- or at least their users -- cannot. )

So don't mix chimps and humans. And creole has full-blown UG, not part
of it. Children learning language (not Genie, who missed the critical
period) eventually exhibit UG, but they do not learn it.

> If 1% UG made a difference to the communication between animals
> in that they were able to string grunts together to mean something, the
> difference would show in greater reproductive success.

What is 1% UG? To tell even a just-so story about its "advantages", you
have to show the specific advantages of UG. The advantages of using
sounds and their order to mean something are very clear, but not the
advantages of UG. (A little word-order is not a little UG.)

> Grammar and syntax arise to make sense of the strings of
> arbitrary symbols.

Some grammatical rules are needed to use strings of symbols as
propositions that can be true or false and that can be used to get and
give the information we convey with sentences. But it's not at all
obvious why the grammar of propositions should be the complex and
baroque one of UG. What was the advantage of that?

> Processing becomes quicker and more efficient. All
> languages discovered are, by the nature of being a language (the ability
> to express concepts),

And propositions. Especially propositions. We have a concept of a "cat"
amd of a "mat," but it takes a proposition to say that "the cat is on
the mat." A concept cannot be true or false. The concept of "witch" is
not false. What is false is the proposition that "there are witches."

> complete with grammar.

Yes, but why UG rather than something much simpler that could do the job
(propositions) just as well, but could also be learned?

> It is a fundamental feature
> of humanity. The search for universals may begin to uncover the basis of
> UG. UG could have evolved in to the complex thing it is because of the
> ingeniousness of language users to create new forms and eradicate old
> ones within even tens of generations.

Can you give examples of what would evolve into what, and why?

Could you say what it is about starred sentences that makes them
"wrong" and unstarred sentences that makes them "right"?

Who did he think went out?
Who did he thank that went out?
*Who did he think that went out?
*Who did he thank went out?

It is not enough to assert that something must have had an adaptive
advantage just because it's there. You have to give some idea of what
that advantage might have been. None of the evolutionary advantages of
LANGUAGE are at issue; they are not the problem. The problem is the
evolutionary advantages of UG.

> To explain UG as an evolutionary adaptation is not difficult.
> (It might be a just-so story but) what is does is enable to combination
> of strings of symbols into complex sentences expressing feelings,
> desires, danger, location of food, a good cave nearby to shelter in.

You have just described the evolutionary advantage of language; that is
not in dispute, and quantities of just-so stories can be told about its
origin and usefulness (the bow-wow theory, the yo-he-ho theory, the
poopoo theory, the ding-dong theory, Babbleluck, etc.). The problem is
specifically with UG.

> UG evolved into the complex (and indefinable) form it is in because of the
> many ways of solving the adaptive problem of how to string words
> together meaningfully. The eye evolved from a patch, generating a
> lens and so on.

Does your UG story not strike you as remarkably vague, compared to the
eye story? Remember to focus on UG, not language as a whole.

> UG evolved like an eye EXCEPT THAT it is intangible?

The fact that it is not a visible structure like the eye or heart is not
the problem. The underlying complexity of our capacity to run is not
evident to visual inspection either, but buried in neuronal mechanisms
that we are only now mapping. So the fact that the "structure" is not a
morphological one, as it is with organs like hearts and eyes, is not the
problem. There is an evolutionary story to be told about our ability to run,
and how it evolved, even though the neuronal basis is "intangible."

> it does not use its constituent parts? (piggy-backing?)

No one has yet found some other function on which the complexity of UG
may be piggy-backing. (They've tried it with language as a whole,
attributing it to hunting, throwing, tool use, tool making, etc., not
very convincingly -- rather like attributing vision to a special form of

> it is all or nothing? (only humans have full language)

The uniqueness of language is no problem for telling just-so stories
about language, so that can't be UG's problem.

> Why can UG not be treated like any other evolved trait?

Because we can't even tell a just-so story about why it, rather than
something much simpler, should have been adaptive.

> Sorry this is so long, but I did not think there was too much of
> a problem with evolutionary theory. Maybe I am wildly off the mark.
> thanks for the help and indeed for the whole course!

Glad you liked it. Best of luck! -- Stevan

This archive was generated by hypermail 2b30 : Tue Feb 13 2001 - 16:23:09 GMT