Re: Natural Language and Natural Selection

From: Stevan Harnad (harnad@cogsci.soton.ac.uk)
Date: Mon Feb 26 1996 - 17:52:52 GMT


> Date: Mon, 26 Feb 1996 14:12:54 GMT
> From: Nick Bollons <nsb195@soton.ac.uk>
>
> I did get a little lost in some of the intricacies of the language
> faculty (i.e. defining language), though as Steve told us: no complete
> language definition suffices to explain this phenomena.

With some things, like language (and consciousness), it's better to just
rely on our ability to POINT to it (that's an example of language,
that's not) rather than committing ourselves to a definition before we
really understand what it is.

> (The Spandrel analogy is used to exemplify that for some structures
> you have to not look at the centre piece (a mosaic) but to what lies
> around it (the pillars) which if removed would decrease the
> 'profoundness' of the centre piece. The spandrel can be considered to
> be an innate language ability which is needed in order to create the
> detailed mosaic of language acquisition. Where this spandrel
> originated from is where the article is directed).

This account of spandrels would fail the kid-sib test! What is a
spandrel? In the Dome of San Marco in Venice, the round top of the
church, there are paintings that are made out of a lot of dome-like
circles, touching one another. Geometry dictates that when three such
circles touch at a point, it makes a kind of concave triangle, called a
"spandrel." Spandrels have since become as important in the dome
symbolism as the circles. The circles were dictated by structure and
function (domes are round). The spandrels were "byproducts" of the
circles, till they took on a life of their own.

Some evolutionists (like Stephen Jay Gould) have used this analogy to
try to explain how biological structures could evolve that do not have
a direct selective advantage, hence, there is no adaptive story to be
told about them: They did not evolve because they made their bearers
fitter or better at reproducing. Some have tried to apply this spandrel
analogy to UG...

> 1- Natural Selection (in this article) states that the only plausible
> explanation for the creation of near perfect organs is by adaptive
> complexity . Where many interacting parts of a system arrange
> themselves to fulfill a specific function. The eye (organ) is a perfect
> adaptation for visual imagery (function) and could only have occurred
> through a process of natural selection and those who did not need the
> function would not develop the eye. This process occurs gradually
> through a series of small mutations and proceeds over a relatively long
> period of time (some sort of hill climbing to adaptation). The
> argument that if it is not natural selection then what other process
> could create such phenomena : Divine Intervention ? (I don t think so
> ?) do not seem plausible. Natural selection seems to be the only method
> (at present) to explain complex organs and organisms adapted to their
> specific function/functioning. NATURAL SELECTION SEEMS PLAUSIBLE.

Yes, but with most organs -- such as heart or eyes or wings -- there is
a plausible story to be told about how they could evolve gradually
because they help their bearers survive and reproduce better. But can
such a story be told about UG?

> 2 - So if natural selection can seem to explain the complex design of
> the eye then could it be used to explain the complex design of the
> cognitive mechanisms used in language? Pinker then goes onto
> characterise the features of a language functioning i.e. Verb Affixes
> e.t.c and it's complexity .Which was even recognised by Darwin himself
> "that perfection of structure (language)... which justly excites our
> admiration" . Not only is the fact that such a complex tool could only
> have been created by natural selection: but the universality of it's
> structures (that there are similar linguistic traits of languages all
> over the world (Chomsky)) also supports this view. Natural selection
> can only explain the cognitive mechanisms such as the arbitrariness of
> the sign (that grammar and language is just the use of some sort of
> systematic symbolism to mean something for everyone) that are needed in
> order to achieve language competency.

Many things are mixed together here: grammar is one thing, arbitrary
signs (names) another.

> This arbitrariness of the sign is important in an achievement of parity
> (same communication standardisation) I.e in computers where one
> computer goes: beep beep and another goes boop boop - these two are of
> different parity and are in compatible). Humans must have some sort of
> parity in order to be compatible and understand one another. The
> arbitrariness of the sign allows communication within a language as well
> as outside to other languages (Steve talked about the fact that you can
> transform any word in one language to any word in another - the word is
> just the arbitrariness of the sign) and we all have this ability and it
> must have been created somewhere.

All true. But parity is not the same as UG. UG is a very complex,
abstract system of rules. It is not at all clear that we needed ALL of
that in order to achieve "parity" -- in fact it's highly unlikely, as
something much simpler would have done just as well. And that is the
evolutionary problem with UG: What GOOD is all that stuff,
evolutionarily speaking, and how/why would it be better than much
simpler alternatives?

> The final part of this section is a difference between language
> evolution (described above) and language acquisition. The fact that
> early infants can distinguish different grammatical utterances and find
> the correct ones relatively quickly would indicate the existence of
> some cognitive mechanism that is innate, and is very much different to
> the evolution of how it got there. The child has this ability that has
> been adapted to it's present complex functioning by natural selection
> and now the child can build upon it. LANGUAGE IS A COMPLEX DESIGN
> SYSTEM.

The poverty of the stimulus implies that the child could not be learning
UG by trial and error based on the data the child encounters. If the
child could not learn them, it is even less likely that "evolution" could
have "learned" them in advance: How would it make us fitter and
reproduce better if we were "selected" by evolution to say "who did he
think went out" and "who did he see that went out" but never "who did
he think that went out"?

> 1 - Natural Selection is a valid method of explaining complex organs
> and there functions at there present state.

Yes, but only when there is a believable evolutionary story for how that
complex structure was selected by evolution in the first place. For wings
and heart there is such a story; for UG there is not.

> 2 - Language fits the bill of being a complex cognitive mechanism and
> passes the test of being adapted to it s present form by a process of
> natural selection.

All true, but not really relevant to the point. (Not your fault, Nik,
this is a critique of the Pinker paper too!)

> The final part of the article concentrates on some sort of steps that
> may have occurred during this evolutionary process. At one stage there
> must have been no language and it must have evolved by some method and
> due to some reasons into it s present form ? Pinker hypothesises these
> reasons such as the fact that early communication must have been among
> kin and maybe certain cognitive mechanisms were created by this,
> specific to that community. He talks later of a cognitive 'arms race'
> between groups in order to accelerate there development. Whereby a
> group with a quicker and more complex language are going to be more
> adept in teaching their youngsters learned 'stuff' and in planning
> hunting/gathering and other exercises. This is a viable hypothesis as a
> group with a greater language capability would have developed quicker
> than ones with a deficit or difference.

All good reasons for believing "language" involved, including the
disposition and ability to learn and use it: But why UG rather than
something a lot simpler that would do all these evolutionarily useful
things much more simply?

> A biological explanation of language evolution is also presented. These
> cognitive language mechanisms are no good if it is only one person who
> has them i.e. if one cave man has one language whilst all the others
> have another no one will be able to understand him. By a process of
> inter-mating (this returns to the kin idea ) certain genetic traits are
> continued along a line leading to stabilising of that trait (not
> perfection as natural selection does not strive towards perfectionism:
> mainly to some form of similarity and stability). Here two points are
> important : that organisms pass on some learned information to their
> offspring genetically, who improve upon this knowledge and pass it onto
> their offspring e.t.c add infinitum. This could explain the passing of
> increased complex cognitive mechanisms used in language by genetic
> mutation which is increased and stabilised over a relatively short
> period of time. (A few million years is nothing in the evolutionary
> time scale of the process of natural selection).

Fine. But for all this "passing on" to be Darwinian (rather than
Lamarckian -- the discredited theory that we can pass on in our GENES
things that we have learnt in our lifetimes), there has to be some
evolutionary advantage to all these bits of structure we pass on, and
the changes we pass on have to be genetic ones, that were passed on
because they made us survive and reproduce faster. Again, this is easy
to envision for, say, water creatures, as the land gets drier, and natural
genetic variation in the size and shape of their fins favours some of
them over others, and so gets passed on. But UG? "Who did he think that
want out?"

> Conclusion
>
> There should be some mention that the article makes no reference to
> being refutable and that all the ideas within it are un-testable
> hypotheses and totally conventional (we cannot go out and find a
> caveman and test the level of his language ability). But Pinker
> produces evidence by Gould, Chomsky, Slobin, Darwin and others which
> supports each of the points that he conveys and the theories would seem
> to be plausible. I will use a sentence from his conclusion that
> summarises the main body of the article : 'Language shows signs of
> complex design for the communication of prepositional structures, and
> the only explanation for the origin of organs with complex design is
> the process of natural selection'(Pinker Behavioral and Brain Sciences
> 1990).

All evolutionary theory, including the story of the evolution of feet
and wings from fins, is historical, hence not testable directly. Direct
tests of evolution are done with fruitflies or mice, showing how
ARTIFICIAL selection pressures can "shape" later generations. It does not
require too much stretch of the imagination to see natural selection
operating the same way. But this works for bodily structures and
functions that are "shaped" by selection on the basis of how they help
their bearers survive and reproduce. It is much harder seeing how they
could have "shaped" UG...

> (Will send my ideas on Monday s tutorial in a few days. Just as soon as
> I can find some time. And I think that I worked out my confusion over
> the poverty of the stimulus problem : it's the wording that I am
> confused about (poverty means deficit) and this is in my view a little
> mis-leading in this context.
 
Consider that the same weakness of the data (lack of negative evidence,
for example) that makes UG unlearnable on the basis of the data a child
ever hears or produces when doing Skinnerian learning that is shaped by
its consequences (reward) should also make it difficult for evolution to
shape UG by its consequences (survival and reproduction).



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