Re: Re: Natural Language and Natural Selection

From: Stevan Harnad (
Date: Mon May 27 1996 - 15:07:00 BST

> From: "Liz Lee" <>
> Date: Mon, 22 Apr 1996 12:10:57 GMT
> Is language a biological capacity or a cultural invention (or both)?
> What is language? The general consensus seems to be that it is a
> rule - based system, using arbitrary symbols to represent
> information which is passed on to others who are able to receive,
> and understand the same system. The fact that symbols are used is
> relevant, this makes language distinct from mime which, although a
> means of communication, depends on mimicry of gestures to convey
> meaning, for instance making a "sad" face to show "sadness" is very
> different from using the completely arbitrary word "sad" to do the
> same.

It is not even clear whether imitation really conveys "meaning" at all:
It certainly does convey meaning to someone who DOES have language,
because they immediately translate a bit of mime into the sentences
describing it, or the concepts (whatever those are) behind the words in
the sentences. But if we were not capable of language, and it was
uncertain what concepts we had, would mime "mean" anything at all?
If a child echoes a smile, does it mean anything? Does an analog copy of
something mean anything? Remember that with a sentence, a proposition,
there is the possibility of saying something FALSE (or incorrect):
"That's not what I meant." But can one say anything false through a
copy? The copy may not be faithful, but does that mean it's false? And
if it can't be true or false, what does it mean?

> It is unique to the human world, with all human children
> having acquired it by the age of six (Pinker, 1994), other animal
> groups may have a means of communicating fear, amorous overtures,
> aggression or the whereabouts of food, but that is the limit of
> their ability.

These are all forms of expression, or even communication, but what
language expresses or communicates is "meaning" (whatever that is),
whereas expressing, say fear, is just emoting.

> Humans are able to contemplate the past, speculate
> about the future, discuss politics, counsel their kin, move an
> audience to tears or make it rock with laughter by using a shared
> language. This is possible even if verbal communication is
> unavailable, as in the case of the deaf community using sign
> language. This essay will look at the origins of language
> acquisition, focussing on whether evolution can be credited with our
> language abilities, or whether language was "invented" ( like
> writing) to meet our cultural needs.
> Biologically, we have been endowed with the detailed structures
> required to be able to produce and receive a spoken language. The
> vocal tract is shaped in a specific way, creating hollow cavities
> for sound resonance, the lips, tongue and mouth are controlled at
> will, the vocal cords produce sound by vibrating when air from our
> lungs is expelled whilst "speaking". The human auditory system is
> "tuned in" to the correct frequency to receive speech sounds.
> Studies of newborns detail their "preference" for human speech over
> other similar sounds, and their "conversation- like" interactions
> with their mothers (Schaffer, 1978). Pinker (1994) suggests that
> the development of the vocal tract has drawbacks, in that it
> compromises breathing and swallowing, this implies there must have
> been some evolutionary advantage in being able to speak. Areas of
> the brain (Broca's and Wernicke's areas in particular), which are
> associated with language but no other known cognitive processes,
> have developed. Proof of this is seen when stroke sufferers may have
> language, but no other cognitive, impairment, alternatively they
> may have intellectual impairment with intact language ability
> depending on the area of brain tissue affected.

But language impairment itself is not unitary -- it can be in
comprehension or production; it may be specific to speech, or
a higher-order problem with word meaning, or naming, or word-finding, or
syntax. All these deficits seem to have different geographies in the
damaged brain. So maybe language is a cluster of abilities, not just one

> However Bates (1994)
> argues against domain specificity pointing out that children who
> suffer large brain lesions in "language" zones still go onto develop
> a normal language ability - albeit after some delay.

The young brain is famous for its "plasticity" -- it's capacity for
reorganisation and recovery from damage.

> Pinker and
> Bloom, (1990) suggest a "critical period" for language acquisition
> which renders language acquisition difficult after a certain period.
> The structures required for language are innate, they are common
> to all humans and are unaffected by environment, but the cultural
> setting is important in language acquisition and will be discussed
> later. Pinker and Bloom deal with the question of whether, given
> that we share a common ancestor, (approx. 6 million years ago), one
> would expect similar language developments in apes, (citing Bates,
> Greenfield and Lieberman) by arguing that since our paths diverged,
> apes have followed their own pattern of selective adaptation which
> need not resemble ours. He cites cases where apes (chimpanzees)
> have been taught a limited signing vocabulary after many teaching
> sessions, and compares this to the ease with which human children
> learn their own language, without any obvious instruction.

And ape signals don't generalise as human ones do: They each seem to
have to be trained one at a time, which suggests that they may not
"mean" what their English translations imply that they mean.

> So there
> being no evidence of language structures in other primates is not a
> bar to human language development being designed by natural
> selection. There was, it seems, some advantage to be gained, at
> some point in our evolution, by being able to pass on information to
> others.

Apes pass on information too; language does it in a specific way:
through PROPOSITION, which are strings of symbols, arbitrary in sheep,
but referring to objects, events and states, and capable of being
combined and recombined into an infinity of further propositions --
and, in principle, able to express any proposition (anything that can be
said in words!).

> There is another view, Noam Chomsky and Steven Gould ( cited
> in Pinker and Bloom, 1990) suggest that language may have evolved
> as a result of some other development or because of an overall
> growth in brain size. What they are proposing is that the specific
> adaptations which have occurred, allowing us to develop and use
> language, were a by-product or "evolutionary spandrel" which just
> "happened" to be. This does not seem likely when compared to other
> complex biological structures which have evolved with a specific
> function, the eye for instance. One reason for this idea was
> Chomsky's belief that evolutionary changes are a long, gradual
> process and as such a new ability (language) is unlikely to emerge.

In general, Chomsky would be right about the gradual shaping of
structure by the variation/retention mechanism of the Blind Watchmaker's
success-machine (reproductive and survival success that is genetically
coded translates into more of the same in the next generation).
Sometimes evolutionary changes happen more quickly, whether because of
increased mutation rate, or environmental change, or the engagement of a
spandrel lurking from the evolutionary past.

> A second reason is the presence of what he calls "universal
> grammar" (UG), this is a grammatical pattern which is present in all
> languages, and obeyed even by young children. Chomsky suggests that
> obeying UG is an innate ability, but one which confers no
> evolutionary advantage, and therefore cannot be the subject of
> natural selection. But UG is one very small part of language and it

Not SUCH a small part, but indeed just a part...

> seems rash to dismiss the whole of language as a by-product for the
> sake of one of it's less obvious constituents.

True, but the origin of UG IS a problem...

> The premise that UG
> is always obeyed and so children are not exposed to these specific
> grammatical errors ( the Poverty of the Stimulus) is argued
> against by Pullum (1996 ), on the grounds that the rules
> surrounding UG are regularly broken by adults in everyday language,
> he gives examples from the "Wall Street Journal" and concludes that
> rather having an innate language ability, children learn in a
> data-driven way.

Pullum is very clever, but I'm not sure how systematic or compelling his
argument is: Sure there are occasional "violations" of UG. They could
happen because I swallow my food wrong, and leave something out that UG
says must be there. But these violations are rare and unsystematic. Do
they really provide the vast body of missing negative evidence a child
needs to have encountered (or produced -- don't forget negative examples
can come from the child, working under the wrong rule) by age four?

> Pinker and Bloom ( 1990 ) gave a more reasoned
> argument against the spandrel idea, they suggest firstly that
> evolutionary changes can, and do occur over relatively short time
> spans, coupled with the facts that it requires only a small
> advantage to initiate change, and there is no proof that language
> has evolved on a random way.

Fine, but do P & B give any evidence that something like this occurred
in the specific case of UG? And what's the advantage?

> Secondly that there is an obvious
> advantage to be had by being able to pass on knowledge to one's
> offspring without them needing to experience incidents first hand.

True. But that's the advantage of language in general: The tricky issue
is UG in particular. Why couldn't we pass on knowledge to offspring with
a much simpler (and learnable) set of grammatical rules?

> They answer the UG question by suggesting that there is an advantage
> to be had by using grammar in an exact way - it is essential to know
> the intended meaning of a sentence in order to act appropriately,
> for instance whether "the child fell over the dog that died" has
> the same meaning as "the child fell over the dog and died" - which
> of course it does not. So there is a good reason for having some
> kind of inbuilt ability in place - to already know the code - this
> reduces the chance of getting it wrong when the learning process
> starts.

Yes, propositional language needs SOME kind of grammar, but why as
complex and unlearnable a one as UG?

> When, and why, did language evolve? Bates ( 1994 ) estimates that
> language is a relatively recent cognitive development, approximately
> 30,000 years ago, although Pinker and Bloom point out that there is
> fossil evidence of the presence of Broca's area 2 million years ago.

Hard to interpret ancient anatomy's FUNCTION, though; the frontal lobes
are involved in other higher areas of cognition too, in humans and
nonhuman primates as well.

> The emergence of language coincided with that of tool this and this
> would seem to be relevant.

The language/tool-use link (like the language/hunting link), though much
mentioned, is not much more than a just-so story...

> Pinker (1994) proposes two reasons for
> the evolution of language, firstly, our ancestors developed
> knowledge both of technology (tool use) and the environment, and
> second, co-operation between community members was significant.

Nonhuman primates use some tools too, and cooperate; worlves cooperate
in hunting. It depends on the KIND of cooperation. These are all really
just speculations.

> Harnad (1995) suggests that language is used by a community to
> classify their world, when referred to a hunter-gatherer existence,
> it makes sense that passing on shared information about distances
> travelled to obtain food supplies, or the location of poisonous
> vegetation should enable those receiving, decoding, and acting on
> the information to be better equipped than those who do not have
> this information source. The better communicators were more likely
> to produce offspring, (being well-nourished and able to keep away
> from danger), and passed on this ability to future generations.

It may all have started within the family, between mother and sibs,
because there the genetic bases and the advantages are most
concentrated (inclusive fitness). Unrelated animals are competitors, so
it's not as clear how the benefits of communication would be felt and
passed on.

> Pinker and Bloom (1990) note that humans have the ability to store
> knowledge across generations, and that this is a "species -
> specific and universal feature".

Yes, propositions made it possible to pass on knowledge by an "oral

> An interesting, but less
> compelling reason for language is covered by the Whorf hypothesis (
> cited by Pinker and Bloom,1990, and Harnad, 1995). Whorf suggests
> that language is just a way of labelling our thoughts, and that this
> influences our view of reality, the way things look to us depends
> on how we name them, so each different language-speaking people will
> look at the world differently. Pinker (1994 ) argues that this
> cannot be the case, since babies are able to distinguish between
> phonetics before they can understand the meaning of the sounds, and
> questions whether they could learn a language without some relevant
> structure in place first.

The fact that some structures are (and must be innate) does not mean
others aren't learned. But you decsribe the Whorf Hypoethesis rather
vaguely. It is that your language (and culture) shapes or influences how
the world LOOKS to you.

> Harnad (1995) goes further and suggests
> that it is probably the other way round - our view of reality
> influences our language.

Actually, I think the influence occurs in both directions...

> How is culture or environment involved?
> Language has been shaped by culture, the particular language,
> dialect, accent and lexicon we use is entirely due to the cultural
> setting. Babies are able to make phonemic distinctions, which is as
> well because these are culturally determined and it would be
> disadvantageous for a child to be unable to differentiate between
> the essential components of his or her language. There is well
> documented evidence of the pattern of neural growth and destruction
> in the fetus, newborn and child, the loss of brain cells in the
> early years coincides with a rapid growth in nerve fibres. The
> ability of a young child to learn any language effortlessly compares
> favourably with the older child or adult trying to learn a second
> language, it is not the learning of a new vocabulary which is hard
> - there is no evidence that this ability reduces with age - but then
> a language as stated earlier is a great deal more than just
> individual words.

> But a culture without a language is unknown,
> linguistic social competence is important in any culture, and the
> language must be able to convey whatever meaning the speaker
> chooses, it must also be readily understood by the recipient. In
> conclusion, the evolution of language by natural selection seems a
> reasonable proposal, whether or not the finer points in any language
> are so readily explained is unproven, however there is no doubt that
> the influence of the cultural setting is as important to the
> language being used, as the ability to use the language itself.

Good essay, but different conclusions for different parts of language
might have been more informative.

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