Re: First language acquisition

Greg Lee wrote:

'Evolving' isn't really 'being subject to drift', is it?

Wow... there's a bag of worms for the evolutionary biologists to
spill and deal with if ever I saw one! You might want to have a
discussion with people like Motoo Kimura and thrash that one out.
I'm not taking sides on it for any price!

If you're arguing that Parameters should vary as other genetically
determined traits do, yes, I can see that. And for all I know,
they do vary. I don't know what this 'transmission fidelity'
you're talking about is, how you measure it, or who has
pegged it at 100% for human language.

It has to be 100% faithful transmission of genetic information from
one generation to the next, or it has to be that any drift that occurs
is 'neutral'. Either will do to live up to the claims that are made
in an innate P&P approach.
In other words, 100% *functional* fidelity. The claims of
uniformitarianism -- that languages are all equal and that every normal
child has the same ability to learn *any* language with ease -- demands
that there be a functional homogeneity within the Parameters otherwise
the uniformists' claims would not be possible.
If you have one child with full functionality in its genetic Parameters
system it can pick up any human language with ease in the right
environment. But a child whose Parameter sequences show signs of
drift *that have visible surface effects* will have a different ability
to pick up any human language with ease. Of course, it might be that
a mutation improves some expressed ability, but it's more likely that
a mutation will have a negative impact if it has any impact at all.

There are two main ways you can achieve this functional homogeneity
that the innate P&P and the uniformitarianism claims require:
1. That the genetic sequences that code for the Parameters remain
stable, somehow circumventing the genetic mutation rate that affects
all other biological systems. In effect this means that the code
for language must remain the same throughout the generations or you
would see some children not able to acquire certain languages because
the crucial bit of their genetic Parameter-Space has drifted to
non-functionality.
2. That the genetic sequences for the Parameters *can* be affected by
mutation -- just like all other biological systems -- but the
mutated versions are functionally no different. I.e., at the the
surface level you end up with no discernible difference in the
expression of those sequences.

Option 1 is the most unlikely. We don't know of any other biological
system that can maintain its fidelity perfectly through many generations.
(In fact, if you make any genome immune to the mutation rate, you
effectively stop evolution. Even Darwin, right at the start of the
scientific study of evolving systems, listed "variation" as one of
the requisites of evolution. Variation is a necessary double-edged
sword.)

Option 2 could happen, and indeed is the basis of the "genetic clock"
methods of dating speciation events from existing species' DNA. You
have neutral drift within, say, some part of species A's mtDNA and
you measure it against the drift that is measured in species B's
mtDNA, and if you work with the assumption that A and B shared a common
ancestor, you can use the amount of difference to work out roughly when
the two lineages split. The mtDNA in each species still works (since
you still have the two extant species to examine now) but there are
neutral variations within it.

So, no, 100% transmission at the specific sequence level is not
necessary, but there must be 100% *functional* fidelity or else you
will have some lineages of people who cannot live up to the claims of
uniformitarianism because their Parameters have drifted too far away
from some particular type of language for them to acquire it by the
innate P&P method.

We can't call on natural selection (of whatever sort) to keep these
from drifting into non-functionality.

Who is calling on natural selection except you? I don't
understand why you think we should be able to observe
any effect of natural selection on the Parameter system
by looking at typological language change for the last
few thousand years. I do understand your inference that
if the Parameters are biological, they should be subject
to natural selection, but that doesn't mean that there will
be any relevant facts for linguists to observe by examining
language history.

I hope you don't take this the wrong way, but I have to wonder if
you are familiar enough with the concepts in evolutionary biology to
know what I mean by "selection" and the roles that it plays. It's
not just something that makes species change into new species; it
also maintains species and their features. Selection as a maintaining
force is absolutely necessary to keep mutation under control. It's
like a filter that lets the "fit" traits through and removes the
harmful "unfit" traits, yes, but a part of the result of that filtering
is that certain traits become a ubiquitous feature among the population.
This stops the mutations building up to such a level that genes no
longer function. You can see the effect of relaxing natural selection
on a trait in my oft-used example of subterranean species that once
had perfect eyesight. Selection doesn't affect an individual with 50%
vision or 0% vision anymore than it does an individual with 100% vision
in such an environment, so selection cannot act as the maintainer of
perfect vision in this species in this situation. That is, you might
have a mutation occur so that you have only 85% vision compared to the
species average. But this makes no difference in the dark environment.
As far as eyes are concerned, you are no better or worse than your
conspecifics, so you will have just as much success at producing
offspring as they do. Therefore, you have just as many kids and they
will all have, on average, your level of 85% vision. But this doesn't
matter, and they will be just as successful at producing their own
offspring as any other individuals in this population, and their kids
will also inherit, on average, that lower level of vision. And so on.
Each mutation in eyesight will not be selected against so the average
eyesight of the population will, on the whole, deteriorate into non-
functionality. If the subterranean species has a 'cousin' species that
it split from to go underground, however, while the cousin remained
above ground, the cousin species retains its eyesight *because of
natural selection* favouring those individuals with 100% vision over
those with less vision.
(And just to make this clear if it isn't already: this has nothing to
do with the specifics of visual processing. Sight and subterranean
environments provide an easy scenario to explain and use as an example.
It could equally be any adaptive feature that I'm talking about.
Wings and the need for flight might make for another example.)

In answer to your "examining language history" point, I don't know
what exactly you're getting at there, since I'm not really basing any of
my arguments solely on the historical evolution/change in a language.
Are you thinking that I'm saying that languages change because the
Parameters evolve? That's not what I'm saying at all. In fact, in
order to explain the genetic basis of P&P, I've been saying that the
Parameters must remain static for that system to work. The innate
P&P approach requires a functionally unchanging structure of Parameters
because one of the things it attempts to explain is how we can use the
proposed genetically-based Parameter structure to acquire *any* human
language without any problems, no matter what our background is.
As I said above, if we have Parameters drifting about and changing
left anr right ("evolving" if you wish, but I dislike using that word
too much because it can lead to definitional debates), then we lose
the uniformitarianism that human language is supposed to possess.

Not at all. The idea is that somewhere in the genome of all humans
is a complex of genes (I won't go along the rather naïve line of
suggesting that there is a gene "for" subject-verb agreement and a
gene "for" head directionality, but rather that these things arise
from the actions of interacting genes) that, when their effects are
taken as a whole, produce the Parameter hierarchy so that a child
can set the Parameters relavant to the linguistic input it receives.
If its linguistic environment is a language that does not need a
particular Parameter to be set to a definite setting (i.e., it can
be ignored, set to some genetic 'default', or simple set randomly
with no effect on the target language) then there is no phenotypic
effect for selection to work on. It's simply an 'invisible' part
of the genetic code, not offering any 'handle' for selection to
grasp and keep in check.

How do you know whether it's invisible?

Because not all languages depend on the entire universal inventory
of proposed Parameters that humans are supposed to carry in their
genome.
There will be some Parameters that, because of the route the learner
must take through Parameter-Space when setting them in the acquisition
stage, will have zero effect on the final product. Or, if there is an
effect, it will be so negligible that any setting will do. If that
Parameter is, therefore, not invisible at the surface level where
selection can use it to filter the phenotypic effects according to
the environment, then what is it? If something exerts no significant
effect on the final language product, is it not invisible? If you
disagree, then how can you look at that language and decide which
setting the Parameter has? This is even more difficult (if not
impossible) if the 'invisible' Parameter has nothing in the final
language that it has control or influence over. Imagine a language
that does not express any form of inflectional morphology. By looking
at the final language, how can you tell which settings all of the
(necessarily universal) Parameters that affect inflectional morphology
have been set to? I would suggest that it's impossible. You can look
and look again, and still not find any clues as to what your inflectional
morphology Parameters have been set to.

I gave (below) the example of devoicing word-final obstruent
consonants in languages without word-final consonants. If humans
evolved holes in their cheeks, there would no longer be a tendency
to devoice obstruents. But the development of cheek holes wouldn't
be invisible to natural selection. This illustrates the problem with
your reasoning that you're implicitly assuming these Parameters are
formally arbitrary fictions with no real physical basis.

But you would have to show that all the proposed Parameters have a
link to physical structures. I'm sure that the majority of the
Parameters suggested for syntactial features have no link to a
physical structure. How does the Polysynthetic Parameter (and, more
importantly, all of the Parameters that are exclusively dependent on
the "Polysynthetic = Yes" setting) ground itself to a physical reality
in the human body and so remain visible even if we lose polysynthesis
from the language? I guess if I wanted to make a version of innate P&Pism
that linked all the features of a language to a bodily/physical correlate
then it might be possible with a lot of effort and ad hocking all over
the place. But has anyone done this? All the P&P structures so far
contain at least some abstract/formal/arbitrary/whatever Parameters
whose domain exists only within the language itself.

Japanese speakers first learning English words ending in voiced
obstruents go through a stage when they devoice them, even though
they do not devoice word-final obstruents in Japanese (there being
none to devoice).

So then a True Believer of innate P&P might re-jig his hierarchy of
Parameters to have "devoice word-final obstruents" as the default.
Or as the 'spandrel'-esque result of how other devoicing environments
are treated in their phonological system (Japanese *do* know how to
devoice, so that's not a problem, and they *do* know how to form
obstruents, so that's not a problem either, and they do know that there
are "word-final" environments since their phonological system includes
them, so that isn't be a problem either, and...).

But this is a highly silly scenario, I agree, and I'm trying my best
to think of a plausible way that innate P&P could set it up. I guess
if I were a True Believer then I could construct something that would
answer you properly, but basically your point rests on a fault that I
agree with and in explaining what I meant I have to defend the thing
that I am ultimately attacking.

Uh, then we agree?

That P&P with a large emphasis on the Parameters and their dependencies/
hierarchy coming purely from an innate, gene-driven source is not a very
sound idea? Yes, sure, that's what I've been arguing against all along.

Out of interest, what objections do *you* have against the gene-based
innate P&P approach, if any?

And I've just noticed, before posting this reply, that Brian M Scott
has addressed some of those points -- and much more succinctly than I
did!

--
johnF
.

Relevant Pages

  • Re: Both SAT and AAT have failed
    ... water was involved with the selection of language. ... voluntary breathing, ... Biological evolution is based on evolutionary theory the central ...
    (sci.anthropology.paleo)
  • Re: Both SAT and AAT have failed
    ... water was involved with the selection of language. ... voluntary breathing, ... hyper-ventilation were preadaptations to voluntary breathing control for ...
    (sci.anthropology.paleo)
  • Re: Non-linear transfer function of a cat and mouse in Laplace domain.
    ... "Natural Selection" as a single phrase on the blackboard while you sat ... This inability to discern SA has lead to the pragmatics communicated ... would influence your general aphobetics you have with language itself. ...
    (talk.origins)
  • Re: The Paradox of Speciation
    ... compatibility can build up over time into big changes that do. ... The selection pressures ... French is the product of intelligent design, ... let me switch to any other language. ...
    (talk.origins)
  • Re: "Its uncertain whether intelligence has any long term
    ... There is ample evidence that "intelligence" has had considerable ... > with other animals run up against a critical problem - we are capable ... > of things that other animals are not (language being the obvious ... Natural selection propels some species along a ...
    (sci.bio.evolution)
Loading