And Rosta wrote:
> Mark P. Line, On 20/06/2008 14:41:
>> And Rosta wrote:
>>> Hmm. Your answer would I think imply (as I had wondered) that cognivism
>>> doesn't go far towards determining what a (formal) model of syntax
>>> would
>>> look like. That is, I can imagine pretty much any formal theory of
>>> syntax
>>> being implemented in wetware.
>>
>> That means that formal syntax doesn't tell us much about what language
>> is like, unless language is not like something that can live in a human
>> brain.
>
> I don't understand your reasoning here.
> How great a constraint on the
> properties of a formal system does implementability in wetware impose?

The short answer is that formal systems presuppose algorithms for parsing
and generation, and that the brain does not have algorithms. (I'm using
'algorithm' as a technical term, not as a synonym for any procedure: it's
a procedure that will halt for any given input.) Note that a formal system
of syntax is meaningless on its own terms if there are no algorithms or
parsing and generation (which is my main criticism of Montague, by the
way), quite apart from any question of implementability in wetware.

The long answer would consist of a very long list of constraints on how
the brain might be able to transduce between acoustic (or visual) signals
and thoughts and how it certainly can't.

The brain has no way to implement an open-ended last-in-first-out list
(called a "stack" when it's a software data structure) of the kind that is
required to implement, say, center embedding. So a formal system that
allows center embedding will not only not tell us anything useful about
center embedding in human language, it will imply that center embedding is
a perfectly reasonable mechanism in human language.

The brain has no way to implement very long-range dependencies, because of
the decay time inherent in short-term memory. A formal syntax that allows
very long-range dependencies cannot be implemented in wetware.

Formal syntax implies that syntax is a separately calculable component of
language. All the evidence from wetware (and from cognition, for that
matter) is that there is no such separate component. Also, the brain is
not a calculator.

And so on.


> I,
> in my naive, and, I've already acknowledged, ignorant, reflections on the
> question, do not see that implementability in wetware poses significant
> constraints on the formal system -- at least not within the limits within
> which formal theories of syntax vary.

That's because formal theories of syntax are part of philosophy, not
science. Scientific models need not be expected to constrain philosophical
speculations arrived at by ignorant reflection. Do they?

That's why I was saying before that formal syntax simply doesn't tell us
anything useful about language. Philosophy of X is very useful when
there's no science of X. As the science grows, the philosophy needs to
move on to something that's still too woolly for scientific attack. Many
linguists are still resisting the transition from the philosophical method
to the scientific method.


>> What would the implementation of Montague Grammar look like, in broad
>> brush strokes, in wetware? Just broad brush strokes to indicate a
>> possible approach. Montague can't even be implemented as a computer
>> model without modification, so it would be interesting indeed if there
>> were a way to implement it with any fidelity at all in a
>> neurocognitively informed way.
>
> I have no idea -- scant knowledge of Montague syntax itself, no idea of
> how to implement it, no idea of what the problems in implementing it are.

Montague Grammar is a (very) formal theory of syntax. You said you
couldn't imagine how any formal theory of syntax might not be
implementable in wetware. If you're not familiar with a wide variety of
formal theories of syntax and you're admittedly ignorant of wetware
constraints on language processing, then your statement seems to have
violated a Gricean implicature or two:

"I can't imagine how a horblebobble could possibly snigglefin a
blurbeyfutt." has implicatures that you know what a horblebobble is, that
you know what a blurbeyfutt is, that you know what it means for something
to snigglefin something else, and that you have some conception that the
way a horblebobble (in particular) might snigglefin a blurbeyfutt (in
particular) is at least not trivially obvious.


> Clearly, neurological constraints on language are relevant to my
> conception of language, since language must be implementable in wetware.
> So all I'm saying is that my state of knowledge is such that this doesn't
> help me rule out one model of syntax in favour of another.

Well, that's okay then. You have a very interesting learning curve ahead
of you.


>> So you start with something that is in the mind and then abstract it?
>>
>> Can you give an example of something that is not abstractable from the
>> mind?
>
> See my previous message, where I spoke of emotions and chess.

So you don't believe that emotions can be described?


>>> and Cognitivism is the doctrine that the tool is shaped by the minds of
>>> its users and is learnt (rather than merely 'acquired') from
>>> observation of usage. To me, then, brains are an irrelevance to
>>> cognitivism, except to the extent that they shape language (by
>>> affecting what is possible and what is preferred).
>>
>> So, you either don't believe that language evolved in human brains, or
>> you don't believe that linguistic theory needs to account for the
>> evolution of language.
>
> How come? I believe that the language faculty evolved in human brains and
> that langue evolved in human brains and human societies. And I believe
> that linguistic theory needs to account for the evolution of language.
> Where is the contradiction in my thinking?

I was assuming that you meant "cognitivism" to refer to a branch of or
approach to linguistic science, where theories are not allowed to
contradict each other in the long run and where therefore every theory is
potentially relevant to every other theory.

If you meant "cognitivism" to refer to something that uses the
philosophical method and not the scientific method, then my inference
wouldn't hold, of course. You could easily hold various beliefs
(scientific or otherwise) about the evolution of language and about how
linguistic *science* needs to operate in accounting for that evolution,
while allowing philosophical speculations of the cognitivist sort to run
wild in a way that brains are an irrelevance to it.


-- Mark

Mark P. Line
Bartlesville, OK