Grammar with a "G"

[Sydney M Lamb]

Rob, Tony, Sherman, Esa, Steve, Brian, and bystanders:

Hi, I'm back!

I'd like to try to clarify a bit, adding to y'all's already interesting
efforts.

On Fri, 19 Mar 1999, Rob Freeman wrote:
> ...
> anything come up yet regarding 'emergent structure' from 'analogy based'
> processing? ...  Basic idea is that grammar is just the observed
> regularities of collections of examples, and analogies to them, which are what
> really control our perceptions. You still get grammar, but because the basic
> mechanism is analogy it has soft edges.

The point I was trying to make takes off mainly from the assertion that
this grammar that "you still get"  "is just the observed regularities ...
and analogies to them ..."

The question is, as asked by one interested in what is in the mind and
what is going on there: Are you distinguishing between (1) the mechanism
being proposed for arriving at "grammar" (or some command of the language)
and (2) that resulting "grammar" or command of the lg?

Put in another way, are you proposing that the result consists just of
remembered examples plus an ability to analogize?  Or are you proposing
that the result of the analogical and other operations has some particular
form other than that?


On Tue, 23 Mar 1999 12:47:57 +0800 Rob Freeman further writes:

> ... In the theories I was referring to the 'real' mechanism is example
> and analogy, Grammar is only the shadows it casts, real but unreal, with
> fuzzy edges.

To this I ask the same question, (ceteris paribus), plus this one:
What/where is this "grammar" you refer to here?  Anywhere real?

> ... Anyway, I was just drawing attention to this 'analogy-based' work

Right - I appreciate this point, but I still ask my question.

> (BTW analogy is very naturally implemented using networks ;-).

Yes, I agree. And my view (now expressed more fully in my new book)
includes a warm spot for analogy (cf. Chap. 14); But it also includes the
position that analogy is a mechanism and that what results from its
operation is new network structure, which in effect incorporates
generalizations that have resulted from the opn of analogy upon observed
and remembered inputs.


On Mon, 22 Mar, Tony A. Wright asks

> If I give the mechanic an account of my car's behavior, namely that it
> dies out any time I let my foot off the gas, which I formalize using the
> following rule:

>     dies /   [- accelerator] __________   (dies immediately following an
> environment                            negatively-specified for
> accelerator pressure).

> is my account of my car's behavior really only about me, and not about
> the car in any sense?

The facts are about the car, the means of stating them are about you.
Those who take an interest in "God's Truth" (which we can reformulate
for modern times as taking an interest in what is in the mind and/or
brain) have to be concerned with the means of stating the facts.  If one
is not concerned with the means but only with getting the fact accurate --
for language this means providing an accurate account of some aspects of
the linguistic productions of people -- then one is doing (what I at least
was calling) hocus pocus linguistics -- a practise to which I have no
objection at all, by the way.  But I do insist on the distinction between
the two kinds of aim. Both are valid, both have useful applications.

> ... All of these would certainly say much about me and my penchant for
> branching nodes, arrows, brackets, etc.

Exactly my point.

> But at the core, these formalisms would all equate to "dies when I let
> off the gas," which let's suppose is exactly, for whatever reason, what
> the car does, with not a single counter-example.

And this is a good example of the usefulness of the "hocus-pocus" approach
to data.

(On Tue, 23 Mar 1999 20:46:10 +0800 Rob Freeman already conveyed approx
the equiv message)


On Tue, 23 Mar 1999 09:05:34 -0700 Sherman Wilcox <wilcox at UNM.EDU>
adds:

> ...
> Maybe linguists' grammars work -- that doesn't make them right. ...

And I wholedheartedly agree, with a qualification: they can be more or
less right in a hocus-pocus sense, but that doesn't in any way make them
right in a neurocognitive sense.


As Sherman further says, evidently motivated by cognitive considerations:

> ... I'd like to figure out why and how and what people do what they do
when they create utterances. When it comes to language, we are all
improvisational geniuses.


Since I was using the terms "God's Truth" and "hocus-pocus" in a more
clearly defined way than those mid-century structuralists, maybe I could
communicate more clearly by using, instead of these terms, the following:

Analytical Linguistics -- This is the familar kind. In this mode one is
mainly concerned with accurately describing linguistic productions
(without concern for the process of production or that of comprehension or
the system that makes those processes possible).

Neurocognitive Linguistics -- The aim here is to understand that system
and those processes.  I use the term 'neurocognitive' rather than just
'cognitive' since many people nowadays who use the term 'cognitive' are
using it to label concerns which are mainly those of analytical lx.  (See
the article by Bert Peeters in WORD, August 98).


Esa Itkonen adds, and I wholeheartedly agree:

> ... Both approaches are equally legitimate ...


Thanx for 'listening',

  -Syd