Rules vs. lists

[Rob Freeman]

David,

Nice papers. Thanks!

When you said "any number" of rules I thought you were going to argue
abstracted rules could be completely arbitrary. Instead I think you
are just saying the number can be very, very large. And that I agree
with.

Most importantly, I think the number of rules which can be abstracted
can be larger than the number of examples they are abstracted from.

If we think about this I believe it is an important result. The
storage of each extra example need not be seen as a cost. For each new
example we store, we can get an even greater number of
rules/generalizations for the system. (Each new example can be related
to multiple others, to get multiple new "rules".)

For N examples, we can have >> N rules.

I'm not sure if your paper is arguing for this. In the light of it
lists seem a very powerful way to specify grammar to me. Not to
mention explaining certain idiosyncratic and inconsistent aspects of
grammar. In practice we have not used lists in this way. Any idea why
not?

Beyond that I don't entirely understand the importance of Cognitive
Grammar for your analysis. Why is it necessary for generalizations to
become entrenched before they can be thought of as being "part of the
grammar"? Couldn't any meaningful generalization which might be
abstracted from a set of examples already be considered to be "part of
the grammar", independently of whether it later becomes entrenched?

-Rob

On Tue, Jul 1, 2008 at 11:09 AM, David Tuggy <david_tuggy at sil.org> wrote:
> If language users list/learn a generalization, why should I deny it them?
>
> As to how I personally do it:
> http://www.sil.org/~tuggyd/Scarecrow/SCARECRO.htm is over twenty years old,
> but I haven't seen any reason to shift from the basic position it sets
> forth. See especially Fig 3, and  Fig 5; note in Fig 5 how the V+O=S rule
> (which Fig 3 shows to be itself a generalization over generalizations) is a
> subcase of at least 6 higher-level generalizations. The same sort of thing
> works with syntax or phonology as well.
>
> (A more recent version, contrasting the English set of rules/generalizations
> with comparable Spanish ones, is available at
> http://www.um.es/ijes/vol3n2/03-Tuggy.pdf ; there's a Spanish version and
> powerpoint available too at www.sil.org/~tuggyd.)
>
> Do I know that all English speakers have abstracted all these rules? I
> don't, in any absolute sense. But if they (any/most/a fortiori all of them)
> have, I want them in my grammar.
>
> Note that there is a globally-general rule: Optionally add something to
> something else to make a word. But the interesting stuff is the not-totally
> general rules (like X+Y=Y "structure with the rightmost element as head"),
> and the specific learned forms that prompt them.
>
> --David T