Theiling Online    Sitemap    Conlang Mailing List HQ    Attic   

Re: "Theory informs practice" - OK?

From:Dirk Elzinga <dirk.elzinga@...>
Date:Thursday, November 13, 2008, 23:58
On Wed, Nov 12, 2008 at 1:16 PM, David J. Peterson <dedalvs@...>wrote:

> > The closest I've seen is an attempt to work with Bochner's > obscure framework. A former colleague of mine had an > interesting account of the stem alternation in Polish.
I've looked briefly at Bochner's model. I'm not sure what advantages it has over other Word and Paradigm models, but I'm generally sympathetic to WP approaches; I always spend at least a week on WP in my phonology/morphology class.
> He > modeled it very well, but then the next question was, with > two likely alternatives, how does a speaker of Polish figure > out which alternation to use--or with novel or nonce forms? > The idea, in my mind, is to attempt to predict how a speaker > will act. The future may be dealing with percentages based > on patterns, but we'll see. To me, the fact that, as a joke, an > English speaker can make up "blunk" for the past tense of > "blink" says something important about language that > theorists need to pay attention to.
The Future is Now. Exemplar models have been around for quite a while. A model I work with from time to time (Analogical Modeling) was first published in 1989, complete with the source code for a Pascal implementation (admittedly not very good; the current software is much better). Exemplar models allow for--and in fact predict--the kind of statistical behavior actually observed in language; and so from the point of view of observational adequacy they outstrip other theories out there. Whether the brain actually works that way is still undecided, but the same is true for any other theoretical model you care to choose. But none of these models really helps me in language creation. For that I rely on my intuition and my good taste. Dirk -- Miapimoquitch: Tcf Pt*p+++12,4(c)v(v/c) W* Mf+++h+++t*a2c*g*n4 Sf++++argh La----c++d++600