LINGUIST List 3.678

Tue 08 Sep 1992

Disc: Parameters

Editor for this issue: <>

Directory

1. , RE: 3.671 Parameter Setting and the Garden of Eden

Message 1: RE: 3.671 Parameter Setting and the Garden of Eden

Professor Givon asks if these observations are new? If I understand them, the
answer is "no!"

(1) Givon wonders if "it is generally realized that such a "family tree"
structure follows from very minimal assumptions and need imply nothing about
the historical development either of languages or of populations."

This seems to be a restatement of the more general principle of numerical
taxonomy or multivariate methods generally of "garbage in-garbage out."(GIGO)

What seems important in the "Garden of Eden" issue, as Swadesh noted in his
utilization of glottochronology, is to find independent evidence pointing in
the same direction. Thus if linguistic reconstruction, fossil evidence,
DNA reconstruction all tell the same story, one ought to tend to believe it.
The current problem, of course, is that the "independent evidence" is weak-
the taxonomic algorithms reliability is in question and there are unresolved
matters concerning the sampling of DNA--not to mention disagreement among
linguists. (Recent articles in Science discuss both of these issues)

(2) Givon also suggests "There may be room in formal linguistics for concepts
 of optimality, not so?"

The answer is "of course"--indeed can there be formal linguistics without some
sense of optimality or economy? The notion of an "evaluation metric" played
a fundamental role in early Halle/Chomsky work. More recently, for example,
--to pick someone at random--Chomsky (1992) in "minimalist Program" says:

"The linguistic expression are the optimal realizations of the interface
conditions, where "optimality" is determned by the economy conditions of UG.
LEt us take these assumptions too to be part of the minimalist program...
it seeems that economy principles of the kind explored in early work play
a significant role in accounting for projperties of language. pp.6-7"

Of course, here too we face the general "GIGO" principle. Counting symbols
in some formalism may be no more informative than minimizing least squares or
some distance measure in a clustering algorithm.

John Limber, Psychology, University of New Hampshire