LINGUIST List 5.179

Sat 19 Feb 1994

Disc: Last Phonological Rule

Editor for this issue: <>


  1. Alex Monaghan, last phonological rule

Message 1: last phonological rule

Date: Thu, 17 Feb 94 01:11:59 GMlast phonological rule
From: Alex Monaghan <>
Subject: last phonological rule

What with messages to linguist and others to me directly, there has been a
fair amount of discussion of the infinity or otherwise of Goldsmith's
model of phonology. Bugs computational and biological have prevented
me from responding earlier: I hope it isn't too late to do so now! here goes:

As various people have explained, there is a difference between infinite
dimensions and infinite points: Goldsmith's model has the latter but not
the former. My original point (?) was that an infinity of points was bad
enough, even though the dimensionality is limited.

There are of course infinite numbers of points in linguistic phenomena,
such as fundamental frequency, and these points can be split up into
small numbers of linguistically distinct areas (e.g. n-tone systems where
n is generally less than ten) in fairly arbitrary ways. this sort of thing
happens all the time in languages.

What bothers me about Goldsmith's approach (and others) is roughly the

a) one physical dimension, say f-zero, is represented by many dimensions
(i.e. many weights) in these models. if one dimension gives you just as
many points to play with, why not stick with one?

b) the values attributed to weights in such models are often completely
arbitrary, and different values could produce the same results. indeed,
an infinite number of different values could produce indistinguishable
results in principle. what, then, is the status of these values?

c) if we found, say, that natural language stress systems could be modelled
by a 12-dimensional space, and in fact that they tended to cluster around
the point (2,4,6,3.5,10,4.5,9,6.5,3,2,1) in that space, what would that
actually teach us about such systems?

d) people do have intuitions about things linguistic, and although these
intuitions may be wrong they rarely (if ever) involve more than four
dimensions: maybe four is enough.

This through a stinking head-cold at 1am here, but i hope it clarifies my
position and sparks some more discussion of what is a fascinating topic.
good night,
Mail to author|Respond to list|Read more issues|LINGUIST home page|Top of issue