LINGUIST List 15.1816

Tue Jun 15 2004

Disc: Re: Comments on things no language does

Editor for this issue: Sarah Murray <sarahlinguistlist.org>

Directory

1. William J Poser, Re: Comments on things no language does

Re: http://linguistlist.org/issues/15/15-1701.html#2 
and Linguist 15.1782

 
I'm not quite sure how the hypothesis that:

 Languages cannot have long contrasting vowels
 unless they have short contrasting vowels.

is intended to be interpreted. The most straightforward interpretation
would be that a language cannot have contrasting long vowels unless it
contrasts some pair of short vowels, allowing any number of long
vowels so long as the language has at least two contrasting short
vowels. As far as I know, that is true.

However, my guess is that the constraint is not correctly stated and
that what is actually meant is that you can't have a contrast between
/X:/ and /Y:/ unless /X/ and /Y/ contrast. This is part of
phonological folklore, but it is false.

Early Modern Japanese, as documented in the late 16th/early 17th
century, had the same five short vowels as now: a i e o u (with the u
not really rounded\ ), but it had six long vowels, a i e o u and
open-o. We are quite certain of this because the Portuguese Jesuits
consistently distinguished the two kinds of long o, e.g. in Joao
Rodriguez _Arte da Lingoa de Japao_ (Nagasaki, Society of Jesus,
1604). As I recall the distinction is also marked in the pedagogical
materials that survive from the Korean government foreign language
school. Long open-o resulted from the contraction of /au/ sequences
which in turn arose due to the loss of intervocalic consonants. The
vowel system eventually regained its senses and merged long open-o
with plain long o.

Another problematic claim is:

 There no languages where the inflectional affix is closer
 to the root than the derivational one.

One problem of course is that "inflectional" and "derivational" have
different meanings for different people. Another is that this isn't
sufficiently formalized. The strongest constraint would be:

 There exists no language in which there exist an inflectional
 affix I and a derivational affix D such that I is closer to the
 root than D.

But the above could also be intended to mean something weaker, e.g.

 There exists no language in which for every
 inflectional affix I and for every derivational
 affix D, I is closer to the root than D.


If the former is intended, on most people's notions of "inflectional"
vs. "derivational", it is falsified by all of the Athabaskan
languages. In Athabaskan languages, the order of morphemes in the veb
is, roughly speaking:

 derivational prefixes - inflectional prefixes - valence prefix - 
		root - suffixes

The position immediately preceding the valence prefix contains the
inner subject prefixes: 1s,2s,2dp, and 1d(p). The reason this doesn't
falsify the latter version of the constraint is the the valence
prefixes might be considered derivational. To some extent they are
lexically determined by the verb root, but they also play a role in
transitivity alternations.

Of course, it also matters what level of representation the constraint
is supposed to hold of. There is, for example, an account of
Athabaskan due to Keren Rice on which the surface order is quite
different from the underlying order. On such an account, the
constraint would hold of underlying representation.


As for:

 No human language lacks a mechanism for ... number

unless I have badly misunderstood the paper he sent me, Dan Everett
says, convincingly as far as I can tell, that this is false for
Piraha.



- Bill Poser