🚀 go-pugleaf

In article <Jan.2.18.28.38.1993.13710@pilot.njin.net> hubey@pilot.njin.net (Hubey) writes:
>In article <1i39avINN5e2@pith.uoregon.edu> delancey@darkwing.uoregon.edu (Scott C DeLancey) writes:
>
>
>> It is most definitely not syntax.  Morphology is very useful, but
>> not in the way that you seem to mean.  What you need is not similar
>> morphological patterns, but cognate morphemes arranged in paradigms
>> (e.g. the verb endings of different Indo-European languages).
>
>We are back to square one. Instead of looking for cognates in
>words we're looking for cognates of word-parts. Are we not ?
>
>Should we expect this in two languages if we cannot find
>word cognates?

They tend to occur together.  Cognate grammatical morphemes, and
especially cognate paradigms, are considered better evidence of
relationship than cognate lexical items, partly because they are
less susceptible to borrowing, and partly because coincidence of
whole paradigms is statistically more impressive than coincidence
of random words.

>> Yes, it's very difficult, and therefore very difficult to prove
>> relationships at this level.  (Conservative historical linguists
>> tend to say "impossible" rather than "very difficult", but that is
>> a controversial point).
>
>Aye, there's the rub. It hinges on what is proof. Suppose you
>use method Ma to find the maximum of a function A. Then you
>use the method Ma to find the maximum of function B. It doesn't
>work. Does it mean that you've proven that function B does not
>have a maximum or that your method is insufficient?

It means that your method is insufficient.

>The question involves whether the use of the same method
>that worked on IE languages (separation or spread which
>started about 6,000 years ago), will work to prove or
>disprove that two languages which might have started
>divergence 12,000 years ago.

It doesn't, at least so far.  That's the state of the art.
The fact that the best methods we have available are not sufficient
to a particular task is good motivation to try and develop other
methods that will be sufficient.  It is not motivation to resort to
other methods that we know are not sufficient.

>> No.  First of all, morphological typology can also spread from
>> one language to another, so similarity doesn't prove common
>> inheritance.  
>
>What then is inheritance. It's a term borrowed from biology

Possibly, but I wouldn't be too dogmatic about it.  The development
of models of linguistic descent and relationship slightly predates
the development of models of genetics and evolution.

>and mammals have two parents. 

Languages are not mammals.  In particular, linguistic evolution
is more Lamarckian than Darwinian, in that acquired characteristics
(acquired, for example, through contact with a neighboring language)
are transmitted from one generation to the next exactly like inherited
characteristics.

>If a language borrowed morphology
>from two languages, is it not a descendant of both ?
>How do you judge ancestry ?

If a language _borrowed_ morphology from two languages, it is not
a descendant of either, as the term is used by linguists.  But you
were talking (as far as I could tell) not about borrowing morphology
per se (i.e. specific morphemes) but morphological typology. Here's a 
concrete example:  Greek, Macedonian-Bulgarian, and Rumanian--representing,
respectively, the Greek, Slavic, and Italic branches of Indo-European--share
certain morphological structures which none of them shares with other 
members of its branch.  But what sense does it make to speak of any
of them as "descendants" of any of the others?
>
>Yep, there's the rub again. Would we not expect all of these
>word orders to be spread all over the world ?

Well, they are, pretty much.  What's the point?

>If not, then is
>there a natural direction of evolution so that some die off
>and others absorb them ?

That's a strange way to talk about it.  Languages can shift from one
word order pattern to another for a range of different reasons,  
including but not limited to areal influence from other languages.

>What then happens to parenthood
>of languages ?

Huh?

>What are the statistical measures which we can
>use to 'disprove' relatedness? Is lack of proof equivalent to
>disproving ? 

There aren't any.  No, it isn't.

>Would we not also expect the pattern of any given
>language to be shared by related languages? 

If we expect this, we will be wrong.  E.g. we know that Irish,
French, and Bengali are related, but Irish is VSO, French SVO,
and Bengali SOV.  This kind of variation would be mildly surprising
among closely related languages, but is completely unsurprising
among languages related at the level of Indo-European.

>Do we always assume
>unrelatedness until proven, which would always heavily bias
>toward what you assume since no criteria other than word
>cognates is acceptable ?

Actual practice may vary, but the correct answer is that, as in
any other discipline, we wait for legitimate evidence before
we start to construct hypotheses.

Scott DeLancey               delancey@darkwing.uoregon.edu 
Department of Linguistics
University of Oregon