🚀 go-pugleaf

In article <C0pHKw.12z@spss.com> markrose@spss.com (Mark Rosenfelder) writes:

> Theories generally aim to explain the data, not to be "logical"; but we'll
> let that pass. 

You don't have to if you don't want to. There's the story of the
statistician who found a correlation between teacher's salaries
and drinking levels. Since drinking probably won't increase one's
salary or work competence, he concluded that high salaries cause
drinking. Is this a good theory ? It explains the observed facts.

 Your proposal is on the table: to test relatedness of
> languages, use the "nano level" (phonology and morphemes) for closely related 
> languages, the "macro level" (typology) for those that are far apart,
> the "micro-midi level" (words, morphology, syntax) for those in between.
> 
> Of course, the first problem here is circularity: "If the languages are
> very close you only try nano level."  How do you know they're close?
> If you already know, what is this procedure intended to accomplish?


There's no circularity. There are priority levels (or weights if you
prefer). If the syntax and other hi-level structure of Americanish,
RP and Australian Engl. did not show extreme closeness you wouldn't
be testing at the nano levels to discern the difference.


> Next, since you have different evaluation procedures for "close" and for
> "far apart" languages, what do you do when the procedures give conflicting
> results? 

Punt.  Linguistics is an experimental science. Simply constructing
equations is not what it's about. I can think of constructing
several different ways right now. They'll have to be tested and be
'logical'. I don't mean formal logic but it can be included. We
all use logic, including linguists. Comparative/Historical method
is based on probabilistic reasoning--just not so formalized--
at least I haven't seen any.


> I'm afraid your proposal very quickly runs into some intractable data here.
> Namely: Typology can change much quicker than phonology does.  The slow
> and overwhelmingly attested change from Latin to French (etc.) shows this.

Sorry to disappoint people but it doesn't. There's a difference
between change of structure and movemements within the structure.
Mathematical objects exist in their own mathematical spaces. The
structure of a language would constrain its phonological word
forms to within specific volumes in n-space. If the words move around
within this volume--which is largely what happened in the Vowel
Shift, it only shows the crowding effects of new words trying to
occupy occupied spaces. The collisions--phonological and semantic-
in this space forced a reorganization and reoccupation of their
spots.

There are statisticians in this group who can explain things but
I'll give another analogy. A probability law of a stochastic
process describes the structure of the flow of the process. A given
initial set of points (like a cloud--a density) moves in the n-space
constrained by its probability laws. It actually obeys fluid like
equations and there are ways of explaining it.

> Latin is largely a verb-final language; French is basically SVO.  (Caveats 
> apply on both sides; Latin had of course fairly free word order; while spoken
> French is arguably turning into a different beast.)  Or consider how 
> the highly inflected Anglo-Saxon turned into highly isolating modern English.
> Hock (_Historical Linguistics_) also has a nice description of important
> typological changes in Sanskrit and its descendants.

Nothing to do with anything I said. Casual word borrowing can be 
accomodated by a language by changing the word to fit into its own
private phonological space (structure). If the word borrowing becomes
more massive, it'll cause crowding and might make cyclical shifts of
the vowels to accomodate. If the borrowing is even more intense, it
might spread to higher level changes. 

Furthermore agglutination, positionality, etc can be all put into
a linear scale based on flexibility. There are trade-offs. If several
germanic languages are vying within themselves and with the underlying
non-germanic population trying desperately to learn the words  so
they can at least get along in some pidgin-language, it doesn't
surprise me in the least that case gets slurred-becomes irregular
to the point of becoming useless. Positionality then becomes important
since it now carries the added burden that the case provided.

I'm only giving these examples because people keep giving examples.
I would like to have kept quiet until I got to write it all up.
I have about 15 pages now. Hopefully before the semester starts
I'll have about 30-70 pages.

Similar comments and arguments can be easily made about shifts
in word order.

The general comment is that the languages try to disambiguate. If
this is done via word order (positionality), case (partially
positional, partially key-morph),and agglutinate (takes care of
both prepositions (key word) but without the problem of word order).
In times of great change, what happens to which depends on who's
top dog, who's doing the learning, at what level between these 
extremes the languages are, and some other things which
I'm sure every linguist already knows about.



> There's also Sprachbunden to worry about; unrelated languages which happen
> to be geographically similar can come to share similar syntactic features,
> while retaining separate vocabularies.  

We don't know which way until we first make some plausible arguments
as to what kinds of things are more general.

Here's a short explanations of the general idea:

When systems get disturbed, the time it takes for them to settle
down is called relaxation time. When a system is continuously being
disturbed (by random forces), you cannot observe the relaxation time
directly since it never settles down.  You can do it indirectly
however by using the 'correlation length'. It comes from a
mathematical procedure first used in turbulence. You compute the
autocorrelation of the phenomenon and the relaxation time will
show up in the correlation length.

What is it?  It has to do with being able to give some general 
indication of the time scale on which certain phenomenon occur.
For example, searching my memory--(I think) some Indo-chinese
languages (Burmese, Cambodian ?) were found to be tonal and
dumped in the Chinese category. Then some other smart fellow
comes along and says; No, they share some features with Hindi (I
think.)

What to do ? Well, based on all the intuition and knowledge that
years of study gave you (impersonal you) try to come up with
some plausible arguments as to which kind of phenomenon persists
over longer time periods and is more resistant to disturbance. This
is the relaxation/correlation time idea. If tonality is a
more persistent phenomena than word order or propensity to pick
up affixation, then it must have been originally more like Chinese.
If it's the other way around, then somehow it picked up tonality
from Chinese at a later date.

Once again, I'd like to remind people, that the idea of a priority
based scale for measuring 'closeness' is being used. Whichever
persists longer gets a heavier weight.

None of this is against any of the principles of either the
historical or the comparative method (implicit or explicitly),
and neither is it anti-logical.


> The comparative method's reliance on lexical items and paradigms in 
> determining the descent of languages is thus eminently reasonable; these are 
> the things that change the slowest in a language ("slowest" here resting not 
> on additional linguistic theory, but on the observed history of real 
> languages).

It looks like we're in almost-agreement. Do the facts justify
the conclusions ?

Did the word borrowing (massive scale) change some mid-level
phenomenon, or was the mid-level phenomenon already there and
all we can do is throw it away because there doesn't seem to
be a pattern.

Final question. Let's look at biology and archeology. Humanoids
probably evolved around 0.5 - 1.5 millin years ago. They must have
started spreading at that time. If they started grunting around
100,000 years ago, then languages would have had to evolve
independently. That means the number of languages must have
been decreasing for the last 100,000 years. But if we look
at the conclusions of present day linguistics (as some people
claim) there are only a few language families (say 30-40)
and all the languages are descended from them. This means
as we go backwards in time we go back to the Tower of Babel
essentially. The Nostraticists apparently want to go back
to Adam and Eve.

It seems to me that languages have been 'converging' not
diverging. This implies that the great IEpeople-whoever they
were--forced somehow unrelated languages to coalesce. And
following up on this, it looks like it's a history of word
borrowing--on a massive scale--which somehow induced whole
groups of languages to be more similar.

Any opinions, from anyone ?
-- 

	mark	

   hubey@amiga.montclair.edu         hubey@apollo.montclair.edu
   hubey@pilot.njin.net               ...!rutgers!pilot!hubey