🚀 go-pugleaf

From alderson@cisco.com (Rich Alderson) Thu Jan  7 19:21:27 1993

>>However, if languages change as a result of some internal effort at some kind
>>of an optimizing principle, then it should be an observable universal. It
>>should be possible, then, to find this univeral principle since there are so
>>many languages that lingusists have studied.

>But they don't.  Exactly why they do change is an area of research.  Clearly,
>the primary mechanism of change is the transmission problem:  Children learning
>a language from their elders--not simply audlts but anyone more advanced in the
>language than they--do not learn to reproduce perfectly the learning of others.

	I don't doubt that it's an area of research. I don't see however
	that it's so clear that the main mechanism is the transmission from
	parent to child.


>(A theoretical question to ponder:  Do neural nets given the same *large* sets
>of stimuli in different orders organize themselves the same way?  Would another
>neural net provided with its stimuli from several such nets organize itself in
>the same way as any of its inputs?)

	It's a research area :-). As long as they output the same
	thing we don't care much to look at its internal organization.
	That's actually not very accurate since we are interested in its
	internal structure, for example in some optimization problems
	we'd like to know how we can drive it to its minimum energy
	level istead of it getting stuck at a local minimum.


>Then please do so.  So far I haven't seen a single inconsistency pointed out.

	Oy vey, or should I say Oh be..	I'm trying, I'm trying :-)..


>But don't you think it behooves you to learn a bit more what the particular
>field entails before making grand denunciations of methodology with which you
>are not familiar?


	There are a couple of fallacies here:

	1 I don't have to walk from NYC to CA to prove I can walk.
	2.I don't have to eat greasy food for 100 days before I decide
	I don't like it. A mouthful will do.
	3. I don't have to memorize every tree in the forest to be
	able to know about the forest.
	4. Being a walking encyclopedia of the minutiae of world's
	languages, I'm quite sure, has its merits. I'm not anti-knowledge.
	But it doesn't imply being able to impose a structure on them.
	This is often called information-overload. Being unable to
	see the forest for the trees is a good analogy.


	Finally let's see if I can try one more time to see if I can
	explain it differently. BTW, I've started to write up some
	things, which should be finished in a month or two. Until then,
	I have to stick to analogies and give only a brief hint of
	where things might be going wrong, as has been hinted at by
	others. Maybe wrong is the wrong word; say, overambitious and
	overoptimistic--NO! I'll say it; 'wrong!'


	I'm forced to use analogies. And I'll cut it rather short.
	
	Phenomoma occur at various levels; there is a kind of
	size-scaling; in time or space.
	Phsysics: quantum, Newtonian, astronomic.
	CS: operation,machine cycle,instruction cycle,procedure,pgm.
	Ling: nano,micro,midi, maxi (macro). Just for fun.

	HL (hi-level) regimes phenomena are based on mass scale
	measurements. Pressure exists at HL--there's no pressure of
	an atom or a molecule. Temperature similarly.
	For CS--HL measurements are things like the final output,
	or the algorith. At HL, we're not concerned about machine
	cycles. 

	For Ling what do we have ? Here's a really rough idea:
	First, there are parallel structures. It all has to do
	with semantics (obviously) and how it's manifested.

1.	Nano: phonemes, morphemes.

2.	Micro: words, meanings of words, morphology and how
	various word-forms are constructed.

3.	Midi: Mid-sized structures; NP, VP etc. and how they're
	constructed i.e. infixing, postifixing, to construct
	other than the simplest word-forms such as Adj-forms, Adv-forms.

4.	Maxi(Macro): General overall property of language; candidates
	being things like vowel harmony, tonality, how sentences are
	constructed from smaller constituents i.e. VSO, SVO etc.
	agglutination, positionality (ie. isolating or key-word
	key-morph), degree of infixing, prefixing, postfixing.


	(If you'd like to make up more levels, be my guest.)

A. How do we tell apart languages which differ by the smallest 
degree i.e. very close dialects or idiolects.
	ANS: nano level, phonemes, formants, phoneme level.
	No need to check micro, midi, macro levels.

B. How do we tell apart languages which differ greatly?
	ANS. macro-level. Chinese is tonal. Uralic-Altaic
	have vowel harmony. Chinese is strictly positional.
	Others may use key-word, key-morph concepts so that
	the VSO, SVO order is very flexible. Semitic derives
	many words by infixing simple three consonant forms.
	Turkish creates Nouns, Adj, Adv by postifixing.
	No need to check for phonemes or words. If the words
	match, they're borrowed. How do we reach this 
	conclusion ? 

There are priority levels. If there's no 'match' at the
upper levels, lower level matches are borrowings.


If the languages are very close you only try nano level. If
they're far apart, macro-level will do. How about if
they somewhat resemble each other or you're not sure ?
Well, it seems that micro-midi level become important.
Doesn't  this seem logical ?


And now this question is for you Mr. Alderson.

AT what level does the vaunted historic/comparative method operate?
nano, micro, midi or macro ?

	[This space left purposely blank.]





To say that infixing, postifixing, prefixing are not important
because Bengali is postfixing and English is both pre and post
and infixing is circular logic. Who says Bengali did not borrow words
and that it wasn't originally agglutinative and maybe derived
from proto-Dravidian ? AT what level are the comparisons being 
made ? Which aspects of language persist longer ?

To say that Irish, French and English differ as to VSO, SVO
etc so that it's unimportant is circular logic. In fact it
neatly divides them up (at the micro-midi level) into
Celtic, Italic and Germanic. Now which level aspects of
language tend to persist longer ? Macro level or micro-nano level ?
Which level constructs does the comparative method use (as if
I didn't know) to posit family trees ?

To say that Finnish has vowel harmony and is postfixing (like
Altaic) but is not Altaic and more closely related to IE
thus neither  vowel harmony nor agglutination (postixing) is important
is putting the cart before the horse ? How did it become IE ?
At what level was the comparison made in the vaunted comparative
method ? Omega Man learned his lesson but some linguists apparently
still haven't.


>Linguists *are* well aware of this.  However, we are also aware, as you seem
>not to be, that some tools are wrong for some jobs.  Without a strong back-
>ground in the history of the field, as you appear, what leads you to believe
>that you know better than the practitioners *what it is that they practive*?

	ONce again, I must interject some caution. The implication
	is incorrect. Something that seems logical will hold up
	on its own merits, mostly. I don't want to get into
	philosophy of science type argument.


	If the hammer is not suited for the job, you have to get
	another tool.

	IS LINGUISTICS DO FOR A PARADIGM CHANGE ?????????
	(WELL, MAYBE JUST UNDERSTANDING THE LIMITS OF THE COMPARATIVE
	HISTORICAL METHOD WILL DO........
-- 

	mark	

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