Thresholds of Comprehensibility

[X99Lynx at aol.com]

In a message dated 7/14/01 2:12:22 AM, anaikio at mail.student.oulu.fi writes:

<< At least of earlier in this discussion, the issue was the "rate of change"
in a *language*, not in the phonological manifestations of some individual
lexical items. Are you suggesting that the rate of language change could be
measured by counting sound changes in individual words? And what "particular
words" are you referring to? >>

This happens.  In the course of these discussions, things switch around so
much, I end up on the other side of the fence.

My position all along has been that "linguistic rate of change" (the words
are Larry Trask's) is not a coherent concept and therefore deserves no
scientific "uniformity through time" status.  And that therefore it says
nothing scientifically about the timeline of IE Language.  My offer of how a
statistical rate of change might go was not so much to advocate one, but
hopefully describe how such an analysis might go - to be scientifically
valid.  Whether it's feasible or not is another matter.

Your post appears to be saying, on the other hand, two different things.

On one hand you write:

<<but these isolated examples of course tell us nothing about "the rate of
change", WHATEVER THAT IS,...>> (Caps are mine.)

Then on the other hand you write:

<<"rate of change" in a *language*, not in the phonological manifestations of
some individual lexical items... Note that counting sound changes in
*individual* word-forms is irrelevant, if the "rate of change" of the
*language* is to be measured...>>

Hopefully, you see the problem in this.  You appear to be saying that you
DON'T know what rate of change in a language is, but you are SURE that it
cannot be measured by changes in individual sounds or words.

Now, I don't know whether the latter is true or not.  I think that one would
have to do some real measuring among recorded languages to know the answer to
that.

But it is not logical to eliminate such a way of defining "rate of change in
a language", when you say you have no definition of those terms yourself.

<<Counting sound changes is usually impossible in practice, because we face
the problem of what should be counted as a sound change. Moreover, all
changes cannot be reconstructed. Note that counting sound changes in
*individual* word-forms is irrelevant, if the "rate of change" of the
*language* is to be measured....  How could the level of phonological
divergence in individual cognate items be used as a measure of the rate of
language change *in general*?>>

BUT "key indicators" are established fact in statistical science.  Yes, you
can measure and predict the movement of whole systems based on certain
"parts" of that system.  The way you arrive at what these indicators are is
not essentially theoretical.  It is empirical.  You do before and after's on
existing systems until you find those elements which best correlate with the
changes of the system as a whole.  (This is the way, for example, we
calculate the age of planetary bodies through the measurement of isotopic
half-lives.)  Then you extend those measures to unattested systems.

The same applies to the "what is a sound change" question.  For these
purposes, relevant sound change definition comes from what actually would
work best in terms of producing an accurate outcome with regard to time of
change in recorded languages.

There may in fact be no averageable correlation between time and particular
sound changes.  But that is not something you decide based on theory.  That
is something you would test, before you come to that conclusion.

So, going back to my point, I think you've reached conclusions that are way
ahead of what you can say on a simple logical basis.

You wrote:

<<It is naturally impossible to "measure" dissimilarity, because the degree
of observed dissimilarity is dependent on subjective matters.>>

I think this is wrong.  What I was suggesting was that "dissimilarity" can be
measured on an objective basis.  (Or as objective as human subjects can be
measured - which is objective enough.)

One definition of dissimilarity that can obviously be measured is the point
at which changes cross over into incomprehensibility.  And this would apply
to sounds, grammar and even syntax.  And though the measure is binary (yes or
no), binary data can carry enormous meaning when viewed cumulatively, across
a language system.

<<Yes, I of course agree - e.g., a change *w > *v obviously creates less
dissimilarity between cognates than a change *w > *q.>>

A good question to ask yourself is why you agree.  Is it because of the
difference in wave patterns on an oscilloscope?  Or is it because of the
musculature it takes to make those sounds?  Is it just "how different" they
sound?  Or how different they sound to the people who actually live with the
language?

I'd suggest that you consider that the real test of difference might be in
the ears of mathematically averaged listeners.  If listeners expect w > q,
perhaps w > v might be more "dissimilar."  (Benji Ward - who I keep going
back to - noted that there is an important difference between how a sound
change originates/developes versus how it spreads.  If w > q were somehow
spreading fashionably, then w > v might sound odd indeed.  In the ears of
statistically sampled listeners.)

Regards,
Steve Long