Rate of Change: A Closer Look

[X99Lynx at aol.com]

In a message dated 6/7/2001 11:19:18 PM, larryt at cogs.susx.ac.uk writes:
<< We can imagine a lower rate of change, but there seems to be no good
evidence for it.  Without some hard evidence, we have no right to assume
that linguistic change was any slower in the remote past than it has been
in the historical period.  Such an assumption would violate our linguistic
Uniformitarian Principle: languages and speakers in the remote past behaved
much as languages and speakers have behaved in the historical period. >>

While I respect the trained linguist's ability to "feel" the relative age of
a language, I must be firm in pointing out that "rate of change," as Prof
Trask is using it, has little scientific validity - at least without
confirmable numbers that anybody can double-check.

That's the price of admission if you are going to start claiming scientific
validity and a "linguistic Uniformitarian Principle."

What specific changes is Prof Trask referring to?  How is the "rate" is
measured in terms of IE languages?  Are all the changes in a language added
together?  Or are all the changes irrelevant and do only certain changes
count?  Which changes and why?  How much do each of these changes equal in
years?  How much is a sound shift worth in years?  How do you measure the
difference between two morphologies and how do you calculate the time it's
been since they were once one morphology?  Why do any of these better reflect
the actual rate of change than let's say rate of borrowings?

Do you count up all the cognates in two "daughter" languages, divide by the
total words in each and then multiply by ten years?  Or do you count the
sound shifts in "core" words, whatever that means?

I helped prepare a presentation before a board of physicists at the National
Science Foundation a few years back.  You got to know if you went in front of
them with phrases like "rate of change", you'd be expected pinpoint exactly
what you were measuring, how you measured it, how many times you measured it
and convey a clear assurance that if anyone else measured it they would come
up with the same result.

If they expect this kind of specificity about "rates of change" in quantum
mechanics and animal behavior experiments, they can expect it in historical
linguistics.

Some observations:

1. The basic rate change formula for function [f] in the time interval [a,a +
h] is:  Rate of Change = f(a + h) - f(a).  This is true in physics,
psychology and pina coladas.  If there is no way to measure the quantitative
difference between languages, there is no way to validly measure "rate" of
change.  ("Rate" being defined in English and in science as a change in unit
quantity per another unit quantity.)

If we are talking about sound changes that separate two daughter languages,
we should be able to show that those specific sound changes took a certain
number of years, days, hours and with historical evidence.  If we are
claiming that that rate is somehow universal - justifying the application of
a uniformity assumption - we should be prepared to show that the same rate
per sound change can be observed in a large enough number of other languages.
 Or explain why they vary.

2.  An example.  In biology, we talk about the rate of mutation.  That is the
rate of particular chromosomal change per unit of time.  And here we actually
count the chromosomes that change per unit of time and can even project from
that correlation.  And it is also understood in what way that rate is
variable.  For example, when we artificially induce mutation, the rate at
which such change occurs is actually correlatable to the strength of the
radiation.  Or, as another example, at the age of fifty a man's sperm cells
will contain three times as many mutated chromosomes as they do when he is
twenty, plus or minus .062.  These are different quantifiable vectors that
affect the rate of change.

2. Others on this list invoke the close "resemblance" of cognates and
inherited morphology to suggest a lack of change.

In point of fact, this might be the worst indicia.

An example: Most mammals had body hair, a distinguishing "inherited"
characteristic.  So what does body hair tell us about "rate of change?"  50
million years ago, most mammals had body hair.  About 20 million years ago,
several branches of mammals developed that did not have body hair.  Elephants
who developed before humans have less body hair than humans. But today 40
million years later, most mammals still have lots of body hair.  So, the last
thing we would use to measure "rate of change" is the identifying "cognate"
of body hair.  Because it is sure to tell us the opposite of what we are
looking for.  It might even lead us to dubious conclusions like Greek and
Latin were in some way mutually comprehensible in 1000BC.

3. How much has English changed in the last thousand years?  Can that rate be
represented in a number?  How is that number derived, what specific feature
in the languages that changed is reflected in the number?  A sound change?
How does one measure sound changes?  How many sound changes? Why are other
features not being measured?  Is the rate of borrowing a more accurate
measure of rate of diversity among related languages?

How does the formula derived from changes in English in 1000 years measure up
when applied to French in 1000 years?  To Old Norse?  To Slavic?

How do you propose to measure the rate of change in modern preliterate
languages?  How do they measure up against written languages?  Do they give
any evidence that rate is affected by the acquisition of writing?

Are these results reproducible?

Can any scientist sit down with the same data and apply the same formula and
come up with the same results?

4. As I said, this is not meant to be disrespectful of linguistic expertise.
It is however what is required before any claim of any Uniformitarian
Principle can be seriously claimed in any scientific sense.  Before you can
make any claims about knowing "rates of change" in the preliterate past of IE
languages, you should at minimum be able to show some coherent, reproducible
formula for present IE languages.

But of course if Prof Trask's use of "Uniformitarian Principle" was
metaphorical or casual, and not meant to imply any real level of scientific
certainty, then I can have no objection to it.

Regards,
Steve Long