Uniformitarian Principle

[Douglas G Kilday]

Phil Jennings (2 Jul 2001) wrote:

>The basic subject here is the "rate of change" of human languages, and
>whether that rate itself changes (accelerates / decelerates) over time, as
>a consequence of other factors that have changed over human history.  In
>physics, we are pleased to know that the rate of (locational) change of
>falling objects increases with mass and decreases with distance, in
>accordance with laws described by Isaac Newton.  Gravity is a constant,
>but it is contingent on mass and distance.

This is a rather severe mangling of Newtonian mechanics, and in fact
contradicts Galileo's observation that gravitational acceleration is
independent of the falling object's mass, which Newton neatly explains. Next
time, consult an introductory textbook of calculus-based physics.

>Philosophically, language belongs in two places, (1) in the individual
>toolbox of each language speaker, and (2) in the community of those who
>use language to understand each other.   If the rate of change of human
>languages is governed by laws to do with individual ears, brains, mouths,
>vocal cords, et cetera, and these physical attributes have not changed
>over thousands of years, then the rate of change of human languages can
>intelligently be assumed to be constant.

That wouldn't be a very intelligent assumption. Nature is full of
discontinuous processes like earthquakes, landslides, and lightning bolts
which do _not_ involve any change in physical laws.

>If the rate of change of human languages is governed by laws to do with how
>often humans talk to others in their community, to strangers, to people with
>different areas of expertise, and so forth, then as society gets more complex
>and populations increase, the rate of change will accelerate or decelerate
>through time.

It has been tacitly assumed that a _single_ rate of change is valid.
Language has many aspects, and there is no _a priori_ reason to suppose that
the rates of change of all subsystems of a given language are functions of a
single parameter. Nor is there an _a priori_ reason to assume or imply that
rates of change should accelerate or decelerate smoothly.

>Larry Trask has given us the history of Basque as an instance where the
>rate of language change decreased as Basque society grew increasingly
>engaged with a complex world.  This is certainly the opposite of what I'd
>expected, in arguing for a gradual increase in the rate of change of
>languages over time, which is what I'd theorized in a prior post.
>However, in a contrary way, it is also evidence for hypothesis (2),
>substituting deceleration for acceleration.  I suspect that Larry Trask
>would like to be armed with a hundred instances of rates of change
>veering one way or the other without respect to any societal factors
>whatever.  This would be evidence that rates of change are truly random,
>and uniformitarianism is the best way to smooth across several millennia
>of random ups and downs.

No, that would merely be evidence that rates of linguistic change are
independent of societal factors. It wouldn't tell us that the rates are
truly random (i.e. independent of _all_ factors).

Larry Trask's example of Basque, John McLaughlin's example of Comanche, and
the well-known Great Vowel Shift of Middle English have two things in
common. They involve systematic shifting of phonemes over a finite interval
(50-100 years) and no apparent correlation with external (non-linguistic)
factors. This type of systematic articulatory shift gives quantifiable rates
of change, since the average position of the relevant articulatory organ for
each phoneme actually moves some slight distance per decade in the standard
speaker's head. These shifts also appear to be practically decoupled from
other ongoing _linguistic_ phenomena like syntactic shifts and lexical
turnover. Hence they provide a good opportunity for adapting a physical
model to one particular class of linguistic changes.

It is qualitatively clear that the rates of change in the phonemic shifts
mentioned are strongly non-uniform: the episodes of rapid shifting have
definite beginnings and endings. Someone has used the term "punctuated
equilibria" from the Eldridge-Gould model of speciation, which in my opinion
doesn't belong here. Analogies between evolutionary biology and linguistic
change are very poor, not least because language is inherently incapable of
reaching equilibrium. Instead, phonemic shifting suggests classic stick-slip
behavior, or Reid's elastic-rebound theory of earthquakes. In a given
language, certain phonemes are under "stress" to change position, but are
held in place by "static friction". At low values of "stress", the "force"
tending to move the phonemes is balanced by an opposing "frictional force".
At some higher value of "stress", this "static frictional force" is overcome
and movement begins. Its indefinite acceleration is prevented by "dynamic
frictional force". As the phonemes move, the "stress" is reduced until its
"force" is overcome by the opposing "dynamic frictional force", bringing
things to a halt, back into the regime of "static frictional force".

I used the annoying quotation marks because the actual quantities in
phonemic shifts are only _analogous_ to the mechanical ones whose names I
have borrowed. Nevertheless the linguistic quantities are real and, in
principle at least, measurable. Instead of trying to beat professional
comparativists at their own game, this is the sort of thing Steve Long
should be doing, since his voice has been the loudest in calling for the
quantification of historical linguistics.

>This is an application of Heisenberg's uncertainty principle to language
>change.  We don't know => we can't know => there are no consistent factors
>out there to be known.  It's odd to find linguists reconciled to
>randomness re. rates of language change, when they delight in systems and
>consistency in all other areas.

This mumbo-jumbo is no application of Heisenberg's principle. Consult a
junior-level or senior-level modern-physics textbook, not watered-down
popular Sunday-supplement hogwash.

>Also, the evidence in Larry's hundred instances is anecdotal in the
>absence of a consensual measuring system, and some anecdotes, however
>entertaining, will be wrong.

Well, I hope the examples of systematic phonemic shifts offer at least a
starting point for an objective measuring system. If it turns out possible
to deal with simple phonemic shifts this way, then similar applications of
physical models _may_ work for some of the other aspects of historical
linguistics.

DGK