LINGUIST List 5.607

Tue 24 May 1994

Disc: Language laws

Editor for this issue: <>

Directory

1. Stephen P Spackman, Re: 5.534 Language laws

Message 1: Re: 5.534 Language laws

aarbakkeisl.uit.no (Jon Hareide Aarbakke) writes:

|Re. b), presumably the reason that language laws do not produce the same
|results everywhere is that they have a different epistemological status
|than for instance physical laws. Language laws are descriptive rather than
|prescriptive, and they have no predictive power. What linguists tend to
|call laws are simply the observation of some uniformity in or across
|languages, and to account for their existence a law is posited, post
|factum. This does not explain why things are as they are, it merely
|introduces an economy of description.

I don't think this is true at all. It is important to remember that
physics is, on the whole, not at all predictive either, and things go
wrong there in similar ways and for similar reasons. The real difference
between the situation in linguistics and that in physics is that in
linguistics we might, if we're lucky, know of a dozen or even a hundred
cases where a given situation arose, and after taking all the other
circumstances into account there are four of five unarguable instances
of a phenomenon left. Then, perhaps we have a 'law'.

In physics, when you (say) drop a handful of gravel and it - quite
predictably - falls to the floor, you are actually dropping something
like ten to the twenty-fifth particles (perhaps a lot more, depending
what quite we're counting), and seeing how the long-term, large-scale
tendencies of the things work out.

When you look at the physics of individual objects, it's just like
linguistics. Last I heard, they *thought* they'd seen the new quark, but
they weren't entirely sure. They'd developed a method that if repeated
enough would definitely almost certainly eventually produce one, perhaps
- though other circumstances would *probably* conspire to prevent our
noticing.

For comparison with bulk physics, observations like "phonological
reduction happens" are still small potatoes in terms of the amount of
data involved - but here you begin to see how predictiveness-in-bulk
emerges.

In the fringe cases, where there aren't enough data for bulk effects to
hold sway, but there are enough to build good models, physics tends to
produce results like estimates of individual _stability_. Perhaps if we
want to reassure ourselves that linguistics _can_ have predictive power,
we should try our hand at predicting the relative stability of some
particular phenomena on the timespan of a few decades, and see how we
do. My guess is that we know enough to score some remarkable successes
(along with a lot of spoiled data and statistically insignificant
results).