Tom Payne tpayne at OREGON.UOREGON.EDU
Sun Sep 28 05:01:34 UTC 2003

Right! A statistical tendency is just an intriguing pattern until you
come up with a principled explanation for it. And if the principled
explanation explains more than one statistical tendency, all the better.
For example, "iconicity" is one general principle that accounts for
several structural statistical tendencies in the world's languages. If
you posit a line of reasoning something like 1) speakers intend their
messages to be interpretable, 2) iconicity makes messages more
interpretable, and 3) language structure evolves under these pressures,
then you can explain (or perhaps "motivate" is a better term) many
statistical universals, such as: the tendency for phonologically larger
structures to express semantically more complex concepts (I'm thinking
of causatives, reflexives, etc.), certain kinds of split ergativity, and
so forth. 

This is why I like to say I do Typological/_Functional_ linguistics.
Typology is the descriptive part -- "What are the intriguing structural
patterns out there?" But the job isn't finished until the patterns are
explained according to some coherent and general principle, anchored in
cognition, neurology, communication or some other domain independent of
language structure itself.

Tom Payne

-----Original Message-----
From: Discussion List for ALT [mailto:LINGTYP at LISTSERV.LINGUISTLIST.ORG]
On Behalf Of Leon Stassen
Sent: Friday, September 26, 2003 3:09 PM
Subject: pseudo-generalizations

The point that Matthew Dryer raises in the current discussion on method
typology is of course a very real one:

>Hopefully, the problem is obvious: it is natural for 5% of
>to come out as statistically significant at the .05 level, not because
>are valid generalizations, but simply due to chance.

Since there is no real way of telling the useful generalizations from
spurious ones, it is very probable that Matthew is also right on the

>I predict that pseudo-generalizations of this sort will be published in
>typological literature, and there will be little way to distinguish
>from valid generalizations.  Perhaps some already have been published.

But then, this is not very different from the real-existing situation in
other sciences. I don't think there's a principled way to exclude
pseudo-generalizations, but fortunately, science itself has a way of
eliminating them, namely by letting them wither gradually. The key-word
"explanation" here: any generalization for which, after a certain amount
time, no explanation in terms of higher, abstract principles can be
will come to be disregarded and will disappear from the theoretical
discussion in the field. Of course, it is possible that this "selection
process" will also sacrifice generalizations that, by later generations
scientists, will be found to be highly relevant; but that's, I fear, the
way the scientific cookie crumbles.

Leon Stassen.

More information about the Lingtyp mailing list