[Q] Universals, Statistics

[Daniel Everett]

Folks,


Everybody is right. Use all the deductive reasoning you want and all
the inductive thinking you can do, up to usefulness.


I don't know a single linguist who reasons deductively or inductively,
though, except on odd-numbered days. The American Pragmatists, sparked
by Peirce, who considered himself a Pragmaticist (to show William James
that he didn't like the fact that James was getting credit for
Pragmatism, a term Peirce introduced, nor the direction that James took
Pragmatism), argued that people - scientists, actors, morticians, and
others - reason their way through the world with the poor tools that
evolution has bestowed upon them, making guesses (sometimes called
abduction or retroduction, if you want to get all latinate) and seeing
how far these guesses will get them towards their goal.

Chomsky's guess is that there is an absolute Truth for Language called
UG (there are plenty of quotes in recent interviews and throughout his
career on this). Some typologists believe that the best we can hope for
are tendencies, statistical generalizations and relative knowledge,
rather than Truth. A few might even agree with me that writing grammars
informed by cultural knowledge and language learning is the core task
of the linguist and that the particulars of each grammar as it emerges
in symbiosis with culture and cognition (assuming that the latter two
are indeed separate) are crucial to understanding our species and
knowing the world better, i.e.   that the differences between languages
are every bit as important and revealing about what we want to know as
the similarities or 'generalizations'. (Some poor souls might even
believe that there is some usefulness in implicational universals, e.g.
"If nasal vowels ---> oral vowels" (universals of this type are of
course an embarrassment: since the prodosis is true in all instances,
the apodosis can be anything, rendering such statements
meaningless/useless or always true, which is the same)).

The danger with deductive reasoning as a goal is that can make the data
irrelevant, leading to a poverty of stimulus for the theory, rather
than the child. The danger of random inductive reasoning is that it
leads to a smorgasbord approach to linguistics research, where we just
go through the line of grammars on our shelves and help ourselves to
what looks good, without caring about where it comes from.

But I think we all know this already and we all just do our best to
understand the things that interest us, using and doing what is useful
to us, rather than worrying terribly (or even needing to worry) about
the subjacent epistemology. Which is to say, again, that neither the
deductive nor the inductive approach is wrong. As Anwar Sadat said
about the Middle East - the conflict involves two rights. Not that
there is any linguistic conflict.

So it may be useful to consider these kinds of things from time to
time, to be able to justify to ourselves at least why we do business as
we do. But it has really very little impact on who we are and,
therefore, on individuals' ways of studying languages.

-- Dan



------------------------------------------

Daniel L. Everett
Professor of Phonology
Postgraduate Programme Director
Postgraduate Admissions Officer
Department of Linguistics
The University of Manchester
Oxford Road
Manchester, UK M13 9PL
http://ling.man.ac.uk/info/staff/de
Fax: 44-161-275-3187
Office: 44-161-275-3158