[Corpora-List] Corpus vs Intuition

[Yannick Versley]

> Dom Widdows wrote:
> > That's just a preamble to my main question - what, if anything, makes
> > language different from any other natural phenomenon?
Martin Wynne wrote:
> That's an easy one. Language is not a natural phenomenon.

I'd like to expand this argument: unlike physics, where we gain no better 
insight by jumping from the tower ourselves than we would by throwing stones, 
language gives us access to introspection data, which offers us 
a "preferential" way to observe facts about language.
This also means that, unlike in physics, a thought experiment can be used to 
derive authoritative facts - even though, as in other disciplines, you have 
to be careful about the experiment you're making (e.g. see different 
constituency tests and where they don't give the right result).
And of course, once you make principled thought experiments, you're no longer 
_completely_ relying on intuition.
But you can fool yourself using all discovery methods that are currently 
available - introspection, corpus search, (psycho)linguistic experimentation, 
as well as implementation/simulation studies, so it's usually important to 
(a) have some basic ideas about plausibility and (b) come up with a way of 
checking a theory using a different mode of experimentation (falsification).
What happens instead of (b) is that people tend to develop theories that fit 
well to their experimental methodology - semantic prosody, for example, if 
only based on corpus data, could end up being a bunch of different unrelated 
phenomena that could be teased apart without problems once you analyse it 
with other means.

-- 
Yannick Versley
Seminar für Sprachwissenschaft, Abt. Computerlinguistik
Wilhelmstr. 19, 72074 Tübingen
Tel.: (07071) 29 77352

_______________________________________________
Corpora mailing list
Corpora at uib.no
http://mailman.uib.no/listinfo/corpora