Perhaps this analogy might be useful. In cognitive psychology, workers
used computational models to test some of their hypotheses. Perhaps,
it might be useful to think of a spectrum from description to
effective procedure, as follows:
Description Effective
Procedure
Linguistic Mathematical Computational Programming
Model <==> Model <==> Model <==>
(Logic/Statistical/...)
with more and more (formal/rigid) details of how hypothesis or models are validated.
For many empiricalists, they can short cut the mathematical/computational models and use
programs as tools for generating observations. However, utimately, these observations and
building these tools might feed back and advance the computational/mathematical/linguistic models?
Without the knowledge of how data are generated would also restrain the investigator from examining certa
aspect of the validity of the data as well as the analysis of the data (e.g. why the data were not
as expected). One can ask the programmer about how data is obtained. But communication demands some
knowledge of programming and the programmer may not have the computational/linguistic model one has
in mind. One needs a scientific programmer - a mathematically more sophisticated programmer that
can build computational/mathematical models.
I don't know whether this discussion arose because of language is thought of as belonging to the
Arts instead of Science. In astrophysics, there are many able scientific programmer who studied
astrophysics. Is there a need to put into the curriculum of linguistics-teaching-modules more rigorours
scientific investigation and model building techniques? Perhaps, I am regarding linguistics
as a "hard science" subject.
My apologies if the above sounds too arrogant.
Regards,
Robert Luk
PS: Note that I am not taking any side on the theory-led or observation-led advances of science.
I think both can happen depending on the current knowledge.