In order to understand what language is and how it works,
research of many types is relevant.
Linguistics of course: Even if you don't buy any of the theories,
you can't ignore the descriptive work which keeps you honest when
you make claims about Language on the basis of 2 languages belonging to the
same family.
Psycholinguistics: How do people recognize words? How do they
access words in production? How are sentences parsed?
What stages do children go through in learning
words? Psycholinguistics by itself can't answer most
of these questions, but at the very least we need their data.
NLP/CL: Most computational models were not designed to tell what
people do, but they can sometimes tell us what is possible or
impossible for a particular kind of device.
Neuroscience: Language takes places in nervous systems.
How much longer can we use the excuse, "Well, we still know almost
nothing about the brain anyway"? "We" may know nothing about the
brain, but somebody out there probably does.
The rest of cognitive science: Try as we might, it seems impossible
to separate language from phenomena like short-term memory,
analogical reasoning, categorization, attention, spatial cognition,
auditory processing in general,
mental models, social cognition, even visual processing.
Dynamical systems: The nervous system, the individual in its
environment, and entire societies are apparently best understood as
dynamical systems. Language (in the brain/mind, within particular
social contexts, as a cultural entity changing through time) can't be
far behind.
Biology: Can we understand what human language is without placing
it within the larger context of communication systems
and within an evolutionary context?
It may not be fruitful to slice off components of language
and study them in isolation.
Syntax and semantics seem inextricably intertwined.
Phonology is motivated by constraints on the processing
of words and, to a lesser extent, phrases, and on the
need to store large numbers of morphemes in long-term memory.
All sorts of information is brought to bear, more or
less simultaneously, in the analysis of sentences.
Out of context, sentences may be uninterpretable to people.
Language is a lot like the elephant being examined by the
blind men, except that each of the men suffers deprivation of
a different sense.
What is to be Done?
Stick close to the periphery.
Take children seriously.
The form of the end-product depends on how it got there.
Children develop, in a more or less continuous fashion,
and children differ dramatically from each other.
Take The World Out There seriously.
At least we can see it, touch it, and hear it.
Take Symbols seriously; you can't deny words.
Question authority.
The other person's primitives (word,
relation, negation, constituency, composition,
etc.) may not be things that should be taken for granted.
The tasks themselves may need to be redefined.
Think vertically, rather than horizontally.
Relate different "levels" to each other without trying to accommodate
everything at a given level.