Grammaticalizaton: Process or result?

[KATZ, A]

  As an addendum to Spike's query, I would like to point out the following
issue:

   Is grammaticalization a process or a result?


If it is a process, then it can be identified by the sorts of changes it
effects on the linguistic material upon which it operates. So, under a process
definition, grammaticalization might be something like:

     The process by which linguistic units become

        (a) reduced phonologically
        (b) bleached semantically
        (c) obligatory instead of optional
        (d) dependent rather than independent
        (e) arbitrary rather than motivated
        and
        (f) recruited as members of grammatical paradigms.

   Under a process definition, every linguistic unit can be judged to have
undergone X degree of grammaticalization between points A and B in time, but
saying that it is grammaticalized is meaningless without reference to change.


On the other hand, if grammaticalization is a result (or condition), then
units can be judged as more or less grammaticalized synchronically, but
this sheds no light on their history. (That is, a unit may become an isolate
after undergoing "maximal" grammaticalization under a process definition.)

   If we opt for the result definition, there is really not much difference
between saying that a linguistic unit is `grammatical' (as opposed to
`lexical') and saying that it is `grammaticalized'.

  If we opt for the process definition, we may find that after undergoing a
considerable amount of grammaticalization, a unit becomes less grammatical
(and more lexical) than when we started observing it.

  Either way, this presents difficulties for us if we hope to use
grammaticalization theory as a tool for recovering earlier stages of language
and for drawing conclusions about history from language typology.