More questions on grammatic(al)ization

[Susanna Cumming]

Folks,

I resonated with all of Spike's questions about grammatic(al)ization. I
don't have any answers, but I have a few more questions of my own. These
came up for me because I'm teaching a historical course for the first time
in a few years this quarter, & we're doing grammaticization now. In
catching up with the large amount of recent literature out there and
trying to present it to my students, I noticed some surprising trends:
notably that people, for the most part functionalists, when writing about
grammaticization seem to display some assumptions and hidden theories of
grammar and meaning that I don't think of as particularly functional. It
may be that I'm just more radical relative to the functional community
than I think I am! But in order to find out if that's true, I wonder what
people think about the following issues:

In my opinion, one of the reasons that there are so many different
understandings of and ways of using the term "grammatic(al)ization" is
that people have different ideas about what counts as "grammar".
Unfortunately, when writing about grammaticalization, people rarely make
it explicit what their theory of grammar is. In the easy cases Spike
cites, grammar=morphology and that's that. For all the cases where the
result isn't an affix, the question is harder. People usually talk about
"open" and "closed" classes here, on the understanding that closed classes
are "more grammatical" because they are listed in the grammar instead of
in the lexicon. But I think many linguists today reject the hard-and-fast
distinction between "lexicon" and "grammar" that we used to operate with,
and it's not far from there to rejecting the distinction between closed
and open classes. On the one hand, there are lots of reasons to want to
let items like prepositions, conjunctions and even complementizers to be
listed in the lexicon -- since they are clearly not simply "structure
markers", but rather they have the kinds of semantics we associate with
lexical items -- and on the other, why should we consider "verb" to be an
open class in English, for instance, when virtually every high-frequency
verb has its own unique set of distributional properties?  The insights
about the highly idiosyncratic grammatical distribution of all kinds of
items brought to us courtesy of construction grammar should also shed some
doubt on the open/closed opposition. So: if you use the term
grammaticization, what's your theory of what is and isn't "in the
grammar?"

On the semantic side, there is of course some disagreement about how to
characterize the changes that have been labeled "grammatic(al)ization",
but many linguists subcribe to a "generalization" / "bleaching" /
"abstraction" approach. This worries me too, since all these terms seem to
presuppose an old-fashioned theory of semantics which I wouldn't otherwise
have attributed to many of the linguists who use these terms: meanings as
feature bundles arranged in a taxonomy, such that features are added as
one moves down the taxonomy from more general to more specific.
Grammaticization then is movement up the taxonomy by removing features:
it results in "less" meaning (as implied in the laundry metaphors, i.e.
weakening bleaching / fading) and ergo more "general" or more "abstract"
meaning. WHile this approach to meaning works pretty well in some areas of
lexicon -- especially biological domains -- there are many others where it
is much more problematic to build taxonomies and/or do feature analysis --
especially anything which isn't a noun. Question: is there an approach to
semantics within which the notion of "bleaching" makes sense that isn't a
features-and-taxonomies theory? (Some linguists, of course, think of the
changes involved in grammaticization as a change in kind rather than
quantity of meaning; they're exempt from this question.)

Finally, here's one for the grammaticization-is-syntax-from-discourse
folks out there: how do you know when grammaticization has happened, that
is, when something that was once a "loose pattern" has become a "tight
construction"? SHould you be able to tell from looking at spoken discourse
data when a particular stretch of speech is to be seen as a token of "a
construction", and if so, how?

Thanks to anyone who can provide some guidance here that I can share with
my students!

Susanna