LINGUIST List 7.1158

Fri Aug 16 1996

Sum: Grammaticalization

Editor for this issue: Ann Dizdar <dizdartam2000.tamu.edu>

Directory

1. Frederick Newmeyer, Grammaticalization: Summary and discussion

Last week I posted a message dealing with certain historical changes
that appear to run counter to theories of grammaticalization. (URL:
http://www.emich.edu/~linguist/issues/html/7-1124.html#2)

In particular, I gave examples from the history of English where it
appears that 'more grammatical' prepositions have developed into 'less
grammatical' verbs, nouns, and adjectives, prima facie contradicting
claims that grammaticalization is 'unidirectional'. I would like to
thank the following people for their replies to me: John Verhaar, Jess
Tauber, Ron Kuzar, Martin Haspelmath, Karl Teeter, Mayrene E Bentley,
J. Arends, Dave Harris, Diego Quesada, Elly van Gelderen, Tom Cravens,
and Sandra Golstein.

There were two general sorts of objections to my points. The first
objection was that some of my examples involved adverbs transforming
into N, A, V, not prepositions, or were not appropriate for some other
reason. (Martin Haspelmath gave me some very plausible alternative
etymologies for some of my examples.) Of course if my etymologies were
wrong, I accept this criticism.

The other type of objection I do not accept. This objection was that I
did not provide counterexamples to unidirectionality because what is
going on is not 'grammaticalization reversing itself', but rather is
some other process, say, 'lexicalization'. My first thought was that
pursuing this line of argumentation, there can in principle be no
counterexamples to unidirectionality, since it appears to be built
into the very *definition* of grammaticalization. But my only interest
is whether, given that in some language we find a homophonous V and P
with related meanings, we can assume that the directionality of
historical change was V > P. Evidently we cannot. This is a crucial
issue for grammaticalization studies, since so much reconstruction is
done on languages with no written history. Many African languages, for
example, have focused prominently in grammaticalization studies, where
assumptions of unidirectionality have led without exception to the P
being taken to be the later development.

In fact, I agree that certain changes that come under the general
rubric of 'grammaticalization' *are* more likely than others, though I
share little of the confidence in their 'near unidirectionality' that
one reads in the literature and the consequent certainty with which
reconstructions are put forward. Here are a couple other points, which
I hope will trigger responses. (I must stress that they have been made
by Brian Joseph and Rich Janda in recent conference presentations, and
are not original with me.)

First, I am very disturbed by the frequency with reconstructions are
used as *evidence*. Let me give an example, practically at random,
from the work that has been done on the development of future tense
morphemes. On the basis of languages for which we have written
records, Bybee and her colleagues have developed quite reasonable
hypotheses about common historical sources for these morphemes. These
hypotheses are then applied to unwritten languages to reconstruct
possible historical sources for future morphemes in those
languages. And then these very reconstructions are taken by many
linguists as evidence -- i.e. established results -- to be applied to
further theorizing about grammaticalization. This goes well beyond
acceptable practice in historical linguistics. It's a little bit like
reconstructing the Indo-European consonantal system on the basis of
notions about the naturalness of sound systems in general and then
including that reconstruction in a sample of sound systems in order to
argue that some particular systems are more natural than others!

Second, I am skeptical that even exists a phenomenon called
'grammaticalization'. There are semantic changes, some of which are
more natural than others. And there are phonological/phonetic
changes, some of which are more natural than others. I see little
value in dignifying the intersection of one subset of the former and
one subset of the latter with a label, and calling it a 'process' that
needs a 'theory' to explain it. Indeed, the fact that the semantic
processes (metaphor, metonymy), the morphosyntactic processes
(reanalysis), and the phonetic reductions can occur independently of
each other -- and often do -- is a powerful argument for taking an
epiphenomenonal approach to grammaticalization -- i.e. not regarding
it as a distinct process. In his reply to me defending the classical
approach to grammaticalization, Diego Quesada refers to 'paths of
evolution, hierarchization of functions (less to more
grammaticalized), etc., which open a word to cognition and to human
categorization schemes in general'. While I admit that I cannot show
this for every case cited in the literature, I suspect that these
properties will all turn out to be consequences of the interaction of
well understood processes from separate spheres of grammar.

One does not need a theory of grammaticalization to explain why
unbound forms are more likely to become bound than vice-versa any more
than you need a 'theory of ice cream' to explain why it is easier to
melt together a scoop of chocolate and a scoop of vanilla in the same
bowl than to reconstitute the original scoops from the melted mass. In
the former case, elementary facts about speech production and
perception will suffice very well.

Unfortunately, I'll be away from e-mail for a few weeks and will not
be able to participate further in this discussion. So reply to the
List, not to me personally. I look forward to reading the replies when
I return!

- fritz newmeyer