diachronic functionalism

[Martin Haspelmath]

Joan Bresnan asks an important question, referring to my earlier
posting:

>> The usual functionalist explanation for both cross-linguistic
patterns and
>> language-particular regularities is that they show the effect of
>> diachronic change.
>
>But precisely how?

The basic idea is that functional factors apply in performance. Take the
devoicing example again: Voiced obstruents are harder to pronounce than
voiceless ones, especially in (syllable- or word-)final position.
Speakers are thus constantly "tempted" to devoice consonants (partially
or completely) under these conditions, and sometimes they give in to the
temptation. This may then spread throughout the speech community and
result in a language change, such as that from Old High German (e.g. tag
'day') to Modern German (Ta[k] 'day'). Since the phonetic factors are
universal, we find languages that have completely lost voiced obstruents
(if they ever had any), as well as languages that have devoicing only in
final position. We find no languages lacking voiceless obstruents,
because there is no possible diachronic change that could give rise to
such languages.

The source of the universals is thus in performance and (hence) in
diachrony, and they are reflected in competence only secondarily. The
prediction that this view makes is that impossible languages (e.g. with
initial devoicing, or with only voiced obstruents) should NOT be
unlearnable, but that they should be diachronically unstable.
Unfortunately, learnability experiments are impractical (and perhaps
unethical), so the issue is so hard to resolve.

Let me cite Joan Bresnan again:

>In a nutshell, the analysis of final devoicing in Dutch predicts the
>existence of the well known typological asymmetry in obstruent voicing
>across languages.

The same applies to the diachronic-functional view, except that we would
talk in terms of "explanation" rather than "analysis". (Of course, in
both approaches the prediction is not literal, because one would not
have adopted this particular analysis/explanation for Dutch without the
knowledge of the typological asymmetry.)

>It is because the OT constraints are universal and not
>language-particular rules or parameter settings, that this deep
>connection between typology and language-internal "grammar" is
>possible.  You literally cannot do the (real) OT grammar of any
>particular language without doing typology.

This would be an advantage only if it turned out that that's the right
way of "doing grammar". And diachronic functionalists (such as Matthew
Dryer, Fritz Newmeyer, Joan Bybee, Joe Greenberg, Bill Croft) would
reject the view that anything is gained by subsuming language-particular
description and explanation of universal tendencies under a single
theory. (Or in Newmeyer's case, he already seems to be convinced that
this is impossible.)

Martin

P.S. Joan Bresnan seems to use the term "evolutionary" where I use
"diachronic". I think this terminological usage should be avoided,
because "evolutionary" is ambiguous between "phylogenetic (Darwinian)
evolution" and "glossogenetic evolution" (i.e. diachronic change").

--
Martin Haspelmath (haspelmath at eva.mpg.de)
Max-Planck-Institut fuer evolutionaere Anthropologie, Inselstr. 22
D-04103 Leipzig (Tel. (MPI) +49-341-9952 307, (priv.) +49-341-980 1616)