historical explanation of language structure

[bwald]

----------------------------Original message----------------------------
I read Ratcliff's message of July 10, 1998 with considerable interest,
since I am much interested in the difference of methods in synchronic (esp
GG, but really all synchronic descriptions) and diachronic analyses.  They
boil down to a data-base difference between *introspective judgments*,
whether the analyst's own or those of another speaker, in the case of
synchronic analyses, and documentary evidence -- usually necessarily
written texts (since voice-recorded data is not old enough to have deep
historical documentary value yet), in the case of diachronic.  These are
differences in the data bases of these approaches, and all further
problems, hypotheses, theories, explanations, assumptions, etc. follow from
them and respond to them.  Both methods have well-known pitfalls.
Documentary evidence is painfully incomplete.  Linguistic methods of
reconstruction arose to supplement the documented data.  The * flags its
undocumented status as data (NB data for something else).  As we know,
application of the comparative methods is often contentious, though they
have proven useful on a large amount of data (though rarely larger than the
data which must be excluded by the methods -- for various reasons).  In the
case of synchronics, the reliability of introspections is often
contentious, again for various reasons, some of which remain unclear.  The
worst case is two synchronic linguists arguing about their own conflicting
judgments to support their theoretical positions.  A gentleman's agreement
to accept both judgments as data, when it occurs, does not resolve the
issue of what causes the disagreement, and what its implications are for
both synchronic and diachronic analyses.   Generally there are two choices,
either we have hit indeterminacy in the shared grammar of the dissenting
speakers, or one reflects an *innovation* that can be passed on to
successive speakers (so indeterminacy means it isn't passed on, speakers
don't form rules of the degree of determinacy to encompass the
disagreement, and they are only fashioned ad hoc as the methods of
synchronic analysis lead to considering the contested form for some
analytical generalisation based on more secure data).
 
Robert's point about ontology repeating phylogeny in synchronic analysis is
tricky.  The topic was much discussed in the 70s, with little to show for
it, in my opinion.  Mainly because few people were interested in sustaining
the discussion.  Obviously synchronic analysis contains some features
directly relevant to historical analysis, because some speakers have
grammars with changes in progress in them.  It is a *shared* problem for
synchronic and diachronic analyses to decide at what point in its
development a change is no longer "in progress", or has no longer left a
trace of its former status as a change in progress, accessible to the
speaker, not just the analyst, in some way.
 
With respect to the above point, Robert's consideration of various devices
proposed synchronic analysis leads to consideration of the difference
between the level of abstractness that these devices represent, and
relatively transparent and straightforward analyses.  Take the example of
representing morphological umlaut as an abstract affix.  Why this device?
Already it is unclear.  It could be because the particular language has
comparable affixes, e.g., a morphological suffix for indicating plural,
e.g., -er, which as a plural suffix synchronically triggers umlaut in
German.  That is a language-specific reason, closer to the concerns of
diachronic analysis.  But another possibility is that the affix is more
abstract and that its position is arbitrary.   The "parameters" approach, I
would guess, would position the segment according to some language
typology, but it remains unclear how to interpret the notion that
"plurality" is a *segment*.  History supports the notion that the German
-er suffix itself is connected with umlaut, and that the process switched
from phonetic to morphophonemic during a certain period of time.  More
difficult are English umlaut pairs, like "man/men" etc.  If you want to
consider the issue of where GG analyses stop being synchronic and start
being diachronic (two separate issues according to what I said earlier),
consider what its methods have in common with the comparative method.  This
is most obvious for morphophonemics (where the problem was immediately
recognised), and more problematic for either morphology (e.g., Semitic
vowel patterns)/syntax or low-level phonology.
(Generally, GG methodology resembles internal reconstruction, not
comparative reconstruction.)
 
Finally, Robert's point against "all changes simplify (the grammar)" has
been discussed many times in many guises on this list.  If the
simplification involves loss of information, it may cause decoding
difficulties on the part of the hearer, burdened with interpreting the
utterance.  Hearer sanctions on speakers helps to contain them, but we're
not sure why hearer sanctions are more effective in some cases than in
others.  The umlaut example is interesting because as a phonological
process it seems to have started as a speaker-oriented simplification of
articulatory gestures (minimizing articulatory movement from the vowel of
one syllable to the next, i.e., a form of vowel harmony).  It is unclear
how this could cause difficulty for the hearer.  On the contrary, it
increases the cues to plural marking.  With regard to first learners, an
empirical source for determining "complexity", learners certainly
regularise English umlaut plurals, but they also learn to suppress those
regularisations as they become more mature (for most dialects).  This is a
hearer-oriented concession, but not obviously for reasons of decoding ease.
Even more interesting is "was/were" instead of "beed" (i.e., be + ed).  I
am not aware of any evidence that "be+ed" is ever constructed by first
learners.  (it does not even seem to exist in African American vernacular,
despite finite use of "invariant be".)  It seems that for many speakers
"was/were" is learned early, earlier than the -ed past tense marker, and
immediately accepted as a portmanteau.  If this is the case, as I think it
is (and second language learners definitely follow suit), English speakers
readily accept the historical conflation of two distinct verbs.  Unusual is
the acceptance of a distinct verb limited to a particular tense (past).
The same is not true of "went", which has its stage "goed" in first learner
speech (as in history).
 
In the final analysis, the abstract plural *segment* that Robert mentions
for some GGists may contain the claim that when speakers first learn that a
"meaning" has a *transparent* formal marker, they immediately generalise it
to all contexts, later withdrawing it in cases which cause social
disapproval (often unconsciously).  The past tense of "be" seems to
contradict that assumption.  Lexical conflation is clearly a possible
change, as in the Victorian (?) prescriptive example: "men perspire, horses
sweat, and women glow" (among others).  However, the "be" vs. "was" case
involves grammatical conditioning, not semantic or pragmatic.