Principles like trade-offs, tunnel vision, and local generalizations

[Richard Janda]

----------------------------Original message----------------------------
 
Regarding Larry Trask's query, and following up on Vit Bubenik's and Theo
Vennemann's responses (especially the latter's term "tunnel vision"):
  Certainly the phenomenon in question overlaps with the principles described
by the statement that languages practice therapy, not prophylaxis (most strong-
ly repeated by Lightfoot 1979, I believe, though the idea goes back at least
to Vendryes, doesn't it?), modulo Theo's locality restrictions, and by Sturte-
vant's Paradox about sound-change being regular but causing irregularity (in
morphology), while analogy is irregular (morpholexically sporadic) but rees-
tablishes regularity.
  In my own work, I have emphasized not only grammatical locality but also
data-related locality:  i.e., any language is truly so vast that no speaker
can keep all subsystems and their interrelated consequences in mind when
tempted to make a particular innovation or to apply a given principle which
turns out to yield what is effectively a hypercorrection.  Graphically, I
once prepared a diagram in which language structure was presented as a leaded
glass window looking out on the world, and a speaker's linguistic conscious-
ness was represented as a flashlight beam directed at something in that out-
side world, but broad enough to illuminate only a small area of the language-
structure window.  As a result, it is not surprising that a speaker may notice
competing parts of two patterns, and so be tempted to restructure part of the
window so as to yield a more consistent local pattern (within the flashlight
beam) that is actually not so consistent when viewed more globally.  (In the
diagram I used at the Linguistic Association of the Southwest meeting in Hous-
ton in 1995, the speaker was equipped with a blowtorch for redoing inconsis-
tent parts of leaded-glass windows, though this suggests more conscious tink-
ering with language than I would like to imply.)
  If correct, this factor (what Brian Joseph and I have called "a limited
window of data" restricting the amount of linguistic structure that a single
speaker/hearer can keep in mind) to which Larry Trask has indirectly pointed
in diachrony has the direct implication for synchrony that linguists probably
overestimate the degree to which speakers "consider all the relevant data"
when arriving at (tacit) analyses--hence the prevalence of what Brian and I
have called "local generalizations".  (E.g., the shift from [i] to [eI] for
the first vowel in English _academia_ _ [still not reflected in some diction-
aries] seems to reflect the local influence of _macadamia_ rather than some
general favoring of foreign-like [e] (_anemia_ in the U.S. never has [e], to
the best of my knowledge); still, there are a lot of nuts in academe.)
  With Vennemann, though, I am tempted to think that such limitations of at-
tention are so pervasive in both linguistic and non-linguistic behavior that
they may not need a special name, though "tunnel vision" and consequence-
blindness are suggestive....
     Richard Janda