[From: Bruce Nevin (Tue 930330 08:36:29)]
[Rick Marken (930329.1230)] --
> "Flying planes can be dangerous" is
>ambiguous only in isolation.
I agree. But I think that the same is true of ambiguous
non-linguistic behavior too (I'm saying that all behavior,
linguistic and non-linguistic, is ambiguous inasmuch as we
can never be positive (without doing the Test) what controlled
variable might be involved).
I'll repeat something from my (Fri, 26 Mar 1993 11:56:11):
Ambiguity is quite apart from the differences in nonverbal
perceptions (non-category perceptions) that we associated with
utterances in an idiosyncratic way that is not socially
standardized. You can't call that ambiguity, because ambiguity
is a choice between structurally defined alternatives.
Now, I was wrong to say "you can't call that ambiguity". You can
call anything you like anything else you like, of course. And in
particular there is a common sense of "ambiguous" meaning that
something is ill defined. So let's use "ill defined" for the
indeterminacy of observed behavioral outputs w.r.t. the reference
perceptions of the observed control system. (Or if you want to
keep the word "ambiguous" add some qualifier, "ambiguous because
of its indefiniteness" or some such.)
The linguistic ambiguity that we see in "flying planes" is not
ill defined at all. The two alternative linguistic structures
(linguistic information) that we perceive in that utterance
are both quite well defined. Even the fact that there is more
than one structure is quite well defined. It is well defined
because those structures are socially available, common knowledge
of anyone who knows the language, by the very nature of what
"knowing the language" is.
Now, when you successfully perform the Test for the controlled
variable, the reference perception back of some observed
behavioral output becomes well defined in two respects at once:
the reference perception becomes socially available, common
knowledge of anyone who has observed (and understands) the Test;
and there is only one reference perception put in correspondence
with the behavioral outputs, as opposed to the indeterminate
range of possible reference perceptions (including none, if the
behavior was incidental, not a consequence of perceptual control).
But notice that the range of possibilities was indeterminate, not
socially available prior to the Test. In language, ambiguity is
structurally determined, that is, the range of possibilities is
socially available prior to disambiguation.
Furthermore, with linguistic ambiguity one need not always
perform the Test to determine which linguistic structure (which
linguistic information) the speaker intended. Frequently, more
careful attention to context suffices. And suffices not in a
probable, statistical sort of way, but precisely and
determinately. The speaker would have had to change the subject
twice in quick succession, once just before the sentence and
again back to the original subject just after. A change of
subject is signalled by structural cues in the discourse (e.g.
"by the way"), as well as by intonation and other gesture in
face-to-face interaction. Such a change would violate
conversational norms. Most importantly, it would be incoherent,
not fitting the informational structure of the ongoing discourse.
If the intention of the speaker was in fact to shift topic and
use the ambiguous utterance in a non-obvious sense, then she
would take steps to make sure that you knew it (structural cues,
intonation, etc.). Prevarication is possible, NB, even with the
Test: the observed person can manipulatively pretend to control a
perception that in fact she doesn't care about. She has to
control that perception in fact at the time of the Test, but
likewise a person exploiting linguistic ambiguity has to speak as
though intending one meaning, while secretly holding an
alternative meaning in reserve.
language is like all other behavior -- you can't tell
(for sure) what constitutes a person's behavior (an intended
consequence of their outputs) without doing "The Test" in order
to determine what perceptual inputs they are controlling.
Language is not behavior. Speech is behavior. Language is
socially standardized reference perceptions, the control of which
has speech as a byproduct. The reference perceptions are
socially available without recourse to the Test because they are
learned (and taught--cf. Bruner) as socially standardized
perceptions.
The observed speech varies from the socially standardized
references in ways that can at best be described statistically.
It is common for people to control for different norms with
respect to different other perceptual control (am I speaking
"correctly" or as a non- alienating member of my group or am I
slumming, etc.). This usually adds to the variability of speech,
as people alternate from one norm to another depending on their
perception of their social situation, but it can result in
conflict.
(I think this perspective, familiar to linguists, answers Bill's
qualms about Labov's work.)
Bruce
bn@bbn.com