From Bruce Nevin (101297.1144)
If two control systems are in conflict, controlling the same complex
variables in the environment (CEV) with different reference signals, the
result is ordinarily somewhere between their two reference signals,
satisfying neither.
What if control of the perception that matters less is simply dropped? But
only dropped for CEVs for which the two perceptions conflict, still
controlled elsewhere.
The context for the question is the idea that the grammar of a language is
not rules, but an effect of universal constraints being ranked in a
particular precedence order.
For example, here are two constraints that have been postulated as universals:
1. A syllable begins with a consonant.
2. The pronunciation faithfully reflects the underlying form.
(What is an "underlying form"? The indefinite article in English has
two forms, a and an. The idea is that one of these, an, is the underlying
form, and that the n gets dropped when a consonant follows. Figuring out
when, why, and how the n gets dropped, in a way that is consistent with
what happens in all languages, is the sort of thing linguistics is about.)
If constraint 1 is ranked higher than constraint 2, then you get a language
like English, where we start a syllable with a glottal stop if there is no
consonant in the underlying form ("Apple, I said."), or we carry over the
consonant at the close of the preceding word if possible ("an apple" =
a.napple). But in some languages, constraint 2 is ranked higher than
constraint 1, and onsetless syllables are OK. (This is like "Give me a
apple" with no n or glottal stop.)
The obvious PCT interpretation is that the constraints reflect the control
of perceptions. If so, then when constraints are in conflict, the outcome is
not midway between the two, rather, control of the lower-ranked constraint
is simply dropped. And there is no effect of error when we fail to pronounce
the n of "an", or when we pronounce a consonant that is not inherently
there. But the lower-ranked perception is still controlled in situations
that do not bring it into conflict with the higher-ranked one.
(The example, chosen for simplicity and accessibility, may not be convincing
because we English speakers don't perceive a glottal stop as a "real"
consonant. Consider, then, that in some languages the inserted consonant is
a t. We could say that the glottal stop is really there in the underlying
forms of words like apple, so that all underlying forms satsify constraint 1
already. The same problem arises: there is no evidence of error when we drop
this initial glottal stop in favor of the n of an--not to mention no
motivation for the a-an alternation, etc.)
Does this sort of discrete switching turn up anywhere else?
Bruce