[From: Bruce Nevin (Wed 930929 14:09:42 EDT)]
( Rick Marken (930929.0730) ) --
Apparently something went wrong with the net on monday
Maybe the things I sent out didn't make it either.
A couple of
purported examples of non-PCT based implementations
of "The Test" have been mentioned recently on the net [...]
Harris' "pair test"
mentioned by Bruce Nevin (I asked Bruce for a description
but have, as yet, not received one).
Stripped-down operationalist-principle version a la Turing test:
Setup: Two native speakers of a given language plus the investigator.
(One native speaker may be in the form of recorded playback.)
Present the hearer with pairs of short utterances (pairs of words, to beg
a question) e.g. English pin/bin and (using the same sound
differentiation) spin/sbin. If they contrast phonemically, the hearer
will judge them different 100% of the time or nearly 100% of the time; if
they do not, the hearer will judge them the same something like 50% or
more of the time (i.e. may discriminate non-contrastive
sound-differences). For an English speaker and hearer, spin/sbin are
repetitions despite their differences; pin/bin contrast.
Many variants are more flexible and naturalistic. For example, with
training in discrimination of phonetic differences, observe in the speech
exchanged by native speakers that a given short utterance is occasionally
pronounced with what you transcribe as segment or phonetic feature x at a
given location relative to other segments or features, though it is
usually pronounced with segment/feature y instead, but that hearers
recognize the one as equivalent to the other. Ask a speaker to say it
again and get a slightly different pronunciation. Substitute one feature
for another in one's own pronunciation of a given word, and observe that
the hearer accepts it as a repetition, or says it is a repetition but
in a peculiar "foreign" way of saying it, or says, no, that's a different
word (or, equivalently, that's not a word but it could be), and this also
either "properly" pronounced or with a "foreign" twist. And so on.
Do tape splicing (or the digital diddling equivalent) to slip a stretch
of one utterance into the equivalent position within another utterance,
which differs by some feature(s) in that stretch, ask hearers if the one
is the same word as the other or a different utterance. Etc. The
non-native quality of the investigator's substitutions or of the result
of tape splicing does not prevent hearers from making these judgements of
phonemic categories.
By substitution tests, one arrives at a segmentation of utterances such
that each segment represents a contrast with other utterances. By
generalizing and regularizing over many utterances, one may arrive at a
more efficient representation of the contrasts. More than one way of
representing the contrasts by segments is possible, and indeed speakers
may differ in how they do this, even within the same family. (The
segmentation may be into concurrent elements instead of or in addition to
sequential elements.)
References:
Harris, Zellig S. 1951. _Methods in Structural Linguistics_. Chicago:
UChi Press.
__________. 1968. _Mathematical Structures of Language_. New York:
Wiley. Brief description here.
__________. 1991. _A theory of language and information_. Oxford:
OxUP. Brief description here.
__________ and Carl Voegelin. 1952. Eliciting in Linguistics. <Sorry,
I don't have a cit. handy. It's in the first collection of
Harris' writings, _Papers in Structural and Transformational
Linguistics_, 1972.>