Pair Test = The Test?

[From Rick Marken (930929.1300)]

Bruce Nevin (Wed 930929 14:09:42 EDT) gives a detailed
description of Harris' "pair test" as an example of
"the test". But the "pair test" looks like standard
IV-DV research to me; you vary the stimuli (pairs)
and note differences in responses (proportion of times
a pair is called "different", in this case). What I don't
see is any hypothetical controlled variable being disturbed.
Maybe you could help me out here, Bruce. What is Harris'
hypothesis about the variable being controlled? How does
the subject influence this variable? How does Harris
disturb the variable?

If they contrast phonemically, the hearer
will judge them different 100% of the time
or nearly 100% of the time;

So is the controlled variable "relationship between
degree of phonemic contrast and judgement of 'same'
or 'different'"? If so, why not have the subject make
a continuous instead of just the binary judgement so
that you get a better measure of the ability to control
this relationship? If subjects really are controlling
this relationship (at a reference level of "perfect")
then you should be able to predict their exact numerical
judgement to any pair of a certain degree of phomemic
contrast. Of course, if you can't predict the judgment
perfectly it's time to continue testing for the controlled
variable. That might involve selecting a new hypothesis about
the reference for the hypothesized variable, or a new
hypothesis about the components of the perception being
controlled (a different measure of "phonemic contrast",
perhaps).

Best

Rick

[From: Bruce Nevin (Thu 930930 14:40:06 EDT)]

My mail is getting disorganized. I saw Martin's comment before seeing
this of Rick's.

( Rick Marken (930929.1300) )

"Degree of phonemic contrast" is not the issue. As Martin says, this is
categorial perception.

The variations are not variations of stimuli, with the investigator
measuring responses. They are disturbances to the perception of a given
utterance. The hearer says "no disturbance" (or equivalent) or else
"Whoops! You're not saying the same thing. That's not a repetition of
<utterance>."

The controlled perception is "repetition of the same words." Differences
of pronunciation ("accent") do not matter so long as the two
pronunciations are recognizable as the same words. Consider the example
I gave you. Saying spin or sbin makes no difference. Saying pin vs. bin
does make a difference. The identical difference in pronunciation. After s,
that difference is not contrastive, that is, it does not make a
difference between words. Saying pode vs bode makes a difference, even
though (so far as I know) pode is not a word. This is an instance where
a hearer might say it's not the same word, bode, and though pode is not a
word it could be. I think that covers the cases I mentioned,
illustrating them within the limits of the sound-differences that are
reliably recognizable by a speaker of English without phonetic training.

    Bruce
    bn@bbn.com