[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
I want to reply to the people who have sent me comments on
my proposed experiment so far. In general I want to thank
you all for your comments and suggestions:
Daniel Levitin and Arty Samuel:
I AGREE THAT THE PROBLEM OF GETTING THE SUBJECTS TO TREAT THE
SIGNAL SYNTHETICALLY IN THE NON-SPEECH BIASING CONDITION WILL
BE HARD, BUT IS TRUE THAT IT IS NECESSARY TO CREATE A WHOLISTIC
BIAS IN THE NON-SPEECH INSTRUCTIONS TO
MATCH THE SYNTHETIC BIAS IN THE
SPEECH INSTRUCTIONS. IF IT TURNS OUT THAT ONLY WITH FAMILIAR
OR FAIRLY INTEGRATED SOUNDS IS THERE THE POSSIBILITY OF A
SYNTHETIC BIAS, THE EXPERIMENT MAY OBSERVE SUPPRESSION OF
COMPONENTS IN THE SPEECH CONDITION NOT BECAUSE THE TASK INVOLVES
SPEECH, BUT BECAUSE IT INVOLVES A GLOBAL PATTERN FOR WHICH THE
SUBJECT ALREADY HAS A SCHEMA AND CAN LISTEN SYNTHETICALLY.
Bruno Repp and Yoshi Nakajima:
It is clear that we can also hear the qualities of the sound(s)
in speech or speechlike signals. Whalen & Liberman's
explanation is that
the speech recognizer only needs a little energy. What is left over
can be used to form other perceptual qualities. This was the implica-
tion of the experiment by Whalen and Liberman in Science a few years
The experiments that you proposed are excellent, Bruno. I hope you
will get around to doing them some time.
Yoshi, the possibility that we may be partitioning the synthetic
speech signal into parts that contribute vs. those that are left
over to supply other percepts is like the Whalen & Liberman propo-
sals mentioned above. Your proposal would allow the speech
recognition to subtract components. theirs allows it to
subtract small amounts of energy from components, making the
test of their proposal even harder.
The difficulty of biasing adult subjects may not be as difficult
as you think. I agree that it would be impossible if the
signals were good exemplars of the patterns in question. However,
sine-wave-analog speech is not always heard as speech, so I am hoping
that it would be possible to prevent it. Another precaution would be
a post-experimental interview in the style of the social psychologists
to determine whether the manipulation was effective. Then reclassify
the subjects who were unaffected.
I agree that the issue is not whether the exact claims of Liberman
are or are not correct, but exactly what the relationship is between
speech perception and other aspects of perception. Perhaps we could
draw up a list of questions. Here are a few. Can any of you think
of more, or of better ways of expressing them?
- Does speech perception act like other processes that extract a
global pattern from data that is susceptible to being perceived at
both a global and a part level?
- Does it matter that the global description and local descriptions
are framed in terms of different concepts (say phonetic vs. tonal)?
- Does a global percept suppress or otherwise interfere with
percepts based on the parts? Does the suppression also go in the
George Mandler once said to me years ago, "good experimental
ideas are a dime a dozen. The important thing is to bring them
to a successful experimental conclusion." I partly agree. I am
not so worried about claim-staking as long as we can avoid any
"chilling" effect. ("The other guy's going to do it, so I'd better
not.") My feeling is that two people are unlikely to do it exactly
the same way, and we will learn something about the issues from
the two slightly different approaches. If the results agree, that
means they are robust and not affected by small details of procedure.
If not, then we face the challenge of reconciling them. Given the
rareness of replication in psychology research, we shouldn't hesi-
tate to tread on each other's toes. Physicists seems to be obsessed
with who thought of something first or did it first (no offense,
Bill Hartmann). I don't think I am going to get the Nobel Prize
anyway, so I'm not worried. Any candidates for the Big One out there
need not share their ideas.
- Al Bregman