A Study in Cross-Cultural Interpretations of Back-Channeling Behavior Yaffa Al Bayyari

advertisement
A Study in
Cross-Cultural Interpretations
of Back-Channeling Behavior
Yaffa Al Bayyari
Nigel Ward
The University of Texas at El Paso
Department of Computer Science
February 22, 2008
Meeting of the Society for Cross-Cultural Research
Back-Channeling
•
•
•
•
•
Short utterances: uh-huh, yeah, mm-hm ...
Show active listening
Frequent: ~4 per minute (English, Arabic)
Below conscious awareness
A turn-taking function
2
Misunderstandings are Common
in Cross-Cultural Communication
culturedependent
interpretations
3
What Causes Problems?
Language-Dependent
Culture-Dependent
Universal
smiles
simple
emotions
non-lexical
utterances
complex
emotions
words
emblematic
gestures
?????????????????? turn-taking mechanisms ??????????????????
Culture-dependent behaviors that we think are universal
can cause deep misunderstandings.
4
Cultural Differences in
Back-Channeling
• frequency (Maynard 89)
• words and non-lexicals used
• timing (this study)
– typically in response to cues by the other
– but the cues differ ...
5
Prosodic Cues for Back-Channels
English
pitch
26th percentile
>110 ms
time
Arabic
< 500 ms
pitch
brief pause
>40 ms, with slope > .7% every 10 ms
time
“When you hear such a cue, respond with a back-channel”
is a good description of listener behavior.
Game 1: listen to the back-channels
Game 2: listen to the cue; what emotional value do you perceive?
6
Initial Hypotheses
Pitch downdash
perceived as
Arab
subjects
American
subjects
a cue for a
back-channel
response
yes
no
an expression
of negative
affect
no
yes
7
Experiment 1 Stimuli
lead-in
prosodic cue
downdash
response
back-channel
cadence
upturn
full turn
silence
subjects asked to judge the naturalness of each response, given the context
resynthesized to obscure the words, retaining pitch contours
Participants
• 18 naive American-English speakers
– students from an introductory CS class
– mostly Spanish-English bilinguals
– no knowledge of Arabic
• 18 Arabic speakers
– 7 living in El Paso Texas, 11 in Qatar
– some to total knowledge of English
• 18 exposed American-English speakers
– with about 25 minutes of training in this aspect of Arabic,
several months before
9
Results of Experiment 1
Ratings of the naturalness of the various pairings
5
4.5
4.7
4.3
4.5
4.2
4.1
4
3.7
naturalness
3.5
3
other pairings
2.5
downslope+BC
2
1.5
1
0.5
0
Arab
N aive American
different*
E xposed American
different*
* matched-pairs t-test
10
Experiment 2
downdash
cadence
upturn
Subjects were asked to judge the emotional state:
“does the speaker sound more positive or more negative?”
11
Experiment 2 Results
6
5
5
4.5
4.2
4.2
4
4
3.2
positive feeling 3
controls
downslope
2
1
0
Arab
Naive American
Exposed
American
different*
different*
* matched-pairs t-test
12
Summary of Experiments 1&2
Pitch downdash
perceived as
Arab
subjects
American
subjects
a cue for a
back-channel
response
yes
no
significantly
different
an expression
of negative
affect
no
yes
significantly
different
13
What Causes Problems?
Language Dependent/
Culture Dependent
Universal
smiles
simple
emotions
non-lexical
utterances
complex
emotions
words
emblematic
gestures
back-channeling
Behaviors that are culture-dependent, but that people think
are universal, can cause deep misunderstandings.
14
Implication and Follow-Up Question
Imagine an Arab happens to use this cue while
talking to an American (in Arabic or English)
• The American is likely to misinterpret it,
without suspecting the danger
(Experiments 1 and 2)
• The Arab may feel the American is not
being a cooperative listener ...
(Experiment 3)
15
Follow-on Hypotheses
Learners of Arabic who back-channel better
will be judged as
• knowing Arabic better
• being nicer
• being more socially effective
and the effect sizes will be large
16
Stimuli
Well Timed Back-Channeling
Well Pronounced
Greeting
Poorly Timed Back-Channeling
Poorly Pronounced
Greeting
Absent Back-Channeling
Absent Greeting
17
Results
How well does the listener know Arabic?
5.5
5.2
knowledge of Arabic
5
4
4.4
3.8
4.3
3.8
6
5.3
3.4
no greeting
3
sloppy greeting
good greeting
2
4
4.7
4.1
3.7
no g reeting
sloppy g reeting
g ood g reeting
2
1
0
0
back-channeling 2
5.3
3
1
no back-channelling back-channeling 1
4.7
4.2
4
5.7
5.4
5.3
5
social effectiveness
6
Is this person likely to succeed in making someone want to help him?
no back-channelling back-channeling 1
back-channeling 2
significantly different (matched-pairs t-tests, 54 pairs)
18
Conclusions
• The prosody of back-channeling is not universal.
(experiment 1).
• It is worthwhile for learners to master its meaning.
(experiment 3).
So they should be taught it, and other turn-taking
patterns, and also in other languages.
• Americans perceive it as negative, but even a brief
exposure reduces this. (experiment 2)
So people likely to hear even sound-bites of Arabs
should also be taught about it.
19
A Study in
Cross-Cultural Interpretations
of Back-channeling Behavior
Yaffa Al Bayyari
Nigel Ward
Thank You
February 22, 2008
Meeting of the Society for Cross-Cultural Research
The Phenomenon
• Back-channel feedback happens when
– One person is explaining something
– The other produces short response indicating he is
paying attention
• Definition
–
–
–
–
Responds directly to an utterance of the speaker
Is optional
Does not require acknowledgement by the speaker
Does not interrupt the flow of the conversation
21
Predictive Value of the Cue
Corpus-based study found that these times are
commonly indicated by a prosodic feature complex
which includes a steep pitch downslope, “downdash”
(Ward & Al Bayyari, 2006, 2007)
• Coverage = 43%
• Accuracy = 13%
22
Experiment 3
• Stimuli preparation
Good Greeting
– 9 audio fragments
Poor Greeting
– 11 sec conversation between
Arabic speaker and a learner No Greeting
– Greeting not synthesized
– Direction-giving and BC synthesized
Good BC
Poor BC
No BC
• Subjects were asked to judge the Arabic learner
23
Previous Work
• In Arabic:
– Statements & wh-questions end with a falling
pitch
(Kulk et al., 2005; Eldin & Rajouani, 1999; Rifaat, 2005)
– yes-no questions generally end with a pitch
rise
(Eldin & Rajouani, 1999; El-Hassan, 1988)
– Back-channels function pragmatically in
Arabic much as in English
• Do not always convey understanding
• Can overlap the speaker’s talk
(Ola Mohamed Hafez, 1991)
24
Results of Experiment 2 – cont.
Answers to question “write 2 or 3 adjectives describing
the speaker” by English speakers:
For the downslope: half or more used “angry”, “scared”,
“sad” or “disgusted”
25
Hypothesis
• The pitch downslope is a cue for back-channel
in Arabic, although it is not perceived as such by
speakers of American English
• The pitch downslope is perceived negatively by
American-English speakers but not by Arabic
speakers
• In Arabic good back-channeling matters & even
more than good pronunciation
26
Does this person sound like a nice person?
7
5.8
6
5.4
5
5
4
4.3
3.3
4.9
4.6
3.6
3.7
no greeting
sloppy greeting
3
good greeting
2
1
0
no back-channelling
back-channeling 1
back-channeling 2
27
Sharp pitch downslope
Most likely we’re gonna stay for a month and a week
OK
OK
To see how life is there and probably uncle & aunt are coming with us,
so it’ d be good chance for you to come visit us
28
Masking
Original
Masked
29
Download