Transcript Psychonomics_Poster_2005

Is Timing Everything? Grounding Costs in Speech and Typing
MICHELLE GUMBRECHT & HERBERT H. CLARK, Stanford University
Method
Abstract
Introduction
• In conversation, speakers adjust their workload to
benefit both themselves and their addressees. They
do so, it is claimed, according to a principle of least
collaborative effort (Clark & Wilkes-Gibbs, 1986).
Participants: Sixty-four university students (34
men and 30 women, M = 19 years).
Design:
Summary: Content questions were those questions in
Summary: As expected, typing-only partners took
Procedure: Pairs of students (labeled director and
matcher) worked on a tangram matching task. In each
of six trials, a director communicated freely with a
matcher about a set of ten tangrams, who was to place
the tangrams in the instructed order. The two partners
communicated with either the same or different media
(speech or typing) in separate rooms. Speakers wore
microphones, and we used iChat to enable either oneor two-way audio through the computers. Typists used
MSN Messenger.
significantly longer to complete the task than pairs in
the other three conditions. Typing takes much more
time and effort than speaking.
Mean Word Count
Conclusions
2500
2000
1500
1000
500
0
1422.4
484
Director
1655.3
150.1
646.3
510.6
813.3
267.1
Matcher
S-director S-director T-director T-director
S-matcher T-matcher S-matcher T-matcher
• People also change their communicative strategies
based on certain costs of the medium (Clark &
Brennan, 1991). The cost of producing a sentence,
for example, is much greater in typing than in
speaking. There are also contrasting costs in
formulating sentences; making repairs; changing
speakers; dealing with asynchronies; and dealing
with delays.
• Most of these costs are far greater in typing than in
speaking (e.g. Fussell et al., 2004; Newlands,
Anderson, & Mullin, 2003; Cohen, 1984; Ford,
Chapanis, & Weeks, 1979). In most previous
studies, the two partners used the same medium to
communicate, so the costs for the two of them were
equal.
the conversations that were about the content of the
tangrams, e.g. “It’s standing on one foot, right?” In all
conditions, content questions were asked more often by
matchers than by directors. It was the matchers
themselves who recognized what they had not
understood.
• As predicted, people immediately adapted to the relative
Mean # Words
How do conversations change when two partners use
different media of communication, one speaking and
the other typing? The principle of least collaborative
effort (Clark & Wilkes-Gibbs, 1986) as applied to the
contrasting costs of speaking and typing (Clark &
Brennan, 1991) makes a clear prediction: Speakers
communicating with typists should take on more of the
communicative load because they expend less effort to
produce utterances. To test this prediction, we had two
partners complete a referential communication task
using either the same or different medium of
communication (speaking or typing). Results followed
the prediction. Two speakers produced many more
words than two typists, but finished faster. When
speakers were paired with typists, they produced more
of the content; they also took on more of the load by
asking the typists questions that allowed brief answers.
Two speakers lessened their loads by exploiting
precision timing in interrupting and responding to each
other. Speakers paired with typists could not, and that
added to their load and completion time.
Results
Condition
Summary: As predicted, speakers (regardless of their
role) took on more of the communicative load when
paired with typists than with other speakers.
Director’s display.
costs of the medium of communication. In speech-typing
pairs, speakers took on more of the effort, regardless of
their role as director or matcher. For example, when
matchers could speak, they took over more than one-third
of the directors’ load when the director could only type.
Speaking directors contributed 92% of the total words
when paired with typing matchers, whereas typing
directors contributed only 56% of the total words when
paired with speaking matchers.
• The content of the conversation itself changed with the
medium of communication. Usually it is the director who
manages these conversations, asking management
questions to check on the matcher’s current state of
understanding. But when only one of the partners could
speak, that partner took over the management role,
whether he or she was the director or the matcher.
Managing takes effort, and speakers could afford that
effort. Likewise, it is usually the matcher who asks the
content questions. When matchers spoke to directors who
were typing, they especially made use of content questions
that yielded brief answers (e.g. yes/no questions).
• In summary, these findings support the principle of least
collaborative effort when there are gross differences in
effort (or cost) between the two partners.
• However, suppose one person speaks to another
person with a motor speech disorder who must type
to communicate. How do the two of them adapt to a
situation in which the costs are unequal? Will they
change the way they communicate?
Summary: Management questions were those
• In our study, we addressed these questions by
comparing speech-typing pairs with speech-only and
typing-only pairs.
Matcher’s display.
questions in the conversations that did not include
information about tangram descriptions, e.g. “Ok, you
got it?” With two speakers or two typists, the directors
did most of the managing. But as predicted, in the
mixed pairs, the management role was taken over by
the speaker. It takes effort to do this management, so
speakers took over that role.
References
Clark, H. H., & Brennan, S. E. (1991). Grounding in communication. In L. B. Resnick, J. Levine, & S. D. Teasley (Eds.),
Perspectives on socially shared cognition (pp. 127-149). Washington, DC: APA.
Clark, H.H., & Wilkes-Gibbs, D. (1986). Referring as a collaborative process. Cognition, 22, 1-39.
Cohen, P.R. (1984). The pragmatics of referring and the modality of communication. Computational Linguistics, 10, 97-146.
Ford, W.R., Chapanis, A., & Weeks, G.D. (1979). Self-limited and unlimited word usage during problem solving in two
telecommunication modes. Journal of Psycholinguistic Research, 8, 451-475.
Fussell, S. R., Kiesler, S., Setlock, L. D., Scupelli, P., & Weisband, S. (2004). Effects of Instant Messaging on the
management of multiple project trajectories. CHI 2004 (pp. 191-198). NY: ACM Press.
Newlands, A., Anderson, A.H., & Mullin, J. (2003). Adapting communicative strategies to computer-mediated
communication: An analysis of task performance and dialogue structure. Applied Cognitive Psychology, 17, 325-348.
Acknowledgments
We thank Teenie Matlock and the members of the Stanford Language User Group (SLUGS) for helpful feedback. We
also thank Aurélie Beaumel, Roma Shah, and Olivia Tam for assistance in running participants and data coding.