Letting the cat out of the box:
problems in the measurement and analysis of human conversational behavior
R.P.H. Vertegaal
Human Media Lab, School of Computing, Queen's University, Kingston ON, Canada
A well known principle of Quantum Mechanics is that the act of observation may determine what you observe. Measuring human group behavior is not unlike determining the state of Schrödinger's Cat inside its box. Here, I will share my experiences in resolving the difficulties of observing cause-effect relationships in multi-modal human group communication. I will discuss how one can set up a complex, real-time measurement environment on five dollars a day, using equipment available at any music store; how one can precisely control the presentation of human conversational cues through the magic of computer mediation; and how co-variance analysis may help when everything else fails.
As an example, I will discuss our attempts to establish the function and effect of eye gaze behavior on human group conversation. In an early experiment, we found that turn-taking behavior is correlated with the amount of eye contact experienced. More recently, we evaluated whether this is because more eye gaze allowed subjects to better observe when they were being addressed. By collecting eye-tracking data during four-person conversations, we established that face-gazing is indeed an excellent predictor of conversational attention.
To establish whether the communication of conversational attention is responsible for the effect of gaze on turn-taking, we performed an experiment to compare speaking behavior between two conditions: (1) in which subjects experienced gaze synchronized with conversational attention; and (2) in which subjects experienced random gaze. The amount of gaze was a covariate, but it did not vary beyond 12% between conditions. Subjects were 22% more likely to speak when gaze behavior was synchronized with conversational attention. However, covariance analysis showed that these results were due to differences in the amount of gaze perceived, rather than to gaze synchronization, with very high correlations (r = 0.62) between the amounts of gaze and subject speech.
These results imply that it is not the exact timing of eye gaze that conveys conversational attention. Instead, it seems that interlocutors sample their environment to see whether they are still receiving sufficient attention from others to estimate whether they should start or continue speaking. We believe that an energy conservation function associated with attention in human speech production and comprehension may be responsible for the above effects.
Paper presented at Measuring Behavior 2002, 4th International Conference on Methods and Techniques in Behavioral Research, 27-30 August 2002, Amsterdam, The Netherlands
© 2002 Noldus Information Technology bv