Figure 1.
Experimental setup and stimulus.
A. The subject is viewing conversation from a screen with a built-in eye tracker. B. A line-art rendering of the original stimulus video. The regions of interest, covering the heads of both conversation participants, are shown as green and blue squares (ROI 1 for P1 and ROI 2 for P2, respectively). C. Turn-transition durations in the conversation. The fast turn transitions (fTT) are shown in dark grey.
Figure 2.
Horizontal gaze locations (x-coordinates) during a 30-s epoch of the conversation.
The ROIs indicate the locations of the speakers’ heads (see Fig. 1B), dark blue and green denoting speech and light colours silence. A. The red trace shows the gaze of one representative subject and the black lines the gazes of each individual subject. B. Median (red line), mean (dark blue line), and standard error of mean (grey belt) of all subjects’ gaze x-coordinates. The time points significantly differing from random distribution of subjects between ROIs are shown along the x = 0 line with yellow, orange, and red dots, corresponding to statistical significance levels of 0.05, 0.01, and 0.001, respectively (binomial test).
Table 1.
Proportions of gazes (in %) to different targets in the screen during speech of P1 (33% of the total time), P2 (44%), both (5%), silence (18%) and during the whole stimulus presentation (total).
Figure 3.
ROI changes around conversational events.
A. Single ROI changes relative to each start-event. Each vertical line of dots represents a 5-s epoch around one start of speech in conversation, and a single dot represents one ROI change of one subject. B. Mean ROI change rate in main experiment around all conversational events: ROI changes per second as a function of time relative to each event, normalized according to number of subjects. C. Mean ROI change rate in video-only condition. D. Mean ROI change rate in sound-only condition.
Figure 4.
Baseline-corrected cumulative distributions of ROI changes around fast turn transitions.
Cumulative distributions of ROI changes, with baseline cumulative distributions of ROI changes subtracted, during a time interval from –1 s to 3 s relative to the fast turn transitions. The black, the grey and the dashed line represent the main experiment, video-only and sound-only conditions.