Personal tools
You are here: Home Dataset ELEA




The corpus was gathered with the aim of analyzing emergent leadership as a social phenomenon that occurs in newly formed groups. For each group in the corpus, the participants performed the winter survival task. The annotations of the corpus include self-reported and perceived personality, concepts related to leadership,  participants’ performance in the survival task, and manual transcriptions from the recordings in English. The full corpus contains approximately 10 hours of audio and video.

Audio recordings were gathered using the Microcone, a commercial microphone array, designed to record small discussion groups (up to 6 individuals) with audio sample rate of 16kHz (2). As shown in Figure 1, the Microcone was placed in the center of the discussion table to capture the interaction. The Microcone automatically segments speakers, and provides audio for prosodic cue extraction. The high quality audio recorded by the device, allows for automatic speech recognition as well.

For the video recordings we used two webcameras (Logitech Webcam Pro 9000, see Figure 1), with video frame at 30 fps.

Figure 1. The recording sensors of the portable setup on the ELEA corpus. Two webcameras and the Microcone.

Figure 1. The recording sensors of the portable setup on the ELEA corpus. Two webcameras and the Microcone.



102 participants. With this population, mixed teams were formed; 21 teams were four-person and 6 teams were three-person.

Winter Survival Task

The participants in the task were supposed to be survivors of an airplane crash in winter (56). They had 12 items that they had to rank in order of their importance, giving 1 to the item considered the most important to survive, 2 to the second most important, and so on. The task was performed first individually (5 min, not recorded) and then we asked them to come up with the group ranking while being recorded (15 min). The average length of the group discussions is 14.61 minutes, ranging from 8 to 19 minutes. Considering that not all the participants might be familiar with the items, we provided them with slides containing a picture and the definition of the item. The slides were consulted only during the individual ranking, to avoid the occlusion of the cameras during the group discussion.


Self reported, Big Five personality traits.
Personality Research Form (PRF), 16 true-false items scoring dominance personality trait.

17 statements that capture how they perceive each participant (including themselves). 16 of the statements (five-point scale), developed by Prof. Marianne Schmid Mast (University of Lausanne), the variables included in these statements are: perceived leadership (PLead: person gets involved, directs the group), perceived dominance (PDom: person is in a position of power, dominates), perceived competence (PCom: person is competent, is experienced) and perceived liking (PLike: person is kind, is friendly). Afterwards, participants provided a dominance ranking (RDom), i.e., participants were asked to rank the group, given 1 to the most dominant participant, and 3 or 4 for the less dominant, such that they have to include themselves in the ranking.

Judgment from external observers

We use the same questionnaire as filled by the participants, focusing only on leadership and dominance. For each meeting, we assigned two external observers, one male and one female, who watched the first five minutes of the meeting video and answered eight questions for each of the participants in the meeting.

When using this data for your research, please cite the following paper in your publication:
D. Sanchez-Cortes, O. Aran, M. Schmid Mast, and D. Gatica-Perez.
"A Nonverbal Behavior Approach to Identify Emergent Leaders in Small Groups", IEEE Transactions on Multimedia, Vol. 14, No. 3-2, Jun. 2012.