Electronic Appendix A
The study site (Sangba base) was carried out in the protected area in Central African Republic. (Universal Transverse Mercator co-ordinates: 34N, 754898N, 2055497E). The fieldwork was carried out during two periods, from December 2004 to January 2005 and May 2005. The study area comprises a treeless basin of saline soil in a forest of native tree species. The outcropping rock allows animals to feed upon mineral elements in the soil of the basin’s floor. The main populations of animals that live in the study site are Buffon’s kob (Kobus kob), Defassa waterbuck (Kobus ellipsiprymnus defassa), red-flanked duiker (Cephalophus rufilatus), bushbuck (Tragelaphus scriptus), African elephant (Loxodonta africana), warthog (Phacochoerus africanus) and anubis baboons (Papio anubis). The density of waterbuck on the study area was estimated to be 0.01 animals.km-2 (aerial census, see Renaud (2006) for more details). The main predators that occur in the study area are leopard and lion. Although it is very difficult to estimate the densities of predators, the first species was known to be the main permanent threat to Defassa waterbuck. During the 2 months video sample, several individuals were preyed upon. We studied groups of Defassa waterbuck which formed mono-specific groups (i. e. were alone in the saline area). Animals were not marked for individual recognition.
Electronic Appendix B
Because of the saline area’s size (50x50m), all observable individuals in the basin at the same time were considered to belong to the same group. Moreover, because of a thickly vegetated edge and a dense forest, the perception area of an individual in the saline area was strongly expected to be limited to this area. Data were collected by a hidden observer in the edge of the saline area. Only groups whose predominant activity was foraging and that did not move far during recording were considered. Only groups whose size and composition did not change during recording were retained in the sample. Most of animals in the saline were clumped separating by short inter-individual distances. So, groups were unambiguously determined. The observer was completely hidden in the edge of the saline area. During the film, the observer did not move to avoid disturbing animals. We began the video sequence when all animals displayed feeding behaviour. When we analysed video sequences, the observer paid special attention to identify if he triggered group members’ scans, in particular by looking at the orientation of animals’ head-up. Thus, we could see that the observer did not interact with groups. A group was filmed once a day to ensure that sampled groups were independent of each other. Using these methods, 46 groups were sampled including a range of group sizes from 1 to 13 individuals (mean ± SE = 4.0 ± 0.43). Statistical analyses were computed using R software (the R Foundation for Statistical Computing, 2005).
Electronic Appendix C
First, from each observed group, the separate vigilant bouts of one individual were randomly permuted, as were the separate anti-scan bouts. This operation was performed for all group members and allowed us to construct an imaginary group in which all individuals scan the environment independently of one another. It was performed 5000 times per observed group, and for each 5000 simulations, we calculated a Pearson’s correlation coefficient for all of the possible pairs of group members. The mean of the 5000 simulated correlation coefficients for each observed group gave us an estimate of the mean of the correlation coefficients expected under the assumption that individuals scan independently of one another.