The Rumble Repertoire of Forest Elephants
Like any large family, an elephant herd can get quite noisy. As social creatures, elephants use calls, rumbles, and roars to communicate with each other. Since elephant herds can cross great distances, it is important for them to be able to hear and understand each other’s calls if they get separated. African forest elephants face a greater challenge, since the dense vegetation of their environment makes it difficult for individuals to see each other. In the forest, it is even more important that members of an elephant group stay in contact with each other. Recognizing the “message” contained in their group’s rumbles — whether it’s to synchronize their movements through the thick underbrush or to warn them of danger — is paramount. In the study of animal vocalizations, the “ambiguity reduction hypothesis” is the idea that because of the lack of visual clues to help distinguish and interpret a call’s “message,” a species in densely vegetated habitat should have less ambiguous calls than a species living in an open area. Based on the “ambiguity reduction hypothesis”, the rumbles of forest elephants should be more discreet, with less chance of ambiguous interpretation, than the rumbles of elephants who find their home on the open savannah, where visual communication can play more of a role in imparting meaning to vocalizations. In a study published last year, ELP researchers Daniela Hedwig, Anahita Verahrami, and Peter Wrege used auditory data recorded in central African forests to test this hypothesis¹.
In studies of animal communication, vocalizations can be classified as either graded or discrete. Graded vocalizations have many similar characteristics and are hard to distinguish from each other, while discrete vocalizations differ substantially and are less ambiguous. If the auditory data reveals that forest elephants have more discrete rumbles, that would support the hypothesis. However, if the forest elephants’ rumbles tend to intergrade or overlap, this would not support the hypothesis.
In order to address these questions, ELP researchers recorded elephant rumbles at four different locations in central Africa: Dzanga Bai in the Central African Republic, Jobo Bai and Ivindo National Park in Gabon, and Noubale-Ndoki National Park in the Congo. Fifty recording devices capable of detecting elephant rumbles up to eight hundred meters away were distributed across these locations. These recording devices were placed in a sort of grid, with the most devices concentrated around bodies of water that attract elephants. Some recording devices were as close together as 3½ kilometers, or as far apart as 50 kilometers. From January 2011 to December 2018, these devices captured all the sounds of the forest. In order to pinpoint the elephant rumbles in eight years of audio, an algorithm was used to “tag” rumbles which were later manually verified.
After careful analysis of the data, the study yielded surprising results. The analysis identified five to eight types of rumbles, but these types were highly integrated, or overlapping, and thus ambiguous. Although these sounds were not very discrete, they did vary enough for a computer to tell them apart with 75% accuracy. No significant difference was found between rumbles recorded in clearings or in forests. However, ELP found that the elephants produced a greater variety of rumbles in clearings than in forest areas. There were also differences in the types of rumbles used, but the significance of this is yet unknown. Overall, the results of the ELP study seem not to support the ambiguity reduction hypothesis. This conclusion opens up even more possibilities, and there are several speculations as to why such interesting results were found.
Reducing vocalization ambiguity may not have been evolutionarily advantageous for forest elephants in the first place. Since acoustic characteristics of elephant rumbles tend to degrade after a hundred meters, forest elephants may not rely on these sounds to communicate over long distances. It is also possible that the elephants don’t find their rumbles as ambiguous as we humans do; perhaps the elephants just listen for different acoustic characteristics or markers than those tested in this study. No one has yet investigated how the elephants actually recognize and perceive these rumbles, and the distinction between “graded” and “discrete” is human-imposed. There is also the limitation of audio recordings, which provide no clues as to the context in which the rumble was made, or which individual made it. This ELP study has provided some interesting insights into the world of elephant communication, but the case has definitely not been closed. If anything, we have more questions than before, and so much remains to be discovered. If forest elephant rumbles are truly ambiguous, how do elephant groups communicate effectively? As we continue to eavesdrop on these forest elephants, we will no doubt uncover even more complexity in their conversations.
References:
¹Hedwig, Daniela, Anahita K. Verahrami, and Peter H. Wrege. 2019. “Acoustic Structure of Forest Elephant Rumbles: A Test of the Ambiguity Reduction Hypothesis.” Animal Cognition, no. 6: 1115. doi:10.1007/s10071–019–01304-y. Retrieved from https://link-springer-com.proxy.library.cornell.edu/article/10.1007%2Fs10071-019-01304-y.
