A three-year search for hidden patterns in the roars, sighs and other noises of Indonesian orangutans has uncovered “a full spectrum” of complex vocalizations.
The breakthrough comes on the heels of other recent discoveries that further reveal the depth of the great ape’s intelligence, including an orangutan’s practice of healing his own wounds with a medicinal herb he prepared himself.
The research team bolstered their analysis by testing artificial intelligence (AI) detection methods against the painstaking work of biologists and bioacoustic scientists, using only their trained ears, intellects, and measurement tools.
The Cornell University-led team assembled a data set of 117 recorded ‘long calls’ made by 13 males of one particular species, the Bornean orangutan, using 46 acoustic measurements on 1,033 different pulses detected within those calls.
“These features appear to greatly increase the potential complexity of this signal,” they wrote, suggesting that humanity might soon know what the great apes are saying.
A Cornell-led team assembled a data set of 117 recorded “long calls” made by 13 male orangutans, deploying 46 acoustic measurements on 1,033 pulses detected within the calls. Results appear to show greater “potential complexity” in ape calls than previously thought
As reported in the new study, published Tuesday in the journal PeerJ Life and EnvironmentThe researchers found “a continuous gradation of sounds across phases and pulses,” suggesting that orangutans can modulate their voices very precisely.
All of these distinct phases and pulses, the team wrote, can be “combined into variable sequences within a single long call vocalization,” meaning that the “long calls” of male apes most likely communicate complex messages to their distant ape peers. .
“Our research aimed to unravel the complexities of orangutans’ long calls, which play a crucial role in their communication over great distances in the dense rainforests of Indonesia,” said study lead author Dr Wendy Erb. in a sentence.
“We are fairly confident that there is still a lot of complexity to be unraveled in the vocal system of this great ape,” Dr. Erb said.
A primatologist at Cornell’s K. Lisa Yang Center for Conservation Bioacoustics, said Dr. Erb Lounge that bioacoustics experts studying the meaning behind animal sounds still lack “a unifying framework for quantifying complexity.”
As part of their effort to detect and classify orangutan noises with AI, the team adopted parallel “supervised” and “unsupervised” machine learning approaches.
As part of their effort to detect and classify orangutan noises with AI, the team adopted parallel “supervised” and “unsupervised” machine learning approaches. Above, waveforms of the ape’s various distinct vocal pulses published by the team on PeerJ Life & Environment.
Above, an endangered Bornean orangutan looks knowingly at the camera as it climbs a tree in Tanjung Puting National Park, Borneo, in a 2021 photograph.
The team used a state-of-the-art unsupervised machine learning algorithm, Uniform Multiple Approximation and Projection (UMAP), which had demonstrated success in decoding ‘animal vocal repertoires’ for the University of California, San Diego, back in 2020.
The UMAP algorithm was further aided by additional statistical algorithms developed in the R coding language.
For their supervised machine learning, R-based code was also used, but in all cases the apes’ 1,033 vocal phases and unique pulses were randomly divided into a 60/40 split, where 60 percent were used to train the AI, while the remaining 40 percent was used to test the accuracy of its newly trained classification ability.
“Through a combination of supervised and unsupervised analytical methods,” as Dr. Erb summarized, “we identified three distinct pulse types that were well differentiated by both humans and machines.”
The names of those three pulse types, given by the researchers to help guide future studies, were: “Roar” to define high-frequency pulses, “Sigh” meaning low-frequency pulses, and “Intermediate” to capture each and every pulse. one of the pulses. that fell between those two previous categories.
Dr. Erb and her colleagues were at pains to emphasize that these were only the limits of their current research, not the limits of what we may one day learn about apes’ vocal communication abilities.
“Although many pulse types were not well differentiated by humans or machines in this study,” they noted, “we do not intend to suggest that other researchers were unable to make these distinctions or that orangutans cannot perceive them.”
This ability to distinguish and differentiate between the seemingly countless unique sounds that these great primates make will ultimately be the next project for those who hope to one day understand and perhaps speak with humanity’s primate cousins.
“Although our study represents an important step forward in understanding orangutan communication, there is still much to discover,” Dr. Erb admitted.
“Orangutans may have a much larger repertoire of sound types than we have described, which highlights the complexity of their vocal system.”
