Researchers at MIT CSAIL and Project CETI believe they have decoded a type of sperm whale alphabet using machine learning technologies. The results of the study, published under the title “Contextual and Combinatorial Structure in Sperm Whale Vocalizations,” indicate important breakthroughs in our understanding of cetacean communication.
The study looks at codas – a series of clicks that perform different linguistic functions. “We found that there are previously undescribed variations in the coda structure,” CSAIL director Daniela Rus told TechCrunch. “We found that coda types are not arbitrary but form a newly discovered combinatorial coding system.”
While whale vocalization has been a central research topic for decades, the teams behind this new research suggest they have discovered a previously unknown level of nuance in the talkative marine mammals. The paper notes that previous research has identified 150 different sperm whale codas.
“A subset of these has been shown to encode information about the identity of the caller and the clan,” it says. “However, almost everything else about the sperm whale communication system, including fundamental questions about its structure and information transmission capabilities, remains unknown.”
The teams drew on the work of Roger Payne, the pioneering marine biologist who died last June. One of Payne’s most influential works was the songs of humpback whales. “He really inspired us to use our most advanced technologies to gain a deeper understanding of whales,” says Rus.
The teams used machine learning solutions to analyze a dataset of 8,719 sperm whale codas collected by researcher Shane Gero off the coast of the small eastern Caribbean island of Dominica.
“We would receive the inputs and then adapt our machine learning to better visualize and understand more,” explains Rus. “And then we would analyze the results with a biologist.”
The team’s method represented a change from previous analyzes that examined individual codas. A fuller picture emerges when the sounds are examined in context, such as an exchange between whales. Contextual details are classified using musical terminology. These include tempo, rhythm, ornamentation and rubato. From there, the team isolated what they call the sperm whale phonetic alphabet.
“This phonetic alphabet makes it possible to systematically explain the observed variability in coda structure,” says Rus. “We believe it is possible that this is the first time outside of human language that a communication represents an example of the linguistic concept of duality of pattern formation.” This refers to a series of individually meaningless elements that form larger meaningful units can be combined, similar to the combination of syllables to form words.”
The meaning of these “words” takes on different meanings depending on the context. The paper adds:
Our results show that sperm whale vocalizations form a complex combinatorial communication system: the seemingly arbitrary inventory of coda types can be explained by combinations of rhythm, tempo, rubato, and ornamentation features. Larger combinatorial vocalization systems are extremely rare in nature; However, their use by sperm whales shows that they are not exclusively human and can arise from completely different physiological, ecological and social pressures.
Although the breakthrough is exciting for everyone involved, there is still much work to be done, first in sperm whales and then potentially expanding to other species such as humpback whales.
“We decided to go to sperm whales because we had a large dataset and have the opportunity to collect many more datasets,” says Rus. “In addition, because the clicks form a kind of discrete communication system, it is much easier to analyze than a continuous communication system. But even Roger Payne’s work showed that humpback whale songs are not random. There are sections that are repetitive and there is an interesting structure there. We just haven’t gotten around to doing an in-depth study.”