Straight Talk

Language has long been held up as one of the characteristics that distinguishes humans from other animal species. However, recent research is suggesting that “language” may not be as unique to us as we believe. Many species also utilize sound for communication, and while many of these sounds can be relatively easily understood, such as warning or territorial signals of many birds and mammals, it is the complex acoustic repertoire of whales which is among the most intriguing. 
 

Whales are unique among mammals in sharing with humans large brains that have a high degree of structural complexity. The level of whale intelligence can only be speculated on, but the sperm whale and orca are almost tied for having the largest brains of any animal, which can each weigh over 9 kilograms (the average human brain weighs 1.4 kilograms). While both species of whales have large bodies, they are not the largest whales. Large body size does not guarantee big brains, as we see in sauropod dinosaurs. 
 

In addition to the large size, the surface of a whale’s higher brain (neocortex) is wrinkled or folded, making it densely packed—a trait shared with humans and one generally correlated with intelligence. In fact, some whales have a higher level of cortical folding (gyrification) than we do. Since the 1960s, humans have tried to understand whether this represents whale intelligence. Some of this research has included examining patterns of variation and motif in whales’ calls and even rudimentary experiments in two-way communication between humans and whales (principally dolphins). 

While some of the research has been fruitful, it has fallen short of understanding true interactive communication. Moreover, the question of whether some whale species may have a language of their own remains unanswered. The rise of artificial intelligence (AI) and complex language modelling software seems poised to make a major contribution to answering this question.

In the last few years, natural language processing in computer science has begun to revolutionize our ideas of what is possible with AI. Tools such as ChatGPT and others can compose multifaceted narratives and even write computer programs based on minimal user in put. Algorithms developed in the last 10 years can simultaneously translate any human language to any other human language without manual input or any need for dictionaries. 

In 2020, Project CETI (Cetacean Translation Initiative) was founded as a non-profit organization with the goal to apply advanced machine learning, AI, and robotics to listen, observe, and eventually translate inter-individual communication between sperm whales.
 

Sperm whales were chosen because, like us, they are highly social and have a complex communication system. 

In the Caribbean, female-led family groups of sperm whales can be found year-round off the coast of Dominica. This means calls from individuals can be repeatedly recorded, along with documenting whatever activity the whale is doing when the call is made. This study population is ideal because it provides the opportunity to gather the large quantities of acoustic and observational information which machine learning requires.

 Unlike humpback whales, with their long melodic songs, sperm whales generate bursts of broadband clicks. In the 19th century, whalers used to refer to the sperm whales as “carpenter fish” because they could hear the click series through the hulls of their wooden ships. While most of sperm whales’ signals function to echolocate, they also produce varying sequences of clicks known as codas, which are believed to play a crucial role in their social interactions. These codas can vary in the number, tempo, and rhythm of clicks, potentially allowing for greater complexity if sperm whales have the capacity to recombine them in different ways.

It turns out that they do. Because these patterns resemble musical components, they are given the musical terms inter-click “interval,” “rubato,” and “ornamentation.” Inter-click interval is a way to measure tempo in a series of notes, or “clicks.” Rubato refers to purposeful change in rhythm, while ornamentation (here) refers to the addition of signal flourishes (think of an add “click” on a coda that occurs in a uniquely expressed way).

Over the last several years, Project CETI has collected massive amounts of bioacoustical data using tethered buoys and other semi-autonomous aquatic robots. CETI published preliminary findings this summer in Nature Communications. Their results show that, while it was previously believed sperm whale populations had only 21 distinct codas, they are much more expressive, with 156 distinct coda patterns. “We found that sperm whale vocalizations are significantly more complex than previously believed—with both previously unknown combinatorial structure and context-dependent call modulation,” says Pratyusha Sharma, the lead author of the paper. CETI has also identified variation in rhythm and flourish, much like the musical terms applied. Minimally, the indication is that sperm whales seem to have a phonetic alphabet—the rudimentary building blocks of language.

This is among the few examples where what was thought to be a strictly human ability has been found in another mammalian species. Researchers acknowledge that their work is just beginning. Future research aims to look for patterns within coda structure and to couple this with associated whale activities, using machine learning and AI. While we may not be ready to talk to sperm whales yet, we are moving a step closer.

Oliver Haddrath is a Collections Technician of birds at ROM. 

Dr. Jacqueline Miller is a Collections Technician of mammals at ROM.