Project CETI aims to translate sperm whale communication. Their most recent paper, published in May, presents evidence for a sperm whale phonetic alphabet. Eloise Trawick discusses these findings, how they might change our understanding of sperm whale communication, and why studies such as this can be controversial.
From attempting to teach primates sign language to classifying calls used by vervet monkeys, scientists have been trying for decades to work out how animals communicate with each other, and how we might communicate with them. However, the rise of machine learning and AI may offer an entirely new way of approaching these studies. Scientists at Project CETI (Cetacean Translation Initiative) are using machine learning to study sperm whales and have found that their communication system is more complex than previously thought.
Sperm whales have the largest brains of any cetacean. They communicate in short bursts of clicks called codas, which are combined into longer sequences of calls. Project CETI aim to use machine learning to work out the meaning of these codas. Computer programmes can classify calls and group them according to similarity much faster than scientists would be able to manually. The results from these analyses provide theories on the function of calls, which can then be tested experimentally.
Sperm whales use sequences of clicks to communicate within their family groups. Image Credit: Gabriel Barathieu via Wikimedia Commons
Analysis of a dataset of almost 9000 codas has revealed that sperm whale codas may be able to communicate an unprecedented amount of information. The study, led by Pratyusha Sharma (MIT) and published in Nature Communications, explains that codas show both “contextual and combinatorial structure”. Their features change according to the context they are produced in, and these features can combine to create a wide range of individual codas.
Each distinct coda may have a unique meaning. Classified into discrete categories based on their musical characteristics, 18 different rhythms and 5 different tempos of coda have previously been observed. This unique study discovered two additional features of coda sequences: sperm whales subtly modify codas by smoothly changing their duration over multiple repeats of the same sequence, or by adding extra, ornamental clicks onto the end. Scientists have termed these changes rubato and ornamentation.
Rubato – named for the musical term - creates subtle, continuous variation in the duration of codas. Analysis has shown this variation is not deliberate, not random. Rubato is often sustained across multiple codas generated by the same whale, and when several whales call at once, the duration of their codas are similar. Individual whales can change the duration of their calls to match other individuals calling at the same time. Scientists believe these slight changes in speed may play a role similar to how changes in the speed of delivery of a word can change its meaning. For example, a sharp “what?” versus a sustained, disbelieving “what?”.
Ornamentation clicks, which were found in 4% of codas, are sometimes added onto the end. They are more likely to be found at the start of call sequences and often precede a change in chorusing behaviour. This means the following whale is likely to begin or stop vocalizing with the leading whale, or to pause its vocalisations after ornamented codas from the leading whale. This is strong evidence for the extra clicks playing a role in conversation.
By changing durations and adding ornamental clicks, sperm whales can create a wide range of codas. It was previously thought the sperm whale communication system was comprised of 21 distinct codas. However, when rubato and ornamentation are considered, there are at least 143. The researchers have dubbed these the sperm whale phonetic alphabet. The different features – tempo, rhythm, rubato, ornamentation – are presented as analogous to phonemes in human language: basic units of sound that can be combined to create meaning.
A sperm whale in action. Image Credit: Marion & Christoph Aistleitner via Wikimedia Commons
Unfortunately, fascinating results such as these highlight the controversy surrounding studies of animal communication and how findings are reported. Scientists often use analogies to human language to explain new discoveries. In this paper, the structure of sperm whale communication is compared to phonetics and, throughout the field, there are often mentions of language. Project CETI has attracted widespread media attention, which often focuses on the possibility that it could eventually allow us to talk to sperm whales. The organisation itself only talks about the possibility of deciphering what sperm whale communication means, but it has long been common when discussing animal communication for people to talk about the possibility of us communicating with them. This focus on relevance to humans is a common criticism of both science and reporting within the field.
In response to the phonetic alphabet study, Dr Luke Rendell (University of St Andrews), who works in cetacean communication, has warned in his own article that scientists need to “drop the stifling and anthropocentric focus on language”. He argues that while the new findings suggest codas may have interesting social functions, a tendency to try to force animal communication to fit the rules of human language (like the phonetic alphabet) will hinder progress and understanding. When we try to understand how other species “talk” to each other, are we blinded by our own assumptions about how communication works? We may be so focused on trying to find similarities to human language that we miss key features of animal communication systems.
In this case, we cannot assume that, just because the features of codas combine in a way that can be compared to a phonetic alphabet, the method of transferring information is the same, nor can we assume that coda sequences function in the same way as sentences. It is important to note that we do not currently know what codas mean – if indeed they mean anything. While this study has proved that sperm whales form codas with tempo, rhythm, rubato and ornamentation, it doesn’t tell us about their ability to communicate information. For this, scientists will need to conduct many more studies. To determine the potential meaning of codas will require detailed observation of the whales, and the context in which these calls are produced. Once there are theories about what each coda signifies, playback experiments will be needed to determine whether the response of the whales to the sound are consistent with its hypothesised significance.
David Gruber, the lead on Project CETI, has said that the project is still at “base camp”, and it will likely take years of dedicated study before we have a comprehensive understanding of sperm whale communication. However, with machine learning programmes allowing scientists to process large amounts of data so quickly, we may be closer to this goal than ever before.
Comments