Latest Research Suggests Whale Songs And Human Music May Share A Deep Commonality

Whale song, a complex and haunting vocalisation, is mainly produced by male whales to attract mates over long distances in the deep ocean.

These songs have been observed to become increasingly complex over a period of time. Perhaps it is the complexity of the songs which makes them more attractive to the females.

Beyond their complexity, scientists have observed that humpback whales improvise on existing song patterns. New songs emerge every few years in the southwestern Pacific and gradually spread across the ocean.

There are two types of transformational changes in whale songs:

—One type of change occurs gradually within a local whale population, evolving season by season. Over time, these gradual shifts transform the song’s themes entirely, creating a completely new pattern within a few years.

—Then there are what are called the ‘song revolutions’. A song from a neighbouring whale community may get adapted in another community and may quickly and totally replace their own song. Such instances have been recorded multiple times.

Clearly song transmission among whales is cultural.

A natural question to ask at this point is: how does the learning of the song happen in whales?

Whales are wild and free-spirited, making laboratory studies impossible. However, recordings made during song transitions—when a male whale is actively changing its song during a "song revolution"—have provided crucial insights into how whales learn new songs.

The whale song displays a remarkable acoustic organisation of nested hierarchy:

--Individual sounds, which can be considered as the fundamental elements of the song, are arranged into recurring patterns. These are the equivalents of "phrases," which are repeated in a specific sequence.

--These sequences then form larger units, or "themes."

--Finally, multiple themes are arranged in a fixed order to create the complete "song."

A similar nested structure exists in human speech, where phonemes form words, and words combine into sentences. This resemblance offers a unique opportunity to study how whales learn their songs.

A team of experts—including psychologists studying language evolution, behavioral ecologists specializing in marine ecosystems, and marine ecologists—identified a fascinating parallel:

Based on this they predicted ‘that statistically coherent subsequences will be present in whale song and the distribution of these subsequences will follow a language-like power-law distribution.’ In simple words, they predicted whale songs would have repeating, meaningful patterns, and these patterns would show up in a way that’s similar to how human language works.

This is a distribution which has been proposed by the linguist George Zipf (1902-1950) in his 1939 book, The Psycho-Biology of Language.

According to Zipf, if you rank words by how often they appear (like first rank for the most common, second for the next, and so on), the rank of a word and how frequently it’s used are inversely proportional. This means that the higher a word’s rank (like 1st or 2nd), the more often it appears, and as the rank gets lower, the frequency drops off predictably.

For example, the second most common word tends to appear about half as often as the first, the third about a third as often, and so on.

Though Zipf proposed this only in the context of language, this phenomenon finds its application in multiple domains. Naturally, this also helps in understanding human language acquisition.

Kids learn common words first because they hear them most. The polymath mathematician Mandelbrot gave a mathematical explanation for Zipf’s word-frequency idea. For example, even if a monkey types randomly, short "words" pop up more often, showing this pattern isn’t just about language.

Studies have been conducted applying the Zipf-Mandelbrot approach to study language acquisition in human children. Now the same method was applied for whale songs.

How does one understand the separators of the sub-sequences in a song? Through years of observations, scholars identified probabilities of occurrences of sound elements in the sequence.

For example, if the song contained the sequence ‘grunt'-‘grunt’-‘ascending moan,’ then the team would estimate the probability of ‘grunt’ coming after ‘grunt’ in the entire song. Then based on the relation between consecutive transitional probabilities which have been observed in previous works, the segmentation boundaries were inferred. In plain words, where one part of the song ends and another begins.

Now the team of Arnon et al. plotted the frequency of humpback whale songs' sub-sequences against their rank on a logarithmic scale. The most common sub-sequence got rank 1, the next most common rank 2, and so on. The line they obtained was a straight line, very similar to the ‘Zipfian distribution’ found in human infants learning words from sounds.

This finding seems to suggest that sound-pattern-based communication in non-human animals follow a law that is also present in human language acquisition.

The paper although cautions that this does not mean the findings make any claim to semantic content or meaning to the whale songs. Just because whale songs have a pattern like human language, it doesn’t mean the songs have meanings or messages like words do.

In other words, the observed structure of whale songs resembles human music more closely than language. (In humans too, songs are an integral part of courtship).

This is an example of convergent evolution, where similar processes develop independently across different branches of life. In this case, culturally transmitted sound structures with nested hierarchies follow similar laws in how they spread across groups.

While we cannot yet determine the meaning of whale songs, it is possible that the way meaning maps to sound structures also follows certain principles, which future research may uncover.

Journal reference: Inbal Arnon et al., Whale song shows language-like statistical structure, Science, 7 Feb 2025 • Vol. 387 Iss. 6734, pp. 649-653