From the musical whistles of a red-whiskered bulbul to the mechanical whirr of a barn swallow song, there is an incredible diversity to avian music. But according to a new study by researchers from France and the U.S., this vast diversity comes from just eight building blocks.
Researchers analysed more than one lakh song recordings from more than 3,000 passerine bird species worldwide using machine learning to reveal the eight core motifs, or basic sounds.
These motifs consist of three types of trills (repetition of the same note) distinguished by their speed, three types of whistles based on their pitch modulation, and two other complex sounds.
“If you go to different countries in the world, you will hear totally different [bird] songs,” said Quentin Bacquelé, one of the paper’s authors and a PhD student at the Universities of Montpellier and Saint-Étienne, France.
The central question of their study was whether it was possible to characterise the songs into broad types, Mr. Bacquelé added.
The researchers also did not just deconstruct birdsong down to their sounds: they also found that where birds reside influences the sounds they use.
Infinite possibilities
By sourcing data from Xeno-canto, a citizen-science repository of verified wildlife audio recordings, the researchers could analyse birdsong recordings from almost everywhere on the planet. But Mr. Bacquelé said it also meant there was data disparity, with more recordings coming from Europe and the Americas than Africa.
First, the team denoised the recordings to focus on the species of interest, then converted them into high-resolution spectrograms — visual maps that show sound frequencies over time. Following this, the researchers computed a modulation power spectrum (MPS) for each song, which reduced the acoustic signals to the constituent patterns of frequency and temporal modulation.
While less precise than a spectrogram, an MPS can provide a broad perspective on timbre, pitch, tonality, and rhythm, Mr. Bacquelé said.
“That’s really useful because when you have thousands of songs, you don’t want to know every detail. Instead, you want to know how similar one song is to another.”
The researchers were then able to explore the variability of these songs inside a 37-dimensional acoustic space (a 3D representation of which is available online). “It’s like all these differently shaped clouds altogether, where proximity reflects spectro-temporal similarities,” Mr. Bacquelé said.
With the help of machine learning, researchers found that all songs concentrate around eight poles in the acoustic space, each captured by a specific motif. These motifs included three types of trills — slow, fast, and ultrafast; three types of whistles — flat, slow-modulated and fast-modulated; atonal chaotic notes, which are harsh, rough sounds; and harmonic stacks, where there are several low- and high-pitched tones at once, like in the song of a crow.
The main finding is that these motifs serve as building blocks, with bird species using just one or all eight in their songs, said Pooja Choksi, co-founder and lead scientist at India Ecoacoustics Network. She was not involved in the study.
“This is a great attempt to classify and generalise across such a large group of species,” Dr. Choksi said. “At the end of the day, motifs are simply characteristics of a song, and there will probably always be a debate on how to characterise something.”
Environmental constraints
Per the study, birds, on average, use five motifs and which ones depends largely on their evolutionary history. “If an ancestor is using a certain motif, there is a huge chance [the current generation] is using the same motif,” Mr. Bacquelé said.
Social organisation, morphology, and mating systems were the other determining factors. Researchers found that birds communicating over long distances tend to use simple motifs like flat whistles and slow trills whereas those signalling at close range favour faster, more complex motifs.
Similarly, simpler sounds are represented better in larger birds and more challenging motifs are predominant in smaller species. Finally, their analysis revealed that groups facing intense sexual selection pressures tend to use complex motifs while species subject to weaker pressures use simpler ones.
“Song motifs are essentially a trade-off between encoding more information and transmitting it over long distances,” Mr. Bacquelé said. “And what we found is this trade-off is different between temperate and tropical forests.”
Upon mapping the global geographic distribution of motifs, they found that in tropical rainforests such as in Southeast Asia, birds prefer simpler songs that travel efficiently through dense vegetation. In temperate regions such as Europe and North America, on the other hand, birds have small territories and short breeding seasons, and complex, information-rich songs.
This is the biggest finding from their research, according to Mr. Bacquelé: it reveals how environmental constraints shape birdsong at a planetary scale.
More to come
“It makes logical sense that species have adapted their songs for optimal transmission in their local environment,” Dr. Choksi said. “But it is interesting to see this pattern in such a large global analysis.” She also questioned whether the patterns would hold at much finer spatial resolutions.
“The [motifs] are broad categories, and within each, there can be a lot of variation in frequency and shape,” said Elodie Briefer, an associate professor at the University of Copenhagen who was not involved with the study. So there is variation among species that share spaces, according to Briefer, whose research focuses on animal cognition and communication.
The birds themselves perceive many more nuances within these eight motifs, Dr. Briefer, who has worked closely with skylarks, added. She was, however, interested in building on this research and exploring how birds perceive these eight motifs and whether any factors influence the order in which motifs are used.
Dr. Choksi said she believes it is worth studying how habitat changes affect birdsong and mating success and reproduction. This is particularly true in India, where invasive species have proliferated and tree-planting is almost out of control, she added.
Nikhil Sreekandan is an independent journalist.
