On the rooftop of a six-story building in Mumbai's Bandra West neighbourhood, a juvenile Oriental magpie-robin sat on the lip of a water tank at 06:18 and produced a soft, formless string of notes. It was the eighth of May.
Inara Khan, who lives in the building, had heard the same bird in the same spot every morning for ten days. The bird was perhaps ten weeks out of the nest, no longer being fed by its parents, holding a position near an active adult male's territory but not yet challenging it.
What the bird was producing is technically called subsong. It is a stage of avian vocal development that has been described in the scientific literature for more than seventy years, but is heard by most people, including most birders, as background.
Peter Marler, again, did the foundational work. In a 1956 paper on chaffinches in England, he described the developmental sequence by which a young male moves from undifferentiated subsong to plastic song to fully crystallized adult song.
The sequence holds, in broad outline, across the songbirds. The young bird first produces quiet, unstructured sound. It then begins to incorporate elements that resemble the species' adult song, but in odd order and with frequent errors. Finally, over a period of weeks or months, the song stabilizes into the adult pattern.
What is happening neurally during subsong is, in part, the construction of a motor template based on what the bird has heard. Songbirds learn their songs by hearing adult conspecifics during a sensitive period in early life, then practicing until their own output matches the memorized template.
This is one of the few clear analogues in the animal world to human language acquisition. The bird hears, internalizes, practices, and converges on the model.
The young magpie-robin on Khan's rooftop is in the late stages of this process. Khan recorded several mornings of his vocalizations on her phone and compared them to the song of the adult male who held the adjacent territory across the lane.
The adult, whom Khan calls the Pali male, sings a distinctive song with seven recognizable phrase types in his rotation. The juvenile, by the third week of May, was producing five of them, recognizably but imperfectly.
Two of the phrases were not yet present in the juvenile's repertoire. One was the long, ascending whistle that the adult uses as the opening of his dawn song. The other was a complex four-note phrase that the adult deploys most often during territorial encounters.
That the juvenile had not yet acquired the territorial phrase is unsurprising. It is a context-specific vocalization that he may not have heard the adult produce often enough to memorize.
What was striking, in the recordings, was the matching of phrase intervals. The juvenile was not just producing approximations of the adult's notes. He was producing them in approximately the right temporal relationships.
This is what crystallization means. The young bird is not assembling a song from scratch. He is converging on a model.
The magpie-robin is a useful species to watch this in, because the adult song is loud and consistent and the juvenile subsong is often produced in the same location. Khan's rooftop is unusually convenient, with the juvenile's regular perch in line of sight of the adult's regular perch.
In quieter species, subsong can be nearly inaudible. Researchers studying it in white-crowned sparrows have had to use sensitive directional microphones at close range. The bird's beak barely opens.
What subsong tells us, at the level of natural history, is that song does not appear. It is built. The bird that sings perfectly in May was, the previous June or July, a fledgling making sounds that no one would call a song.
For the urban birder, this is a useful corrective. The mockingbird in your tree did not arrive fully formed. The bulbul on the wire learned its song the same way you learned your first sentences, by imitating what you heard, badly at first, until you converged on the model.
The Bandra juvenile has continued his rehearsal through the first week of June. Khan expects he will be in adult song by July. If he secures a territory of his own, he will keep that song, with minor modifications, for the rest of his life.
The Pali male, his model, was banded in 2022 by a researcher at the Bombay Natural History Society. He is now in his fifth year on the same territory. He sings, at 05:45 on most mornings, the same seven-phrase rotation he has sung since he claimed the lane.
The juvenile is learning, in effect, an inherited document. The transmission is quiet. Most mornings, on most rooftops, no one is listening for it.
