Wild Parrot Duets Follow Syntax Rules, 10-Year Field Study Reveals

The videos are easy to find: parrots mimicking human speech, corvids solving puzzles, cockatoos dancing to pop songs. Bird cognition research has spent decades on captive animals demonstrating impressive feats of learning and association. What remains far murkier is what birds say to each other in the wild, on their own terms, without a human in the room.

Christine Dahlin, professor of biology at the University of Pittsburgh at Johnstown, has spent the last decade trying to answer that question for one specific bird: the Yellow-naped Amazon parrot (Amazona auropalliata), a critically endangered species ranging from southern Mexico to southern Costa Rica. Her team's analysis of wild territorial duets, published in the Journal of Avian Biology, finds these vocalizations are not random noise - they follow consistent rules of structure that parallel, in several measurable ways, the grammar of human language.

Warble Duets vs. Standard Duets

Yellow-naped Amazons produce different types of vocalizations for different contexts. Dahlin had previously studied what she calls "standard duets" - relatively simple paired calls with only four call types - and found these had syntax. The warble duet is something different. Longer, louder, and far more complex, warble duets occur specifically during territorial disputes when pairs of parrots confront rival pairs. They sound, Dahlin said, like the birds are getting in each other's faces.

"The warble duets would get really fast and really loud when there was a territory dispute," she said. "You could tell the context was changing." Every physical altercation between parrots that Dahlin observed in the wild was preceded by warble duets.

The question the new study addressed was whether these complex, high-stakes vocalizations also have syntactic structure, or whether the variety of calls is simply expressive noise - a display of vocal range rather than organized communication.

Three Years of Fieldwork, Ten Years of Analysis

Data collection required multiple research trips to Costa Rica over three years. Dahlin and field assistants stationed themselves in cattle pastures with video cameras and directional microphones, waiting for parrots to arrive at their breeding territories and monitoring interactions. From hundreds of recorded duets, 50 were warble duets, produced by 13 mated pairs. Those 50 warble duets contained more than 450 individual calls.

Transcribing and categorizing that volume of vocal data took the research team - primarily Dahlin and undergraduate students at UPJ - more than ten years of iterative work alongside other projects. The resulting catalog identified 36 distinct call types. More exist: some call types appeared so rarely, in some cases only once, that they could not be assigned to a reliable category.

Standard duets contain four call types. Warble duets contain at least 36. The question was whether those 36 types were used randomly or according to rules.

A Text-Analysis Tool Finds Grammar in Bird Song

The methodological breakthrough came from an unexpected direction. Owen Small, a UPJ undergraduate, was using Voyant Tools - software designed to analyze patterns in literary texts - for a humanities course. He suggested trying it on the bird call data.

"Voyant was able to run the same analysis as if it were a body of writing," Dahlin said. "And the results show the parallels between these complex signals the birds are giving and our own language."

The analysis found syntax - consistent rules governing the sequence in which call types appear. It found a broad lexicon: 36 identifiable call types is a large repertoire for a wild bird. And it found collocates - pairs of call types that consistently appear together, analogous to how words like "grass" and "green" or "eat" and "food" co-occur in human language with above-chance frequency. The team identified more than 20 such syntactic patterns.

Collocates are significant because they suggest semantic relationships rather than purely sequential rules. If two call types regularly co-occur, it may mean they convey related information, or that one call primes the context in which the other makes sense. Whether that is true for parrot warble calls requires further research, but the statistical pattern itself is clear.

What This Does and Does Not Tell Us

Finding language-like structure in bird communication does not mean Yellow-naped Amazons have language in the full human sense. Syntax without semantics - rules about sequence without established meanings for the individual units - is a structural parallel, not equivalence. The study identifies that these calls follow consistent organizational rules; it does not decode what specific calls mean, or whether the birds intend their vocalizations in anything like the way humans intend words.

The study is also limited by sample size. Fifty warble duets from 13 pairs is a modest dataset for statistical analysis of complex combinatorial structure. Rare call types could not be categorized at all. The data come from one population in one geographic region; Yellow-naped Amazon populations across their range are known to have regional dialects in their simpler calls, and whether warble duet structure varies geographically is unknown.

Dahlin is clear about what comes next: understanding whether and how individual pairs match or diverge from each other's calls during confrontations, which requires knowing the call inventory first. "That started this whole process," she said of the years spent simply cataloging what sounds exist. Now that the inventory exists, studying the dynamics of territorial confrontations - who copies whom, when, and to what effect - becomes tractable.

The species itself faces pressure beyond academic interest. Yellow-naped Amazon parrots are critically endangered, their populations reduced by habitat loss and capture for the pet trade. Understanding their communication systems may have conservation value: territorial vocalization plays a role in breeding success and habitat defense, and disruption of those systems - from habitat fragmentation or population decline - could affect reproductive outcomes in ways not yet quantified.