Storing Data in a Starling: Where Bioacoustics Meets Information Theory
Benn Jordan, the musician and sound designer who runs the YouTube channel of the same name, has produced a piece that is part nature documentary, part DIY hardware guide, and part fever dream about turning songbirds into biological storage media. The video -- transcribed here from its original format -- follows Jordan as he hauls ultrasonic recording equipment across the American Southeast, builds automated bird-monitoring stations, and ultimately travels to Louisville, Kentucky to record a rescued European Starling named "the Mouth" that can replicate virtually any sound it hears, including the iPhone camera shutter, human speech patterns, and ambient room reverb.
The central stunt is audacious. Jordan drew a picture of a bird inside a spectral synthesizer, converting the image into sound, then played that sound for the Starling on the off chance that the bird would learn to reproduce it. If successful, the Starling would effectively be storing and transmitting image data using nothing but its syrinx and lungs.
The Syrinx: A Biological Synthesizer
The most technically fascinating material in the piece concerns avian vocal anatomy. Humans produce sound by pushing air through a single set of vocal folds in the larynx. Birds, particularly songbirds, use a syrinx -- a structure that places two pairs of muscles on each passage coming from the lungs, with some species possessing as many as fourteen of these muscles. The result is a vocal instrument of staggering precision and speed.
They can go from this note to this note dozens if not hundreds, sometimes even thousands of times per second.
Jordan draws an elegant parallel to radio technology. When a bird modulates the volume of its two airstreams rapidly enough, it produces amplitude modulation. When it modulates frequency, it produces FM synthesis -- the same principle behind 1980s pop keyboards and, more fundamentally, behind how radio waves carry audio signals. The bird's syrinx is not merely an instrument; it is a signal processor operating on the same physical principles as telecommunications hardware.
This comparison is not mere analogy. When Jordan records the Starling with ultrasonic microphones capable of capturing frequencies up to 192 kilohertz and then slows the audio down by a factor of nine, the spectrograms reveal layers of modulation that are invisible at normal playback speed. The bird is doing things with sound that human perception cannot track in real time.
The Data Transfer Experiment
The headline claim -- that Jordan "saved a PNG image to a bird" -- is deliberately provocative, and he is honest enough to acknowledge its limitations. After playing the spectral bird-drawing sound for the Starling, he did not observe the bird reproduce it during his visit. It was only later, while reviewing gigabytes of recorded audio, that he found something in the spectrogram.
When I got home, when going through my many gigabytes of Starling sounds in this recording session, I noticed a little bird in the spectrogram. At first, I thought I was just simply seeing my phone playing the sound to the Starling, but this was much later in my visit, and it was combined with another type of vocalization.
The Starling's reproduction was not perfect -- it was approximately 50 to 60 hertz flat, and it was embedded within other vocalizations rather than delivered as a clean isolated signal. Jordan calculates that the successful reproduction transferred roughly 176 kilobytes of uncompressed information, and extrapolates that a hypothetical audible file transfer protocol using 10:1 compression could move nearly 2 megabytes per second.
This is where scientific caution should temper enthusiasm. A single observation of a captive bird reproducing a novel sound, discovered after the fact in a spectrogram, does not constitute a reliable data storage mechanism. The bird chose what to learn, when to reproduce it, and with what fidelity -- none of which a storage protocol can tolerate being nondeterministic. Jordan acknowledges the caveats but still frames the result as a proof of concept, which is generous to the concept.
Starlings Versus Crows: An Intelligence Reframe
Jordan makes a compelling argument that the popular fascination with corvid intelligence reflects anthropomorphic bias more than objective measurement. Crows solve puzzles collaboratively, which humans find relatable because it resembles their own problem-solving. Starlings, meanwhile, perform murmurations -- coordinated flocking patterns that require individual birds to track the movements of their seven nearest neighbors and replicate flight adjustments with precise latency.
I just don't think that humans are very good at measuring the intelligence of a class of species that have been evolving for well over 150 million years before humans became a thing.
This is a fair point, though it risks overcorrecting. The research on corvid cognition -- tool use, causal reasoning, episodic-like memory -- is robust and well-replicated across multiple species and laboratories. Starling murmurations are extraordinary, but they may reflect distributed computation (each bird following simple local rules) rather than individual cognitive sophistication. The two forms of intelligence are not necessarily comparable on a single scale. Jordan's instinct to question anthropocentric rankings is sound; the specific comparison he draws is less so.
The DIY Bioacoustics Kit
Roughly half the piece functions as a practical hardware guide, and this is where Jordan's background as an audio engineer shows most clearly. He walks through ultrasonic recording setups at three price points, evaluates specific microphones and recorders by name, and flags common purchasing mistakes -- including a warning about suspiciously cheap 384 kilohertz audio interfaces on Amazon and Alibaba that almost certainly do not deliver their advertised specifications.
But if you do not own any audio gear whatsoever, you should be able to get started recording high-quality professional scientific stereo ultrasonic recordings for around 500 bucks.
The BirdNET Pi section is particularly strong. Jordan describes a system built on a Raspberry Pi that continuously monitors, records, classifies, and organizes bird songs without requiring internet connectivity, subscriptions, or personal data. He provides three build configurations ranging from $40 to $250, each with specific component recommendations. The enthusiasm is infectious, and unlike many technology recommendations, these are clearly born from extended personal use rather than spec-sheet reading.
His automated monitoring station, left running during a field recording trip, captured 31 bird species on his property, including chimney swifts and common grackles approaching endangered status, as well as red-headed woodpeckers whose populations are in sharp decline. This is citizen science at its most effective -- passive, continuous monitoring that generates data no human observer could produce through intermittent attention.
The Temporal Resolution Gap
One of the piece's most thought-provoking observations concerns how birds experience time. The critical fusion flicker frequency of passerine birds -- the rate at which they can detect changes in light patterns -- exceeds 150 hertz. For context, most human displays refresh at 60 hertz. A smartphone lock screen that appears static to a human would look like a flickering strobe to a Starling.
Jordan extends this principle to sound. If a Starling perceives time at higher temporal resolution, then seconds effectively last longer for the bird, and imperfections in pitch and timing that humans cannot detect become glaringly obvious in "bird time." This reframes the Starling's mimicry not as approximate imitation but as high-fidelity reproduction operating at a perceptual resolution humans simply cannot access without specialized equipment.
The implication is humbling. When researchers evaluate bird song as "close enough" to the original, they may be applying a perceptual standard that is irrelevant to the bird. The ultrasonic microphone does not just reveal hidden detail -- it reveals hidden standards of precision.
Bottom Line
Benn Jordan's piece is sprawling, digressive, and occasionally undermined by its own enthusiasm -- the "storing data in a bird" framing, while memorable, overstates what a single anecdotal observation can support. But the underlying material is genuinely remarkable. The syrinx-as-FM-synthesizer analysis is elegant and technically precise. The temporal resolution discussion reframes avian cognition in ways that challenge comfortable assumptions about perception. And the practical hardware guides lower the barrier to bioacoustic research to a point where a curious hobbyist with $40 and a Raspberry Pi can contribute meaningful observational data. The piece is at its best when Jordan lets the science speak louder than the clickbait -- which, fortunately, is most of the time.