Where the syrinx is
The sits at the base of the trachea where it bifurcates into the two bronchi leading to the two lungs. Membranes (the tympaniform membranes) vibrate as air flows past them; the bird controls airflow, membrane tension, and the syrinx's surrounding muscles to produce a remarkable variety of sounds. The two sides of the syrinx — one on each bronchus — can be controlled semi-independently, which is the anatomical basis for the 'two-voice' capability that some songbirds display. The architecture is distinct from the mammalian larynx (which sits at the top of the trachea and has vocal cords that vibrate similarly but with a single source). The syrinx is a bird invention; mammals don't have anything like it.
Several songbird species — particularly the more vocally-sophisticated songbirds — can produce two pitches simultaneously, with each side of the syrinx generating an independent frequency.
What the two-voice capability enables
Several songbird species — particularly the more vocally-sophisticated songbirds — can produce two pitches simultaneously, with each side of the generating an independent frequency. Wood thrushes, song sparrows, brown thrashers, and many others use this for the structural complexity of their songs. The capability isn't universal; not all songbirds use both syringeal sides equally. American crow appears to use the two-voice capability less than typical singing songbirds, with most crow vocalizations using primarily one side or producing harmonically-related sound that doesn't fully exploit the two-source architecture. This shapes the corvid acoustic repertoire toward shorter, more impulsive calls rather than the long sustained two-voice songs of, e.g., a thrush.
Coen Elemans's work
Coen Elemans at the University of Southern Denmark is one of the leading researchers on physiology. His work uses high-speed video and biomechanical modeling to understand how the syringeal membranes actually vibrate, what muscles control which parameters, and how the system's degrees of freedom map to acoustic output. The work is technical but important: it provides the physical-mechanism understanding that turns 'birds produce songs' into 'birds produce specific frequencies via specific muscle contractions in specific anatomical structures.' The biomechanical model of bird vocalization is now well-developed enough to predict acoustic output from physiological parameters in some cases.
Why this matters for AI bioacoustics
Bird vocalizations carry signatures of the physiological system that produced them. The frequency range a species can produce, the call-shape patterns it can create, the timing precision it can achieve, the spectral richness it can generate — all of these are constrained by syringeal anatomy and the muscles surrounding it. AI bioacoustic models learn these patterns implicitly: a classifier doesn't know about syringes, but it learns acoustic features that ultimately reflect syringeal physiology. Understanding the underlying biology helps researchers know when an AI model is finding real structure (consistent with what the bird can produce) versus finding artifacts (acoustic patterns that don't have a plausible biological generator). It also helps with cross-species generalization: species with similar syringeal anatomy plausibly share vocal capabilities in ways that species with different anatomy don't.
What crows specifically can produce
Crow syringeal anatomy supports a wide vocal range with strong emphasis on impulsive, harmonically-rich calls rather than sustained tonal singing. The 'caw' family of vocalizations is the most-recognizable example: short broadband bursts with characteristic harmonic structure. The rattle call, juvenile-begging vocalizations, and various contact calls all use the same anatomical machinery in different configurations. The species can also produce remarkable mimicry of other species' sounds, of mechanical sounds, and of human speech in captive settings — the syringeal machinery is flexible enough to support significant out-of-typical-repertoire vocal production. What it cannot easily do is produce the long sustained two-voice songs of thrushes or warblers; that's not in the crow vocal repertoire and the anatomy biases the species away from it.