For centuries, naturalists have suggested that the tonal elements of pigeon wing sounds may be sonations (non-vocal acoustic signals) of alarm. However, spurious tonal sounds may be produced passively as a result of aeroelastic flutter in the flight feathers of almost all birds. Using mechanistic criteria emerging from recent work on sonations, we sought to: (1) identify characteristics of rock pigeon flight feathers that might be adapted for sound production rather than flight, and (2) provide evidence that this morphology is necessary for in vivo sound production and is sufficient to replicate in vivo sounds. Pigeons produce tonal sounds (700±50 Hz) during the latter two-thirds of each downstroke during take-off. These tones are produced when a small region of long, curved barbs on the inner vane of the outermost primary feather (P10) aeroelastically flutters. Tones were silenced in live birds when we experimentally increased the stiffness of this region to prevent flutter. Isolated P10 feathers were sufficient to reproduce in vivo sounds when spun at the peak angular velocity of downstroke (53.9-60.3 rad s-1), but did not produce tones at average downstroke velocity (31.8 rad s-1), whereas P9 and P1 feathers never produced tones. P10 feathers had significantly lower coefficients of resultant aerodynamic force (CR) when spun at peak angular velocity than at average angular velocity, revealing that production of tonal sounds incurs an aerodynamic cost. P9 and P1 feathers did not show this difference in CR. These mechanistic results suggest that the tonal sounds produced by P10 feathers are not incidental and may function in communication.