Summary Velocity distributions within three models of the human larynx, namely, a rigid plexiglas model, an excised canine larynx, and a computational model are investigated with experimental and theoretical analyses. A plexiglas wind tunnel with interchangeable glottal constrictions was used as a two-dimensional steady-flow model to measure velocity and pressure for various glottal shapes. A canine excised larynx was used as a prototype pulsatile flow model to study pressure and velocity variations during phonation. Results of the plexiglas modelling indicated a parabolic laminar velocity profile upstream of the glottal constriction and turbulent and asymmetric velocity profiles downstream of the glottal constriction. The time-averaged velocities of the excised larynx had similarities with the plexiglas model results, and instabilities and asymmetries were also demonstrated by the computational method.