In their study of time‐dependent thermocapillary convection in cylindrical liquid bridges Veltenetal. [Phys. Fluids A3, 267 (1991)] determined the azimuthal flow structure by measuring the phase of the temperature oscillations at three different azimuthal angles. Their explanation of the data in terms of modes with wave numberm=0 is not always compatible with the normal‐mode decomposition of the disturbances at the threshold. A new interpretation is given that explains the peculiar phase measurements in terms of pairs ofm≠0 counterpropagating waves.