An analysis of the evolution of the amplitude spectrum over many seasons of the DA pulsating white dwarf G29-38 has been performed. Neither beating nor resonant mode coupling can account for the observed appearance and disappearance of modes, although some of them clearly grow while others get damped. Therefore some unknown non-adiabatic, non-linear process has to be invoked that affects both the mode selection mechanism and the driving efficiency on a time-scale as short as a day.