The invention relates to a method for monitoring the vibrational state of a wind turbine comprising detecting a plurality of acceleration values that represent accelerations which a component of the wind turbine is subject to at different respective points in time within a specified time interval, storing the acceleration values in an acceleration data set; generating a sum frequency distribution on the basis of the acceleration data set, comparing a first sum frequency value of the sum frequency distribution with a second sum frequency value of a threshold sum frequency distribution for at least one acceleration value, and outputting a warning signal if the second sum frequency value is greater than the first sum frequency value.