The vibration data from defective gearboxes generally exhibit non- stationary and nonlinear property. As a result,it seems hard to effectively monitor the running status of gearbox by using the conventional time domain statistical parameters method which is based on stationary and linear theory. In order to solve this problem,the method of time series statistical linguistic analysis is adopted to examine gearbox vibration data and a novel method for gearbox condition monitoring based on time series statistical linguistic analysis is proposed. Firstly,an original series is mapped into a word- occurrence- frequency series,afterwards,a correlation coefficient of two word- occurrence- frequency series for the initial condition and another condition is used as a characteristic parameter for detecting a change of gearbox conditions. Finally,the proposed method is applied to condition monitoring of a realistic gearbox. The results indicated that the proposed method can both effectively detect changes of gearbox conditions and unravel a natural evolutionary process of gearbox conditions.In addition,the proposed method has a clear advantage over the conventional time domain statistical parameters method in condition monitoring of gearbox.