This paper introduces the theory of cyclic statistics as a powerful tool for the diagnosis of gear faults. More precisely, a new method based on cyclic bispectrum, a third order cyclic statistics, is used for monitoring. This estimator furnishes a valuable means of detecting and characterising non-linear coupling effects as well as periodic correlation between different parts in machines. Therefore, cyclic bispectrum provides much more combined information than classical methods such as spectrum and cepstrum analysis. Application in diagnosis of spalling in gear teeth of U.S. Navy helicopter demonstrates the effectiveness of this new parameter for a good diagnosis.