The aim of this study was to develop and validate under laboratory conditions an algorithm for a time-frequency analysis of rhythmic masticatory muscle activity (RMMA). The algorithm baseband demodulated the electromyographic (EMG) signal to provide a frequency versus time representation. Using appropriate thresholds for frequency and power parameters, it was possible to automatically assess the features of RMMA without examiner interaction. The algorithm was first tested using synthetic EMG signals and then using real EMG signals obtained from the masticatory muscles of 11 human subjects who underwent well-defined rhythmic, static, and possible confounding oral tasks. The accuracy of detection was quantified by receiver operating characteristics (ROC) curves. Sensitivity and specificity values were >/=90% and >/=96%, respectively. The areas under the ROC curves were >/=95% (standard error +/-0.1%). The proposed approach represents a promising tool to effectively investigate rhythmical contractions of the masticatory muscles. Muscle Nerve, 2009.