Purpose: The aim of this study was to test the utility and validity of a newly developed EMG-based bruxism recording system, which allowed high-resolution digital recordings of the masseter EMG in the patient's home environment and systematic discrimination of artifact signals with the aid of semiautomated software. Materials and Methods: First, simulated bruxism and nonbruxism signals were recorded in two subjects; a total of 72 signals were obtained. Two independent scorers were shown these signals and asked to differentiate bruxism from nonbruxism signals. Second, EMG recordings were actually performed in 56 subjects for six consecutive nights. Using 10 randomly selected recordings, data cleaning was performed by the two scorers to remove any EMG artifact signals, and five bruxism event outcome variables (duration per event, hour, and night, and number per hour and night) were obtained twice with a 7-day interval. Results: The signal differentiation test using signals recorded during simulated tasks revealed that 97% of simulated bruxism and 87% of nonbruxism signals were recognized accurately by the scorers. A total of 317 night recordings, 94% of the scheduled recording nights, were successfully completed by 56 subjects without any substantial difficulty. The results of analysis of intraclass correlation coefficient of outcome variables scored by the two scorers revealed that intrascorer reproducibility varied between .70 and .97. The intraclass correlation coefficients for interscorer reliability were >= .84. Conclusion: This EMG-based bruxism recording system has high utility and reasonable accuracy and precision.