The aim of the study was to assess the reproducibility of vibrations recorded from the temporomandibular joint (TMJ) in a group of healthy subjects. The vibrations from TMJ were recorded bilaterally from 34 healthy subjects by electrovibratography in three sessions at intervals of 3 min and again after 1 week. The total integral of the vibration energy, the ratio of the integral between frequencies above 300 Hz and below 300 Hz (ratio of >300 Hz/<300 Hz), peak frequency, median frequency, peak amplitude and distance to centric occlusion position were calculated. Data were analysed with intraclass correlation coefficients (ICC) and two-way anova for repeated measures. All variables showed good to excellent reliability across different sessions (ICCday1 : 0·935-0·987; ICCday2 : 0·910-0·992) and across different days (ICC: 0·738-0·907). According to anova for repeated measures, all variables showed good reproducibility (P > 0·05) between sessions at the same day. There was no significant difference between the 2 days for the frequency-related variables including peak frequency (P = 0·083), median frequency (P = 0·188) and ratio of >300 Hz/<300 Hz (P = 0·26). There was a statistical difference between the 2 days for the intensity-related vibration variables including total integral (P = 0·045) and peak amplitude (P = 0·026). The wave patterns of the power-frequency spectra were qualitatively similar over both the sessions and days. Joint vibration analysis could provide a fast, non-invasive, and repeatable method to record the status of TMJ. Further studies are needed to identify the characteristic waveforms for different subgroups of temporomandibular disorders and to evaluate the possibility of diagnostic value.