Very little has been done to quantify temporomandibular joint (TMJ) sound amplitudes and background noise and to determine the spectrum from healthy TMJs. Thus, the aim of this study was to record acoustically the sounds emitted by healthy TMJs with and without mandibular movements, for determination of baseline spectra. TMJ sounds were recorded bilaterally from 40 subjects with healthy joints by means of a self-developed recording system using miniature capacitor microphones inserted into the earpieces of a medical stethoscope placed in the meatus of the auditory canal. The recordings were performed without mandibular movements and during three consecutive opening and closing movements. The signals were high-pass-filtered at 50 Hz, low-pass-filtered at 2 kHz, and analyzed by fast Fourier transform computation on a 1024-point window (fs = 5 kHz). The linearly weighted average baseline spectrum recorded without motion showed maximum values of 31 dBSPL (sound pressure level) with a standard error of +2 to -3 dB. The linearly weighted average movement spectrum had a peak of 66 dBSPL at 156 Hz and decreased almost linearly by about 40 dB/decade to 25 dBSPL at 2000 Hz with a standard error of +/- 2 dB. Thus, the TMJ sound spectrum of mandibular movements in asymptomatic subjects differed at low frequencies by up to 35 dB from the baseline spectrum in absence of motion.