Journal of Oral Rehabilitation 1998 25; 765–769
Subjective assessment of temporomandibular joint sounds
J . F. P R I N Z Department of Anatomy, University of Hong Kong
SUMMARY If it could be shown that the human ear
was sufficiently sensitive to describe TMJ sounds,
there would be no need to use sophisticated
electronic equipment to analyse the sounds. To
test this, the ability of normal listeners to
distinguish the subtle changes in position, pitch,
duration and latency present in TMJ sounds is
measured using triangle tests to determine the
just-noticeable differences. The results suggest that
Introduction
The human ear is a very sensitive detection system,
with a very large (140 dB) dynamic range and frequency
sensitivity ranging from below 20 Hz to over 20 kHz.
Subjective assessment is the preferred assessment
method for comparing hi-fi audio equipment, since
small differences, which cannot be determined using
measuring instrumentation, can be perceived by the
ear. The pattern recognition capabilities of the human
brain are unsurpassed; for instance, it is possible for a
person to be identified on the telephone from a few
spoken words. However, discrimination of the
amplitude of sounds is poor. For example, when walking
towards a radio playing music, the perceived loudness
remains almost constant, even though the power of the
signal from the loudspeaker follows an inverse square
relationship with the distance. It is difficult to quantify
features of a sound subjectively and auditory memory
for most types of sound is poor. Although it is possible
to compare two sounds presented in a short space of
time, this becomes progressively more difficult as the
time interval between them increases. Thus, although
it might be possible to assess the reproducibility of a
series of TMJ sounds during a single session, it would
© 1998 Blackwell Science Ltd
the human ear is a rather poor instrument for
describing subtle differences in the position,
duration and latency of TMJ sounds, but is capable
of detecting small differences in frequency. It is
therefore doubtful that the human ear can
distinguish the reciprocal click associated with disc
displacement with reduction from clicks due to
defects of form on the basis of their relative
position in the envelope of movement.
be difficult to say if the sounds were louder than they
were the year or even the week before. The same type
of difficulty relates to the verbal description of sounds.
For instance, TMJ sounds are usually described in
terms relating to their aural characteristics. However,
opportunities for confusion abound, since authors
variously refer to high or low amplitude, hard or
soft, loud and soft, and fine or coarse sounds without
rigorously defining these terms.
When listening to music in order to judge the quality
of a hi-fi system, several minutes may be needed to
form a judgement of sound quality but TMJ sounds are
very short – of the order of 30–40 ms, and only one or
two examples of the sound may be heard. Thus,
although some people claim to be able to hear subtle
improvements in recorded music achieved, for instance,
by using gold-plated connectors on their speaker leads,
when listening to TMJ sounds the performance of the
human ear does not achieve anything like these levels
of sensitivity.
Just as drawings such as a Necker cube (Fig. 1) where
the orientation flips between two gestalts can induce
visual illusions (Rucker, 1986), some sounds produce
auditory illusions. Two common illusions are: (1) when
a low amplitude sound precedes a high amplitude
765
766 J . F. P R I N Z
Fig. 1. The Necker cube.
Fig. 2. Watt’s hard crepitus. Watt provided a tape recording of
TMJ sounds with his book (Watt, 1981). This 300 ms sample is
an example of what he termed hard crepitus – described as ‘foot-
steps walking on gravel’. This sound appears continuous to the
ear. However, as can be seen, visually it consists of a series of
three separate impulses. This sound would be described as a creak
in this study.
sound, the first sound is either not perceived at all, or
is perceived as following the high amplitude sound
(Watt, 1981), and (2) when the interval (latency)
between a series of short pulses becomes short, only a
single long pulse is perceived.
Although TMJ sounds are typically divided into two
groups, clicks and crepitus, Prinz & Ng (1997) suggested
that a classification into three basic groups could be
more appropriate. Single short-duration sounds were
termed clicks, multiple short-duration sounds were
labelled creaks and all long-duration sounds were called
crepitus. Watt provided a tape recording of TMJ sounds
with his book (Watt, 1981). Figure 2 shows a 300 ms
sample of what he termed ‘hard crepitus’. This sound
appears continuous to the ear, but as can be seen,
visually it consists of a series of three separate impulses,
and could be interpreted as a series of three closely
spaced clicks, or, as a (triple) creak. Figure 3 shows an
example of Watt’s soft crepitus, this is similar to the
© 1998 Blackwell Science Ltd, Journal of Oral Rehabilitation 25; 765–769
Fig. 3. Watt’s soft crepitus. This 300 ms sample, taken from Watt’s
tape, is also perceived as a continuous sound, and was described
by Watt as a gentle rubbing sound.
sample shown in Fig. 2 in terms of its duration, but the
visual difference is obvious.
This potential conflict between the visual and aural
perception of Watt’s two types of crepitus is resolved by
the proposed classification scheme and it has been
argued that the three types of sound are the result of
different underlying mechanisms (Prinz, 1996; Prinz &
Ng, 1997; Prinz, 1998).
In clinical practice, TMJ sounds are currently usually
assessed using a stethoscope. In a research environment,
sonography, i.e. visual assessment of the sound
waveform, can also be used for description, with or
without computer assistance. These two methods are
not necessarily equivalent, i.e. the method used to
assess a sound can affect the perception of it. For
instance, the hard crepitus described by Watt appears
to the ear to be a long duration burst of continuous
sound, whereas visually it appears to consist of a series
of distinct pulses (Fig. 2), in contrast to the soft crepitus
which appears to be continuous both visually and
audibly (Fig. 3).
Observer reliability in detecting even the presence or
absence of TMJ sounds using non-instrumental
measures is poor (Hardison & Okeson, 1990; Westling,
Helkimo & Mattiason, 1992). Eriksson, Westesson &
Sjoberg (1987) recorded TMJ sounds on tape and then
replayed them to observers who were asked to classify
each sound as either click, crepitus or silence. Even
though Erikson used high quality equipment, observers
frequently mistook tape hiss for crepitus.
Psychophysical perceptions follow the Webber–
Feschner law which states that the response to a
stimulus follows a power law relationship (Cameron,
1975). With visual perception, differences of less than
1% in brightness, colour and size can readily be detected,
but auditory perception is far less sensitive (Pope, 1984).
Two thresholds are important in hearing – the absolute
threshold, i.e. the level at which a stimulus can be
detected (0 dB), and the just-noticeable difference, i.e.
the level at which an observer can detect that one
sound is louder than another.
 SUBJECTIVE ASSESSMENT OF TMJ SOUNDS
Loudness is normally expressed in dB(A) relative to
the threshold of human hearing adjusted to take account
of the ear’s frequency response (10–12 pa 5 0 dB(A) at
100 Hz). The smallest difference in loudness (i.e. the
just-noticeable difference) that can be discriminated is
reported to be about 10% for music and speech
(Cameron, 1975). However, no thresholds have been
reported for brief click-like sounds, such as those from
the TMJ.
Two factors determine the perceived origin of a sound:
firstly, the amplitude difference between the two ears
and secondly, the phase relationship at the two ears.
Thus, a sound appears to originate from the loudest
side and from the side at which it appears first (Blauert,
1982; Hashimoto et al., 1989, 1990). In normal listening,
a sound is transmitted through the air with a velocity
of 300 m/s, and the difference in the time of arrival at
each ear, i.e. the phase relationship, is used to locate
the origin of the sound in space. However, TMJ sounds
detected with a stethoscope, or with microphones, travel
through the tissues of the head at velocities ranging
from 64 m/s to 5000 m/s (Franke & von Gierk, 1952;
Potts, Chrisman & Buras, 1983; Prinz, 1985, 1991,
1998). The human brain is not designed to resolve the
location of acoustic signals of this type. If the phase and
amplitude relationships conflict, it is the amplitude
relationship that dominates the subjective assessment
of location (Blauert, 1982).
Materials and methods
When assessing TMJ sounds by ear, it is important to
monitor the frequency of the sound, its duration and
the interval between events, since it is these factors
which distinguish types of TMJ sound. The just-
noticeable difference can be assessed using a triangle
test. In this type of test, the observer is presented with
three items and asked to identify the odd one out.
Where there is no discernible difference, the observer
is forced to guess. By varying the size of the difference
between the test samples, the just-noticeable difference
can be determined by observing the magnitude of the
difference at which a set proportion of the observers
correctly identify the ‘odd one out’ (O’Mahoney, 1988).
A series of triangle tests were used to determine a
typical listener’s sensitivity to duration, frequency and
latency. Sounds were produced using the computer’s
internal loudspeaker to produce bursts of sound. The
© 1998 Blackwell Science Ltd, Journal of Oral Rehabilitation 25; 765–769
767
amplitude of the sound was constant for all of the
following three types of triangle tests:
(i) Single sounds where only the frequency of one
sound differed.
(ii) Single sounds where only the duration of one
sound differed.
(iii) Triplets of 30 ms bursts of sound where only the
latency (i.e. the interval between the components of
the triplet) was varied between 10 and 300 ms.
The three items for comparison in each test type were
presented in random order to 10 normal (on self-report)
listeners who were asked to identify the odd one out.
Each judge performed 100 triangle tests with an interval
of 0·5 second between items. The sensory thresholds for
the smallest noticeable difference in duration, frequency
and latency were determined from the results. The
threshold for a just noticeable difference was set at
60% (i.e. the level at which 60% of assessments were
correct). Note that 33·3% of correct assessments can be
attributed to chance.
Aural assessment of TMJ sounds
The ability of five normal listeners to discriminate
between the three types of TMJ sounds, single events
(clicks), multiple events (creaks) and bursts of
continuous sound (crepitus), and to determine their
position in the cycle (opening, middle or closing) and
their location (unilateral left or right, or bilateral) was
tested. In addition, observers were asked to classify the
sounds as high, medium or low amplitude.
To allow precise control of the sounds to be presented
to the listeners, a set of test sounds were created. This
was done by splicing together digital samples from
actual TMJ sounds, recorded using microphones placed
in the ears, that had been digitised at 44·1 kHz with 16
bit resolution (i.e. CD quality) using a Sound-Blaster
16 sound card*. For instance, a typical synthetic cycle
might consist of a tooth contact sound followed by
300 ms of silence, a 300 ms sample containing a click,
300 ms of silence and finally another tooth contact
sound.
Three examples of each type of TMJ sound were used
in constructing the cycles. The three clicks used were
of different durations (40 ms, 20 ms and 15 ms) but
were modified so that they had the same peak
amplitude. The three examples of crepitus were formed
*Creative Inc., Singapore.
768 J . F. P R I N Z
from a single recording where the duration of the
crepitus was adjusted by editing the sample to create
crepitant sounds of 50 ms, 100 ms and 150 ms in length.
Three creaks were likewise formed from a single actual
recording, the interval between the events being
adjusted by adding or removing silence between the
individual events. Thus, it was possible to create three
creaks with an inter-event spacing (latency) of 20 ms,
50 ms and 100 ms.
During playback, the amplitude relationship (balance)
of the left and right channels was adjustable in steps.
Ten 2 dB steps were used from a maximum left
(–10 dB) to maximum right (110 dB). The volume of
each section of the cycle could also be precisely
controlled in 255 steps. Synthetic cycles were
constructed by splicing together the samples of tooth
contact sounds, silence, clicks, creaks and crepitus in
different orders, based on a randomized Latin squares
design. In these tests, although the balance, amplitude
and character of the sounds varied, only three positions
were presented, at 25, 50 or 75% of the cycle (early,
middle and late).
The synthetic cycles were played back from the
computer via headphones. First, a series of three
examples of each sound type were presented to the
listener as a training and calibration exercise. Following
a verbal description, each sound in the training set was
then repeated five times. The computer then presented
a series of 25 different synthetic sound sequences to
the listener, each sequence being repeated five times.
The listener was then asked to indicate the origin of
the sound (left, right or bilateral), the position in the
cycle (early, middle or late), a subjective evaluation of
the amplitude of the sound (high medium or low) and
finally to categorize the sound as either a click, creak,
crepitus or silence. Listeners’ assessments were
automatically recorded by a program, which allowed
the listeners to code responses. For the assessment of
the origin of the sound, responses were also categorized
as central or non-central since it appeared that several
of the listeners may have confused left and right when
responding.
Results
In the evaluation of the computer-generated sounds,
observers were very sensitive to frequency differences,
and were able to detect a difference of 2 Hz. For duration
differences, the just noticeable difference was 60 ms
© 1998 Blackwell Science Ltd, Journal of Oral Rehabilitation 25; 765–769
while for latency observers could detect differences of
30 ms (Fig. 6). Note that in this experiment the bursts
of sound generated by the loudspeaker had a sharp
attack and decay in contrast to the more gradual attack
and decay of clicks and creaks occurring in TMJ sounds;
hence discrimination of the individual bursts was easier
in the artificial situation than when listening to actual
TMJ sounds. There was a positive correlation of the
number of correct assessments with both frequency
(r 5 0·48; P , 0·05) and duration (r 5 0·58; P , 0·05),
i.e. at higher frequencies and with longer duration, the
discriminatory ability of the listeners improved. There
were no significant differences between listeners.
Ignoring the ‘don’t know’ responses in the evaluation
of the TMJ sounds, 52% of the clicks, 82% of the
creaks, 75% of the crepitus and 80% of the silent sound
sequences were correctly identified. All 250 mV sounds
were classed by listeners as high amplitude. Those
having an amplitude between 150 mV and 200 mV
were generally classed as medium amplitude. Those
below 100 mV were classed as low amplitude or silence.
Observers performed poorly on locating the sounds
in space, correctly classifying only 42% of the TMJ
sounds as left, right or bilateral; however, this may have
been due to left–right confusion since ANOVA showed
significant differences for listener. After a re-
classification of the sounds as central or non-central,
performance improved to 77% and there were no
longer any significant differences between listeners.
Performance was poor in classifying sounds as early,
middle or late, with only 49% of the sounds being
correctly classified.
Discussion
The results presented here cast some doubt on the
validity of the previous literature on TMJ sounds where
data were collected using a stethoscope and/or
palpation. Of course, in the clinical situation the dentist
has a great deal more information than just sound.
There are also visual and tactile sources. Although
listeners recognized the difficulty of the triangle test,
they expressed some surprise when presented with the
results of their evaluations, most believing that they
had performed better than was the case.
In the tests using synthetic cycles, every cycle was
constructed with identical tooth contact sounds at the
start and finish and all had the same overall duration
(1 s). This test was in some ways simpler than the real
 SUBJECTIVE ASSESSMENT OF TMJ SOUNDS
life situation where both cycle times and tooth contact
sounds are variable. Therefore, it might be expected
that listeners would perform better in this type of test
than they would in the clinical situation.
In the categorization of the sounds into the three
groups (clicks, creaks and crepitus) subjects performed
well, correctly classifying 72% of the examples
presented to them, In a similar test using a visual
classification of the sounds (Prinz & Ng, 1997) 82% of
a set of 42 TMJ sounds were correctly classified.
Viewers performed better than listeners at classifying
sounds into the three groups, even though for the visual
assessment all of the sounds were different whereas for
the auditory assessment, only a small subset of sounds
were used throughout the tests. The ability of observers
to correctly identify the three groups of sounds suggests
that there are three (clicks, creaks and crepitus) rather
than just two (clicks and crepitus), natural groupings
of TMJ sounds. Listeners were also surprisingly
consistent at classifying the sounds as high, medium or
low amplitude, even though no training was given as
to this categorization.
Although listeners performed well at distinguishing
the type of sound and its amplitude, they were poor at
distinguishing a sound’s position (early, middle or late)
in the envelope of movement. In this test only 49% of
classifications were correct and of these 33% could be
attributed to chance.
The difficulty experienced in classifying the position
of sounds (early, middle or late) has important
implications. The main diagnostic criteria used to
differentiate defects of form from reciprocal clicking
(disc displacement with reduction) is that the defect of
form occurs in the same place on opening and closing
and the reciprocal click does not. These differences are
small and the results suggest that it is doubtful that an
observer could distinguish the two types of sounds on
this basis alone. However, both opening and closing
clicks due to defects of form are of similar frequency
whereas in reciprocal clicking the opening click is of
higher frequency than the closing click and this
difference in frequency can be detected by the human
ear.
The results presented here demonstrate the difficulties
encountered by clinicians attempting to classify sounds
using a stethoscope alone and also serve to underline
© 1998 Blackwell Science Ltd, Journal of Oral Rehabilitation 25; 765–769
769
the advantages of objective instrumental recording of
TMJ sounds and automated classification by computer.
References
BLAUERT, J. (1982) Binaural localisation. In: Binaural Effects in
Normal and Impaired Hearing (eds O.J. Pederson & T. Poulsen),
Scandinavian Audiology. Supplement. 15.
CAMERON, O. (1975) Medical Physics. Springer Verlag, Berlin.
ERIKSON, L., WESTESSON, P.L. & SJOBERG, H. (1987) Observer
performance in describing temporomandibular joint sounds.
Cranio, 5, 33.
FRANKE, E. & VON GIERK, H. (1952) The motions relative to the
skull and their influence on hearing by bone conduction. Journal
of Acoustic Society of America, 24, 142.
HARDISON, J.D. & OKESON, J.P. (1990) Comparison of three clinical
techniques for evaluating joint sounds. Cranio, 8, 307.
HASHIMOTO, K., TAKEHANA, S., ITO, Y., TAKENAKA, M., SUZUKI, N. &
ABE, T. (1990) A study of how to distinguish the side on which
TMJ noise is occurring. Aichi Gakuin Daigaku Shigakkai Shi,
28, 1151.
HASHIMOTO, K., TAKEHANA, S., TAKENAKA, M., YAMAMOTO, T., MITUYA,
H. & SUZUKI, N. (1989) A study of how to distinguish the
occurring side of TMJ noise multiple click. Nihon Ago Kansetsu
Gakkai Zasshi, 1, 11.
O’MAHONEY, M. (1988) Sensory differences and preference testing.
In: Applied Sensory Analysis of Foods (ed. H.R. Moskovitz) p. 153.
POPE, J.A. (1984) Medical Physics. Heinemann Educational
Publications, London.
POTTS, R.O., CHRISMAN, D.A. & BURAS, E.M. (1983) The dynamical
properties of human skin in vivo. Journal of Biomechanics, 16, 365.
PRINZ, J.F. (1985) An instrument for the assessment of dental
occlusion. UK patent No 8509307.
PRINZ, J.F. (1991) Computer Aided Gnathosonic Analysis
M.Med.Sci Thesis. University of Sheffield.
PRINZ, J.F. (1996) Physical Mechanisms in the Pathogenesis of
Temporomandibular Joint Sounds. PhD Thesis. The University
of Hong Kong.
PRINZ, J.F. (1998) Correlation of the characteristics of
temporomandibular joint and tooth contact sounds. Journal of
Oral Rehabilitation, 25, 194.
PRINZ, J.F. & NG, K.W. (1997) Classification of TMJ Sounds. Archives
of Oral Biology, 41, 631.
RUKKER, R. (1986) The Fourth Dimension and How to Get There.
Penguin Books, London.
WATT, D.M. (1981) Gnathosonics and Occlusal Dynamics. Praeger,
New York.
WESTLING, L., HELKIMO, E. & MATTIASSON, A. (1992) Observer
variation in functional examination of the temporomandibular
joint. Cranio, 6, 202.
Correspondence: Dr J.F. Prinz, Department of Prosthetic Dentistry,
London Hospital, Dental Institute, Turner St, London E1 2AD,
U.K. E-mail: jfprinz@mds.qmw.ac.uk