Monday, July 5, 2010

Noise Hazard Evaluation, U.S. Army 1975, Sound Level Data of Noise Sources

For the Army's Noise Level of common items[Helicopters, tanks, etc.] look at this 1975 U.S. Army guide for "Noise Hazard Evaluation":

Found at TG 040 (3,105 KB)

* Noise Hazard Evaluation - Sound Level Data of Noise Sources (USACHPPM POC: Army Hearing Program (51) at 410-436-3797)
* Published – January 1975

Noise Exposure Levels of M16 and M60 Machinge gun fire

Given that acoustic trauma is a given of military life and that this is the single most awarded VA disability[hearing loss /tinnitus]we felt that this article and its references to dB levels are important because they show M16 and M60 to produce dB levels above 120dB, where permanent damage occurs. While the VA may concede acoustic trauma, the veterans needs to make sure that the VA concedes permanent damage as well, especially given that hearing loss is what is called a slow moving disease, one slowly progresses [get worse] with time after the initial exposure.
==========================================================


Between test sessions, the noise-exposed group received
impulse-noise exposures (M-16 rifle, M-60 machine gun, and
C-4 explosives) as prescribed by standard operating
procedures.5 A 5%–10% incidence of permanent hearing loss
was expected, despite mandatory hearing protection.

===========

All recruits spent six days on outdoor rifle ranges, where each individual fired 340 rounds with an M-16 rifle (157 dB pSPL, US Army Center for Health Pro-
motion and Preventive Medicine, 2008). Next, they made three 20–30 min runs through a combat obstacle course where they were exposed to M-60 machine-gun fire
(155 dB pSPL, US Army Center for Health Promotion and Preventive Medicine, 2008) and simulated artillery using C-4 explosives. Noise measurements performed with the Quest M-27 noise logging dosimeter revealed both the simulated artillery and C-4 explosions produced levels in excess of 146 dB pSPL (maximum limits of the M-27). Noise sources were located 5–20 ft from volunteers’ ears depending on where individuals were located on the course at the time of an impulse-noise presentation.

===============


Detecting incipient inner-ear damage from impulse noise with
otoacoustic emissions
Lynne Marshall,a) Judi A. Lapsley Miller, and Laurie M. Hellerb)
Naval Submarine Medical Research Laboratory, Groton, Connecticut 06349-5900
Keith S. Wolgemuthc)
Naval Medical Center San Diego, San Diego, California 92134-5000
Linda M. Hughes
Naval Submarine Medical Research Laboratory, Groton, Connecticut 06349-5900
Shelley D. Smith
University of Nebraska Medical Center, Omaha, Nebraska 68198
Richard D. Kopked)
DOD Spatial Orientation Center, Naval Medical Center San Diego, San Diego, California 92134-5000
(Received 13 June 2008; revised 20 November 2008; accepted 21 November 2008)
Audiometric thresholds and otoacoustic emissions (OAEs) were measured in 285 U.S. Marine
Corps recruits before and three weeks after exposure to impulse-noise sources from weapons’ fire
and simulated artillery, and in 32 non-noise-exposed controls. At pre-test, audiometric thresholds for
all ears were
25 dB HL from 0.5 to 3 kHz and
30 dB HL at 4 kHz. Ears with low-level or
absent OAEs at pre-test were more likely to be classified with significant threshold shifts (STSs) at
post-test. A subgroup of 60 noise-exposed volunteers with complete data sets for both ears showed
significant decreases in OAE amplitude but no change in audiometric thresholds. STSs and
significant emission shifts (SESs) between 2 and 4 kHz in individual ears were identified using
criteria based on the standard error of measurement from the control group. There was essentially
no association between the occurrence of STS and SES. There were more SESs than STSs, and the
group of SES ears had more STS ears than the group of no-SES ears. The increased sensitivity of
OAEs in comparison to audiometric thresholds was shown in all analyses, and low-level OAEs
indicate an increased risk of future hearing loss by as much as ninefold.
Ä™ 2009 Acoustical Society of America. [DOI: 10.1121/1.3050304]
PACS number(s): 43.64.Jb, 43.64.Wn [BLM]
Pages: 995–1013
I. INTRODUCTION
Otoacoustic emissions (OAEs) are more sensitive than
pure-tone audiometric thresholds in detecting the early stages
of permanent noise-induced inner-ear damage in humans.
Typical results for noise-exposed groups followed longitudi-
nally show a decrease in OAE amplitudes, but no change in
audiometric thresholds (Engdahl et al., 1996; Murray et al.,
1998; Murray and LePage, 2002; Konopka et al., 2005;
Seixas et al., 2005a, 2005b; Lapsley Miller et al., 2006).1
Most longitudinal studies do not last long enough to also see
hearing loss in the noise-exposed group. A recent finding is
that low-level or absent OAEs in noise-exposed individual
ears may be a risk factor or predictor for hearing loss in their
near future for continuous noise overlaid with impact noise
(Lapsley Miller et al., 2006). It is of interest to know for both
theoretical and clinical reasons whether this finding general-
izes to impulse noise.
Impulse noise is a common occupational and recreational hazard (Clark, 1991; Humes et al., 2005). The waveform of the impulse as received at the ear is shaped by the
individual pinna, ear canal, and middle ear, which may have an activated middle-ear reflex, sometimes even prior to the noise exposure in the case of an anticipatory reflex (e.g., Marshall et al., 1975). To add to the complexity, higher-level
sounds may result in less stapes motion and thus less damage than lower-level sounds (e.g., Price, 2007). In a work setting with considerable impulse-noise exposure (both selfgenerated and from other sources in the environment), as is
the case for some military jobs, the impulse-noise exposure for any one individual can be difficult to quantify (unless one has the luxury of a microphone in the ear canal). We expect
more variability in the noise exposure to the cochlea across
these individuals than for individuals working in steady-state
background noise, such as an engine room, where the noise
exposure is more homogenous.
If the noise exposure reaching the cochlea is more vari-
able across individuals and if a single exposure can cause
inner-ear damage, we expect that the state of the inner ear
a)Author to whom correspondence should be addressed. Electronic mail:
lynne.marshall@med.navy.mil
b)Present address: Department of Cognitive and Linguistic Sciences, Brown
University, Providence, RI 02912.
c)Present address: Department of Communicative Disorders, University of
Redlands, Redlands, CA 92373.
d)Present address: Hough Ear Institute, Oklahoma City, OK 73112.
J. Acoust. Soc. Am. 125 (2), February 2009
Ä™ 2009 Acoustical Society of America
995
0001-4966/2009/125(2)/995/19/$25.00
Page 2
prior to the noise exposure will not be as predictive of in-
cipient risk for this group of people as for a group of people
with more uniform noise exposure.
Noise exposure from live-fire training can produce hear-
ing loss very quickly. Permanent threshold shifts (PTSs)
have been reported in 10% or more of military personnel
during weapons’ training, including 300 rounds of M-16
live-fire training,2 Army special forces undergoing routine
weapons’ training,3 and Israeli army recruits firing an aver-
age of 420 M-16 rounds (Attias et al., 1994), all in spite of
wearing hearing protection. In the current study, Marine re-
cruits undergoing basic training were chosen because some
PTS within a short amount of time was expected, and the
overall amount of noise exposure for each volunteer would
be very similar.
The current study is complementary to the study re-
ported in Lapsley Miller et al. (2006). The same experimen-
tal protocol was used, with each volunteer receiving both
pure-tone audiometry and OAE tests before and after a sig-
nificant multiday noise exposure in a military operational
setting. The primary difference between the two studies is
the type of noise exposure—with impulse-noise exposure
(from weapons’ fire) in the current study, in contrast to con-
tinuous noise overlaid with impact noise (from aircraft and
machinery noise) in the previous study. The current study is
also a subset of a larger interdisciplinary study investigating
the auditory and genetic determinants of susceptibility to
noise-induced hearing loss (NIHL).
II. METHOD
A. Participants
The study participants were 401 male volunteers who
had just begun mandatory basic training as U.S. Marine
Corps recruits.4 No female volunteers were available, as this
military installation provided training to male recruits only.
Two experimental groups were formed from the volunteers
who met the screening criteria and completed the study: the
noise-exposed group (N=285; age at enrollment: range
17.4–28.1 years, median=19.2 years); and the control group
(who were not exposed to noise between pre- and post-tests;
N=32; age at enrollment: range 18.2–27.1 years, median
=20.0 years).
Each group was tested twice with an identical protocol.
Test times were determined by the recruit training schedule.
The pre-test measurements occurred one to six weeks prior
to the noise exposures (all volunteers had been noise-free for
at least a day), and the post-test measurements occurred three
weeks after the noise exposures. At pre-test, volunteers were
screened for clear ear canals, audiometric thresholds of
25 dB HL from 0.5 to 3 kHz and
30 dB HL at 4 kHz,
and peak immitance within the range of
50 daPa atmo-
spheric pressure, with grossly normal amplitude, slope, and
smoothness of the tympanogram. Volunteers who met these
screening criteria proceeded to OAE testing. Volunteers who
did not meet the screening criteria did not enroll in the study.
At post-test, volunteers were checked for clear ear canals
(cerumen was removed if present) and peak immitance
within the range of
50 daPa atmospheric pressure, with
grossly normal amplitude, slope, and smoothness of the tym-
panogram.
Between test sessions, the noise-exposed group received
impulse-noise exposures (M-16 rifle, M-60 machine gun, and
C-4 explosives) as prescribed by standard operating
procedures.5 A 5%–10% incidence of permanent hearing loss
was expected, despite mandatory hearing protection.
The
volunteers in the control group were recruits who were in a
“medical hold” status for minor nonauditory injuries, and a
few U.S. Navy medical personnel. They were tested on two
occasions within a 24–48 h period without any weapons’
noise exposures before or between tests.
Data collection occurred from January to September
2000.
B. Noise exposures
During basic training, the noise-exposed group under-
went approximately three and a half weeks of training that
involved weapons’ noise exposures at Marine Corps Base,
Camp Pendleton, CA. All recruits spent six days on outdoor
rifle ranges, where each individual fired 340 rounds with an
M-16 rifle (157 dB pSPL, US Army Center for Health Pro-
motion and Preventive Medicine, 2008). Next, they made
three 20–30 min runs through a combat obstacle course
where they were exposed to M-60 machine-gun fire
(155 dB pSPL, US Army Center for Health Promotion and
Preventive Medicine, 2008) and simulated artillery using C-4
explosives. Noise measurements performed with the Quest
M-27 noise logging dosimeter revealed both the simulated
artillery and C-4 explosions produced levels in excess of
146 dB pSPL (maximum limits of the M-27). Noise sources
were located 5–20 ft from volunteers’ ears depending on
where individuals were located on the course at the time of
an impulse-noise presentation.
Finally, there were several
days of simulated-combat exercises where each recruit fired
an additional 50–75 M-16 rounds, as well as exposure to
more M-60 machine-gun fire and simulated artillery using
C-4 explosives. Because other recruits were simultaneously
firing M-16 rounds on the rifle range and during combat
exercises, each individual was exposed to more than the
390–415 rounds they fired from their own weapons. The
exact number of M-16 rifle-fire exposures for each volunteer
was not measured. The recruits fired from three positions:
lying down in a prone position, in a sitting position, and
standing up. The majority of the rounds fired were in the
prone position (80%). The noise exposures were very similar
for each volunteer for the rifle range and obstacle course, but
there was a lot more variability in exposures through the
simulated-combat exercises. The rest of the time, the re-
cruits’ activities were severely restricted, and they were not
exposed to any significant levels of nonmilitary noise.
The recruits were provided with foam, disposable
EÄšAÄšR Classic (Aearo Corporation) earplugs each day on
the rifle range, obstacle course, and during combat exercises.
These earplugs come in one size only and have a noise re-
duction rating of 29 dB (Berger, 2000), but only if properly
and deeply fitted. The drill instructors informed large groups
996
J. Acoust. Soc. Am., Vol. 125, No. 2, February 2009
Marshall et al.: Detecting inner-ear damage with otoacoustic emissions
Page 3
of recruits how to use the earplugs. Because individual fitting
was not done and only one size was provided, the actual field
attenuation no doubt was much less than optimal.
C. Audiometric equipment, stimuli, and testing
Audiometric testing was performed using either a Maico
MA-1000 PC audiometer or a Grason-Stadler G-117 portable
audiometer. Pure-tone stimuli were delivered through Tele-
phonics TDH-49 supra-aural earphones in MX41/AR cush-
ions. Audiometers were calibrated (ANSI, 1996), and daily
calibration and listening checks were performed each day of
testing (Navy Occupational Health and Safety Program,
1999). Audiometric testing was performed one volunteer at a
time in double-walled sound-attenuating chambers (ANSI,
1991). Earphone placement was checked by the examiner. At
the pre-test audiogram, the recruits had been without much
sleep for one to two nights. This necessitated the examiner
being in the same room as the volunteer, testing him simi-
larly to a pediatric patient (e.g., frequent animated verbal
interaction and encouragement) to maintain alertness.
Audiometric thresholds were measured in both ears (the
left ear was always tested first), using the standard U.S. Navy
hearing-conservation program test protocol, which is an as-
cending, modified Hughson–Westlake procedure, with a
5 dB step size and frequencies tested in the order of 1, 0.5, 1,
2, 3, 4, and 6 kHz (Navy Occupational Health and Safety
Program, 1999). All audiograms were collected manually by
qualified technicians or audiologists.
Note that we did not use the results of the group hearing
testing typically done for marine recruits as they enter basic
training, because it was not reliable enough for our purposes
(automated audiometry with up to eight recruits at a time in
the booth).
D. Tympanometry equipment, stimuli, and testing
Middle-ear pressures were estimated from the peak of an
immitance tympanogram with a 226 Hz tone using a Grason-
Stadler GSI 33 version 2 analyzer at a sweep speed of
12.5 daPa/s to minimize hysteresis.
E. Otoacoustic emission equipment, stimuli, and
testing
Two types of OAEs were measured: transient-evoked
otoacoustic emissions (TEOAEs) and distortion-product
otoacoustic emissions (DPOAEs). Both OAE types were
measured with the ILO292 Echoport system (Otodynamics
Ltd., England), using the DPOAE probe. To allow better
placement and manipulation in the ear canal, an acoustic-
immitance probe tip (which had been enlarged using a grind-
ing tool) was inserted onto the DPOAE probe. The size of
the probe tip was matched to the size of the ear canal, and
was noted so that the same size could be used for the pre-
and post-tests. Individual in-the-ear calibration was used for
both TEOAE and DPOAE measurements. OAE testing was
performed two volunteers at a time, with two testers also
present, in a double-walled sound-attenuating chamber
(ANSI, 1991).
An identical test battery was used as for Lapsley Miller
et al. (2006). Before OAE testing (for both pre- and post-
tests), an otoscopic examination was conducted with ceru-
men removal if necessary, and peak immitance was mea-
sured to ensure it was within 50 daPa atmospheric pressure
in both ears.
TEOAEs were evoked with a 74 dB pSPL click, pre-
sented in nonlinear mode, where responses to three clicks at
one polarity and one click with opposite polarity and 9.5 dB
higher were added together to reduce linear artifact from the
stimulus (Bray, 1989). At pre-test, every attempt was made to
get a flat stimulus spectrum during calibration by manipulat-
ing the depth and angle of the probe tip in the ear canal. At
post-test, every attempt was made to get the same stimulus
pattern during calibration as in the pre-test by referring to a
screenshot printed out after the first test. TEOAEs were col-
lected and averaged until 260 low-noise averages were ob-
tained. The results were windowed (2.5 ms onset delay,
20.5ms duration, with 2.56ms rise/fall) and filtered
(0.683–6.103 kHz bandpass filter), then analyzed into half-
octave bands (0.7, 1, 1.4, 2, 2.8, 4, and 5.6 kHz).
In order of presentation, DPOAEs were measured with
stimulus levels L1/L2=57/45, 59/50, 61/55, and 65/45 dB
SPL (abbreviated herein to DP57/45, DP59/50, DP61/55, and
DP65/45). For all stimulus levels, the f2/ f1 ratio was 1.22,
with f2=1.8, 2.0, 2.2, 2.5, 2.8, 3.2, 3.6, 4.0, and 4.5 kHz.
F. Data definitions, cleaning, and reduction
The short testing time available for each volunteer
meant that it was not always possible to obtain clean data. As
in the previous study, OAE data were affected by electrical
noise when running on line power (it was not possible to
always run with batteries). Data points and/or test conditions
contaminated with off-target stimulus levels, poor calibra-
tions, high noise-levels, large differences in noise level be-
tween tests, or many unexplained outliers were removed
from the data set in an objective fashion, using the same
elimination rules across the entire dataset of all volunteers
(see footnote 5, Lapsley Miller et al., 2006).
A TEOAE was considered present if its amplitude was
greater than the noise floor. A DPOAE was considered
present if its amplitude was greater than the noise floor,
which was redefined as the average noise floor from the three
frequency bins above and below the 2f1-f2 frequency bin
plus two standard deviations. For some ears, a pre-test OAE
was present, but the post-test OAE was absent. The noise-
floor level accompanying an absent post-test OAE was sub-
stituted for the absent OAE providing the noise-floor level
was lower than the pre-test OAE (see Lapsley Miller and
Marshall, 2001, pp. 6 and 7; Lapsley Miller et al., 2004, p.
311; and Lapsley Miller et al., 2006, Sec. II E). Thus some
OAE changes were potentially underestimated, but this was
considered preferable to not using the data at all. The substi-
tution was not done if the post-test noise-floor level was
higher than the pre-test OAE, because a high noise-floor
level could masquerade as an increase in OAE amplitude.
For the susceptibility analyses, it was of interest to know if
low or absent OAEs at pre-test increased the chance of STS
J. Acoust. Soc. Am., Vol. 125, No. 2, February 2009
Marshall et al.: Detecting inner-ear damage with otoacoustic emissions
997
Page 4
at post-test. Absent pre-test OAEs were estimated where pos-
sible by substituting the noise floor for the absent OAE, pro-
viding the noise floor was sufficiently low, defined here as
being in the tenth percentile of OAE amplitude (see Lapsley
Miller et al., 2006, Sec. II E).
To reduce the impact of unusable data, subsets of test
frequencies and levels were used in the analyses: TEOAEs at
1, 1.4, 2, 2.8, and 4 kHz, or just the frequencies 2, 2.8, and
4 kHz for some analyses; and DP65/45 and DP59/50 at 2.5, 2.8,
3.2, 3.6, and 4.0 kHz, or just the frequencies 2.8, 3.2, and
4.0 kHz for some analyses. The TEOAE frequency bands at
0.7 and 5.6 kHz were excluded due to low amplitude result-
ing from the windowing and filtering used to extract the
TEOAE. The DPOAE frequencies 1.8, 2.0, 2.2, and 4.5 kHz
were excluded (a) due to electrical noise artifacts that (usu-
ally) elevated DPOAE amplitudes and/or noise-floor levels at
2.2 and 4.5 kHz, and (b) for noise-floor levels that were on
average much higher than those at 2.5, 2.8, 3.2, 3.6, and
4.0 kHz. It was not possible to sensibly average across the
remaining DPOAE frequencies or to compute DPOAE
growth functions due to unusable data. As such, the DP lev-
els DP57/45 and DP61/55 were not examined further.
We believe that we were measuring permanent changes
in audiometric thresholds and OAEs because the post-tests
were performed long enough after the noise exposure (three
weeks) that any temporary threshold shifts (TTSs) or tempo-
rary emission shifts should have resolved. Nevertheless, be-
cause it was not possible to confirm the significant audiomet-
ric threshold and OAE shifts in individual ears with a
follow-up test at a later time, we are careful here to refer to
significant threshold shifts (STSs), rather than PTSs, and
likewise for OAEs we refer to significant emission shifts
(SESs), rather than permanent emissions shifts.
III. RESULTS
Table I provides an overview of the number of volun-
teers and ears in each group contributing to each analysis.
Depending on the analyses, the noise-exposed group is fur-
ther split into groups of ears with and without STSs (STS
TABLE I. The number of ears and the total number of volunteers in each group that contributed to each
analysis, listed by the section. The numbers varied at each test frequency, OAE level, and OAE type, because
only valid data were used. The exception was for the ANOVAs where volunteers were required to have
complete OAE data sets for both ears.
Section and
analysis
Group
No. of
ears (range)
Total
No. of
volunteers Notes
II.A. Volunteers
completing study
Noise
570
285
369 noise-exposed volunteers were
enrolled, but only 285 completed the
study. Not all noise-exposed volunteers
were available for post-testing.
All control volunteers completed
the study.
Control 64
32
III.A. ANOVA
Noise
120
60
Ear was a factor in the ANOVAs.
III.B. Forming STS
and SES criteria
Control STS: 64
32
Ears were pooled (Tables II and III).
SES: 36–54
32
III.C. Identifying
and describing STS
and no-STS ears
Noise
STS: 42
36
STSs were detected in 15 left ears
only; 15 right ears only; 6 bilateral.
No-STS: 528
279
III.D. Identifying
and describing SES
ears
Noise
SES: 42–49
32–43
The number of ears varied across the
left/right ear, frequency and OAE
type. There were many ears where
SES status could not be determined.
Also see Table VI.
No-SES: 256–398 85–151
III.E. Comparing
STS and SES
status
Noise
540
285
See Table V for the SES ears broken
down by OAE type and STS status.
III.F.
Susceptibility
Noise
STS: 17–21
21
Ear was a factor. The number of ears
contributing to the analysis also
varied across the OAE type, level, and
frequency as all valid data were used.
No-STS: 217–263 264
IV.G. Susceptibility
for worst ear
Noise
STS: 35–36
36
Ear with the lowest OAE amplitude
was used as a predictor.
No-STS: 234–249 249
IV.E. Comparison to
Lapsley Miller et al. (2006)
Noise
STS: 37–42
36
Ears were pooled.
No-STS: 439–523 279
998
J. Acoust. Soc. Am., Vol. 125, No. 2, February 2009
Marshall et al.: Detecting inner-ear damage with otoacoustic emissions
Page 5
and no-STS) and/or SESs (SES and no-SES). Volunteers may
have had unilateral or bilateral significant shifts.
A. Changes in group OAE and audiometric thresholds
after noise exposure
Separate repeated-measures analyses of variance (ANO-
VAs) were conducted on audiometric threshold, TEOAE, and
DPOAE data for the subgroup of 60 volunteers (median age
19 years) from the noise-exposed group with complete data
sets. A volunteer had a complete data set if, for both ears and
for both pre- and post-tests, there was a set of audiometric
thresholds (no missing data for any volunteer) and a set of
measurable (or estimated) OAEs (TEOAEs at 1, 1.4 2, 2.8,
and 4 kHz; and DP65/45 and DP59/50 at 2.5, 2.8, 3.2, 3.6, and
4.0 kHz; see Sec. II F). As described earlier, some absent
post-test OAEs were estimated using the noise floor. Data
from volunteers with incomplete data sets were not used for
this analysis. Incomplete data sets were attributable to mea-
surement errors, high noise, and/or absent OAEs at pre-test.6
By selecting volunteers with complete data sets, a bias may
have been introduced, because those volunteers with unus-
able data may have lower or absent OAEs from noise-
induced damage—the lower OAEs being harder to detect
from the noise floor. However, by using complete data sets,
comparisons across OAE stimulus types, frequencies, and
ears could be made more fairly.
A three-way repeated-measures ANOVA was conducted
for audiometric thresholds (test: pre and post; ear: left and
right; and frequency: 0.5, 1, 2, 3, 4, and 6 kHz). There was
no significant change in audiometric thresholds (main effect)
between pre- and post-tests (F1,59=0.03, ns). There were,
however, significant differences between ears (F1,59=4.4,p
0.05) and across frequency (F5,295=14.8,p 0.05). There
was also a two-way interaction for test by frequency (F5,295
=3.2,p 0.05). Bonferroni post hoc t-test comparisons were
used to establish whether any same-frequency pairs contrib-
uted to this interaction. The familywise significance level
was p 0.05, so, for six comparisons, p 0.008 was used.
None were significant.
A three-way repeated-measures ANOVA was conducted
for TEOAE amplitude (test: pre and post; ear: left and right;
and frequency: 1, 1.4, 2, 2.8, and 4 kHz). All three factors
showed significant main effects. Particularly, there was a
0.94 dB decrease in TEOAE amplitude between pre- and
post-testing (F1,59=14.4, p 0.05). Ears also differed (F1,59
=37.38,p 0.05) as did frequency (F4,236=14.45,p 0.05).
There was one significant two-way interaction: test by fre-
quency (F4,236=2.9,p 0.05). Bonferroni post hoc t-test
comparisons were used to establish whether any same-
frequency pairs contributed to this interaction. The family-
wise significance level was p 0.05, so, for five compari-
sons, p 0.01 was used. The TEOAE amplitudes at the
frequencies 1.4, 2, and 2.8 kHz contributed to the interaction
with significant decreases between pre- and post-testing of
1.1, 1.3, and 1.0 dB, respectively.
A four-way repeated-measures ANOVA was conducted
for DPOAE amplitude (test: pre and post; ear: left and right;
level: stimulus levels of 65/45 and 59/50 dB SPL; and fre-
quency: 2.5, 2.8, 3.2, 3.6, and 4.0 kHz). All four factors
showed significant main effects. Particularly, there was a
0.84 dB decrement in DPOAE amplitude between pre- and
post-testing (F1,59=8.6,p 0.05). There were also main ef-
fects for ear (F1,59=4.9,p 0.05), level (F1,59=140.3,p
0.05), and frequency (F4,236=68.6,p 0.05). There were
three significant two-way interactions: test by level (F1,59
=6.6,p 0.05), ear by level (F1,59=4.1,p 0.05), and level
by frequency (F4,236=10.7,p 0.05). Bonferroni post hoc
t-test comparisons were used to establish which levels con-
tributed to the test-by-level, two-way interaction. The fami-
lywise significance level was p 0.05, so, for two compari-
sons, p 0.025 was used. Neither was significant.
B. Significant threshold shift „STS… and significant
emission shift „SES… criteria
Criteria for the detection of STS and SES in individual
ears were developed using the same method as in Lapsley
Miller et al. (2006),7 which was based on the standard error
of measurement (SEmeas) derived from the control-group
data. Because the control group of 36 volunteers (64 ears)
had not been exposed to noise between tests, the SEmeas rep-
resents the amount of variability attributable to other sources
(i.e., fluctuations in the OAE level over time, differences in
probe position or movement, etc.). Any OAE or audiometric
threshold in a noise-exposed ear that exceeds these SES or
STS criteria can be interpreted as being due to noise expo-
sure (although there is always the possibility of a false posi-
tive).
Table II shows the STS criteria. STSs detected at 2, 3, or
4 kHz and the averaged shifts at 2 and 3 kHz, 3 and 4 kHz,
and 2, 3, and 4 kHz were used to define the group of STS
ears for subsequent analyses. Averaged shifts were included
as they are commonly used by regulatory agencies to detect
and define threshold shifts. As a crosscheck, no STSs were
detected in any ear in the control group. Shifts at 0.5 and
1 kHz were not considered because on their own they are not
diagnostic of NIHL, and we were mindful that each look
increases the false-positive STS rate. We wanted to focus on
those frequencies most likely to show NIHL. Shifts at 6 kHz,
however, also were not considered because the STS criterion
at 6 kHz was deemed less reliable for detecting noise-
induced STS, based on the distribution of negative STSs in
the noise-exposed group (there were more negative STS
cases than positive).
Table III shows the SES criteria. The criteria for the
TEOAEs ranged from approximately 4 to 6 dB, and tended
to be smaller than for the DPOAE criteria. The criteria for
DP59/50 ranged from approximately 6 to 10 dB, and for
DP65/45 ranged from approximately 7 to 8 dB. The criteria
tended to be smaller for the DP65/45 compared with DP59/50.
C. STSs detected in the noise-exposed group
A total of 36 out of 285 volunteers in the noise-exposed
group (12.6%) were classified with a STS in at least one ear
three weeks after the noise exposure (median age
19.1 years). When considering ears rather than volunteers,
42 out of 570 ears (7.4%) were classified with a STS.8 There
J. Acoust. Soc. Am., Vol. 125, No. 2, February 2009
Marshall et al.: Detecting inner-ear damage with otoacoustic emissions
999
Page 6
were 15 left STS ears, 15 right STS ears, and 6 bilateral STS
ears. Table IV summarizes the STS and no-STS ears by left
and right ears.
Figure 1 shows the average pre- and post-test audio-
grams for the STS ears (42 ears) compared with the no-STS
ears (528 ears). The STS ears’ average thresholds increased
while the no-STS ears’ average thresholds stayed the same.
Although there were the same number of left STS ears and
right STS ears, the left ear STSs on average were larger and
broader-band than the right ear STSs. The largest average
increases in threshold were 13.3 dB at 4 kHz for the right
ears and 11.7 dB at 4 kHz for the left ears. The largest indi-
vidual increases in threshold were 40 dB STSs at 4 kHz in
both ears of one volunteer.
D. SESs detected in the noise-exposed group
SES status was determined for each ear of each volun-
teer in the noise-exposed group, separately for each OAE
type. For comparison with the three frequencies used to as-
sess STS, three frequencies/frequency bands were considered
for each OAE type within the 2–4 kHz range. For TEOAEs,
this was 2, 2.8, and 4 kHz half-octave bands. For DPOAEs
this was 2.5, 3.2, and 4.0 kHz.9 Three types of SES status
were defined.
• No-SES. No OAE decrements at any of the three frequen-
cies.
• SES. At least one significant decrease in OAE amplitude at
any of the three frequencies. Other frequencies could have
either no shifts or unusable data.
• Unknown-SES. At least one frequency band where SES
status could not be determined10 and no SES shifts at the
other frequencies. In other words, it was unknown whether
the ear should be a no-SES or a SES. No distinction is
made here between OAEs below the noise-floor criterion
and data loss due to measurement problems.
Summarizing from Table IV, which provides a break-
down of SES status by ear and OAE type, 42 out of 285
noise-exposed group volunteers (14.7%) showed a DP59/50
SES, 32 volunteers (11.2%) showed a DP65/45 SES, and 43
volunteers (15.1%) showed a TEOAE SES. This included
two, ten, and six volunteers with bilateral SES, for each OAE
type, respectively. When considering ears rather than volun-
teers, 44 out of 570 noise-exposed ears (7.7%) showed a
DP59/50 SES, 42 ears (7.4%) showed a DP65/45 SES, and 49
ears (8.6%) showed a TEOAE SES. These percentages are
similar to those seen for STS, but are underestimated due to
the large number of ears with unknown-SES status. The true
SES rate is likely to be much higher.11
To estimate the true SES rate (for individual frequen-
cies), the percentages were recalculated (separately for each
OAE type) for the group of volunteers where status was
known for both ears (see Table IV). For these subgroups, 28
out of 158 volunteers (17.7%) showed a DP59/50 SES, 29 out
of 180 volunteers (16.1%) showed a DP65/45 SES, and 28 out
of 113 volunteers (24.8%) showed a TEOAE SES. When
considering ears rather than volunteers, 30 out of 316 ears
(9.5%) showed a DP59/50 SES, 39 out of 360 ears (10.8%)
showed a DP65/45 SES, and 34 out of 226 ears (15.0%)
showed a TEOAE SES. There was a tendency for more left
TABLE II. STS criteria based on the standard error of measurement (SEmeas) from the control group (32
volunteers/64 ears) for individual audiometric-threshold frequencies and for averaged frequencies. Shown is the
frequency, mean shift between post-testing and pre-testing, SEmeas, and the resulting STS criteria (see footnote
7). Note that although the STS criteria were calculated for all frequencies, only frequencies from 2 to 4 kHz
and the averaged frequency bands were used to determine the STS status.
Frequency (kHz)
Average shift (dB)
SEmeas (dB)
STS (dB)
0.5
−1.4
4.0
20
1
−1.8
3.6
15
2
−2.0
3.4
15
3
−2.3
3.6
15
4
−2.0
3.9
15
6
−1.5
4.5
20
Mean 2 and 3
−2.2
3.1
10
Mean 3 and 4
−2.2
3.1
10
Mean 2, 3, and 4
−2.1
2.7
8.3
TABLE III. SES criteria based on the standard error of measurement
(SEmeas) from the control group (32 volunteers/64 ears). Shown are the OAE
type, f2 frequency for DPOAEs or half-octave frequency band for TEOAEs,
the number of ears contributing to the calculation, SEmeas, and the resulting
SES criteria (see footnote 7). Note that although the SES criteria were cal-
culated for all valid frequencies, only some frequencies were used to deter-
mine the SES status (2.5, 3.2, and 4 kHz for DPOAEs, and 2, 2.8, and
4 kHz for TEOAEs).
OAE type
Frequency (kHz)
Ears
SEmeas (dB)
SES (dB)
DP59/50
2.5
40
2.0
6.0
2.8
40
2.5
7.6
3.2
48
3.3
9.9
3.6
45
2.8
8.5
4.0
51
2.9
8.7
DP65/45
2.5
44
2.5
7.5
2.8
48
2.5
7.5
3.2
54
2.6
7.7
3.6
54
2.3
7.0
4.0
54
2.2
6.7
TEOAE
1.0
45
2.0
5.8
1.4
48
2.0
6.0
2.0
43
2.0
6.1
2.8
42
1.7
5.2
4.0
36
1.3
4.0
1000
J. Acoust. Soc. Am., Vol. 125, No. 2, February 2009
Marshall et al.: Detecting inner-ear damage with otoacoustic emissions
Page 7
ears to show DP SESs (61% left ears and 39% right ears)
and for more right ears to show TEOAE SESs (38% left ears
and 62% right ears).
Figures 2–4 show the average pre- and post-test OAE
amplitudes for the SES ears compared with the no-SES ears,
by left and right ears for each OAE type, without regard to
STS status. Error bars are 95% confidence intervals. Note
that SES status was determined from only three frequencies
for each OAE type. To maximize the amount of data going
into each point, there was no requirement for an ear to have
TABLE IV. Breakdown of the 285 volunteers in the noise-exposed group by STS status and SES status for the
left/right ear and the measurement type. The first number is the count, and the number in parentheses is the
overall percentage. The Unknown category represents those ears for which a SES determination could not be
made, usually due to unusable data. See text for summaries of STS and SES rates for volunteers and ears.
Audiometric thresholds
Right ears
STS
No-STS
Unknown
Left ears
STS
6 (2)
15 (5)
0 (0)
No-STS
15 (5)
249 (87)
0 (0)
Unknown
0 (0)
0 (0)
0 (0)
DP59/50
Right ears
SES
No-SES
Unknown
Left ears
SES
2 (1)
10 (4)
6 (2)
No-SES
16 (6)
130 (46)
39 (14)
Unknown
8 (3)
32 (11)
42 (15)
DP65/45
Right ears
SES
No-SES
Unknown
Left ears
SES
10 (4)
5 (2)
1 (0)
No-SES
14 (5)
151 (53)
41 (14)
Unknown
2 (1)
36 (13)
25 (9)
TEOAE
Right ears
SES
No-SES
Unknown
Left ears
SES
6 (2)
15 (5)
7 (2)
No-SES
7 (2)
85 (30)
26 (9)
Unknown
8 (3)
38 (13)
93 (33)
-5
0
5
10
15
20
25
a) Non-shifting left ears
Average Hearing Thresholds
-5
0
5
10
15
20
25
0.5
1
2
3 4
6
Frequency (kHz)
Hearing
Threshold
(dB
HL
)
b) Non-shifting right ears
Pre-test
Post-test
c) STS left ears
0.5
1
2
3 4
6
d) STS right ears
Pre-test
Post-test
FIG. 1. Average pre-test and post-test audiometric thresholds for the noise-
exposed group by STS status. Average pre-test thresholds for the 42 STS
ears (21 left and 21 right ears) were essentially the same as for the 528
no-STS ears (286 left ears and 286 right ears). Post-test audiograms show
that the average thresholds for the STS ears increased up to 13.3 dB (left
ears, 4 kHz), while the no-STS ears stayed essentially the same. The error
bars are 95% confidence intervals.
-9
-6
-3
0
3
6
9
a) Non-shifting left ears
Average TEOAE Amplitudes
-9
-6
-3
0
3
6
9
1
1.4
2
2.8
4
Frequency (kHz)
TE
O
AE
Amplitude
(dB
S
PL
)
b) Non-shifting right ears
Pre-test
Post-test
c) SES left ears
1
1.4
2
2.8
4
d) SES right ears
Pre-test
Post-test
FIG. 2. Average pre-test and post-test TEOAE amplitudes for the noise-
exposed group by significant TEOAE shift (SES) status. Average pre-test
amplitudes for the SES ears (22–27 left ears and 14–21 right ears) were
slightly higher than for the no-SES ears (112–118 left ears and 129–138
right ears). Average post-test TEOAE amplitudes decreased by approxi-
mately 4 dB from pre-test for the SES ears, while the post-test average for
the no-SES ears stayed essentially the same. The error bars are 95% confi-
dence intervals. The number of ears contributing to the average at each
frequency varied because of some unusable data.
J. Acoust. Soc. Am., Vol. 125, No. 2, February 2009
Marshall et al.: Detecting inner-ear damage with otoacoustic emissions 1001
Page 8
usable data at all displayed frequencies; however, an ear
needed both a valid pre- and post-test measurement at a fre-
quency to be included. There were more unusable data for
the SES ears because these ears needed only one SES at one
frequency to be considered a SES ear, whereas the no-SES
ears were required to have no SESs at all three frequencies.
The SES ears showed on average an 4 dB broadband
TEOAE decrement and
6 dB DPOAE decrements across
2.5–4 kHz in average OAE amplitude between pre- and
post-testing, whereas the no-SES ears showed essentially no
change. There was a tendency, especially for TEOAEs, for
average pre-test amplitudes to be higher for the SES ears
compared with the no-SES ears, but this may be due solely
to the small N. In general, these graphs indicate that the
method used to determine the SES status was appropriate.
E. Comparison of STS and SES
Table V shows the resulting 23 matrices for the STS
and no-STS ear versus the SES, no-SES, and unknown-SES
ears. The amount of data is small in some cells, and is also
unevenly balanced, so it is important to not overinterpret the
findings. The count of the SES ears is likely to be an under-
estimate. First, many potential SES ears are in the unknown-
SES category because the SES is masked by noisy measure-
ments. Second, ears with low-level or absent OAEs at pre-
test cannot show a SES at post-test; these ears are examined
in more detail in Sec. III F.
The nonparametric phi coefficient (Siegel, 1956) was
used as a measure of association for the 22 matrices to
determine whether STSs and SESs tended to occur together
in the same ear (the unknown-SES category was not in-
cluded). The phi coefficient can be interpreted similarly to a
correlation coefficient and can be used for small data sets.
Coefficients below 0.35 are considered to indicate no more
than trivial associations (Fleiss et al., 2003). There was es-
sentially no association between the STS status and the SES
status for TEOAEs (left ears, phi=0.25; right ears, phi
=0.22), DP65/45 (left ears, phi=0.11; right ears, phi=0.10), or
DP59/50 (left ears, phi=0.18; right ears, phi=0.02).
To further assess whether STSs and SESs were associ-
ated, conditional probabilities were considered for the ears in
Table V. As shown in Table VI, in general, the probability
(P) of a SES in an ear was higher than the probability of a
STS. Further, P(STSSES), which is the conditional prob-
ability of the STS in the subgroup of the SES ears, was
higher than the P(STS), which is the STS base rate, indicat-
ing that STS ears were overrepresented among the SES ears.
Going the other way, P(SESSTS) which is the conditional
probability of the SES in the subgroup of the STS ears, was
higher than P(SES), which is the SES base rate, indicating
that SES ears were overrepresented among the STS ears. In
the STS ears, TEOAEs tended to show SES more than did
DPOAEs. Because of the small numbers in some of the cells,
any further analysis would be inappropriate.
F. OAE predictors of susceptibility to NIHL
Pre-test OAE amplitudes were used as predictors of the
STS status in the noise-exposed ears.12 There were two
groups of interest: the 42 STS ears and the 528 no-STS ears.
Due to potential differences in the NIHL susceptibility in the
left and right ears, they were kept separated in the analyses.
As described earlier, where possible, pre-test OAE ampli-
tudes below the noise floor were estimated with the noise-
floor level, providing the noise floor was sufficiently low.
Between 17 and 21 STS ears and 217–263 no-STS ears con-
tributed to the analysis for each frequency, OAE type, and
ear.
The positive predictive value (PPV) (Zhou et al., 2002)
is the conditional probability of an ear from the noise-
-9
-6
-3
0
3
6
9
a) Non-shifting left ears
Average DP59/50 Amplitudes
-9
-6
-3
0
3
6
9
2.5 3.2 4.0
Frequency (kHz)
DP
O
AE
Amplitude
(dB
S
PL
)
b) Non-shifting right ears
Pre-test
Post-test
c) SES left ears
2.5 3.2 4.0
d) SES right ears
Pre-test
Post-test
FIG. 3. Average pre-test and post-test DP59/50 amplitudes for the noise-
exposed group by significant DP59/50 shift (SES) status. Average pre-test
amplitudes for the SES ears (16–17 left ears and 25–26 right ears) were
slightly higher than the average for the no-SES ears (180–185 left ears and
166–172 right ears). Average post-test DP59/50 amplitudes decreased by ap-
proximately 6 dB from pre-test for the SES ears, while the post-test average
for the no-SES ears stayed essentially the same. The error bars are 95%
confidence intervals. The number of ears contributing to the average at each
frequency varied because of some unusable data.
-9
-6
-3
0
3
6
9
a) Non-shifting left ears
Average DP65/45 Amplitudes
-9
-6
-3
0
3
6
9
2.5 3.2 4.0
Frequency (kHz)
DP
O
AE
Amplitude
(dB
S
PL
)
b) Non-shifting right ears
Pre-test
Post-test
c) SES left ears
2.5 3.2 4.0
d) SES right ears
Pre-test
Post-test
FIG. 4. Average pre-test and post-test DP65/45 amplitudes for the noise-
exposed group by significant DP65/45 shift (SES) status. Average pre-test
amplitudes for the SES ears (15–16 left ears and 24–26 right ears) were
slightly higher than the average for the no-SES ears (201–206 left ears and
188–192 right ears). Average post-test DP65/45 amplitudes decreased by ap-
proximately 6 dB from pre-test for the SES ears, while the post-test average
for the no-SES ears stayed essentially the same. The error bars are 95%
confidence intervals. The number of ears contributing to the average at each
frequency varied because of some unusable data.
1002
J. Acoust. Soc. Am., Vol. 125, No. 2, February 2009
Marshall et al.: Detecting inner-ear damage with otoacoustic emissions
Page 9
exposed group with STS after basic training, given a test
result of a low-level OAE. A low-level OAE is defined as an
OAE amplitude that is less than a cutoff value. By varying
the cutoff value over the entire range of OAE amplitudes, the
entire PPV function may be generated. Once the entire PPV
function is known, an optimal cutoff point may be chosen to
define a “low-level” OAE (which may vary depending on the
purpose, and the outcomes associated with the diagnosis).
The PPV is also known as the a posteriori conditional prob-
ability: P(STSOAEcutoff).
Figure 5 shows the PPV as a function of OAE amplitude
for each ear, OAE type, and frequency. Without knowledge
of the OAE level in a given ear, there was a probability of
around 0.07–0.08 that the ear would be classified with STS.
With knowledge of the OAE level, the probability an ear
would be classified with STS rose to a maximum of 0.67,
indicating an eight- to ninefold increased risk for STS.
TEOAEs at 2.8 and 4 kHz tended to be better predictors for
the left ears, with OAE amplitudes below approximately
−5 dB SPL indicating an increased risk for STS. DPOAEs at
4 kHz tended to be better predictors for the right ears, with
OAE amplitudes below approximately −5 to −10 dB SPL
indicating an increased risk for STS. To summarize the risk,
Table VII provides the maximum increased risk (maximum
PPV/base-rate over the OAE amplitude) across all OAE
types by frequency.
To relate these figures to the percentage of volunteers at
increased risk, the PPV functions are replotted in Fig. 6 for
each ear and each OAE type at 4 kHz after transforming
OAE amplitudes into percentiles. For the left ear (but not the
right ear), TEOAEs at 4 kHz show an increased risk for STS
in the bottom quartile, whereas for the left ear (but not the
right ear), DPOAEs at 4 kHz show an increased risk for STS
in the bottom decile.
G. Susceptibility to NIHL for volunteers rather than
ears using “worst ear” as a predictor
In a clinical situation, the focus is more on the risk of
STS for an individual person, rather than individual ears.
One way to use the information from both ears is to take the
results from the worst ear (the ear with the lowest OAE
amplitude) and use that as the predictor for the STS risk.
Figure 7 shows the results of such an analysis for the two
best DP frequencies and the three best TEOAE frequencies.
For each noise-exposed volunteer and for each OAE type
TABLE V. STS vs SES matrices. The first number is the count, and the number in parentheses is the overall
percentage. For the left and right ears separately, ears were grouped by whether they were classified as STS
and/or SES ears. The unknown-SES category is for those ears where there were unusable data; these ears were
not used in the analysis of the matrices.
OAE type
STS status
SES status
Left ear
Right ear
No-SES
SES
Unknown-SES
No-SES
SES
Unknown-SES
DP59/50
No-STS
174 (61)
14 (5)
76 (27)
161 (56)
24 (8)
79 (28)
STS
11 (4)
4 (1)
6 (2)
11 (4)
2 (1)
8 (3)
DP65/45
No-STS
191 (67)
13 (5)
60 (21)
178 (62)
22 (8)
64 (22)
STS
15 (5)
3 (1)
3 (1)
14 (5)
4 (1)
3 (1)
TEOAE
No-STS
111 (39)
21 (7)
132 (46)
130 (46)
16 (6)
118 (41)
STS
7 (2)
7 (2)
7 (2)
8 (3)
5 (2)
8 (3)
TABLE VI. STS ears are over-represented in the group of SES ears, compared with the probability (P) of a STS in general. Likewise SES ears are
over-represented in the group of STS ears, compared with the probability of a SES in general. This finding holds over the left and right ears and for all three
OAE types, except for DP59/50 in the right ears, where representation was proportional, and where OAEs had the highest variability. The pooled category
represents the results pooled over the ear and the OAE type. The true SES rate is underestimated because there are likely to be some unidentified SES ears
in the large group of unknown-SES ears, where SES status could not be determined at all three frequencies. (For the underlying cell counts, see Table V.)
OAE
type
Ear
P(STS)
P(STSSES)
STS represented
in SES group
P(SES)
P(SESSTS)
SES represented
in STS group
P(unknown-SES)
DP59/50
Left
0.07
0.22
Over
0.09
0.27
Over
0.29
Right
0.07
0.08
Proportionally
0.13
0.15
Proportionally
0.31
DP65/45
Left
0.08
0.19
Over
0.07
0.17
Over
0.22
Right
0.08
0.15
Over
0.12
0.22
Over
0.24
TEOAE
Left
0.10
0.25
Over
0.19
0.50
Over
0.49
Right
0.08
0.24
Over
0.13
0.38
Over
0.44
Pooled
0.08
0.19
Over
0.12
0.27
Over
0.33
J. Acoust. Soc. Am., Vol. 125, No. 2, February 2009
Marshall et al.: Detecting inner-ear damage with otoacoustic emissions 1003
Page 10
and frequency, the lowest OAE amplitude of the two ears
was chosen, or if there were valid data for only one ear then
that ear constituted the worst ear. If there were no valid data
for either ear, the volunteer was not included in the analysis
at that test point. Note that unlike the analyses so far, the
contralateral ears of the 30 volunteers with unilateral STS
were included with the STS ears rather than with the no-STS
ears (in the cases where that ear had the lowest OAE ampli-
tude). When choosing the worst ear, TEOAEs, especially at
4 kHz, were the best predictor of incipient NIHL.
IV. DISCUSSION
A. OAEs are more sensitive than audiometric
thresholds to noise exposure
The repeated-measures ANOVA indicated that OAEs
were more sensitive to noise-induced changes to the inner
ear than were audiometric thresholds. Both DPOAEs and
TEOAEs showed significant decreases in OAE levels after
the noise exposure, but there was no change in audiometric
thresholds for the subgroup of 60 noise-exposed volunteers
with complete data sets. The 22 matrices for SES versus
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Left Ears
Positive Predictive Values
a) DP
59/50
PPV 2.6 kHz
PPV 2.8 kHz
PPV 3.2 kHz
PPV 3.6 kHz
PPV 4.0 kHz
Ave Prior Prob
Right Ears
b) DP
59/50
PPV 2.6 kHz
PPV 2.8 kHz
PPV 3.2 kHz
PPV 3.6 kHz
PPV 4.0 kHz
Ave Prior Prob
Probability
c) DP
65/45
PPV 2.6 kHz
PPV 2.8 kHz
PPV 3.2 kHz
PPV 3.6 kHz
PPV 4.0 kHz
Ave Prior Prob
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
d) DP
65/45
PPV 2.6 kHz
PPV 2.8 kHz
PPV 3.2 kHz
PPV 3.6 kHz
PPV 4.0 kHz
Ave Prior Prob
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
-15
-10
-5
0
5
10
15
OAE Amplitude (dB SPL)
e) TEOAEs
PPV 1.0 kHz
PPV 1.4 kHz
PPV 2.0 kHz
PPV 2.8 kHz
PPV 4.0 kHz
Ave Prior Prob
-15
-10
-5
0
5
10
15
f) TEOAEs
PPV 1.0 kHz
PPV 1.4 kHz
PPV 2.0 kHz
PPV 2.8 kHz
PPV 4.0 kHz
Ave Prior Prob
FIG. 5. (Color online) PPV, for the left and right ears separately, as a function of the PPV criterion, which is OAE amplitude (in dB SPL). PPV is the
probability that an ear was classified with a STS given an OAE amplitude less than the criterion. As the OAE amplitude decreased, PPV tended to increase
for the higher-frequency bands, but not for all OAE types and frequencies. [(a) and (b)] DP59/50,[(c) and (d)] DP65/45, and [(e) and (f)] TEOAEs. The thin solid
horizontal line represents the prior probability of a STS averaged over the displayed frequencies.
TABLE VII. Maximum increased risk for STS (PPV/base-rate) for each
OAE type and frequency, by ear. Each number represents how many times
more likely a STS is given a low pre-test OAE result relative to the base
rate.
OAE type
Frequency
(kHz)
Maximum increased risk
Left ears
Right ears
DP59/50
2.5
2.6
1.8
2.8
1.1
1.6
3.2
1.2
1.5
3.6
4.7
4.9
4.0
1.7
4.7
DP65/45
2.5
1.9
1.4
2.8
2.7
1.4
3.2
2.3
1.7
3.6
2.3
1.9
4.0
2.7
9.2
TEOAE
1.0
2.6
3.4
1.4
2.2
1.1
2.0
1.4
4.4
2.8
4.6
1.0
4.0
8.1
1.0
1004
J. Acoust. Soc. Am., Vol. 125, No. 2, February 2009
Marshall et al.: Detecting inner-ear damage with otoacoustic emissions
Page 11
STS status indicated a higher SES rate compared with the
STS rate in the noise-exposed ears, and there was a tendency
for the STS ears to also have SESs and for SES ears to also
have STSs. TEOAE SES status showed more consistency
with the STS status than with the DPOAEs, despite the larger
amount of unusable data with the TEOAEs. These findings
are consistent with Lapsley Miller et al. (2006), where the
same experimental protocol was used, and with many other
longitudinal studies that also showed changes in OAEs
without accompanying changes in audiometric thresholds
(Engdahl et al., 1996; Murray et al., 1998; Murray and
LePage, 2002; Konopka et al., 2005; Seixas et al., 2005a;
Duvdevany and Furst, 2006).
These results could be due to on-frequency inner-ear
damage in the 2–4kHz range that causes subclinical
changes insufficient to affect audiometric thresholds but to
which OAEs are sensitive. This is consistent with observa-
tions in animals that damage to outer hair cells (OHCs) can
be extensive with no concomitant change in audiometric
thresholds (Hamernik et al., 1989; Hamernik et al., 1996),
and consistent with the theory that there is OHC redundancy
(LePage et al., 1993), where it is thought that there are many
more OHCs than what is required for normal hearing. The
OHC loss therefore shows up in OAE measurements before
audiometric threshold measurements because OAE measure-
ments more directly measure OHC activity. Alternatively, the
results could be due to unmeasured higher-frequency inner-
ear damage (that may or may not affect high-frequency hear-
ing thresholds) that affects OAEs measured at lower frequen-
cies, but not audiometric thresholds at those lower
frequencies. This higher-frequency damage might influence
the transmission of a lower-frequency OAE out to the middle
ear (Lonsbury-Martin and Martin, 2007). Furthermore, with
some OAE stimulus configurations (containing high-
frequency energy), high-frequency damage could lessen the
distortion-component OAE from the high-frequency place
that creates a lower-frequency stimulus-frequency OAE
(SFOAE),13 which can interact with OAEs generated at that
lower frequency. It was not possible in the current study to
measure high-frequency hearing thresholds or high-
frequency OAEs, making it difficult to disentangle the two
theories; however, the results of others offer some clues.
For DPOAEs in humans, diminished OAE amplitudes
without an accompanying hearing loss in the same frequency
region have been associated with a hearing loss at higher
frequencies (Arnold et al., 1999; Dorn et al., 1999). Arnold
et al. (1999) used 50 subjects, the majority of whom were
males, with normal hearing
(20 dB
HL
from
0.25 to 8 kHz) and ages from 17 to 37 years. They reported
minimal noise exposure, but people, particularly males, liv-
ing in modern civilizations do tend to accumulate damage
from noise exposure. The multivariate analyses of Dorn et al.
(1999) included hundreds of subjects, age 1–96 years, and
hearing levels from −5 to 120 dB HL at 0.75–8 kHz. In con-
trast, Schmuziger et al. (2005) minimized the effects of pre-
vious noise exposure by using younger subjects (age
16–19 years), and fewer males (38%) in the group (all with
normal hearing, as well as reports of minimal previous noise
exposure). The high-frequency (8–16 kHz) thresholds for
this group were only minimally related to lower-frequency
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
PPVs for Worst Ear
a) DP59/50
PPV 3.6 kHz
PPV 4.0 kHz
Ave Prior Prob
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
P
ro
b
a
bility
b) DP65/45
PPV 3.6 kHz
PPV 4.0 kHz
Ave Prior Prob
0.0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
-15 -10
-5
0
5
10
15
20
OAE Amplitude (dB SPL)
c) TEOAEs
PPV 2.0 kHz
PPV 2.8 kHz
PPV 4.0 kHz
Ave Prior Prob
FIG. 7. (Color online) PPVs as a function of PPV criterion, which in this
case is OAE amplitude for the worst ear, which is the way it would be
implemented in occupational audiology programs. For each volunteer, the
ear with the lowest OAE amplitude was used as the predictor. (a) DP59/50 at
3.6 and 4 kHz, (b) DP65/45 at 3.6 and 4 kHz and (c) TEOAEs at 2, 2.8, and
4 kHz. The thin solid horizontal line represents the prior probability of a
STS averaged over the displayed frequencies.
0.0
0.2
0.4
0.6
P
ro
b
a
bility
Positive Predictive Value
at 4 kHz for OAE Percentiles
a) Left Ears
TEOAEs
DP59/50
DP65/45
0.0
0.2
0.4
0.6
0 10
25
50
75
100
OAE Amplitude (percentile)
b) Right Ears
TEOAEs
DP59/50
DP65/45
FIG. 6. (Color online) PPV at 4 kHz (from Fig. 5). for the left and right ears
separately, for each OAE type replotted as a function of OAE amplitude in
percentiles.
J. Acoust. Soc. Am., Vol. 125, No. 2, February 2009
Marshall et al.: Detecting inner-ear damage with otoacoustic emissions 1005
Page 12
DPOAEs. In rodents, Withnell and Lodde (2006) did not see
lower-frequency DPOAE amplitude losses when higher-
frequency regions were damaged by noise. These results sug-
gest that the more likely explanation for the influence of
high-frequency thresholds on much lower-frequency
DPOAEs is subclinical damage at the lower frequency.
For TEOAEs in humans, decreased OAE amplitudes
without an accompanying audiometric-threshold decrement
in the same-frequency region also were associated with a
hearing loss at higher frequencies (Avan et al., 1997;
Konopka et al., 2005). The subjects of Avan et al. (1997)
(nearly half of whom were males) had normal hearing up to
4 kHz, and were older—ages 24–54 years. The group from
Konopka et al. (2005) consisted of 92 young, male, noise-
exposed soldiers. In a young population with less noise ex-
posure, and with normal hearing up to 8kHz, high-
frequency (8–16 kHz) thresholds were not related to lower-
frequency TEOAEs (Schmuziger et al., 2005).
Yates and Withnell (1999), using a novel measurement
technique that allowed the measurement of high-frequency
TEOAES, observed that TEOAEs evoked by a high-pass
click included frequencies lower than those in the stimulus in
guinea pigs. The generation of new frequencies that were not
present in the stimulus implies that the OAEs were generated
from a distortion mechanism; these OAEs also act as a
stimulus that elicits reflection-component SFOAEs at lower
frequencies. After noise exposure, which damaged the high-
frequency region in guinea pigs, lowered TEOAE amplitudes
were found not only at the higher frequency where the
eighth-nerve compound-action-potential (CAP) thresholds
were lowered, but also at lower frequencies where the CAP
thresholds were not lowered (Withnell et al., 2000). In hu-
mans, however, not only is the distortion mechanism rela-
tively smaller than it is in guinea pigs (Shera and Guinan,
1999), but the method used for clinical TEOAE measure-
ments windows out the first few milliseconds of the TEOAE
to reduce stimulus artifact (Bray and Kemp, 1987). This
leaves only the TEOAE reflection component (Knight and
Kemp, 1999; Kalluri and Shera, 2007; Sisto et al., 2007;
Withnell et al., 2008). Furthermore, behavioral thresholds at
ultrahigh frequencies would not influence most TEOAE
measurements with humans because the TEOAE stimulus
usually does not have much energy above 5 kHz. With a
TEOAE stimulus that extends up to 5 kHz, the TEOAE and
SFOAE spectra in individual ears are nearly identical, at
least up to 2.4 kHz, implying that with TEOAEs, the lower-
frequency SFOAE that gets generated due to the higher-
frequency distortion component does not have much effect
on the lower-frequency SFOAE that is solely generated from
that place (e.g., Kalluri and Shera, 2007). The results from
these TEOAE studies also suggest that on-frequency sub-
clinical damage is the predominant reason why OAE ampli-
tudes can decrease when hearing levels remain unchanged in
humans.
Studies that have shown TEOAE decrements in indi-
vidual ears to be broader than DPOAE decrements may be
indicative of TEOAEs being more sensitive to subclinical
damage than DPOAEs (e.g., Lapsley Miller et al., 2004;
Lapsley Miller et al., 2006), consistent with Shera’s (2004)
view that the reflection component should be more sensitive
to noise damage. TEOAEs as we typically measure them are
primarily reflection mechanism, and DPOAEs are a mix of
the two mechanisms. Therefore, it is not surprising that
DPOAEs and TEOAEs often are equally sensitive for
groups, but TEOAEs tend to edge out DPOAEs in sensitivity
for individual ears.14
There are some studies that do not show OAEs as being
more sensitive than audiometric thresholds. Lapsley Miller
et al. (2006) found significant changes in group audiometric
thresholds along with changes in OAEs, but there was little
consistency between changes in thresholds and OAEs in in-
dividual ears. Duvdevany and Furst (2007) measured hearing
and TEOAEs in the same individuals annually three times—
neither hearing nor TEOAEs changed in the second year, but
both did in the third year. Their TEOAE stimulus was 84 dB
pSPL, which would not be maximally sensitive to noise
damage.15
Aging and sex differences can be discounted in the cur-
rent study because all the participants were young men and
the study duration of 13 weeks was too short for aging to
have any measurable impact. Nor is audiometric resolution
an explanation for the greater sensitivity of OAEs to noise-
induced changes in the inner ear. The standard clinical pro-
tocol, which produces a resolution of 5 dB, may hinder the
detection of small changes in audiometric thresholds, even in
the group average. However, if the only reason for the dif-
ference between OAEs and audiometric thresholds is reso-
lution, all the STS ears that were identified should also show
SESs, but this was not the case. Even within the subset of
ears with both SESs and STSs, there was not much consis-
tency across frequency and OAE type (results not reported
here in detail).
The typical finding for the STS ears was either an ac-
companying SES (not necessarily across all OAE types) or
low-level or absent OAEs. For a couple of ears where there
was STS but no SES and normal OAEs, the STS was small
and possibly a false positive (it was not possible to do a
confirmation audiogram).
B. STS and SES criteria
The SEmeas values underlying the STS criteria were
smaller here than in Lapsley Miller et al. (2006), perhaps
because the current study used double-walled sound-
attenuating chambers, whereas the earlier study used single-
walled chambers in a noisier shipboard environment. The
current STS criteria were identical to those developed in
Lapsley Miller et al. (2004), where the testing environment
was similar.
If possible, it is important to derive STS (and SES) cri-
teria from a control group tested in the same environment so
that there is some certainty that the shifts are significantly
different from test-retest variability. For instance, in the cur-
rent study, using all the derived STS criteria (Table II), STS
was detected in 36 noise-exposed volunteers (42 ears) or
12.6% of volunteers (7.4% of ears). If the strict clinical cri-
terion that is commonly used by regulatory agencies was
used instead, which is an average shift at 2, 3, and 4 kHz of
1006
J. Acoust. Soc. Am., Vol. 125, No. 2, February 2009
Marshall et al.: Detecting inner-ear damage with otoacoustic emissions
Page 13
at least 10 dB (Mining Safety and Health Administration,
1999; Department of Defense, 2004; Federal Railroad Ad-
ministration, 2006; Occupational Safety & Health Adminis-
tration, 2007), STS would have been detected in only 17
volunteers (18 ears) or 5.9% of volunteers (3.2% of ears).
The opposite can also happen, where the criteria chosen are
too lax. For instance, if someone arbitrarily chose a STS
criterion of a 10 dB shift at any single frequency (without
any reference to a control group), in our current study STS
would have been detected in 87 volunteers (109 ears) or 30%
of volunteers (19.1% of ears). Most of these shifts are false-
positive STSs because the criterion was less than the test-
retest variability.
The detection of a STS does not necessarily mean the
ear had a hearing loss as there is a chance that the STS was
a false positive. An upper limit for our false-positive rate (for
STS at three single frequencies and three averaged frequen-
cies; assuming independence of frequencies) is approxi-
mately 11% (for both positive and negative STSs). The prob-
ability of a false positive across six frequencies is one minus
the probability of no false positives at any frequency. Be-
cause our STS criteria were based on a 98% confidence in-
terval (see footnote 7), this is 1−0.986. The actual false-
positive rate is likely to be lower because (a) we looked only
at positive STS, so the false-positive rate would be only
5.5% at most, (b) the frequencies are correlated, and (c) the
5 dB resolution of the audiogram meant that we rounded up
the raw criteria based on multiples of the SEmeas to the next
largest available step so the probability of a false positive at
each frequency was lower. This is borne out when applying
the STS criteria to the control group where there were no
STSs detected. In the larger noise-exposed group, however,
we would expect some of the STSs to be false positives, and
indeed there are some shifts that are not compelling from a
clinical viewpoint (e.g., a shift only at 2 kHz). A stricter STS
criterion, however, would have meant missing more true
STSs. It was unfortunate that the recruits’ schedule did not
allow time for immediate retesting of STSs, which would
have decreased the false-positive rate.
NIOSH (National Institute for Occupational Safety and
Health, 1998) suggested a different STS criterion—a 15 dB
shift at any tested frequency (0.5, 1, 2, 3, 4, 6, or 8 kHz),
with an immediate retest being optional. ASHA (American
Speech-Language-Hearing Association, 1994) also suggested
a significant change criterion (for monitoring ototoxic hear-
ing loss) greater than 15 dB at any one frequency or greater
than or equal to 10 dB at two or more adjacent frequencies.
The current study, as well as previous ones (e.g., Marshall
and Hanna, 1989; Lapsley Miller et al., 2004) found that the
SEmeas at a single frequency varies as a function of fre-
quency, with lower and higher frequencies having a larger
SEmeas. Shaw (1966) demonstrated that supra-aural
earphone-placement variations have the largest effect at these
frequencies. For our data, the NIOSH criterion was the same
as ours at 1–4 kHz (National Institute for Occupational
Safety and Health, 1998), but too lax at 0.5 and 6 kHz. The
ASHA criterion was too strict for 1–4 kHz, but applicable
above and below that, as well as for the 10dB two-
frequency average criterion.
The NIOSH suggestion of retesting immediately follow-
ing a STS is a good one. For example, in the current study,
we estimated an upper bound for the false-positive STS rate
to be 5.5%. If a STS is retested, then we would expect the
STS rate to diminish to 0.5% (probability of a STS over six
frequencies multiplied by the probability of a STS at one
frequency, 0.0550.01; assuming only positive STSs are of
interest). Basing the STS criterion on the known test-retest
reliability of a test situation is necessary for control of false-
positive STSs. Our STSs are a better estimate of true STS
than may be the case when arbitrary values are chosen.
The SEmeas values underlying the TEOAE SES criteria
were similar to Lapsley Miller et al. (2004), but were slightly
larger than in Lapsley Miller et al. (2006), where the equip-
ment was run on battery power more often, which tended to
produce a lower noise floor. The SEmeas values underlying
the DPOAE SES criteria were also slightly larger than in
Lapsley Miller et al. (2006), but could not be directly com-
pared to Lapsley Miller et al. (2004) because here they were
based on measurements at individual frequencies, rather than
averaged within half-octave bands. In general, the SEmeas
values were mostly comparable to those reported elsewhere
(Franklin et al., 1992; Beattie and Bleech, 2000; Beattie,
2003; Seixas et al., 2005b; Wagner et al., 2008).
C. Susceptibility to NIHL from impulse noise
In the analyses discussed so far, all ears used had pre-
test OAE amplitudes that were measurable. Many of the ears
that ended up in the unknown-SES category had OAE signal-
to-noise ratios that did not meet the criteria for presence. The
second thrust of the analyses showed that low-level pre-test
OAE amplitudes were predictive of subsequent STS status
for some OAE types, frequencies, and left and right ears
(Fig. 5). The increased risk of a STS for those ears with low
pre-test OAE levels cannot be explained by pre-test
audiometric-threshold differences between the STS ears and
the no-STS ears as there was essentially no difference be-
tween pre-test audiometric thresholds for these groups, as
shown in Fig. 1. To further illustrate this point, the analysis
underlying Fig. 6 for TEOAEs at 4 kHz, for left ears, was
rerun after excluding all ears with audiometric thresholds
15 dB HL at 4 kHz (see Fig. 8). Seventeen no-STS ears
and zero STS ears were excluded using this new criterion.
0.0
0.2
0.4
0.6
0.8
1.0
0 10
25
50
75
100
Positive
Predictive
Value
Percentile
Left Ear: TEOAE 4 kHz
HLs ≤15 dB
all HLs
FIG. 8. (Color online) PPV as a function of OAE amplitude in percentiles
for TEOAEs at 4 kHz for the left ears from Fig. 6 (solid line) compared with
the same data after excluding ears with audiometric thresholds
15 dB HL
(dashed line).
J. Acoust. Soc. Am., Vol. 125, No. 2, February 2009
Marshall et al.: Detecting inner-ear damage with otoacoustic emissions 1007
Page 14
The exclusion of these ears only enhances the finding that
pre-test OAE levels can be predictive of subsequent STS.
This is supported by an animal study by Perez et al. (2004)
that showed that ears with existing hearing loss were less
likely to get further hearing loss. This implies that ears with
low-level OAEs and hearing loss would be less likely to get
further hearing loss. It is those ears with low-level OAEs and
normal hearing that are at risk.
The OHC redundancy theory described earlier (LePage
et al., 1993) is consistent with these findings. Earlier sub-
clinical damage to some of the OHCs would show up as
low-level or absent OAEs, but not necessarily as hearing
loss. Further noise exposures damaging further OHCs would
then be more likely to lead to hearing loss compared to the
ears with more intact OHCs.
Animal studies also provide some clues as to why low-
level OAEs could be associated with an increase in the like-
lihood of future PTS. Chinchilla studies have shown that
small amounts of OHC loss have a more significant effect
(reduction) on DPOAE amplitude levels than on measures of
threshold sensitivity, suggesting that OAEs may also indicate
an early onset of cochlear damage in humans (Davis et al.,
2005). These results also suggest that OAEs be considered,
within the context of hearing-conservation practices, as a
complement to existing hearing-threshold tests in detecting
OHC loss resulting from noise exposure. The results indicate
that on the basis of threshold information alone, without in-
formation about the OAEs, one might underestimate the sen-
sory cell loss. This conclusion is supported by the results of
others, which show up to 30% OHC loss in subjects with less
than 10 dB of PTS (Hamernik et al., 1989; Hamernik and
Qiu, 2000; Davis et al., 2004). Bohne et al. (1987) also
showed that 20%–30% OHC loss in the low frequencies was
often not accompanied by corresponding behaviorally mea-
sured threshold shifts in the chinchilla. They also explained
that a relatively large reduction (12–15 dB) in DPOAEs in
the presence of smaller OHC losses at some frequencies may
be accounted for not only by the OHC loss but also by mor-
phological changes (e.g., cilia defects or intracellular
changes) that can affect the function of cells that are present
and for which the cochleogram provides no information.
There were differences between the ears, with no one
frequency consistently being a good predictor of the STS
status across ears. TEOAEs appeared to be a better predictor
of the STS status for the left ear, and DPOAEs appeared to
be a better predictor of the STS status for the right ear. Ex-
cept for TEOAEs at 4 kHz in the left ear, STS risk did not
increase greatly until the OAE amplitude moved into the
bottom decile.
It is unclear why DPOAEs would be a better predictor of
a STS in the right ears and TEOAEs a better predictor of a
STS in the left ears. The most likely explanation is that the
small number of STS ears contributing to each analysis gives
some spurious results. There are other possible contributing
factors. The recruits did not get in-depth training on proper
insertion of hearing-protection devices; indeed, they some-
times reported that the earplug fell out during the live-fire
exercise. Duvdevany and Furst (2007) showed large in-
creases in PTS during a time period when hearing-protection
devices apparently were not worn much. The variability in
noise exposure across volunteers could be large compared to
the number of Marine recruits that were in this study.
Although we did not keep track of handedness, we
would expect approximately 95% of the group to be right
handed.16 Two studies indicated that the left ear is more sus-
ceptible to NIHL than is the right ear, irrespective of hand-
edness (Job et al., 1998; Nageris et al., 2007). In the current
study, there was an equal number of STSs in the left ear and
the right ear (21 ears for each, including 6 ears with bilateral
STS). This indicates that there was considerable noise in the
environment (more than just from firing one’s own gun), as
well as the poor quality of the hearing protection.
Figure 1 indicates that on average the left ear STSs were
broader than the right ear STSs, suggesting that the left ear
suffered more extensive damage from the impulse-noise ex-
posures than the right ear. The implication is that low-level
TEOAEs could be a better predictor of broadband STS and
that low-level DPOAEs might be a better predictor of nar-
rowband STS. If this is true, it may be due to higher-order
physiological asymmetries (e.g., efferent innervations) that
somehow treat tonal stimuli differently from click stimuli
depending on which ear the stimuli are presented to (e.g.,
Sininger and Cone-Wesson, 2004), but any mechanism is
speculative at best. Furthermore, the DPOAE results may be
greatly affected by the measurements and data analysis being
seen at single widely-spaced frequencies, making it difficult
to determine if a low-level DPOAE is just due to the test
frequency coinciding with a null in the DPOAE microstruc-
ture in what is otherwise a strong DPOAEgram.
Even though we suspect that the apparent differences are
due to the relatively small amount of data, the results are
suggestive enough that future studies should continue to look
at ear differences. If the ear differences seen in this study are
also found in future studies, it will be important to parse out
whether the differences are due to external factors (e.g.,
noise exposures), innate factors (e.g., efferent innervation),
existing preclinical damage, or methodological idiosyncra-
sies.
D. Susceptibility to NIHL from impulse noise
compared with continuous noise
It is of interest to compare the results of the susceptibil-
ity analysis to that in Lapsley Miller et al. (2006), as it is the
only other known analysis that considers if OAE amplitude
is a predictor of NIHL. Likelihood ratios were used to make
this comparison, rather than PPVs, because the PPV is de-
pendent on the prior probability of STS/PTS, which differed
across the two studies. The likelihood ratio is a ratio of two
probabilities: the probability of a particular test result among
patients with a condition to the probability of that particular
test result among patients without the condition (Zhou et al.,
2002). In the current context, the likelihood ratio indicates
the relative probability that a pre-test OAE amplitude was
below a given percentile in the group of ears that subse-
quently were classified with STS, relative to the same result
in the group of ears that did not.
To ensure a fair comparison with Lapsley Miller et al.
(2006), ears for the current study were combined, but only
1008
J. Acoust. Soc. Am., Vol. 125, No. 2, February 2009
Marshall et al.: Detecting inner-ear damage with otoacoustic emissions
Page 15
for those test frequencies which showed little to no differ-
ence in OAE amplitude percentiles between ears.17 These
included TEOAEs at 4 kHz, DP65/45 at 4 kHz, and DP59/50 at
3.2, 3.6, and 4 kHz. Figure 9 shows likelihood ratio as a
function of OAE amplitude. For the current study, likelihood
ratios decreased when ears were combined, due to the asym-
metry of results between ears. Compared with Lapsley Miller
et al. (2006), likelihood ratios for the DPOAEs were compa-
rable at 4 kHz. At lower frequencies, the likelihood ratios in
the earlier study were higher than for the current study. The
biggest difference was for TEOAEs at 4 kHz, with the like-
lihood ratios in the earlier study showing increased risk for
ears with OAE amplitudes below the 50th percentile (ap-
proximately 0 dB SPL), whereas for the current study, risk
did not increase until about the 5th percentile (approximately
−8.5 dB SPL). The maximum likelihood ratio, however, was
essentially the same at around 8.
Overall, the general trend across both studies is for low-
level OAEs to be predictive of subsequent PTS and/or STS.
In both studies, OAEs—particularly TEOAEs—in the 4 kHz
region were the best predictors. Larger studies with many
more PTS/STS ears are essential to better establish this rela-
tionship. We expect this predictive power to be greater in
situations with continuous noise than in situations with im-
pulse noise due to greater variability of the sound power of
the sound source reaching the inner ear especially in envi-
ronments where much gunfire is in the general environment.
This supposition is supported by the comparison shown in
Fig. 9 where NIHL risk for impulse noise showed up for
much lower TEOAE amplitudes, compared with continuous
noise. Further, if hearing protectors are not worn (or if they
fit poorly) for impulse-noise exposures, we expect the pre-
dictive power to diminish even more because there may be
sufficient damage over a relatively short amount of exposure
time to cause hearing loss irrespective of whether or not
there were previously missing OHCs.
When considering both the current and the earlier study,
in general, TEOAEs were better predictors than DPOAEs
(however, the ear asymmetry shown in the current study in-
dicates that DPOAEs cannot be discounted). There are also a
number of mostly practical pros and cons for choosing one
OAE type over another. If only a few DPOAE frequencies
are tested (which makes the test faster—an important consid-
eration in field studies with humans), it is possible that the
0
2
4
6
8
a) TEOAEs at 4 kHz
Comparison with Lapsley Miller et al. (2006)
Impulse Noise Exposure
Continuous+Impact Noise Exposure
Like
lih
oo
dR
atio
b) DP59/50 at 3.2 kHz
0
2
4
6
8
c) DP59/50 at 3.6 kHz
0
2
4
6
8
-15
-10
-5
0
5
10
15
20
OAE Amplitude (dB SPL)
d) DP59/50 at 4 kHz
-15
-10
-5
0
5
10
15
20
e) DP65/45 at 4 kHz
FIG. 9. (Color online) Comparisons of likelihood ratio as a function of OAE amplitude (in dB SPL), indicating susceptibility to noise-induced hearing loss,
between the current study (Marine recruits exposed to impulse noise, solid line) and Lapsley Miller et al. (2006) (deployed aircraft carrier sailors exposed to
continuous noise overlaid with impact noise, dashed line) for the OAE test frequencies where there were no large differences in the amplitude distributions
between the ears: (a) TEOAEs at 4 kHz (half-octave band), (b) DP59/50 at 3.2 kHz, (c) DP59/50 at 3.6 kHz, (d) DP59/50 at 4 kHz, and (e) DP65/45 at 4 kHz.
J. Acoust. Soc. Am., Vol. 125, No. 2, February 2009
Marshall et al.: Detecting inner-ear damage with otoacoustic emissions 1009
Page 16
test point will fall into a null of the DPOAE microstructure.
To obtain results that are independent of the microstructure,
many points per octave need to be measured and averaged
(Kemp, 2007), and there are further issues when it comes to
combining across frequencies if there are unusable data, par-
ticularly when comparing measurements where data may be
unusable at different frequencies in different measurements.
Furthermore, the DPOAE measurement is a mix of both
reflection-source and distortion-source OAEs (Shera and
Guinan, 1999), whereas TEOAEs as measured in humans are
essentially reflection source (Withnell et al., 2008). Tech-
niques to separate out the two sources for DPOAEs exist
(e.g., Long and Talmadge, 1997; Heitmann et al., 1998; Tal-
madge et al., 1999; Konrad-Martin et al., 2001; Dhar and
Shaffer, 2004; Shaffer and Dhar, 2006; Long et al., 2008),
but either are not yet implemented and tested for clinical
applications or else have inherent limitations for clinical ap-
plication (Dhar and Shaffer, 2004). Although in our studies
there were more unusable TEOAE data compared with
DPOAE data, modern instruments have lower noise floors
and faster data collection allowing for more averaging, so we
anticipate unusable data to be less of a problem in the future.
E. Concluding remarks
OAEs are predictive of incipient NIHL. It is unknown
whether prior noise exposures or innate factors explain why
some normal-hearing ears had low-level or absent OAEs.
Most recruits indicated that they had prior noise exposures
typical of a modern lifestyle, including weapons’ fire, ampli-
fied music, and machinery noise. Regardless, having a test
that indicates ears susceptible to noise-induced hearing loss
is a boon for hearing conservation and audiology in general.
The current study extends the earlier findings to include im-
pulse noise.
If identifying those individuals and groups most at risk
for hearing loss from noise exposure in their near future is
possible by detecting the early stages of inner-ear changes,
then steps can be taken to prevent or mitigate further dam-
age. While the auditory medial-olivocochlear-bundle reflex
(MOCR) may be another way to assess future risk (Maison
and Liberman, 2000; Backus and Guinan, 2007), there is at
present no test in humans that sufficiently differentiates large
and small AERs within the test time available for clinical
testing. Furthermore, such a test requires OAEs with reason-
able amplitude, thereby precluding the use of such a test in
many noise-exposed individuals who do not have strong
OAEs. In the future, a very powerful predictive OAE test
battery might consist of both OAE level and MOCR
strength.
ACKNOWLEDGMENTS
Thanks to Linda Westhusin, Michael McFadden, Denise
Cline, Jackie Adler, Joy Houston, and Brian Ferris for their
assistance with data collection. A special thanks to the staff
and recruits of the Marine Corps Recruit Depot (MCRD) San
Diego and Charles Jackson of the Naval Medical Center San
Diego Occupational Audiology Department. Thanks to Tom
Taggart for his input into the overall experimental design,
help with logistics, and feedback on preliminary analyses.
Thanks to Chris Shera for helpful discussions on the theoret-
ical aspects. Thanks to the two anonymous reviewers whose
considered opinions substantively improved the manuscript.
This research was supported primarily by grants from the
Office of Naval Research. The views expressed in this article
are those of the authors and do not reflect the official policy
or position of the Department of the Navy, the Department of
Defense, or the United States Government.
1In laboratory studies on humans, only TTSs can be studied, and there is
typically a close relationship between changes in OAEs and changes in
audiometric thresholds (Marshall and Heller, 1998; Marshall et al., 2001).
However, TTSs and PTSs are physiologically different (Saunders et al.,
1985; Slepecky, 1986; Nordmann et al., 2000), so the results from TTS
experiments cannot be expected to generalize to PTS. Furthermore, labo-
ratory experiments examining PTS in animals may not generalize to hu-
mans (see summary in Lapsley Miller et al., 2004, p. 308). Therefore, to
understand PTS in humans, there is no substitute for actually measuring
PTS in humans. These human PTS experiments invariably have to be
conducted in field settings where there is usually neither the time nor
facilities to make measurements comparable in quality to those made in
the laboratory. Nevertheless, the stated results have been found repeatedly
across a range of studies.
2Data from K. S. Wolgemuth from a 1998 study on NIHL from Marine
infantry training at Camp Pendleton.
3Data from N. Vausse from a 1994 study on NIHL from Army training at
Fort Bragg.
4Informed consent briefings, conducted by one of the study principal inves-
tigators, took place immediately prior to the Marine Corps recruits begin-
ning their medical evaluation on day 2 of basic training. There were ap-
proximately 80 recruits in each briefing and Informed Consent forms were
passed out prior to beginning the briefing. The recruits were given ample
opportunity to ask any questions about the study, and participation was
voluntary. The voluntary aspect was made very clear to them given this
was a military basic training facility where most activity is mandatory.
Approximately 10% of the recruits declined to participate in the study. The
volunteers in the control group were also briefed, reviewed the informed
consent form, and were asked to participate. The informed consent form
was approved under Naval Medical Center, San Diego-approved research
protocol No. S-99-085.
5It was not possible to control for potential pharmacological influences
across subjects, which may have included over-the-counter and prescrip-
tion medications such as Erythromycin, Motrin, cold medications, and
aspirin.
6It was not possible to do ANOVA on the control group, because only four
volunteers had complete data sets.
7As described in Lapsley Miller et al. (2006, footnote 6), the SEmeas can be
used to specify the magnitude of a statistically significant change within
an individual (Ghiselli, 1964), and is defined as SEmeas= 1
2 (s1
2 +s2
2)(1−r)
where s1
2 and s2
2 are the pre- and post-test variances, and r is the correlation
between pre- and post-tests. Because the focus here is on the difference
between pre- and post-tests, SEmeas is defined as 2SEmeas (Beattie, 2003;
Beattie et al., 2003). Multiplying
SEmeas by an appropriate multiplier
then gives the desired confidence interval. Here a multiplier of 2.12 is
used, which gives a 98% confidence interval.
813 ears were classified with a STS in just a two- or three-frequency aver-
aged band, and 23 ears were classified with STSs at both individual fre-
quencies and across averaged frequency bands. Six right ears were classi-
fied with a STS at only one individual frequency (with no shifts in the
contralateral ear). No left ear was classified with a STS at just one
frequency.
9The other two frequencies in this range (2.8 and 3.6 kHz) were not used as
it would increase the false-positive SES rate when detecting SESs for
DPOAEs; it was decided that having three frequencies/frequency bands
for each OAE type and audiometric threshold would be a fairer balance.
Shifts in averaged frequency bands were not considered because too many
ears had unusable data at one or more frequencies.
10SES status could not be determined if (a) the OAE was below the noise
floor on the pre-test; (b) the OAE was below the noise floor on the post-
test, and the post-test noise floor was high so that the OAE level could not
1010
J. Acoust. Soc. Am., Vol. 125, No. 2, February 2009
Marshall et al.: Detecting inner-ear damage with otoacoustic emissions
Page 17
be estimated; or (c) data loss on the pre-test or post-test.
11In addition, if we had used additional criteria based on averages across
frequencies, as we did with STS, the SES rate would be expected to
increase, especially as the SES criteria for averaged bands are likely to be
smaller thereby allowing smaller wide-band shifts to be detected (Lapsley
Miller et al., 2004).
12It is not possible to use pre-test audiometric thresholds as predictors here
because the volunteers were prescreened for hearing levels. As shown in
Fig. 1, pre-test audiometric thresholds were essentially the same for the
STS ears compared with the no-STS ears.
13SFOAEs are OAEs generated with the same frequency as the evoking
tonal stimulus (e.g., Shera and Guinan, 1999).
14Our implementation of DPOAEs may have also been a factor. We were
trying to capitalize on growth functions (testing various stimulus levels at
specific frequencies) when, in hindsight, using more frequencies might
have been a better bet. By testing with a sparse frequency spacing it is
possible that for some ears, some of the test points fell into a null in the
DPOAE microstructure (Shaffer et al., 2003). If the DPOAEs had been
tested at a sufficient number of frequencies, then we could average them to
get the total energy in a frequency band, which would be equivalent in that
way to TEOAEs. Of course, it would be a much slower test than the
TEOAE test, at least with current instrumentation. In the future, new
methods to measure DPOAEs may enable easier comparisons with
TEOAEs, and DPOAEs might be less dependent on their implementation
(e.g., Long et al., 2008).
15TEOAEs from lower-level stimuli are more sensitive to cochlear changes
due to quinine (Karlsson et al., 1991). Data of ours (to be published) from
a study using similar methodology to Marshall and Heller (1998) indicate
that TEOAEs evoked with a lower stimulus level show greater sensitivity
to noise-induced TTS.
16For left-handed shooters, an adaptor was used on the rifle range so the
recruit would not get hit in the face by a very hot brass shell casing.
17There were some differences between the two studies: In Lapsley Miller
et al. (2006), confirmed PTS ears were compared with no-PTS ears, and
ears with unconfirmed STS were not included. Ears were combined in
analyses, because there were not enough PTS ears (13 left ears and 5 right
ears, including 3 bilateral) to analyze ears separately. The noise exposures
tended to be continuous noise overlaid with impact noise. For the current
study, unconfirmed STS ears were compared with no-STS ears. There
were differences between ears in OAE amplitude at many frequencies, so
ears were analyzed separately. The noise exposures were impulse noise.
OAE amplitude criteria were created from binning into percentile catego-
ries, although the results are plotted as a function of OAE amplitude, not
percentile. Further, the study was part of a larger study investigating ge-
netic factors of hearing loss, with only a subset of volunteers receiving
OAE testing. A blood sample was taken at study enrollment and used to
identify those volunteers with the Connexion 26 (35delG) GJB2 polymor-
phism. Those volunteers with this polymorphism were asked back for
OAE testing if they had not already been tested. Thus, the sample had a
higher percentage of volunteers with 35delG (13 out of 285; 4.6%) than
the general population. Initial analyses did not indicate that the 35delG
group was in any way different to the other volunteers, so they were
included in all analyses. Only two 35delG volunteers were classified with
a STS, which was not sufficient to establish if 35delG was a predictor of
STS. Recently, evidence has come to hand showing no relationship be-
tween 35delG and NIHL (Van Eyken et al., 2007).
American Speech-Language-Hearing Association (1994). “Audiologic man-
agement of individuals receiving cochleotoxic drug therapy [guidelines],”
ASHA 34, 11–19.
ANSI (1991). Maximum Permissible Ambient Noise Levels for Audiometric
Test Rooms (ANSI S3.1) (American National Standards Institute, New
York).
ANSI (1996). Specifications for Audiometers (ANSI S3.6-1996, R1973)
(American National Standards Institute, New York).
Arnold, D. J., Lonsbury-Martin, B. L., and Martin, G. K. (1999). “High-
frequency hearing influences lower-frequency distortion-product otoacous-
tic emissions,” Arch. Otolaryngol. Head Neck Surg. 125, 215–222.
Attias, J., Weisz, G., Almog, S., Shahar, A., Wiener, M., Joachims, Z.,
Netzer, A., Ising, H., Rebentisch, E., and Guenther, T. (1994). “Oral mag-
nesium intake reduces permanent hearing loss induced by noise exposure,”
Am. J. Otolaryngol. 15, 26–32.
Avan, P., Elbez, M., and Bonfils, P. (1997). “Click-evoked otoacoustic emis-
sions and the influence of high-frequency hearing losses in humans,” J.
Acoust. Soc. Am. 101, 2771–2777.
Backus, B. C., and Guinan, J. J. (2007). “Measurement of the distribution of
medial olivocochlear acoustic reflex strengths across normal-hearing indi-
viduals via otoacoustic emissions,” J. Assoc. Res. Otolaryngol. 8, 484–
496.
Beattie, R. C. (2003). “Distortion product otoacoustic emissions: Compari-
son of sequential versus simultaneous presentation of primary-tone pairs,”
J. Am. Acad. Audiol 14, 471–484.
Beattie, R. C., and Bleech, J. (2000). “Effects of sample size on the reliabil-
ity of noise floor and DPOAE,” Br. J. Audiol. 34, 305–309.
Beattie, R. C., Kenworthy, O. T., and Luna, C. A. (2003). “Immediate and
short-term reliability of distortion-product otoacoustic emissions,” Int. J.
Audiol. 42, 348–354.
Berger, E. H. (2000). “Hearing protection devices,” in The Noise Manual,
5th ed., edited by E. H. Berger, L. H. Royster, J. D. Royster, D. P. Driscoll,
and M. Layne (American Industrial Hygiene Association (AIHA) Press,
Fairfax, VA).
Bohne, B. A., Yohman, L., and Gruner, M. M. (1987). “Cochlear damage
following interrupted exposure to high-frequency noise,” Hear. Res. 29,
251–264.
Bray, P., and Kemp, D. T. (1987). “An advanced cochlear echo technique
suitable for infant screening,” Br. J. Audiol. 21, 191–204.
Bray, P. J. (1989). “Click evoked otoacoustic emissions and the develop-
ment of a clinical otoacoustic hearing test instrument,” Ph.D. thesis, Insti-
tute of Laryngology and Otology, University College and Middlesex
School of Medicine, London.
Clark, W. W. (1991). “Noise exposure from leisure activities: A review,” J.
Acoust. Soc. Am. 90, 175–181.
Davis, B., Qiu, W., and Hamernik, R. P. (2004). “The use of distortion
product otoacoustic emissions in the estimation of hearing and sensory cell
loss in noise-damaged cochleas,” Hear. Res. 187, 12–24.
Davis, B., Qiu, W., and Hamernik, R. P. (2005). “Sensitivity of distortion
product otoacoustic emissions in noise-exposed chinchillas,” J. Am. Acad.
Audiol 16, 69–78.
Department of Defense (2004). Department of Defense Instruction 6055.12:
DOD Hearing Conservation Program (HCP) (USD/AT&L).
Dhar, S., and Shaffer, L. A. (2004). “Effects of a suppressor tone on distor-
tion product otoacoustic emissions fine structure: Why a universal sup-
pressor level is not a practical solution to obtaining single-generator DP-
grams,” Ear Hear. 25, 573–585.
Dorn, P. A., Piskorski, P., Gorga, M. P., Neely, S. T., and Keefe, D. H.
(1999). “Predicting audiometric status from distortion product otoacoustic
emissions using multivariate analyses,” Ear Hear. 20, 149–163.
Duvdevany, A., and Furst, M. (2006). “Immediate and long-term effect of
rifle blast noise on transient-evoked otoacoustic emissions,” J. Basic Clin.
Physiol. Pharmacol. 17, 173–185.
Duvdevany, A., and Furst, M. (2007). “The effect of longitudinal noise
exposure on behavioral audiograms and transient-evoked otoacoustic
emissions,” Int. J. Audiol. 46, 119–127.
Engdahl, B., Woxen, O., Arnesen, A. R., and Mair, I. W. (1996). “Transient
evoked otoacoustic emissions as screening for hearing losses at the school
for military training,” Scand. Audiol. 25, 71–78.
Federal Railroad Administration (2006). “Occupational noise exposure for
railroad operating employees; final rule (49 CRF Parts 227 and 229),” Fed.
Regist. 71, 63066–63168.
Fleiss, J. L., Levin, B. A., and Paik, M. C. (2003). Statistical Methods for
Rates and Proportions (Wiley, Hoboken, NJ).
Franklin, D. J., McCoy, M. J., Martin, G. K., and Lonsbury-Martin, B. L.
(1992). “Test/retest reliability of distortion-product and transiently evoked
otoacoustic emissions,” Ear Hear. 13, 417–429.
Ghiselli, E. E. (1964). Theory of Psychological Measurement (McGraw-
Hill, New York).
Hamernik, R. P., Ahroon, W. A., and Lei, S. F. (1996) “The cubic distortion
product otoacoustic emissions from the normal and noise-damaged chin-
chilla cochlea,” J. Acoust. Soc. Am. 100, 1003–1012.
Hamernik, R. P., Patterson, J. H., Turrentine, G. A., and Ahroon, W. A.
(1989). “The quantitative relation between sensory cell loss and hearing
thresholds,” Hear. Res. 38, 199–211.
Hamernik, R. P., and Qiu, W. (2000). “Correlations among evoked potential
thresholds, distortion product otoacoustic emissions and hair cell loss fol-
lowing various noise exposures in the chinchilla,” Hear. Res. 150, 245–
257.
Heitmann, J., Waldmann, B., Schnitzler, H., Plinkert, P. K., and Zenner, H.
P. (1998). “Suppression of distortion product otoacoustic emissions
J. Acoust. Soc. Am., Vol. 125, No. 2, February 2009
Marshall et al.: Detecting inner-ear damage with otoacoustic emissions
1011
Page 18
(DPOAE) near 2f1-f2 removes DP-gram fine structure—Evidence for a
secondary generator,” J. Acoust. Soc. Am. 103, 1527–1531.
Humes, L. E., Joellenbeck, L. M., and Durch, J. S., eds. (2005). Noise and
Military Service: Implications for Hearing Loss and Tinnitus (Institute of
Medicine: Medical Follow-Up Agency, Washington, DC).
Job, A., Grateau, P., and Picard, J. (1998). “Intrinsic differences in hearing
performances between ears revealed by the asymmetrical shooting posture
in the army,” Hear. Res. 122, 119–124.
Kalluri, R., and Shera, C. A. (2007). “Near equivalence of human click-
evoked and stimulus-frequency otoacoustic emissions,” J. Acoust. Soc.
Am. 121, 2097–2110.
Karlsson, K. K., Berninger, E., and Alvan, G. (1991). “The effect of quinine
on psychoacoustic tuning curves, stapedius reflexes and evoked otoacous-
tic emissions in healthy volunteers,” Scand. Audiol. 20, 83–90.
Kemp, D. (2007). “The basics, the science, and the future potential of otoa-
coustic emissions,” in Otoacoustic Emissions: Clinical Applications, 3rd
ed., edited by M. S. Robinette and T. J. Glattke (Thieme, New York), pp.
7–42.
Knight, R. D., and Kemp, D. T. (1999). “Relationships between DPOAE and
TEOAE amplitude and phase characteristics,” J. Acoust. Soc. Am. 106,
1420–1435.
Konopka, W., Pawlaczyk-Luszczynska, M., Sliwinska-Kowalska, M., Gr-
zanka, A., and Zalewski, P. (2005). “Effects of impulse noise on tran-
siently evoked otoacoustic emission in soldiers,” Int. J. Audiol. 44, 3–7.
Konrad-Martin, D., Neely, S. T., Keefe, D. H., Dorn, P. A., and Gorga, M. P.
(2001). “Sources of distortion product otoacoustic emissions revealed by
suppression experiments and inverse fast Fourier transforms in normal
ears,” J. Acoust. Soc. Am. 109, 2862–2879.
Lapsley Miller, J. A., and Marshall, L. (2001). “Monitoring the effects of
noise with otoacoustic emissions,” Semin. Hear. 22, 393–403.
Lapsley Miller, J. A., Marshall, L., and Heller, L. M. (2004). “A longitudinal
study of changes in evoked otoacoustic emissions and pure-tone thresh-
olds as measured in a hearing conservation program,” Int. J. Audiol. 43,
307–322.
Lapsley Miller, J. A., Marshall, L., Heller, L. M., and Hughes, L. M. (2006).
“Low-level otoacoustic emissions may predict susceptibility to noise-
induced hearing loss,” J. Acoust. Soc. Am. 120, 280–296.
LePage, E. L., Murray, N. M., Tran, K., and Harrap, M. J. (1993). “The ear
as an acoustical generator: Otoacoustic emissions and their diagnostic po-
tential,” Acoust. Aust. 21, 86–90.
Long, G., Talmadge, C., Prieve, B., and Lahtinen, L. (2008). “Extraction of
DPOAE generator and reflection components in the time domain in adults
and infants,” Assoc. Res. Otolaryngol. Abstr. 31, 61.
Long, G. R., and Talmadge, C. L. (1997). “Spontaneous otoacoustic emis-
sion frequency is modulated by heartbeat,” J. Acoust. Soc. Am. 102,
2831–2848.
Lonsbury-Martin, B., and Martin, G. K. (2007). “Distortion product otoa-
coustic emissions in populations with normal hearing sensitivity,” in Otoa-
coustic Emissions: Clinical Applications, 3rd ed., edited by M. S. Robin-
ette and T. J. Glattke (Thieme, New York), pp. 107–130.
Maison, S. F., and Liberman, M. C. (2000). “Predicting vulnerability to
acoustic injury with a noninvasive assay of olivocochlear reflex strength,”
J. Neurosci. 20, 4701–4707.
Marshall, L., Brandt, J. F., and Marston, L. E. (1975). “Anticipatory middle-
ear reflex activity from noisy toys,” J. Speech Hear Disord. 40, 320–326.
Marshall, L., and Hanna, T. E. (1989). “Evaluation of stopping rules for
audiological ascending test procedures using computer simulations,” J.
Speech Hear. Res. 32, 265–273.
Marshall, L., and Heller, L. M. (1998). “Transient-evoked otoacoustic emis-
sions as a measure of noise-induced threshold shift,” J. Speech Lang.
Hear. Res. 41, 1319–1334.
Marshall, L., Lapsley Miller, J. A., and Heller, L. M. (2001). “Distortion-
product otoacoustic emissions as a screening tool for noise-induced hear-
ing loss,” Noise Health 3, 43–60.
Mining Safety and Health Administration (1999). “Health standards for oc-
cupational noise exposure; final rule (30 CFR Parts 56 and 57),” Fed.
Regist. 64, 49548–49634.
Murray, N. M., and LePage, E. L. (2002). “A nine-year longitudinal study of
the hearing of orchestral musicians,” paper presented at the International
Auditory Congress, Melbourne, Australia, March.
Murray, N. M., LePage, E. L., and Mikl, N. (1998). “Inner ear damage in an
opera theatre orchestra as detected by otoacoustic emissions, pure tone
audiometry and sound levels,” Aust. J. Audiol. 20, 67–78.
Nageris, B. I., Raveh, E., Zilberberg, M., and Attias, J. (2007). “Asymmetry
in noise-induced hearing loss: Relevance of acoustic reflex and left or
right handedness,” Otol. Neurotol. 28, 434–437.
National Institute for Occupational Safety and Health (1998). Criteria for a
Recommended Standard: Occupational Noise Exposure (Revised Criteria
1998). No. 98–126 (U.S. Department of Health and Human Services
(NIOSH), Cincinnati, OH).
Navy Occupational Health and Safety Program (1999). OPNAVINST
5100.23E: Hearing Conservation and Noise Abatement (Chief of Naval
Operations, Washington, DC).
Nordmann, A. S., Bohne, B. A., and Harding, G. W. (2000). “Histopatho-
logical differences between temporary and permanent threshold shift,”
Hear. Res. 139, 13–30.
Occupational Safety & Health Administration (2007). “Occupational noise
exposure standard (29 CFR 1910.95),” Code of Federal Regulations (U.S.
Department of Labor, Washington, DC).
Perez, R., Freeman, S., and Sohmer, H. (2004). “Effect of an initial noise
induced hearing loss on subsequent noise induced hearing loss,” Hear.
Res. 192, 101–106.
Price, G. R. (2007). “Validation of the auditory hazard assessment algorithm
for the human with impulse noise data,” J. Acoust. Soc. Am. 122, 2786–
2802.
Saunders, J. C., Dear, S. P., and Schneider, M. E. (1985). “The anatomical
consequences of acoustic injury: A review and tutorial,” J. Acoust. Soc.
Am. 78, 833–860.
Schmuziger, N., Probst, R., and Smurzynski, J. (2005). “Otoacoustic emis-
sions and extended high-frequency hearing sensitivity in young adults,”
Int. J. Audiol. 44, 24–30.
Seixas, N. S., Goldman, B., Sheppard, L., Neitzel, R., Norton, S. J., and
Kujawa, S. G. (2005a). “Prospective noise induced changes to hearing
among construction industry apprentices,” Occup. Environ. Med. 62, 309–
317.
Seixas, N. S., Neitzel, R., Brower, S., Goldman, B., Somers, S., Sheppard,
L., Kujawa, S. G., and Norton, S. (2005b). “Noise-related changes in
hearing: A prospective study among construction workers,” paper pre-
sented at the 30th Annual NHCA National Hearing Conservation Confer-
ence, Tucson, AZ, 26 February.
Shaffer, L. A., and Dhar, S. (2006). “DPOAE component estimates and their
relationship to hearing thresholds,” J. Am. Acad. Audiol 17, 279–292.
Shaffer, L. A., Withnell, R. H., Dhar, S., Lilly, D. J., Goodman, S. S., and
Harmon, K. M. (2003). “Sources and mechanisms of DPOAE generation:
Implications for the prediction of auditory sensitivity,” Ear Hear. 24, 367–
379.
Shaw, E. A. (1966). “Earcanal pressure generated by circumaural and su-
praaural earphones,” J. Acoust. Soc. Am. 39, 471–479.
Shera, C. A. (2004). “Mechanisms of mammalian otoacoustic emission and
their implications for the clinical utility of otoacoustic emissions,” Ear
Hear. 25, 86–97.
Shera, C. A., and Guinan, J. J. (1999). “Evoked otoacoustic emissions arise
by two fundamentally different mechanisms: A taxonomy for mammalian
OAEs,” J. Acoust. Soc. Am. 105, 782–798.
Siegel, S. (1956). Nonparametric Statistics for the Behavioral Sciences
(McGraw-Hill, New York).
Sininger, Y. S., and Cone-Wesson, B. (2004). “Asymmetric cochlear pro-
cessing mimics hemispheric specialization,” Science 305, 1581.
Sisto, R., Moleti, A., and Shera, C. A. (2007). “Cochlear reflectivity in
transmission-line models and otoacoustic emission characteristic time de-
lays,” J. Acoust. Soc. Am. 122, 3554–3561.
Slepecky, N. (1986). “Overview of mechanical damage to the inner ear:
Noise as a tool to probe cochlear function,” Hear. Res. 22, 307–321.
Talmadge, C. L., Long, G. R., Tubis, A., and Dhar, S. (1999). “Experimental
confirmation of the two-source interference model for the fine structure of
distortion product otoacoustic emissions,” J. Acoust. Soc. Am. 105, 275–
292.
US Army Center for Health Promotion and Preventive Medicine (2008).
“Noise levels of common army equipment,” retrieved from http://
usachppm.apgea.army.mil/hcp/noiselevels.aspx on 1 May 2008 (Aberdeen
Proving Ground, MD).
Van Eyken, E., Van Laer, L., Fransen, E., Topsakal, V., Hendrickx, J. J.,
Demeester, K., Van de Heyning, P., Maki-Torkko, E., Hannula, S., Sorri,
M., Jensen, M., Parving, A., Bille, M., Baur, M., Pfister, M., Bonaconsa,
A., Mazzoli, M., Orzan, E., Espeso, A., Stephens, D., Verbruggen, K.,
Huyghe, J., Dhooge, I., Huygen, P., Kremer, H., Cremers, C., Kunst, S.,
Manninen, M., Pyykko, I., Rajkowska, E., Pawelczyk, M., Sliwinska-
Kowalska, M., Steffens, M., Wienker, T., and Van Camp, G. (2007). “The
1012
J. Acoust. Soc. Am., Vol. 125, No. 2, February 2009
Marshall et al.: Detecting inner-ear damage with otoacoustic emissions
Page 19
contribution of GJB2 (Connexin 26) 35delG to age-related hearing impair-
ment and noise-induced hearing loss,” Otol. Neurotol. 28, 970–975.
Wagner, W., Heppelmann, G., Vonthein, R., and Zenner, H. P. (2008). “Test-
retest repeatability of distortion product otoacoustic emissions,” Ear Hear.
29, 378–391.
Withnell, R. H., Hazlewood, C., and Knowlton, A. (2008). “Reconciling the
origin of the transient evoked otoacoustic emission in humans,” J. Acoust.
Soc. Am. 123, 212–221.
Withnell, R. H., and Lodde, J. (2006). “In search of basal distortion product
generators,” J. Acoust. Soc. Am. 120, 2116–2123.
Withnell, R. H., Yates, G. K., and Kirk, D. L. (2000). “Changes to low-
frequency components of the TEOAE following acoustic trauma to the
base of the cochlea,” Hear. Res. 139, 1–12.
Yates, G. K., and Withnell, R. H. (1999). “The role of intermodulation
distortion in transient-evoked otoacoustic emissions,” Hear. Res. 136, 49–
64.
Zhou, X.-H., Obuchowski, N. A., and McClish, D. K. (2002). Statistical
Methods in Diagnostic Medicine (Wiley-Interscience, New York).
J. Acoust. Soc. Am., Vol. 125, No. 2, February 2009
Marshall et al.: Detecting inner-ear damage with otoacoustic emissions 1013