Fine Structure of Hearing Threshold and

Fine structure of hearing threshold and loudness perceptiona)
Manfred Mauermannb)
Medizinische Physik, Universität Oldenburg, D-26111 Oldenburg, Germany
Glenis R. Longc)
Hanse-Wissenschaftskolleg (HWK), 27753 Delmenhorst, Germany
Birger Kollmeier
Medizinische Physik, Universität Oldenburg, D-26111 Oldenburg, Germany
~Received 8 August 2003; revised 24 March 2004; accepted 19 April 2004!

Hearing thresholds measured with high-frequency resolution show a quasiperiodic change in level
called threshold fine structure ~or microstructure!. The effect of this fine structure on loudness
perception over a range of stimulus levels was investigated in 12 subjects. Three different
approaches were used. Individual hearing thresholds and equal loudness contours were measured in
eight subjects using loudness-matching paradigms. In addition, the loudness growth of sinusoids
was observed at frequencies associated with individual minima or maxima in the hearing threshold
from five subjects using a loudness-matching paradigm. At low levels, loudness growth depended on
the position of the test- or reference-tone frequency within the threshold fine structure. The slope of
loudness growth differs by 0.2 dB/dB when an identical test tone is compared with two different
reference tones, i.e., a difference in loudness growth of 2 dB per 10-dB change in stimulus. Finally,
loudness growth was measured for the same five subjects using categorical loudness scaling as a
direct-scaling technique with no reference tone instead of the loudness-matching procedures.
Overall, an influence of hearing-threshold fine structure on loudness perception of sinusoids was
observable for stimulus levels up to 40 dB SPL—independent of the procedure used. Possible
implications of fine structure for loudness measurements and other psychoacoustic experiments,
such as different compression within threshold minima and maxima, are discussed. © 2004
Acoustical Society of America. @DOI: 10.1121/1.1760106#
PACS numbers: 43.66.Cb @GK# Pages: 1066 –1080
I. INTRODUCTION emissions—OAE! could be due to the same mechanisms as

those responsible for the threshold fine structure ~Kemp,
When absolute hearing thresholds are evaluated with 1979!. These OAEs are generated by the processes respon-
small frequency increments, consistent quasiperiodic pat- sible for the remarkable sensitivity of the human ear, and
terns of threshold change with frequency ~threshold fine
even small amounts of hearing loss correspond to signifi-
structure or microstructure1! can be obtained. Regions of
cantly reduced levels of OAEs. All types of OAEs in humans
relatively stable poor sensitivity ~hearing threshold maxima!
are separated by narrow regions of greater sensitivity ~thresh- are characterized by strikingly similar fine-structure patterns
old minima!. In this study, we investigated in detail the rela- in the frequency domain with a frequency spacing between
tion between hearing-threshold fine structure and fine struc- adjacent maxima or minima on the order of 0.4 bark ~e.g., He
ture in loudness perception of sinusoidal signals for and Schmiedt, 1993; Mauermann et al., 1997; Zwicker and
frequencies around 1800 Hz to receive a broader base for the Peisl, 1990; Zweig and Shera, 1995!. Threshold minima are
understanding of loudness perception at low levels. Isolated associated with frequencies near spontaneous OAEs
investigations ~Elliot, 1958; van den Brink, 1970; Thomas, ~SOAEs! or large evoked OAEs ~EOAEs! ~Zwicker and
1975! of threshold fine structure were described in the litera- Schloth, 1984; Long and Tubis, 1988a, b; Horst and de
ture prior to 1979, establishing that the frequency spacing of Kleine, 1999!. The depth of the minima is not simply related
the threshold fine structure appeared to be a constant fraction to the level because high-level emissions can interact with
of estimates of the frequency resolution capacity of the ear
the stimuli and elevate thresholds ~Long and Tubis, 1988a, b;
~the critical band!. In 1979, Kemp noted that the capacity of
Smurzynski and Probst, 1998!. Changes in the emission fre-
the healthy ear to generate sounds ~known as otoacoustic
quency are associated with changes in the threshold-fine-
a!
structure frequency ~Long and Tubis, 1988a; He, 1990; Furst
Parts of this study were presented at the 23rd Midwinter Research Meeting
of the Association for Research in Otolaryngology 2000 in St. Petersburg
et al., 1992!. Overall, the spacing of the threshold fine struc-
Beach, Florida @Mauermann et al., Abstract No. 984, p. 284#. ture is very similar to that of otoacoustic emissions. Further-
b!
Electronic mail: [email protected] more, it has been observed that normal-hearing subjects with
c!
Current affiliation: City University of New York; Graduate School and
University Center, 365 Fifth Avenue, New York, N.Y. 10016-4309. Elec- weak otoacoustic emissions show a reduced ‘‘audiogram
tronic mail: [email protected] ripple’’ ~Kapadia and Lutman, 1999!.
1066 J. Acoust. Soc. Am. 116 (2), August 2004 0001-4966/2004/116(2)/1066/15/$20.00 © 2004 Acoustical Society of America
Experimental evidence for a close link between OAE diplacusis. He found ‘‘some recruitment at low SPLs for the
fine structure and threshold fine structure has been supported extreme peaks,’’ i.e., a flattening of threshold fine structure at
by different cochlear models. Any sound generated in the 40 dB. In the example shown ~van den Brink 1970, Fig. 10!,
cochlea must be conducted through the middle ear if it is to he linked maxima and minima in threshold with extrema in
be detected in the ear canal as an OAE. But, not all sound the isophons up to 80 dB SPL, which suggests a preservation
generated in the cochlea will be transmitted through the of a reduced fine structure in isophons ~with fluctuations of
middle ear. Due to the impedance mismatch at the stapes, about 2–3 dB! for reference tones up to 80 dB SPL.
some sound will be reflected back into the cochlea ~e.g., In studies taking the fine structure into account, the vari-
Shera and Zweig, 1993!. The returning reflection will either ance at each frequency is comparable to the within-subject
enhance or partially cancel any energy at the original reflec- variance seen in most other psychoacoustic research with
tion site depending on the round-trip travel time. If the sound well-trained subjects. The across-frequency variance is, how-
was initially reflected ~and is not canceled by the returning
ever, much larger and comparable to the between-subject
echo!, it will be reflected again unless the properties of the
variance seen in many experiments. One interpretation of
cochlea have changed. Multiple internal reflections of co-
these results is that much of the between-subject variance
chlear traveling waves will occur ~Zweig and Shera, 1995!,
seen might depend on the position of the stimuli within the
leading to the resonance behavior of the cochlea originally
cochlear fine structure. Hellman and Zwislocki ~1961! found
suggested by Kemp ~1980! as an explanation for fine struc-
ture in stimulus frequency OAEs ~SFOAE!. This resonance that presenting the stimuli at equal SL in comparison to equal
will naturally enhance the response of the basilar membrane SPL reduces the variability among listeners’ loudness judg-
to sounds at some frequencies, and reduce its response to ments, especially at low levels. Any effects of cochlear fine
sounds at others. Frequencies at which the basilar-membrane structure on psychoacoustic research will increase the vari-
response at CF is maximal will result in threshold fine- ance and thus limit our ability to evaluate the underlying
structure minima, and frequencies at which this basilar- impact of some stimulus manipulations. Nevertheless, with a
membrane response is minimal will result in threshold fine- few exceptions, most psychoacoustic research with sinusoids
structure maxima. This same resonance behavior can be used is done at discrete, widely spaced frequencies chosen without
to explain the origin of the pseudoperiodicity observed in all any attempt to determine whether these tones lie in a mini-
OAE fine structure with a single origin, and thus provides a mum, a maximum, or a transition region within the cochlear
common origin of the OAE fine structure and threshold ~and fine structure. For studies interested in average effects over a
other psychoacoustic! fine structures ~see Talmadge et al., sufficiently large number of subjects, it is certainly reason-
1998!. Even though OAE and psychoacoustic fine structure able to ignore potential effects of cochlear fine structure. But,
are based on the same underlying mechanisms, and the peri- it might be necessary to strengthen the consideration of co-
odicity is similar from the model point of view, the pattern of chlear fine-structure effects on psychoacoustic measurements
both does not necessarily match over all frequencies ~Tal- ~a! to get a more detailed understanding of hearing mecha-
madge et al., 1998!. nisms; ~b! to investigate interindividual variations; and ~c! to
In addition to the theoretical implications of cochlear possibly increase the potential value of audiological tools for
fine structure, there are practical implications for psychoa- more precise and individual diagnoses, or for diagnoses at
coustic research. Except for the investigation of the relation very early stages of cochlea injury. The aim of this study is
between threshold fine structures and OAEs, there are only a to further examine the potential influences of the cochlear
small number of psychoacoustic studies investigating the de- fine structure on perceived loudness of sinusoidal signals. Do
pendence of suprathreshold psychoacoustic data on threshold the interference effects of cochlear mechanics most probably
fine structure. Variations of psychoacoustic observations re- responsible for threshold fine structure, affect loudness per-
lated to threshold fine structures have been found in ~a! the ception? To get a broader base of detailed data on loudness
perceived loudness of low-sensation-level tones ~Kemp,
perception taking the threshold fine structure into account,
1979!; ~b! temporal integration ~Cohen, 1982!; ~c! masked
the following experiments were performed: Experiment 1 in-
thresholds ~Long, 1984!; ~d! amplitude-modulation thresh-
vestigated the range of stimulus levels for which loudness
olds ~Zwicker, 1986; Long, 1993!; ~e! monaural diplacusis
fine structure is preserved in eight subjects. Thresholds and
~Kemp, 1979; Long, 1998!; and ~f! binaural diplacusis ~van
equal-loudness contours for different stimulus levels were
den Brink, 1970, 1980!. The effects of cochlear fine structure
get smaller as stimulus level is increased in most paradigms, measured with a high-frequency resolution. Experiment 2
and there is some indication that the spacing of the psychoa- observed loudness growth at low to moderate levels using
coustic fine structures can change at the highest levels tested test-tone frequencies associated with adjacent minima and
~Long, 1984!. In addition to screening of loudness maxima maxima in individual thresholds in an attempt to describe the
and minima at about 10 dB SL in 42 normal-hearing ears, change in response with level for these different conditions.
Kemp ~1979! showed equal-loudness contours for one sub- We also examined whether the position of the reference tone
ject around a pronounced threshold minimum, which flattens within threshold fine structure affected the measured loud-
out with increasing levels at about 35 dB SPL. Van den Brink ness growth, to check whether threshold fine structure may
~1970! measured thresholds and isophons with a high- influence loudness-matching procedures at low levels.
frequency resolution, using a 1-kHz reference tone at 40, 60, In experiment 3, loudness growth was again measured
and 80 dB SPL, so they could be compared with patterns of for different frequencies—matching either a minimum or a
J. Acoust. Soc. Am., Vol. 116, No. 2, August 2004 Mauermann et al.: Fine structure of threshold and loudness 1067
maximum of threshold—but using a different paradigm. In III. EXPERIMENT 1: PRESERVATION OF THRESHOLD
this experiment, a categorical loudness scaling was used as a FINE STRUCTURE IN EQUAL-LOUDNESS
direct scaling technique in addition to the loudness-matching CONTOURS
procedures used in experiments 1 and 2. By comparing When Kemp ~1979! investigated the equal-loudness
across experiments, we can determine how the effects of fine contours around a pronounced sensitivity maximum in one
structure depend on the measurement procedures. subject, the threshold fine structure could be observed when
the reference tone was about 35 dB SPL. To determine how,
and up to what levels, loudness perception depends on
II. GENERAL METHODS threshold fine structure, data were gathered in detail from
more subjects and a wider frequency region. Hearing thresh-
A. Subjects
olds were measured with a high-frequency resolution from
Ten normal-hearing subjects ~GL, GM, JO, KW, MO, 1600–2000 Hz and compared with equal-loudness contours.
MW, RH, TB, RM, SU—six male, four female! with thresh- While Kemp ~1979! used test tones fixed in level, we con-
olds better than 15 dB HL at the standard audiometric fre- centrated on a paradigm with the reference tone kept fixed in
quencies from 125 Hz to 8 kHz, and two subjects with a level.
slight hearing loss ~MM—male, 25 dB HL at 8 kHz; DS—
male, 25 dB HL at 3 and 4 kHz! participated in this study. A. Methods
The subjects TB, GL, JO, KW, MM, and MO had been
scanned for spontaneous otoacoustic emissions ~SOAE! in 1. Subjects
another study ~Mauermann et al., 1999!. Three of these sub-
jects ~GL, TB, and MO! show weak SOAE within the fre- Subjects were GL, GM, JO, KW, MO, MW, RH, and TB
quency regions observed here. The frequencies of detectable ~four male, four female!. Subject GL ~second author! per-
SOAE are always indicated as asterisks in the first figure that formed measurements on hearing thresholds and extended
shows a threshold fine structure of the respective subject ~GL measurements on equal-loudness contours using adaptive in-
in Fig. 1~a! TB in Fig. 3~a!, and for subject MO in Fig. 8!. terleaved paradigms ~see the procedures below!. Hearing
All frequencies of detected SOAE correspond to minima in thresholds and equal-loudness contours using an adaptive-
threshold perception. The subjects were the authors, mem- interleaved paradigm were also measured in subjects GM,
bers of the medical physics group at the University of Old- MW, and to some extent in subject KW. Adjustment methods
enburg, and students getting paid for the measurements. For were used in subjects JO, KW, MM ~first author!, MO, TB,
all subjects, the measurements were conducted on one ear. and partially in subject MW.
During the measurements, the subjects were seated in a
double-walled sound-insulated booth ~IAC!.
2. Procedures
a. Hearing thresholds (adaptive). Measurements of hear-
ing thresholds were obtained from subjects GL, KW, MW,
B. Instrumentation and software
and GM using a 3 alternative forced choice ~AFC! adaptive
Adaptive loudness matches, and most of the threshold ~1-up, 2-down! paradigm with feedback. Sinusoids of
measurements, were controlled by the signal-processing soft- 250-ms duration were used as stimuli ~including 25-ms
ware SI running on an Indy computer system ~Silicon Graph- Hanning-shaped ramps!. The three observation intervals in
ics!. The signals were generated digitally at a sampling rate each trial were marked optically and were separated by 500
of 44 100 Hz by the SI software, and converted by the 16-bit ms of silence. The subject’s task was to indicate the interval
DA converters of the computer. They were attenuated by a in which the tone was presented. The frequencies ~1600–
computer-controlled audiometer and presented via an Ety- 2000 Hz in 12.5-Hz steps! were divided into three blocks
motic Research ER2 insert ear phone. The presentation of including 11 frequencies. Each of these blocks was measured
each observation interval in one trial was marked optically in separate sessions.
by an LED, attached to the side of the computer display in Every track started with level steps of 8 dB. At each
the booth. reversal, the step size was reduced to 4-, 2-, and finally 1-dB
A PC/MATLAB-controlled setup was used for some of the steps. The median of the final eight reversal points ~at 1-dB
interleaved adaptive threshold measurements, all of the ad- steps! was taken as the preliminary estimate of the threshold.
justment measurements, and the categorical loudness scaling. These measures were obtained in three sessions ~exceptions
The signals were generated digitally at a sampling rate of are indicated in the figure captions!. The mean of the three
44 100 Hz in the MATLAB programs and sent through an median values provided the final estimate of the threshold at
RME Digi96/8 PAD digital I/O card to a SEKD 2496 24-bit each frequency.
DA converter. After amplification ~Behringer headphone am- b. Equal-loudness contours (interleaved adaptive). The
plifier Powerplay II! the signals were presented via an ER2 instrumentation and signals were the same as those used for
insert phone. Each observation interval was marked optically the loudness measurements. A frequency resolution of 25 Hz
on the computer screen instead of using the LED marker. was used. During one session, all test-tone frequencies for
The ER2 were calibrated using an artificial ear for insert one level were measured as 16 interleaved tracks ~subject
phones ~Bruel & Kjaer type 4157!. GL, GM! for the frequency range from 1600 to 2000 Hz. For
1068 J. Acoust. Soc. Am., Vol. 116, No. 2, August 2004 Mauermann et al.: Fine structure of threshold and loudness
the subjects KW and MW, the measurements were divided in clicked on the play button after a level adjustment, the tone
two sessions with eight interleaved tracks. The order of the pair was presented with the test tone at the new level. The
1000-Hz reference tone and the respective test tone were slider control jumped back to the zero point when a new
randomized in each trial. The subjects had to decide which of frequency was presented to avoid anchor effects stemming
the two sinusoids presented in consecutive intervals was from the optical position of the slider. The two observation
louder. The reference tone was held fixed in level while the intervals in each trial were marked optically on the computer
test tones followed an adaptive 1-up, 1-down procedure con- screen. When the subject was sure that both tones were
verging to the 50% point of the psychometric function ~this equally loud, he/she was advised to press an ‘‘is equal’’ but-
is the traditional isoloudness procedure!. The measurements ton to proceed to the next frequency. The order of frequen-
for each frequency started with level steps of 8 dB, which cies was randomized. These measures were obtained in three
were decreased to 4 and finally to 2-dB steps at each rever- sessions. The mean of three adjusted levels was taken as the
sal. The median from the final four reversals ~with constant estimate for the PSE. The same frequency range was mea-
level changes of 2 dB! was taken as the preliminary estimate sured as for the adaptive interleaved paradigm, but with a
of equal loudness. Each session was repeated three times higher a frequency resolution of 12.5 Hz instead of 25 Hz.
~exceptions are indicated in the figure captions! with differ- e. Hearing thresholds (adjustment). Subjects who per-
ent start levels of the test tones. In each session, the start formed the adjustment method for loudness matching also
levels for the stimuli were either the same as the reference used an adjustment paradigm to measure the hearing thresh-
tone level L ref in dB SPL level or L ref610 dB ~1/3 of the old in quiet. The subject could replay a single tone pressing
stimuli at each start level!. In each session, a different start the play button and change its level by adjusting the slider on
level was chosen for each frequency so that all the frequen- the computer screen. The subject’s task was to report a just-
cies were tested with each start level. The order of test and noticeable level. So, the subjects performed a kind of self-
reference tone in each trial was randomized. The mean of the controlled audiogram with high-frequency resolution. The
resulting median values was taken as the final estimate for signals ~sinusoids of 500-ms duration including 25-ms
the level of equal loudness, referred to as the point of sub- Hanning-shaped ramps! were identical to the ones used for
jective equality ~PSE!. the loudness matching used in the adjustment paradigm.
For subject GL, the 1000-Hz reference tone was close to Again, the order of frequencies was randomized. Each ses-
a threshold minimum. Additional equal-loudness contours sion, containing all frequencies, was repeated three times.
were measured using a reference tone at 1800 Hz, which is The mean of the three adjusted levels was taken as the esti-
close to a threshold maximum in this subject. A set of equal mate of threshold.
loudness contours was measured in comparison to a refer-
ence tone at 1000 Hz, as well as for a reference tone at 1800
Hz. All other parameters were the same as for the measure-
ment of the equal-loudness contours. The position of the B. Results
1000-Hz reference tone within the fine structure was not Most of the subjects participating in this study had
evaluated in the other subjects. threshold fine structure, i.e., maxima and minima in hearing
c. Inverse loudness contours. In addition to the loudness thresholds with level differences of more than 5 dB. The only
matches using a reference tone with fixed level, subject GL exception was subject RH ~not shown here!, who showed
provided ‘‘inverse’’ loudness contours with a variable refer- limited threshold and isophon variation ~up to 65 dB SPL for
ence tone level, while the test-tone level was kept fixed. In the reference tone! for the frequencies investigated here.
this procedure, the stimuli are not set to be equally loud. Subjects MO, KW, and JO showed reduced or no fine struc-
Instead, an estimate of the loudness in phons at each fre- ture for frequencies between 1800 and 2000 Hz. The fre-
quency and level is provided. quency range investigated in subject MO ~data not shown
d. Equal loudness contours (adjustment). Due to the ex- here! was changed to 1400–1800 Hz, where he showed a
cessive measurement time and numerous critical remarks more pronounced fine structure.2 In frequency regions with
about the interleaved adaptive method from the majority of fine structure, we mostly see a characteristic quasiperiodic
the subjects, we decided to find a quicker paradigm for the pattern for the fine structure of hearing thresholds. The dif-
loudness measurements. This was done to avoid problems ferences of the adjusted threshold levels for adjacent maxima
with massive reduction of subject motivation, and to permit and minima varied individually from about 5 dB up to about
higher frequency resolution with acceptable measurement ef- 15 dB.
fort. Consequently, a different setup was designed. The sub- The shape of the threshold fine structure was visible in
jects’ task was to adjust the test tone to be equally loud as the the equal-loudness contours of all subjects up to a reference
reference tone at 1 kHz. The subjects were allowed to hear a tone of at least 25 dB SPL ~Figs. 1, 2, 3!, and for subject GL
pair of reference and test tones ~in fixed order—reference ~the one with most practice at the task! even up to 50 dB SPL
tone first! as often as they wanted by clicking on a ‘‘play’’ ~see Figs. 1, 4, 5!. Overall, the pattern flattened out with
button. Each tone had a duration of 500 ms. They could increasing level. In some subjects, the patterns tended to shift
change the level of the test tone by adjusting a slider on the or change shape at intermediate levels before becoming
computer screen. This permitted a maximum change in level smooth ~see, e.g., subject GL in Fig. 1 at 45 to 55 dB for the
of 66 dB during one step, not indicated to the subjects reference tone and subject GM in Fig. 2~b! for 35 and 45
~minimal possible step size was 0.5 dB!. When the subject dB!. These observations hold for both paradigms used.
FIG. 1. ~a! Hearing threshold and equal-loudness contours from subject
GL—right ear, for a reference tone fixed at 1 kHz, measured with the inter-
leaved adaptive paradigm ~data are averages of 6 thresholds!. From bottom
to top: hearing threshold ~gray line with open circles!, equal loudness con-
tours with reference level from 10 to 70 dB SPL in steps of 5 dB ~alternating
thin lines without symbols and thick lines with filled circles!. The reference-
tone levels of the thick lines are indicated by labels next to the lines. Two
independent sets of thresholds showed similar patterns. ~b! shows the results
from ‘‘inversed’’ equal-loudness contour measurements keeping the test tone
fixed in level while varying the level of the 1-kHz reference tone. In order to
compare the difference in dynamic ranges for the two different paradigms,
data with 10-dB step size are indicated with thick lines and symbols in both
~a! and ~b!. Shaded areas give the standard deviation from one set of three
repetitions of the loudness matches. While in ~a! the most sensitive frequen-
cies are represented by minima in the curves, in ~b! the most sensitive
FIG. 3. Hearing threshold ~gray! and equal-loudness contours ~black! from
frequencies are given by the maxima of the adjusted level; note, further-
~a! subject TB—left ear; ~b! subject KM—right ear; ~c! subject MM—left
more, the different dynamic range for the same reference-tone levels or
ear; and ~d! subject JO—left ear measured with the adjustment paradigm.
test-tone levels, respectively ~compare thick black lines with filled circles in
The curves are labeled with the reference levels used or ‘‘threshold,’’ respec-
the left and right panel!. Within the observed frequency range subject GL
tively. The shaded areas give the standard deviations of three repetitions.
has two SOAE at 1762 Hz ~29.9 dB SPL! and 1874 Hz ~212.3 dB SPL!.
The black asterisks in panel ~a! indicate the SOAE frequencies found in
The frequencies of the SOAE are indicated as black asterisks with labels
subject TB—left ear. The levels of the SOAE are indicated as labels beside
that give the SOAE level in the sealed ear canal.
the asterisks. There were no SOAE detected in the other three subjects.
Figure 6 shows a comparison of threshold measurements and

equal loudness contours ~reference tone level at 15 dB SPL! presented as minima in the equal-loudness contours, while
from both paradigms, i.e., the interleaved adaptive and the maxima line up with the maxima in the threshold fine struc-
adjustment paradigm. High consistency between the patterns ture. The level of the louder probe was reduced to match the
from the two paradigms can be seen. loudness of the fixed-level reference. Varying the level of the
When the reference tone was fixed in level, the fre- reference tone while keeping the test tone fixed in level gave
quency regions with the lowest thresholds ~most sensitive! the reverse pattern with maxima in the loudness function
associated with minima in the threshold fine structure. In this
FIG. 4. Loudness growth functions from subject GL—right ear. The data are
extracted from the data shown in Fig. 1 for reference tones fixed in level
FIG. 2. Hearing threshold and equal-loudness contours from ~a! subject ~solid lines, crosses! and ~b! for the reference tone varied in level ~dashed
MW—left ear and ~b! subject GM—left ear, measured with the interleaved- lines, open circles! both for two test-tone frequencies 1600 Hz ~black lines!
adaptive paradigm. The curves are labeled with the reference levels used or at a threshold minimum and 1950 Hz ~gray lines! at a threshold maximum.
‘‘threshold,’’ respectively. The shaded areas give the standard deviation of While the data from the two paradigms are very consistent, a different
three repetitions. The hearing threshold shown for subject GM is the result position of the reference-tone frequency within fine structure results in dif-
from only one measurement. ferent loudness slopes.
FIG. 5. Loudness growth functions ~a! from subject GL—right ear ~ex-
tracted from data shown in Fig. 1! and ~b! TB—left ear @extracted from data FIG. 7. Equal-loudness contours from subject GL—right ear for reference
shown in Fig. 3~a!# at different frequencies. While the data in ~a! were tone levels of 25, 40, 55 dB SPL ~indicated by labels within the plot!. The
collected using the adaptive interleaved paradigm, the data for ~b! stem from black lines with circles show data from measurements using a reference tone
measurements using the adjustment procedure. Both show similar loudness at 1 kHz; the gray lines with triangles show data for a reference tone at 1.8
growth for stimuli from threshold maxima at three different frequencies kHz. While 1 kHz lies close to a local hearing threshold minimum, the
~gray lines, for frequencies see the legend! that is different from the loud- reference at 1.8 kHz matches a threshold maximum. Note that the equal-
ness growth from minima ~black lines!. loudness contours referenced to a tone near a threshold minimum ~black
lines! are closer together, i.e., indicate less loudness compression than the
ones referenced to a tone near a maximum ~gray lines!.
paper, these are called inverse loudness contours. The level
of the reference tone was increased to match the probes,
which were loudest near threshold minima. Figure 1 com-
different for the two frequencies, the two matching strategies
pares results for these two strategies of loudness matching
are identical within the limits of measurement error. The pro-
from the same subject. The pattern of the fine structure is
nounced fine-structure pattern of thresholds flattens out to-
reversed in Fig. 1~b!. The level of the adjusted reference tone
wards equal-loudness contours at higher levels. This leads,
shows maxima at frequencies near threshold minima ~the
obviously, to a difference in loudness compression of tones
level of the reference tone is increased to match the louder
at frequencies of threshold maxima or minima, respectively
probe and reduced near threshold maxima!. However, the
~see also Fig. 5!. At low to moderate levels, the growth of the
width of the fine structure is the same as for the equal-
level adjusted to sound equally loud as the reference tone has
loudness contours @compare, e.g., Figs. 1~a! and ~b!#. To il-
considerably ~loudness-level-growth function! steeper slopes
lustrate the consistency of the data, Fig. 4 shows test-tone
for tones at very sensitive frequencies ~threshold minima!
level as function of the reference tone level extracted from
than for less responsive regions. This is equivalent to a
both the equal loudness contour data from Fig. 1~a! as well
higher loudness compression of the more sensitive frequen-
as from the inverse equal-loudness contours shown in Fig.
cies for which a larger range of levels is mapped to the same
1~b! for one sensitive frequency at 1600 Hz and a more in-
range of loudness than for less sensitive frequencies. To il-
sensitive one at 1950 Hz. While the growth functions are
lustrate the differences of loudness-level-growth functions
from frequencies at threshold maxima and minima, Fig. 5
shows loudness-level-growth functions extracted for fre-
quencies at three different threshold maxima and three
minima for subject GL @Fig. 5~a!# and three different thresh-
old maxima and two minima for subject TB @Fig. 5~b!#.
In Fig. 7, equal-loudness contours for a reference tone at
1 kHz ~fixed in level! are compared with equal-loudness con-
tours obtained with a reference tone at 1.8 khz. While 1 kHz
falls within a threshold minimum of the subject ~GL—
threshold at 1 kHz is 3.25 dB SPL!, 1.8 kHz lies near a
threshold maximum. The equal-loudness contours referenced
to the tone near a threshold minimum ~1 kHz! from the same
subject and the same frequency range, are closer together,i.e.,
indicating less loudness compression than the ones refer-
enced to a tone near a maximum ~1.8 kHz!. Due to the flat-
FIG. 6. Comparison of hearing threshold and equal-loudness contour mea-
surements for two different paradigms. ~a! From subject KW—right ear; ~b! tening of fine structure with increasing stimulus level, a ref-
subject MW—left ear. The black lines with triangles show hearing thresh- erence tone at a threshold maximum needs a smaller range of
olds and equal-loudness contours using the interleaved-adaptive paradigm, stimulus levels for the same change in loudness ~less loud-
while the gray lines with open circles show results from adjustment mea-
ness compression! than a reference tone at a threshold mini-
surements. The shaded areas give the standard deviation from three repeti-
tions. Note the qualitative and quantitative correspondence between the two mum. That means a tone near a threshold minimum shows a
paradigms. higher dynamic range, or larger amount of compression than
FIG. 8. Hearing-threshold fine structures ~top row! for four subjects, from left to right: MO—right ear; TB—left ear; SU—right ear; and RM—right ear. The
panels in the medium row show the associated loudness growth functions from a loudness-matching procedure for an individually selected test tone in a
threshold maximum compared to a neighboring reference tone within a threshold maximum ~black lines and asterisks! or minimum ~gray lines and triangles!,
respectively. The lower panels show the loudness growth functions of a test tone at a threshold minimum in comparison to a reference frequency at a
maximum ~gray lines and asterisks! and a minimum ~black lines and triangles!. The symbols indicate each data point measured. Each loudness match was
measured three times. A least-squares linear regression was computed for each data set. The slope m and its standard deviation for each fitted function is given
in the legend of each panel as indicator for the different loudness growth behavior ~compare Table I!. The asterisk in the top left panel ~subject MO! indicates
the frequency of a SOAE in this subject.
a tone at an adjacent threshold maximum. Therefore, loud- ence tone within the hearing-threshold fine structure. One
ness matching of test signals with a reference tone near a may argue that the effect observed for different reference
threshold minimum ~relatively high compression! will lead tones in subject GL might be mainly influenced by the dif-
to a shallower loudness-level-growth function ~indicating ferent distance of the reference tones to the test tones or due
less compression!, than loudness matching of the identical to an overall difference in the dynamic characteristics of
test tones matched with a reference tone at a frequency close loudness growth around 1 kHz vs 1.8 kHz. Consequently, we
to a threshold maximum ~relative low compression!, as can investigated loudness growth functions for test-and reference
be seen in Fig. 7. tones from different maxima or minima in thresholds which
are closer in frequency. The different character of loudness
IV. EXPERIMENT 2: LOUDNESS GROWTH AT growth is clearly reflected in loudness growth functions at
FREQUENCIES IN THRESHOLD MAXIMA AND single frequencies, although some of the equal-loudness-
MINIMA—LOUDNESS MATCHING contour patterns ~experiment 1! show slight shifts in fre-
The results from the equal-loudness-contour measure- quency with increases in level. This can be seen in Fig. 5 for
ments in experiment 1 indicate different loudness growth two sets of representative growth functions for different test-
with increasing level depending on the position of the refer- tone frequencies extracted from the equal loudness contour
data of experiment 1. Consequently, loudness growth func- size started at 8 dB and was halved at each reversal, ending
tions with level were measured directly in five subjects. An with 2-dB steps. The median of the final four reversals with
interleaved-adaptive loudness-matching procedure was used 2-dB steps was taken as an estimate of the PSE. The mea-
to investigate the different patterns of loudness growth ~using surements were repeated three times with different test-tone
a level resolution of 5 dB! for frequencies near threshold start levels ~in one session the start levels were the same as
maxima and minima. The reference frequency was matched the reference tone level L ref in dB SPL level, and in the two
to two different positions on the individual’s measured fine further sessions a start level of L ref610 dB was used!. The
structure: ~a! to a maximum and ~b! to a minimum. This was order of test- and reference-tone presentation was random-
done to investigate the influence of the different position of ized. A linear function was fitted to all estimates of the PSEs
the reference-tone frequency on the measured loudness for each reference/test-tone condition. These slopes give in-
growth. The loudness growth was evaluated for reference dicators for the growth behavior for the different test-tone/
conditions paired with test-tone frequencies from both a reference-tone conditions.
threshold maximum and minimum.
It is assumed that the loudness growth is almost the
same for pure tones with adjacent frequencies and similar B. Results
thresholds. Therefore, placing both reference and test-tone The slopes of the linear functions fitted to the data indi-
frequencies at either similar pronounced maxima or at simi- cate the loudness growth for the different test-tone/reference-
lar minima in threshold should, ideally, give loudness match- tone conditions. Figure 8 shows the results from four of the
ing ~plotting the adjusted test-tone level as function of the five subjects tested. The results for all five subjects are sum-
reference tone level! which has an almost linear growth close marized in Table I. The slopes when two adjacent sinusoidal
to 1 dB/dB. When the reference tone is near a maximum but tones are compared vary around 1 dB/dB as expected. The
the test-tone frequency falls near a threshold minimum this comparison of either two minima ~mean 0.96, standard de-
should lead to an expansive growth .1 dB/dB. However, viation 0.04! or two maxima ~mean 0.98, standard deviation
when a reference-tone frequency is within a minimum and 0.05! results in slopes quite close to 1 dB/dB, while the
the test tone near a threshold maximum we would expect a comparison of frequencies, one at a maximum and the other
compressive growth with a slope ,1 dB/dB. When probe at a minimum, have slopes that depart from 1 dB/dB.
tones are from both maxima and minima, the loudness For all subjects, the slope of loudness growth at a test-
growth function for a specific test-tone frequency is expected tone frequency chosen to fall near a threshold maximum or
to be steeper for reference tones near threshold maximum minimum is steeper ~indicating less loudness compression!
than for reference tones at a threshold minimum. when the reference tone is near a threshold maximum ~mean
A. Methods slope 1.11, standard deviation 0.08! than when the reference
1. Subjects tone is near a threshold minimum, i.e., a more sensitive place
~mean 0.82, standard deviation 0.05!. Thus, the results of
Four normal-hearing subjects ~TB, MO, RM, SU! with loudness-matching paradigms using single sinusoids as ref-
thresholds of 15 dB HL or better in the clinical audiogram erence stimuli are influenced by the position of this tone
~125 Hz– 8 KHz! and one subject ~DS! with 25 dB HL at 3 within threshold fine structure. The differences in loudness
and 4 kHz participated in this study. Subjects TB and MO growth for a specific test frequency referenced to either a
also participated in experiment 1.3 frequency within a threshold minimum or maximum, respec-
tively, ranged from 0.13 to 0.30 dB/dB. Only subject DS
2. Procedures ~who has a slight hearing loss at 3 and 4 kHz! showed
a. Hearing thresholds. The same adaptive paradigm de- smaller slope differences but his data followed the same
scribed in experiment 1 was used for measuring hearing trend. The position of the reference tone within the fine
thresholds. Two frequency ranges from 1600–1800 and structure and the position of the test tone are both important
1812.5–2012.5 Hz were each measured as 17 interleaved- for the estimates of loudness growth as determined by loud-
frequency tracks with a frequency resolution of 12.5 Hz. ness matching. For a fixed reference tone, the measured
Subjects MO and DS were also tested from 1387.5 to 1587.5 loudness growth of two adjacent test-tone frequencies is dif-
Hz to scan for regions with more pronounced fine structure. ferent if one is near a threshold minimum and the other
b. Loudness growth functions. The stimuli were identical around a maximum. That means results on loudness growth
to the stimuli used in the interleaved-adaptive equal- obtained with loudness-matching paradigms using sinusoidal
loudness-contour experiments. The frequencies of individual test signals may be influenced by the position of the test tone
threshold maxima and minima, as well as the appropriate SL in threshold fine structure. When the test tone is fixed in
conversion, were determined from the hearing threshold level, the observed dynamic in loudness growth, dependin-
measurements. Loudness matching with a reference tone gon the reference or test tone frequencies, is reversed @com-
near a threshold maximum or a minimum was paired with a pare Figs. 1~a! and ~b!#.
test tone close to a maximum and a test tone near a mini- The cochlear fine structure is unique to each individual.
mum. This gave four reference/test-tone conditions. Eleven This means that the same reference/test tone may fall near a
different reference levels ~5 to 55 dB SL in 5-dB steps! were minimum in one subject, while in another subject it will fall
presented using an interleaved-adaptive 2 AFC 1-up, 1-down near a threshold maximum. This may—in part—explain the
procedure as previously described for experiment 1. The step intersubject differences in loudness growth in similar psy-
TABLE I. Frequencies of test- and reference tones and individual hearing thresholds for frequencies used in experiment 2. The next to last column shows the
slopes of loudness growth for each test-tone frequency when compared with a reference tone at a threshold minimum or maximum, respectively. The last
column shows the difference of loudness growth for a test tone when compared either to a reference at a threshold minimum or maximum.
Frequency Hearing thres. of Frequency Hearing thres. of

Subject of test tone test tone of ref. tone ref. tone Slope m Difference of
side in Hz in dB SPL in Hz in dB SPL in dB/dB slopes
RM test Min. 1875 9.67 ref. Min. 1675 7.75 0.94Á0.04 0.1660.08
right ref. Max. 1625 15.50 1.1Á0.04
test Max. 1825 18.5 ref. Min. 1675 7.75 0.78Á0.04 0.1660.07
ref. Max. 1625 15.50 0.94Á0.03
TB test Min. 1875 3.75 ref. Min. 1725 1.75 1Á0.05 0.2460.11
left ref. Max. 1662.5 16.00 1.24Á0.06
test Max. 1775 14 ref. Min. 1725 1.75 0.76Á0.04 0.3060.09
ref. Max. 1662.5 16.00 1.06Á0.05
SU test Min. 1975 7.33 ref. Min. 1725 7.00 0.93Á0.05 0.2460.1
right ref. Max. 1600 13.80 1.17Á0.05
test Max. 1887.5 14 ref. Min. 1725 7.00 0.85Á0.04 0.1460.07
ref. Max. 1600 13.80 0.99Á0.03
MO test Min. 1612.5 5.17 ref. Min. 1550 6.33 0.92Á0.04 0.1360.07
right ref. Max. 1587.5 16.50 1.05Á0.03
test Max. 1525 14.83 ref. Min. 1550 6.33 0.87Á0.03 0.1660.06
ref. Max. 1587.5 16.50 1.03Á0.03
DS test Min. 1612.5 4.33 ref. Min. 1512.5 3.33 1.01Á0.05 0.0560.08
left ref. Max. 1462.5 11.66 1.06Á0.03
test Max. 1650 14.5 ref. Min. 1512.5 3.33 0.85Á0.03 0.0960.07
ref. Max. 1462.5 11.66 0.94Á0.04
choacoustical experiments done at levels near the absolute sion!. Stimuli were sinusoids of 500-ms duration including
threshold. 50-ms Hanning-shaped ramps. The measurements from four
individually selected frequencies, two from maxima and two
V. EXPERIMENT 3: LOUDNESS GROWTH AT from minima of the threshold fine structure ~the same as for
FREQUENCIES IN THRESHOLD MAXIMA AND the loudness growth measurements from loudness matching
MINIMA—CATEGORICAL LOUDNESS SCALING described in experiment 2! were interleaved. Two different
Loudness growth functions were measured using a level ranges were used: ~a! Stimuli from a subject indepen-
loudness-scaling paradigm to investigate the effects of co- dent SL range ~in 2-dB steps! were presented randomly for
chlear fine structure when measured with a direct scaling each frequency investigated in a subject. To control for pos-
technique in addition to the loudness matching procedures sible effects of stimulus range due to differences in thresh-
used in experiments 1 and 2. olds ~b! Stimuli with identical SPLs were presented ran-
domly for all frequencies for each subject. Where
A. Methods appropriate, the specific ranges used were adjusted depend-
1. Subjects ing on the subject’s thresholds at that frequency ~the level
Same as in experiment 2. ranges used are presented in Table II!. Each measurement
was repeated in three sessions in which all levels were
2. Procedures shifted by 60.5 dB from above levels to aid smoothing func-
tions to the data.
a. Categorical loudness scaling. A two-step loudness
Since categorical scaling is not necessarily expected to
scaling procedure was implemented, which is similar to the
provide equal interval scales, the assumptions underlying a
one proposed by Heller ~1985! and Hellbrück and Moser
parametric fit to the data may be violated. Therefore, a robust
~1985!. In the first step, the subject had to choose a response
locally weighted scatter plot smooth ~robust lowess—
alternative out of the verbal categories ‘‘very soft,’’ ‘‘soft,’’
‘‘medium,’’ ‘‘loud,’’ ‘‘too loud,’’ or ‘‘inaudible’’ after hearing
the stimulus. In the second step, the subject had to refine TABLE II. Individual level ranges used for the categorical loudness scaling
his/her judgment using a finer scale by using numbers around in experiment 3.
the previously chosen category @very soft ~1–10!, soft ~11–
20!, medium ~21–30!, loud ~31– 40!, too loud ~41–50!#. Us- SL range SPL range SPL range
Subject in dB SL in dB SPL in dB SL
ing this procedure, loudness is mapped by the subjects to a
numerical scale from 0 ~inaudible! to 50. We refer to these MO 24 to 56 24.83 to 71.17 210 to 66
numbers as categorical units ~cu! ~Brand and Hohmann, RM 28 to 72 20.25 to 89.75 28 to 82
SU 28 to 72 21.00 to 85.00 28 to 78
2002!.4 The ‘‘cu’’ are directly used as loudness indicators for
TB 26 to 74 26.25 to 87.75 28 to 86
further analysis ~i.e., in contrast to the loudness-level func- DS 28 to 72 24.67 to 85.33 28 to 82
tions, a shallower slope indicates higher loudness compres-
FIG. 9. Loudness growth functions from categorical loudness scaling mea-
surements for two subjects MO—right ear, panel ~a! and ~c!; TB—left ear
~b! and ~d!. The upper panels ~a! and ~b! show the results when the stimuli
covered the same SL range for each frequency, while the lower panels ~c! FIG. 10. Each panel shows the loudness growth functions obtained by a
and ~d! give the results, from the same subjects, when the stimuli covered robust ‘‘lowess’’ procedure from the categorical loudness scaling data for
the identical SPL range for all frequencies. The symbols indicate the scaled frequencies at two individual fine-structure maxima ~thick gray lines! and
points from three measurement sessions ~asterisks for measurements within minima ~thin black lines!. From top to bottom: subject MO—right ear; sub-
a threshold minimum, triangles at a frequency close to a threshold maxi- ject RM—left ear; subject SU—right ear; and subject TB—left ear. Only the
mum!. The lines are smoothed functions obtained from a robust ‘‘lowess’’ curves for the SL-range measurements are shown. The SPL-range measure-
procedure. Note the different growth behavior at low stimulus levels for all ments have very similar properties.
subjects and level ranges depending on whether the frequency of the scaled
tone lies near a threshold minimum ~black lines! or maximum ~gray lines!.
The asterisk in the top left panel ~subject MO! indicates the frequency of a The scaling data obtained from three measurements ~slightly
SOAE in this subject; the adjacent label indicates the level of this emission shifted in stimulus levels! are fitted by a robust locally
measured in the sealed ear canal. weighted scatter plot smoothing procedure. The top panels in
Fig. 9 show the data when the range of stimulus levels was
MATLAB 6.5, Curve Fitting Toolbox! was used to extract the based on SL, while the lower panels show the results for the
trends from the categorical scaling data ~e.g., Cleveland, SPL range. The data from the two level ranges show a very
1979!. This nonparametric locally weighted regression similar pattern.
method computes linear least-squares regression of a first- All loudness functions show a more compressive region
degree polynomial to the locally weighted data points.5 The from about 30 to 60 dB SPL, and a steeper growth for higher
whole range of presented levels was used in the smoothing levels. Although the loudness curves for adjacent frequencies
procedure. tend to converge at higher levels ~up to at least 20 dB SPL in
The presented levels were extended to negative SL to subject SU, up to 40 dB SPL in subject TB!, frequencies near
allow loudness judgments of 0 cu, i.e., ‘‘not audible.’’ Since threshold maxima had steeper loudness growth ~see Fig. 10!.
a threshold estimated by a 3-AFC 1-up, 2-down procedure Overall, the results are consistent with the results from ex-
gives the level of a tone heard 70.7% of the time, occasion- periments 1 and 2. The curves from threshold maxima and
ally tones below the threshold estimate will be heard and will minima converge at levels which are similar to the levels at
have a loudness greater than 0 cu. which the fine structure of the equal loudness contours flatten
out. In almost all cases, a categorical loudness of 1 cu ~which
B. Results is always quite close to hearing threshold! for frequencies
associated with threshold maxima or minima is clearly dif-
Due to technical restrictions ~maximum level of the
ferent. The curves for the different frequencies converge at
ER2! the loudness could not be scaled over the whole dy-
around 30 dB SPL. Therefore, the loudness-scaling data also
namic range of each subject. This may bias the shape of the
show different slopes of loudness growth at low levels for
perceived loudness growth functions. However, a ‘‘true’’
frequencies at threshold maxima or minima.
loudness function was not the aim of this study; the aim was
to investigate differences in adjacent frequencies falling ei-
VI. DISCUSSION
ther near a minimum or near a maximum of cochlear fine
structure. There could have been range effects modifying the Different paradigms were used to investigate how much
differences near threshold. If there was any major effect it cochlear fine structure typically affects loudness perception.
should have been observable as a difference in the results Three consistent effects were found: ~a! The fine structure of
from the two measured level ranges. We found no such dif- equal-loudness contours flattens out at reference-tone levels
ference. around 30– 40 dB SPL; ~b! the slope of loudness growth
Figure 9 shows typical scatter plots and fitted curves differs up to 0.3 dB/dB for reference tones in maxima versus
obtained by loudness scaling of tones from a threshold maxi- adjacent threshold minimum; ~c! loudness curves from cat-
mum and a threshold minimum for two subjects ~MO, TB!. egorical loudness scaling converge at levels of about 30 dB
SPL but show different slopes at lower levels. Overall, it can matching experiment, the slopes for test tones from maxima
be concluded that cochlear fine structure affects loudness differ on average by about 1.4 dB/dB from the slopes for
perception of sinusoids up to levels in the order of 40 dB tones from minima matched to the same reference tone. This
SPL ~for the frequency range investigated here!. The aim of indicates 6 dB less compression at threshold maxima com-
this study was not to obtain exact loudness functions but to pared to threshold minima over the range of 40 dB relevant
evaluate differences in loudness perception stemming from for fine structure.
cochlear fine structure. Consequently, we did not try to avoid The absolute slopes of the loudness-matching functions
all known bias effects, e.g., in the loudness-scaling measure- depend on the position of the reference tone within the fine
ments, we did not test the whole dynamic range of the sub- structure. If the reference tone is fixed to a frequency close to
jects. However, the differences in loudness perception from a minimum ~more compressive site!, loudness matching will
minima and maxima in the fine structure were reliable. Bias lead to shallower slopes in the loudness-growth function
effects, which may influence the differences with frequency, (L test /L ref representation! than matches to a reference tone at
were avoided, e.g., range effects in the categorical loudness a threshold maximum ~less compressive site!. This means
scaling due to different minimal levels for the different fre- that differences in loudness-matching functions stemming
quencies. One consequence of the loudness fine structure is from the choice of reference-tone position within threshold
that the slope of loudness growth at low levels depends on fine structure, most probably, depends on differences in com-
whether the reference tone lays within a threshold minimum pression at threshold maxima or minima. For example, Fig. 5
or at a threshold maximum. Although this effect was only shows the loudness-growth function extracted from the fine-
investigated for sinusoidal test tones, it is reasonable to as- structure measurements of subjects GL @Fig. ~5a!# and TB
sume that this is of relevance for all other kinds of test sig- @Fig. ~5b!#. Threshold measurements around 1 kHz from sub-
nals whenever a sinusoid is used as reference tone. The fine ject GL indicate that the reference tone at 1 kHz is located
structure can influence the results of loudness matches be- near a minimum leading to gradual slopes at low levels. The
tween different reference frequencies ~a! within one subject steeper slopes in subject TB indicate that the reference tone
and ~b! between subjects since the fine structure is unique to at 1 kHz is most probably placed close to a threshold maxi-
each individual. mum in this subject.
Since the isophons, or equal-loudness contours, flatten Several authors have suggested a link between basilar-
out at about 40 phons, a larger dynamic range has to be membrane compression and loudness. Florentine et al.
mapped to the same loudness range for stimuli with frequen- ~1996!, for instance, have derived a loudness function from
cies near threshold minima than for stimuli with frequencies loudness temporal integration data which shows similar char-
near threshold maxima. That means that loudness compres- acteristics to basilar-membrane measurements ~e.g., Ruggero
sion is greater for stimuli in threshold minima than stimuli et al., 1997!. These functions are relatively linear near
related to threshold maxima. This is consistent with the re- threshold and more compressive at moderate levels. The dif-
sults for low-level stimuli in both loudness matching ~experi- ferent loudness compression at low levels for frequencies
ment 2! and categorical loudness scaling ~experiment 3!. related to threshold maxima or minima, respectively, is also
Methodological differences between the two experiments most probably closely related to BM compression. A class of
mean that there are differences in the patterns of loudness cochlea models, which explain a wide range of fine-structure
growth. In the categorical loudness-scaling procedure, the effects of otoacoustic emissions, can also explain fine struc-
subject is evaluating each tone in isolation, providing a direct ture in hearing threshold ~Talmadge et al., 1998!. In this kind
estimate of subjective loudness. However, in the loudness- of model, the threshold fine structure is obtained by a local
matching experiment, the subject is comparing the loudness enhancement or reduction of BM excitation due to construc-
of two sets of stimuli, and the slope of loudness growth tive or destructive interference of the initial traveling wave
depends on the position of both the test and reference tones and multiple reflections between the characteristic site of the
within threshold fine structure. When the reference tone is stimulus frequency and the oval window. In the case of con-
fixed in frequency ~e.g., at a threshold minimum! and the structive interference, the BM excitation will be enhanced
data are plotted as a function of the reference-tone level around the characteristic site for the stimulus frequency and
~loudness-growth function!, the different compression for so leads to a reduced threshold. The magnitude of the initial
varying test-tone frequencies ~e.g., one close to a threshold and subsequent reflections depends on local properties
maximum while the other is close to threshold minimum! is ~roughness! on the basilar membrane. Therefore, the size of
indicated by distinct slopes of the matching functions. Com- the reflection depends on the basilar-membrane input/output
pare the black lines ~reference tone at a maximum! or gray function. We assume that the BM compression is ~a! a func-
lines ~reference tone at a minimum!, respectively, in the sec- tion of BM excitation amplitude and the same over a suffi-
ond and third rows of Fig. 8. The matching function is ciently wide frequency range and ~b! close to linear at low
steeper, i.e., indicating more compression, if the same BM excitation and compressive at moderate BM excitation
reference-tone frequency is matched with a test tone in a ~as suggested from physiological data, e.g., Ruggero et al.,
minimum than if it is matched with a test tone in a maximum 1997!. Under these assumptions, there will be a difference in
~see also Table I!. Similarly, the results from the categorical the magnitude of vibration for the same stimulus levels at
loudness scaling at low levels for frequencies close to thresh- BM sites with in-phase reflections and out-of-phase reflec-
old minima are more compressive than for frequencies close tions. Consequently, sites with in-phase reflections will be
to threshold maxima ~Figs. 9 and 10!. In the loudness- affected by BM compression at lower stimulus levels than
BM sites with reduced BM excitation due to interference. more important ~2! insufficient time to build up a stable in-
BM sites with reduced excitation ~threshold maxima! are terference pattern within the cochlea. Cohen questioned
kept in the more linear range of BM characteristics for a whether this effect of temporal integration at threshold holds
wider range of stimulus levels and, therefore, show less loud- for higher stimulus levels. Due to the preservation of fine
ness compression. In this model, the different loudness com- structure in equal-loudness contours up to at least 40 dB SPL
pression for low to moderate levels at frequencies from ~see experiment 1!, it is probable that the influence of co-
threshold maxima and minima is closely, but not directly, chlear fine structure on temporal integration will hold for
linked to the BM compression. The locally different loudness levels up to 40 dB SPL, at least for sinusoidal signals.
compression is not the result of locally different BM com- Zwicker ~1986! found a negative correlation between the
pression but caused by a shift of the operating point due to level of hearing thresholds of the carrier frequency and the
internal cochlea reflections. just-noticeable degree of amplitude modulation ~JNDAM!
This kind of cochlea model also explains most of the prop- using modulation frequencies of 1, 4, 16, and 64 Hz. These
erties of SOAE ~Talmadge et al., 1998; Shera, 2003! and are two examples of the influence of cochlear fine structure
their close relation to threshold fine structure ~Talmadge on psychoacoustical measurements in addition to loudness
et al., 1998!. This relation can also be seen from experimen- measurements. However, the majority of psychoacoustical
tal comparisons of threshold fine structure and SOAE fre- studies has not considered potential effects of cochlear fine
quencies ~e.g., Zwicker and Schloth, 1984; Long and Tubis, structure even when measuring loudness at low levels and
1998a, b!. Independent SOAE always occur at threshold when using sinusoidal signals. There are only a few studies
minima ~there are some SOAE that are distortion products of which partially regard possible effects. For example, Buus
other SOAE, and there may be no minima at these frequen- et al. ~1998! measured loudness of tone complexes at low
cies!. Large evoked OAE are seen at every threshold mini- levels in comparison to a sinusoidal reference. Therefore,
mum. At frequencies where the multiple reflections between they selected complex components individually to avoid fre-
characteristic BM sites and the oval window interfere con- quency components within a pronounced threshold maxi-
structively and lead to an enhancement of BM excitation, the mum or minimum, respectively. All components were ad-
resulting standing wave in the cochlea is preserved by the justed to individual sensation levels ~SL!. Reckhardt et al.
amplification of the active cochlea mechanisms and stabi- ~1999! measured loudness matches at low frequencies from
lized by BM compression. When a fraction of the energy of 200 Hz up to 1 kHz in comparison to a 1-kHz reference tone
the standing wave is transmitted through the middle ear, it at 30 and 50 dB SPL. They found nearly half the interindi-
can be detected as a SOAE. Consequently, frequencies of vidual variation in equal-loudness matches when the indi-
enhanced BM excitation are directly linked to SOAE fre- vidual hearing threshold was taken into account. Corrections
quencies ~for a detailed description of SOAE mechanism see, to the SL may compensate for fine structure of low-level
e.g., Shera, 2003!. Several studies investigating the relation loudness perception. Based on Reckhardt et al. ~1999!, it is
of SOAE and fine structure showed that the interaction of a even possible that the fine structure below 1 kHz is preserved
strong SOAE and an external tone can sometimes be per- up to higher SPLs than in the frequency region observed in
ceived as beating ~discussed, e.g., in Long, 1998!, which the current study. However, a simple SL correction will lead
may give additional cues for signal detection at threshold. to an overcompensation for higher stimulus levels when fine
When the external tone is close to the frequency of the emis- structure in loudness perception flattens out ~see experiment
sion, no beating is detected because the SOAE becomes en- 1!.
trained by the external tone and oscillates at the same fre- Sinusoids, such as those used in the current study, are a
quency and phase ~reviewed in Long, 1998!. High-level very special type of stimuli. However, they are well defined
emissions are only entrained by tones very close to the and for that reason often used for technical acoustic measure-
SOAE, while low-level emissions are more easily entrained. ments, in audiology ~e.g., tone audiogram!, and in many psy-
Under these conditions, the acoustic stimulation in the co- choacoustic experiments. The question arises whether co-
chlea will be the in-phase sum of the external tone and the chlear fine structure influences only the perception of this
SOAE. Consequently, with fixed stimulus level, the stimula- very special type of stimulus, or if it also affects the percep-
tion on the basilar membrane will be greater at places near tion of a wider range of signals. Long and Tubis ~1988a!
SOAE than at other places. We do not know how large the found that narrow-band noise stimuli had little effect on
stimulus is in the cochlea, but previous research has shown threshold fine structure until the bandwidth reached the
that little beating is detected at the levels of SOAE observed bandwidth of the fine structure ~in their study, 100 Hz!. They
in this study ~Long, 1998!. Consequently, we assume that observed an overall flattening out of threshold fine structure
effects of beating with SOAE have minimal effect on the for signals of increasing bandwidth, leading to increased
threshold and loudness fine structure observed in this paper. thresholds near threshold fine-structure minima and de-
Psychoacoustic tasks other than loudness measurement creased thresholds for stimuli near threshold maxima. In
are also influenced by fine structure. Cohen ~1982! showed these circumstances, the threshold is probably due to integra-
the temporal integration function at threshold to be consid- tion of neural activity across the bandwidth of the stimuli
erably steeper for more sensitive frequencies, i.e., a fading and, thus, may provide an indication of growth of loudness
out of threshold fine structure for shorter stimuli. The fine uncontaminated by threshold fine structure. Using narrow-
structure for short sinusoidal signals probably fades out for band noises, instead of sinusoids, as test and reference
two reasons: ~1! Spectral smearing for short stimuli and even stimuli in the measurement of equal-loudness contours
would, most probably, decrease intersubject variability. jects and to find methods for early diagnosis of incipient
There are multiple effects influencing the variability of cochlear damage. Such methods might be based on psychoa-
equal-loudness contours between studies, such as measure- coustic experiments or on OAE measurements. Before the
ment room and choice of test-tone levels ~e.g., Gabriel, 1996; properties of cochlear fine structure can serve as an early
1997!. The individual fine structure most probably has only a indicator of hearing loss, further research on cochlear fine
small effect on the differences. Since most studies investi- structure and its effects on perception is necessary.
gated about ten or more subjects ~e.g., Betke and Mellert,
1989; Suzuki et al., 1989; Fastl et al., 1990; Poulsen and VII. SUMMARY
Thøgersen, 1994; Takeshima et al., 1997!, the individual fine Fine structure of hearing thresholds and loudness per-
structure should be averaged out in the mean data.6 Since ception was investigated in detail for frequencies around
most studies on equal-loudness contours are performed in a 1800 Hz, using different measurement paradigms. The fol-
free field, the individual head-related transfer functions lowing experiments were carried out: measurements of iso-
~HRTF! will probably determine more of the across-subjects phons with a high-frequency resolution ~experiment 1!, mea-
variability than threshold fine structure. surement of loudness growth functions at frequencies either
Fine structure is possibly only a side effect of cochlea around a threshold maximum or minimum, using a loudness-
mechanics, but the existence of more sensitive resonance matching paradigm ~experiment 2!, and categorical loudness
points may even provide some gain for near-threshold sig- scaling ~experiment 3!. In all experiments, the results are
nals with broader bandwidth. Fine structure might be useful affected by the position of the frequency within threshold
as an indicator of a healthy cochlea whether or not it leads to fine structure for levels up to 40 dB SPL. These fine-
any gain in hearing, influences our perception significantly, structure variations in threshold and loudness perception for
or is a minor side effect. The fine structure in the so-called adjacent frequencies are probably one reason that there is
‘‘normal’’-hearing subjects in the current study is variable. significant intersubject variability in several psychoacoustic
While in some subjects it is pronounced, in others it was experiments on loudness at low to moderate levels, e.g., this
hard to measure. Therefore, the question arises whether the fine structure influences loudness-matching paradigms when
absence of fine structure may indicate the beginning of co- using sinusoids either as reference or test signals.
chlear damage, or conversely, a pronounced fine structure Most probably, fine structure in hearing thresholds and
may be an early sign of cochlea damage. There are some loudness perception is caused by interference effects of in-
indications that cochlear fine structure is a property of a coming and reflected traveling waves within the cochlea
healthy ear. For example, fine structure is very sensitive to closely linked to the mechanisms responsible for the fine
cochlea insult. DPOAE fine structure reappears at a very late structure observed in otoacoustic emissions.
stage of recovery after a sudden hearing loss ~Mauermann While fine-structure effects are observable in most of the
et al., 1999!. Ototoxic aspirin consumption leads to a loss of normal-hearing subjects, there are some listeners with no
fine structure in OAE as well as in threshold fine structure. pathological findings who do not show noteworthy fine
Shortly after beginning aspirin consumption, there is an im- structure. Therefore, in future studies it might be valuable to
provement of thresholds near threshold maxima while investigate whether the presence of fine structure indicates a
thresholds associated with threshold minima are only el- very healthy ear or, perhaps, initial damage already observ-
evated following 3 or 4 days of aspirin consumption ~Long able in most normal-hearing adults.
and Tubis, 1988a, b!. Although McFadden and Platsmier
~1984! claimed that there was no consistent trend in thresh- ACKNOWLEDGMENTS
olds with aspirin consumption, threshold shift due to aspirin
consumption was negatively correlated with the initial This study was supported by Deutsche Forschungsge-
meinschaft, DFG Ko 942/11-3.
thresholds ~a rough estimate of the position of the tones in
A fellowship from Hanse-Wissenschaftskolleg, Delmenhorst
the cochlear fine structure!. Similar effects can be observed
partially supported the second author while on sabbatical in
for DPOAE fine structure in ears with noise-induced tempo-
Germany. We gratefully acknowledge Stefan Uppenkamp,
rary threshold shift ~Furst et al., 1992; Engdahl and Kemp,
Gerald Kidd and two anonymous reviewers for the valuable
1996!. Overall, these effects are in agreement with results
comments on the manuscript. We want to thank Monica
from cochlea modeling ~e.g., Talmadge et al., 1998! which
Wagner for proofreading the manuscript.
indicate that damage affecting the cochlear amplifier will
cause a reduction in fine structure. On the other hand, they 1
In the literature, the term ‘‘threshold microstructure’’ is often used instead
also say that the model predicts damage, which does not of fine structure. For reason of convenience, here we consistently use the
directly affect the cochlear amplifier but which causes an term fine structure.
2
enhanced roughness in the mechanical parameters of the co- This was done because subject MO also participated in a different study for
chlea. Such damage would result in a more pronounced fine which a frequency region with a sufficient pronounced fine structure was
needed.
structure. The amount of fine structure differs between indi- 3
The results for subject TB from study 1 are about 2 years older than from
viduals, and the differences most probably depend on the study 2/3, i.e., the thresholds measured with the simple adjustment para-
health of the cochlea, the properties of the basilar membrane, digm in study 1 and the adaptive interleaved in study 2/3 are not necessarily
identical.
and the condition of the middle ear. The high sensitivity of 4
Brand and Hohmann used a one-step procedure with 11 categories mapped
fine structure to cochlear damage may offer the opportunity to the same range of numbers ~0–50!.
5
to further categorize the group of the ‘‘normal-hearing’’ sub- The initial weights w i for the adjacent data points within the span are given
by the tricube function w i 5(12 u (x2x i )/d(x) u 3 ) 3 , where x is the predictor Kapadia, S., and Lutman, M. E. ~1999!. ‘‘Reduced-audiogram ripple in nor-
value associated with the response value to be smoothed, the x i are the mally hearing subjects with weak otoacoustic emissions,’’ Audiology 38,
adjacent points to x within the selected span of data points around x, d(x) 257–261.
gives the distances between x and the most distant predictor value within Kemp, D. T. ~1979!. ‘‘The evoked cochlear mechanical response and audi-
the span. The data point to be smoothed has the largest weight, i.e., the tory microstructure—Evidence for a new element in cochlear mechanics,’’
most influence on the fit, while data points outside a selected span have Scand. Audiol. Suppl. 9, 35– 47.
zero weight, i.e., no influence on the fit. For the current data an empirically Kemp, D. T. ~1980!. ‘‘Towards a model for the origin of cochlear echoes,’’
selected span of 31 data points was used to receive a smoothing function Hear. Res. 2, 533–548.
that keep details of the data while avoiding an oscillating curve. To reduce Long, G. R. ~1984!. ‘‘The microstructure of quiet and masked thresholds,’’
the influence of outliers for the robust lowess procedure, the residuals are Hear. Res. 15, 73– 87.
computed from the initially obtained smooth function as described above. Long, G. R. ~1993!. ‘‘Perceptual consequences of otoacoustic emissions,’’ in
Then, the residuals are used to compute robust weights for each data point Contributions to Psychological Acoustics: Results of the 6th Oldenburg
in the span. The robust weights are given by wi Symposium on Psychological Acoustics, edited by A. Schick ~University of
(12(r i /6MAD) 2 ) 2 if u r i u ,6MAD Oldenburg Press, Oldenburg, Germany!, pp. 59– 80.
5$ 0 if u r u >6MAD
, where r i is the residual of the ith data point
i Long, G. R. ~1998!. ‘‘Perceptual consequences of the interactions between
produced by the initial smoothing procedure, and MAD5median( u r u ) is the spontaneous otoacoustic emissions and external tones. I. Monaural dipla-
median absolute deviation. If r i is greater than 6 MAD, the robust weight is cusis and aftertones,’’ Hear. Res. 119, 49– 60.
0; if the residual r i is small compared to 6 MAD, the robust weight is close Long, G. R., and Tubis, A. ~1988a!. ‘‘Investigations into the nature of the
to 1. The data are smoothed again using both the local regression weight association between threshold microstructure and otoacoustic emissions,’’
and the robust weight for computation. Again, the residuals from the Hear. Res. 36, 125–138.
smoothed function to the original data are calculated and new robust Long, G. R., and Tubis, A. ~1988b!. ‘‘Modification of spontaneous and
weights are derived. This procedure is repeated for a total of five iterations evoked otoacoustic emissions and associated psychoacoustic microstruc-
to obtain the final smoothed function. ture by aspirin consumption,’’ J. Acoust. Soc. Am. 84, 1343–1353.
6
If we assume amplitude variations due to fine structure of 67 dB, and ~as Mauermann, M., Uppenkamp, S., van Hengel, P. W. J., and Kollmeier, B.
a raw guess! that the amplitude of averaged fine structures from different ~1999!. ‘‘Evidence for the distortion product frequency place as a source
subjects goes down with 1/AN ~similar to noise!, and if we have a group of of distortion product otoacoustic emission ~DPOAE! fine structure in hu-
ten subjects, then the variation due to fine structure will be approximately mans. II. Fine structure for different shapes of cochlear hearing loss,’’ J.
62 dB, i.e., on the order of intraindividual variation of equal-loudness Acoust. Soc. Am. 106, 3484 –3491.
contour measurements ~see Gabriel, 1996! Mauermann, M., Uppenkamp, S., and Kollmeier, B. ~1997!. ‘‘Periodizität
und Pegelabhängigkeit der spektralen Feinstruktur von
Verzerrungsprodukt-Emissionen @Periodicity and dependence on level of
Betke, K., and Mellert, V. ~1989!. ‘‘New measurements of equal-loudness the distortion product otoacoustic emission spectral fine-structure#,’’ Au-
level contours,’’ in Proceedings of Inter-noise 89, 793–796. diol. Akustik 36, 92–104.
Brand, T., and Hohmann, V. ~2002!. ‘‘An adaptive procedure for categorical McFadden, D., and Plattsmier, H. S. ~1984!. ‘‘Apirin abolishes spontaneous
loudness scaling,’’ J. Acoust. Soc. Am. 112, 1597–1604. oto-acoustic emissions,’’ J. Acoust. Soc. Am. 76, 443– 448.
Buus, S., Musch, H., and Florentine, M. ~1998!. ‘‘On loudness at threshold,’’ Poulsen, T., and Thørgersen, L. ~1994!. ‘‘Hearing threshold and equal loud-
J. Acoust. Soc. Am. 104, 399– 410. ness contours in a free sound field for pure tones from 1 kHz to 16 kHz,’’
Cleveland, W. S. ~1979!. ‘‘Robust locally weighted regression and smooth- in Proc. Nordic Acoust. Meeting, 195–198.
ing scatterplots,’’ J. Am. Stat. Assoc. 74, 829– 836. Reckhardt, C., Mellert, V., and Kollmeier, B. ~1999!. ‘‘Factors influencing
Cohen, M. F. ~1982!. ‘‘Detection threshold microstructure and its effect on equal-loudness contours,’’ in Psychophysics, Physiology and Models of
temporal integration data,’’ J. Acoust. Soc. Am. 71, 1719–1733.
Hearing, edited by T. Dau, V. Hohmann, and B. Kollmeier ~World Scien-
Elliot, E. ~1958!. ‘‘A ripple effect in the audiogram,’’ Nature ~London! 81,
tific, Singapore!, pp. 113–116.
1076.
Ruggero, M. A., Rich, N. C., Recio, A., Narayan, S. S., and Robles, L.
Engdahl, B., and Kemp, D. T. ~1996!. ‘‘The effect of noise exposure on the
~1997!. ‘‘Basilar-membrane responses to tones at the base of the chinchilla
details of distortion product otoacoustic emissions in humans,’’ J. Acoust.
cochlea,’’ J. Acoust. Soc. Am. 101, 2151–2163.
Soc. Am. 99, 1573–1587.
Shera, C. A. ~2003!. ‘‘Mammalian spontaneous otoacoustic emissions are
Fastl, H., Jaroszewski, A., and Zwicker, E. ~1990!. ‘‘Equal loudness con-
amplitude-stabilized cochlear standing waves,’’ J. Acoust. Soc. Am. 114,
tours between 100 and 1000 Hz for 30, 50, and 70 phon,’’ Acustica 70,
244 –262.
197–201.
Florentine, M., Buus, S., and Poulsen, T. ~1996!. ‘‘Temporal integration of Shera, C. A., and Zweig, G. ~1993!. ‘‘Order from chaos: Resolving the
loudness as a function of level,’’ J. Acoust. Soc. Am. 99, 1633–1644. paradox of periodicity in evoked otoacoustic emissions,’’ in Biophysics of
Furst, M., Reshef, I., and Attias, J. ~1992!. ‘‘Manifestations of intense noise Hair Cell Sensory Systems, edited by H. Duifhuis, J. W. Horst, P. van Dijk,
stimulation of spontaneous otoacoustic emissions and threshold micro- and S. M. van Netten ~World Scientific, Singapore!, pp. 54 – 63.
structure: Experiment and model,’’ J. Acoust. Soc. Am. 91, 1003–1014. Smurzynski, J., and Probst, R. ~1998!. ‘‘The influence of disappearing and
Gabriel, B. ~1996!. ‘‘Equal-loudness Level Contours: Procedures, Factors reappearing spontaneous otoacoustic emissions on one subject’s threshold
and Models,’’ Ph.D. thesis, University of Oldenburg, Germany. microstructure,’’ Hear. Res. 115, 197–205.
Gabriel, B., Kollmeier, B., and Mellert, V. ~1997!. ‘‘Influence of individual Suzuki, S., Suzuki, Y., Kono, S., Sone, T., Kumagai, M., Mirura, H., and
listener, measurement room, and choice of test-tone levels on the shape of Kado, H. ~1989!. ‘‘Equal-loudness level contours for pure tone under free
equal-loudness level contours,’’ Acta Acust. ~Beijing! 83, 670– 684. field listening condition ~I!—Some data and considerations on experimen-
He, N.-J. ~1990!. ‘‘Frequency shift in spontaneous otoacoustic emission and tal conditions,’’ J. Acoust. Soc. Jpn. ~E! 10, 329–338.
threshold fine structure,’’ Ph.D. thesis, University of Iowa. Takeshima, H., Suzuki, Y., Kumagai, M., Sone, T., Fujimori, T., and Mirura,
He, N.-J., and Schmiedt, R. A. ~1993!. ‘‘Fine structure of the 2 f 1 – f 2 acous- H. ~1997!. ‘‘Equal loudness level measured with the method of constant
tic distortion product: Changes with primary level,’’ J. Acoust. Soc. Am. stimuli—Equal loudness contours for pure tone under free-field listening
94, 2659–2669. condition ~II!,’’ J. Acoust. Soc. Jpn. ~E! 18, 337–340.
Hellbruck, J., and Moser, L. M. ~1985!. ‘‘Hörgeräte-Audiometrie: Ein com- Talmadge, C., Tubis, A., Long, G. R., and Piskorski, P. ~1998!. ‘‘Modeling
puterunterstüztes psychologisches Verfabren zur Högeräteanpassung,’’ otoacoustic emission and hearing threshold fine structures in humans,’’ J.
Psychologische Beiträge 27, 494 –508. Acoust. Soc. Am. 104, 1517–1543.
Heller, O. ~1985!. ‘‘Hörfeldaudiometrie mit dem Verfahren der Kategorie- Thomas, I. B. ~1975!. ‘‘Microstructure of the pure-tone threshold,’’ J.
nunterteilung ~KU!,’’ Psychologische Beiträge 27, 478 – 493. Acoust. Soc. Am. Suppl. 1 57, S26 –27~abstract!.
Hellman, R. P., and Zwislocki, J. ~1961!. ‘‘Some factors affecting the esti- van den Brink, G. ~1970!. ‘‘Experiments on binaural diplacusis and tone
mation of loudness,’’ J. Acoust. Soc. Am. 33, 687– 694. perception,’’ in Frequency Analysis and Periodicity Detection in Hearing,
Horst, W. J., and de Kleine, E. ~1999!. ‘‘Audiogram fine structure and spon- edited by R. Plomp and G. F. Smoorenburg ~Sijthoff, Leiden!, pp. 362–
taneous otoacoustic emissions in patients with Meniere’s disease,’’ Audi- 374.
ology 38, 267–270. van den Brink, G. ~1980!. ‘‘Cochlear mechanics as the possible cause of
binaural diplacusis,’’ in Psychological, Physiological and Behavioural quiet, and just noticeable amplitude modulation at low levels,’’ in Auditory
Studies in Hearing, edited by G. van den Brink and F. Bilsen ~Delft Uni- Frequency Selectivity, edited by B. C. J. Moore and R. D. Patterson ~Ple-
versity Press, Delft!, pp. 64 – 67. num, New York!, pp. 49–59.
Zweig, G., and Shera, C. A. ~1995!. ‘‘The origins of periodicity in the Zwicker, E., and Peisl, W. ~1990!. ‘‘Cochlear processing in analog models,
spectrum of evoked otoacoustic emissions,’’ J. Acoust. Soc. Am. 98, in digital models, and in human inner ear,’’ Hear. Res. 44, 206 –216.
2018 –2047. Zwicker, E., and Schloth, E. ~1984!. ‘‘Interrelation of different oto-acoustic
Zwicker, E. ~1986!. ‘‘Spontaneous oto-acoustic emissions, threshold in emissions,’’ J. Acoust. Soc. Am. 75, 1148 –1154.

Fine Structure of Hearing Threshold and

Uploaded by

Copyright:

Available Formats

Fine Structure of Hearing Threshold and

Uploaded by

Document Information

Original Description:

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Fine Structure of Hearing Threshold and

Uploaded by

Copyright:

Available Formats

Fine structure of hearing threshold and loudness perceptiona)

~Received 8 August 2003; revised 24 March 2004; accepted 19 April 2004!

I. INTRODUCTION emissions—OAE! could be due to the same mechanisms as

Figure 6 shows a comparison of threshold measurements and

Frequency Hearing thres. of Frequency Hearing thres. of

You might also like