effect stimuli were within 3 to 4 dB of each other for both AC and BC testing. ... S. Army Research Laboratory, Human ... 10. Train chugging along. 250. 11. Thunder cracking. 250. 12. Drum beating. 500 ... m 70 -- m 60 -. 60--. J 50 l a1 50--. J. 40 t -. 1 i. 40. 100. 1000. 10,000. 100 ... Short breaks were provided about every 20.
J Am Acad Audiol 10 : 422-428 (1999)
Comparison of Air-Conduction and Bone-Conduction Hearing Thresholds for Pure Tones and Octave-Band Filtered Sound Effects Kim S. Abouchacra*t Tomasz Letowskit
Abstract The purpose of this study was to measure air-conduction (AC) and bone-conduction (BC) hearing thresholds with pure-tone and filtered sound effect stimuli using standard audiometric equipment. A group of 20 young, normal-hearing listeners participated in the study. Puretone stimuli were 250, 500, 1000, 2000, and 4000 Hz . Sound effect stimuli were 12 natural sounds that were spectrally limited to an octave bandwidth centered at either 250, 500, 1000, 2000, or 4000 Hz . The AC and BC detection thresholds were measured using a clinical audiometer (Madsen Orbiter 922) with a B-71 bone oscillator and TDH-50 earphones. Results indicated that detection thresholds for the pure-tone and corresponding octave-band sound effect stimuli were within 3 to 4 dB of each other for both AC and BC testing. The findings support the notion that octave-filtered sound effects are a viable alternative to pure-tone stimuli for use in audiology clinics . Key Words: Air- and bone-conduction thresholds, detection, sound effect stimuli ecently, Myers et al (1996) described a set of octave-filtered sound effects that R could be used as alternative stimuli to pure tones when testing children and special older populations. The filtered sound effects, which include various instrumental, animal, and environmental sounds, were found to be useful for both sound detection (threshold of hearing) and sound recognition (threshold of identification) tasks . Specifically, Myers et al (1996) reported detection and recognition thresholds for 25 sound effects that were octave-band filtered (25 dB/octave slope) at either 250, 500, 1000, 2000, or 4000 Hz . The sound effects were presented in quiet and in 20-voice multitalker noise (Frank and Craig, 1984) at 60 dBA. Thresholds for the sound effects were compared to thresholds obtained with pure-tone stimuli at 250, 500, 1000, 2000, and 4000 Hz . Results indi-
*Department of Otolaryngology-Head & Neck Surgery, American University of Beirut-Medical Center, Beirut, Lebanon ; tU .S . Army Research Laboratory, Human Research and Engineering Directorate, Aberdeen Proving Ground, Maryland . Reprint requests : Tomasz Letowski, U .S . Army Research Laboratory, Human Research & Engineering Directorate, AMSRL-HR-SD, Bldg . 520, Aberdeen Proving Ground, MD 21005-5425
422
cated that octave-band sound effects can be reliably detected and easily identified by young, normal-hearing adults in both quiet and noisy conditions . More importantly, detection thresholds (DTs) for a number of sound effects were within 2 to 3 dB of their corresponding pure-tone thresholds for the same subject . Myers et al (1996) concluded that octave-band sound effects, with the 25 dB/octave filtering, would produce DTs equivalent to pure-tone DTs for listeners with normal-hearing, flat, and gently sloping audiometric configurations . For listeners with other audiometric configurations, however, the authors suggested that the octave-band sound effects should be filtered with a steeper filter slope. Steeper filtering reduces the risk of listeners responding to energy outside of the test octave, which could result in erroneous hearing thresholds . Myers et al (1996) also cautioned that their data were obtained using laboratory equipment and a test protocol that differed from standard clinical practice . In order to develop a clinical test using filtered sound effects, the filtered sound effect thresholds need to be measured using standard clinical equipment and protocol . The goal of the present study was to address both of the clinical methodological issues raised by Myers et al (1996) . Twelve sounds that were
AC and BC Thresholds Using Filtered Sound Effects/Abouchacra and Letowski
reported by Myers et al (1996) as promising alternatives for pure-tone audiometry were refiltered using steeper (40 dB/oct) octave filters (Abouchacra and Letowski, 1996) . These sounds were used to measure both the air-conduction (AC) and bone-conduction (BC) thresholds in young adults with normal hearing using a standard audiometric booth, equipment, and proce-
dures. The new set of data resulting from this study has been compared to the data reported by Myers et al (1996) . METHOD Subjects
Twenty adults (7 males, 13 females) participated in the experiment . The subjects ranged in age from 18 to 28 years with a mean of 21 .2 years and a standard deviation of 2 .9 years . The subjects were selected from a group of volunteers on the basis of the following audiomet-
ric criteria : (a) AC and BC hearing thresholds at or better than 15 dB HL at all octave frequencies from 250 to 8000 Hz in each ear (re : ANSI S3 .6-1996) and (b) air-bone gaps of 5 dB or less . One ear of each subject was chosen at random for further testing .
Test Stimuli Five pure tones and 12 filtered sound effects were used as stimuli in the experiment . Puretone frequencies were 250, 500, 1000, 2000, and 4000 Hz . A total of 12 sound effects were selected as test stimuli : (1) 10 octave-band sounds identified by Myers et al (1996) as rendering DTs similar to those for respective pure-tone stimuli and (2) two additional low-frequency sound effect
stimuli : one sound effect centered at 250 (thunder) and another at 500 Hz (drum) . I All octaveband sound effects were filtered so that their spectral slopes were approximately 40 dB/octave . The duration of the sound effects ranged from 1 .2 to 3 .9 seconds . Sound effect labels and corresponding center frequencies of the octaveband filters used for signal processing are listed in Table 1 . Spectral characteristics of the filtered sounds are shown in Figures 1 and 2 .
'A subset of the listeners participating in the Myers et al (1996) study performed more reliably with the additional two sound effects than with the low-frequency sound effects in the original set of 10 sounds recommended by the authors .
List of the Sound Effects Table 1 and Their Octave-Band Center Frequencies Used in this Study Sound Number
Sound Name*
1
Airplane passing
3 4
Bird singing Cow mooing
2
5 6
7 8 9 10 11 12
Center Frequency (Hz) 250
Baby crying
2000
Cuckoo clock sounding Dog barking
1000 1000
Coyote howling Glass shattering Baby rattle shaking Train chugging along Thunder cracking Drum beating
2000 500 500 4000 4000 250 250
500
*Italicized words identify the names of the sound effects used throughout the text .
Compact disc (CD) recordings of the 12 filtered sound effect stimuli were used to establish DTs (Abouchacra and Letowski, 1996). To allow for different testing strategies, the CDs contained recordings of each sound effect on two consecutive tracks . On the first track, the sound effect was repeated continuously for 3 minutes without any pause between repetitions . The second track included a 3-second pause between subsequent repetitions of the sound effect to allow time for a response after each presentation . To establish DTs in this experiment, the first track of each sound effect was used so that thresholds could be established in a manner similar to conventional pure-tone audiometry. That is, a continuous repetition of the sound effect stimulus is generated as long as the audiometer presentation bar/button is depressed and the signal ceases when the presentation bar/button is released . Instrumentation The experiment was conducted in a standard audiometric test booth (IAC) having ambient noise levels acceptable for soundfield testing (ANSI S3 .1-1991). Adiagnostic audiometer (Madsen, Orbiter 922 v.2) calibrated to ANSI specifications (ANSI S3 .6-1996) was used to collect AC and BC thresholds for both pure-tone and sound effect stimuli. Pure-tone stimuli were generated by and gated from the audiometer and directed to either earphones (Telephonics, TDH-39 ; MX-41/AR cushions) or a bone oscillator (Radio Ear, B-71). The sound effects were gen423
Journal of the American Academy of Audiology/Volume 10, Number 8, September 1999
Airplane
100 -
90
100
--
90
80 ~
70 .
m
60 -
J
50 l
40 t 100
90 80
m
/"'~
1
i 1000
Frequency (Hz)
a1 J
70 -60-50-40
100
100 -
90 --
--
m
v 70 --
_ 60 m 50 J 40 A P 100 1000 Frequency (Hz)
1000
10,000
Frequency (Hz)
--
100 90 80
80--
10,000
Train
100 -
Cow
-
m
am J
10,000
-------
Coyote
80 --
70 60 50 40
---
-1
100
i
1
Ii
1000
10,000
Frequency (Hz)
Thunder
m 'v 70 m 60 50 40
100
1000
10,000
Frequency (Hz) Figure 1
Spectral characteristics of sound effects filtered at 250 Hz and 500 Hz (see Table 1) .
erated from a portable CD player (JVC PCX100), routed through the audiometer, and delivered to either the earphones or the bone oscillator. Routine calibration of the earphones was performed prior to the experiment and verified at the end of the experiment using a sound level meter (B&K 2231) and NBS 9A coupler. Similarly, bone vibrator calibration was performed prior to the experiment and verified at the end of the experiment using a measuring amplifier (B&K 2610) and an artificial mastoid (B&K 4930). 424
Procedure For each subject, AC and BC thresholds were obtained for both pure-tone and sound effect stimuli during a single 1.0- to 1.5-hour session . Short breaks were provided about every 20 minutes. The order of stimulus presentation was random with a counterbalanced schedule of stimulus (tones, sound effect) and transducer type (earphones, bone vibrator). After fitting the subject with the transducer, the experimenter gave instructions for responding at
Effects/Abouchacra and Letowski AC and BC Thresholds Using Filtered Sound
Dog
100 90 80 70 -76 60 0) 50 _r 40
100
1000
Frequency (Hz)
10,000
Rattle 100 r-._ ._.-_._._-_ ._ .r ._...-...________.__ .a. ....... 90 ~ 80 m N
70 60 -
> 50 J 40
100
1000
Frequency (Hz)
10,000
Baby
60 50 40
Figure 2
100
1000
Frequency (Hz)
10,000
1) . Spectral characteristics of sound effects filtered at 1000, 2000, and 4000 Hz (see Table
threshold, as it is normally done for pure-tone audiometry. Threshold determination consisted of initially presenting a given test stimulus at 40 dB HL (i .e ., at a comfortable listening level) . The level was decreased in 10-dB steps until no response occurred and then increased in 2-dB steps until the subject responded. This procedure continued until three responses were obtained at the same level on the ascent . The above procedure (10 down and 2 up) departs slightly from the ASHA Recommended Clinical Procedure (10 down and 5 up I Carhart and Jerger, 19591) and
was used to increase the precision of threshold measurement and validity of presented conclusions. If two hearing thresholds do not differ significantly when measured with 2-dB steps, then they will certainly not differ when measured with 5-dB steps .
RESULTS AND DISCUSSION
M
can DTs for the pure-tone and sound effect
stimuli presented via AC and BC are shown in dB HL in Table 2 and in dB SPL in Fig425
Journal of the American Academy of Audiology/ Volume 10, Number 8, September 1999
Table 2 Air-Conduction and Bone-Conduction Detection Threshold Data for Pure Tones (PT) and Octave-Band Sound Effects Test Frequency (Hz) 250
500
Air-Conduction Thresholds (in dB HL)
Stimulus
Mean
Median
SD
Mean
Median
SD
Airplane Thunder Train PT Cow
-4 .0 -4 .6 -5 .1 -5 .1 2 .0
-2 .5 -5 .5 -4 .5 -3 .5 2 .5
5 .9 6 .8
-3 .7 -4 .7
-3 .5 -5 .5
6. 1 6.3
7 .4 5 .1
-4 .9 2 .1
-5 .5 2 .5
Coyote PT Cuckoo Clock Dog PT Baby Bird PT Glass Rattle PT
-3 .7 -1 .1 1 .6 0 .9 -1 .9 0.4 -0 .6 -0 .1 2 .1 1 .3 -0 .5
-3 .5 -1 .5 1 .0 1 .0 -2 .0 0 0 -1 .0 1 .5 0 .5 -1 .5
3 .8 5 .5 4 .1 5 .1 4 .4 5 .3 4 .7 5 .6 4 .1 4 .3 4 .6
Drum
1000 2000 4000
0.5
6.4
-1 .5
ures 3 and 4. In Figures 3 and 4, average puretone DTs obtained in this study are shown by points connected by a solid line, whereas the standardized audiometric zero curve is shown as a dashed line . The DTs for the filtered sound effects are represented by an x (train, coyote) or filled circles (all other sounds). The data for the sounds of train and coyote differed slightly from
30
Hearing Threshold (dB SPL)
-7 .5
0.7
6. 2 5. 5 4 .9 5. 5 4. 7
1 .5
-4 .3 -1 .0 0 1 .1 -2 .2 -0 .6 -2 .1 -0 .9 -0 .5 -1 .4 -1 .5
-4 .5 0 .5 0 2 .0 -2 .0 -1 .0 -2 .0 -2 .0 -0 .5 -1 .5 -1 .5
3.8 3.6 4.0 3 .8 4.9 4. 3 5 .3 4.2 4.8 5 .3
all other data and therefore were identified by different symbols in Figures 3 and 4. All of the AC thresholds as well as all of the BC thresholds for sound effects filtered at the same octave frequency were typically within 1 to 2 dB of each other, with the exception of the sound effects filtered at 500 Hz . At 500 Hz, mean DTs for the three sound effects (cow, coy-
Hearing Threshold (dB SPL)
25
20
20
15
15
10
10
5
5 0.25
0.5
1
2
4
Frequency (kHz)
Figure 3 Group mean detection thresholds (DTs) for air-conduction testing . The dashed line represents the standardized audiometric zero curve for pure-tone stimuli . The solid line represents actual pure-tone DTs obtained in this study. Sound effect thresholds are displayed as an x or a filled circle (see text) at the frequency corresponding to the center frequency of the filter used.
426
-8 .2
4.5
30
25
0
Bone-Conduction Thresholds (in dB HL)
0
0.25
0.5
1
2
4
Frequency (kHz)
Figure 4 Group mean detection thresholds (DTs) for bone-conduction testing . The dashed line represents the standardized audiometric zero curve for pure-tone stimuli . The solid line represents actual pure-tone DTs obtained in this study. Sound effect thresholds are displayed as an x or a filled circle (see text) at the frequency corresponding to the center frequency of the filter used.
AC and BC Thresholds Using Filtered Sound Effects/Abouchacra and Letowski
ote, and drum) varied over a range of 5 .7 and 6 .4 dB for AC and BC conditions, respectively . Although this range is still not large, it results from a 5-dB lower hearing threshold for the sound of coyote than for the two other 500-Hz sounds (cow, drum). Additionally, hearing thresh-
olds measured via BC for sound effects filtered around 250 Hz varied over a range of 4 .5 dB . This variability resulted from a relatively low threshold level for the sound of train . Both of these differences were not statistically significant (p < .05)
but can be explained by specific sound energy distributions within the filtered band of these two sounds . The sound of coyote had most of its energy concentrated in the center of the octave band, whereas the two other 500-Hz sounds
(cow, drum) had their energies distributed more evenly across the octave bandwidth . Since the octave band exceeds the width of the auditory filter (critical band) at 500 Hz, the coyote sound was a more effective (louder) signal than the
other two sounds . On the basis of this observation, one may conclude that the filtered sound effects should have bandwidths similar to those of the specific auditory filters . However, most of the common sound effects would be much harder to recognize at the threshold level when filtered with the auditory filter bandwith, which is typically much narrower than the octave band . A
large difference between detection and recognition thresholds would defeat the purpose of using filtered sound effects as the audiometric signals .
The other problem signal, train, had sound energy concentrated at the upper end of the bandwith rather than distributed evenly across the whole octave band or concentrated at the center frequency. This imbalance in the region of 250 Hz is not critical for AC testing but may affect the results of BC testing due to the resonant properties of the B-71 bone vibrator in the region of 250 Hz . Both reported problems may be eliminated in the future by a slight spectral equalization of the train and coyote sound effects. Standard deviations of hearing threshold data obtained with pure-tone stimuli varied between 4.4 and 7.4 dB for AC measurements and between 3.8 and 5.5 dB for BC measurements. The standard deviations for the octavefiltered sound effects ranged from 3.8 dB (coyote) to 6.8 dB (thunder) for AC measurements and from 3.6 dB (cuckoo) to 6.3 dB (thunder) for BC measurements . This means that the octave-filtered sound effects result in hearing threshold uncertainty, that is, variability resulting from repeated measurements, similar to or better
than that caused by pure-tone test stimuli. Thus, the proper selection of sound effects may result in improved threshold hearing data reliability. The mean difference between AC and BC thresholds for filtered sound effects was slightly larger than the difference between AC and BC thresholds for pure tones. However, none of the differences were larger than 1.2 dB (except 2.6 dB difference at 4000 Hz), indicating that existing audiometer calibration curves should be applicable to sound effect testing in both AC and BC modes. Specific mean differences between AC and BC thresholds for the filtered sound effects are shown in Table 3 . The DTs reported here were based on the presentation of filtered sound effects in 2-dB upward increments . This lengthy procedure was necessary to study DTs for each of the filtered sound effects. In the clinic, however, thresholds are typically measured using the standard adaptive 5-dB up/10-dB down procedure (Carhart and Jerger, 1959). Using this method should decrease test time considerably and result in even smaller threshold differences .
Since the primary goal of this study was to confirm validity of the Myers et al (1996) data using standard clinical equipment and a standardized semiclinical procedure, it is important to relate obtained findings to those of the other study. The outcome of this comparison was very encouraging, despite many methodological differences between the two studies . Mean AC hearing thresholds measured in both studies for the same test stimulus differed by no more than 5 .0 dB and typically by less than 3 .0 dB, despite different subject groups and different signal filtering techniques used in both studies . More importantly, the deviations observed at specific test frequencies followed the same pattern, that is, the changes in all of the thresholds had the same direction, indicating that a large part of the data differences between both stud-
Table 3 Mean Differences between AC and BC Hearing Thresholds for
Pure Tones and Filtered Sound Effects Pure-Tone Frequency or Center Frequency of the Filtered Sound Effect (Hz)
AC-BC Difference Pure-Tone Signals
Filtered Sound Effects
250
0 .2
0 .9
1000 2000 4000
0 .3 0 .8 1 .0
0 .9 1 .2 2 .6
500
0 .1
0.3
427
Journal of the American Academy of Audiology/ Volume 10, Number 8, September 1999
ies was due to natural hearing threshold differences between the groups of subjects. This further supports viability of the tested methodology for widespread clinical use.
CONCLUSIONS
T
he results of this study confirm Myers et al's (1996) conclusion that the octave-filtered sound effects should provide clinicians with complex stimuli suitable to obtain frequency-specific AC and BC thresholds that are comparable to pure-tone thresholds . Moreover, the filtered sound effects can be used with existing clinical audiometers without any modifications or a need for correction factors. Putting the filtered sound effects on CD provides the clinician with calibrated and level-controlled natural sound stimuli that can be rapidly accessed in any sequence . The steep slope filtering applied to the sound effects should allow for using these stimuli to test listeners with a relatively large range of audiometric configurations . However, this study also indicated that proper care should be taken to select sound effects with maximum energy at the center of the filtered band . Such well-filtered and easy to recognize filtered sound effects may be applicable for testing both children and adults with special needs, independent of language and cultural differences . However, further research is needed to validate this assumption . Initial data for normal-hearing older children reported by Besing et al (1998) support the feasibility of the Filtered Sound Effect Test (FSET) for this population . Further studies should include young children and both children and adults with differing audiometric
configurations . A search for more narrow-band sound effects that are common and easily recognizable without training should continue . The feasibility of using filters with bandwidths corresponding to auditory filters (critical bands) rather than octave bands should also be explored . Acknowledgment . The Human Research & Engineering Directorate of the U.S . Army Research Laboratory supported this work . The authors would like to acknowledge LaShawna Wright for her assistance with data analysis and Janet Koehnke and Bob Karsh for their insightful comments on earlier drafts of the paper.
REFERENCES Abouchacra KS, Letowski T. (1996) . Filtered Sound Effects for Audiological Testing [compact disc]. Aberdeen Proving Ground, MD : U.S . Army Research Laboratory, Human Research and Engineering Directorate . American National Standards Institute . (1991). Maximum Permissible Ambient Noise Levels for Audiometric Test Rooms. (ANSI S3 .1-1991.) New York : ANSI.
American National Standards Institute . (1996) . Specifications for Audiometers . (ANSI S3 .6-1996 .) New York : ANSI . Besing J, Koehnke J, Abouchacra K, Letowski T. (1998) . Contemporary approaches to audiological assessment in young children. Top Lang Disord 18 :52-71 . Carhart R, Jerger JF. (1959) . Preferred method for clinical determination of pure-tone threshold. JSpeech Hear Disord 24 :330-345 . Frank TA, Craig CH . (1984) . Comparison ofAuditec and Rintelmann recordings of NU-6 lists. J Speech Hear Disord 49 :267-271 .
Myers LL, Letowski TR, Abouchacra KS, Kalb JT, Haas EC . (1996). Detection and recognition of octave-band sound effects . J Am Acad Audiol 7 :346-357 .
