Combined effects of noise and reverberation on speech recognition performance of normal-hearing children and adults.

Abstract

The purpose of this study is to determine how combinations of noise levels and reverberation typical of ranges found in current classrooms will affect speech recognition performance of typically developing children with normal speech, language, and hearing and to compare their performance with that of adults with normal hearing. Speech recognition performance was measured using the Bamford-Kowal-Bench Speech in Noise test. A virtual test paradigm represented the signal reaching a student seated in the back of a classroom with a volume of 228 m and with varied reverberation time (0.3, 0.6, and 0.8 sec). The signal to noise ratios required for 50% performance (SNR-50) and for 95% performance were determined for groups of children aged 6 to 12 yrs and a group of young adults with normal hearing.
This is a cross-sectional developmental study incorporating a repeated measures design. Experimental variables included age and reverberation time. A total of 63 children with normal hearing and typically developing speech and language and nine adults with normal hearing were tested. Nine children were included in each age group (6, 7, 8, 9, 10, 11, and 12 yrs).
The SNR-50 increased significantly with increased reverberation and decreased significantly with increasing age. On average, children required positive SNRs for 50% performance, whereas thresholds for adults were close to 0 dB or <0 dB for the conditions tested. When reverberant SNR-50 was compared with adult SNR-50 without reverberation, adults did not exhibit an SNR loss, but children aged 6 to 8 yrs exhibited a moderate SNR loss and children aged 9 to 12 yrs exhibited a mild SNR loss. To obtain average speech recognition scores of 95% at the back of the classroom, an SNR > or = 10 dB is required for all children at the lowest reverberation time, of > or = 12 dB for children up to age 11 yrs at the 0.6-sec reverberant condition, and of > or = 15 dB for children aged 7 to 11 yrs at the 0.8-sec condition. The youngest children require even higher SNRs in the 0.8-sec condition.
Results highlight changes in speech recognition performance with age in elementary school children listening to speech in noisy, reverberant classrooms. The more reverberant the environment, the better the SNR required. The younger the child, the better the SNR required. Results support the importance of attention to classroom acoustics and emphasize the need for maximizing SNR in classrooms, especially in classrooms designed for early childhood grades.

Full text

Page 1

Combined Effects of Noise and Reverberation on Speech
Recognition Performance of Normal-Hearing Children
and Adults
Arlene C. Neuman,1Marcin Wroblewski,1Joshua Hajicek,1and Adrienne Rubinstein1,2
Objectives: The purpose of this study is to determine how combinations
of noise levels and reverberation typical of ranges found in current
classrooms will affect speech recognition performance of typically
developing children with normal speech, language, and hearing and to
compare their performance with that of adults with normal hearing.
Speech recognition performance was measured using the Bamford-
Kowal-Bench Speech in Noise test. A virtual test paradigm represented
the signal reaching a student seated in the back of a classroom with a
volume of 228 m3and with varied reverberation time (0.3, 0.6, and 0.8
sec). The signal to noise ratios required for 50% performance (SNR-50)
and for 95% performance were determined for groups of children aged
6 to 12 yrs and a group of young adults with normal hearing.
Design: This is a cross-sectional developmental study incorporating a
repeated measures design. Experimental variables included age and
reverberation time. A total of 63 children with normal hearing and
typically developing speech and language and nine adults with normal
hearing were tested. Nine children were included in each age group (6,
7, 8, 9, 10, 11, and 12 yrs).
Results: The SNR-50 increased significantly with increased reverbera-
tion and decreased significantly with increasing age. On average,
children required positive SNRs for 50% performance, whereas thresh-
olds for adults were close to 0 dB or ?0 dB for the conditions tested.
When reverberant SNR-50 was compared with adult SNR-50 without
reverberation, adults did not exhibit an SNR loss, but children aged 6 to
8 yrs exhibited a moderate SNR loss and children aged 9 to 12 yrs
exhibited a mild SNR loss. To obtain average speech recognition scores
of 95% at the back of the classroom, an SNR ?10 dB is required for all
children at the lowest reverberation time, of ?12 dB for children up to
age 11 yrs at the 0.6-sec reverberant condition, and of ?15 dB for
children aged 7 to 11 yrs at the 0.8-sec condition. The youngest children
require even higher SNRs in the 0.8-sec condition.
Conclusions: Results highlight changes in speech recognition perfor-
mance with age in elementary school children listening to speech in
noisy, reverberant classrooms. The more reverberant the environment,
the better the SNR required. The younger the child, the better the SNR
required. Results support the importance of attention to classroom
acoustics and emphasize the need for maximizing SNR in classrooms,
especially in classrooms designed for early childhood grades.
(Ear & Hearing 2010;31;336–344)
INTRODUCTION
Classroom instruction takes place in noisy, reverberant
environments. The importance of good classroom acoustics is
reflected in the recent adoption of guidelines for classroom
design in the United States (ANSI S12.60–2002), the United
Kingdom (Department for Education and Skills 2003), and
Australia and New Zealand (Australian/New Zealand Standard
2000). The American National Standards Institute (ANSI)
standard specifies that for a classroom with an enclosed volume
?283 m3(?10,000 ft3), the maximum allowable steady
background noise level (1 hr average) is 35 dBA and the
maximum reverberation time (RT60) is 0.6 sec (average 500,
1000, and 2000 Hz octave bands). For larger classrooms
(enclosed volume ?283 m3and ?566 m3, i.e., 10,000 to
20,000 ft3), the allowable background noise level is 35 dBA,
and the allowable reverberation time is 0.7 sec. The standards
in the United Kingdom and Australia/New Zealand are in
general agreement with the ANSI standard, with the exception
that longer reverberation times are allowable in secondary
schools. In addition, in the United Kingdom, a shorter rever-
beration time is recommended for classrooms for hearing-
impaired students.
A recent survey of 32 classrooms, which sampled the
acoustics of unoccupied classrooms in urban, suburban, and
rural settings (Knecht et al. 2002), revealed that a large number
of the classrooms had noise levels and reverberation times
exceeding ANSI recommended values. Background noise lev-
els in the majority of the classrooms exceeded recommended
values. Only four of the 32 unoccupied classrooms had
background noise levels lower than the recommended 35 dBA.
In general, background noise levels of these classrooms were 5
to 15 dB higher than recommended. The quietest classrooms
were those where the heating and ventilation system was off
during testing, thus underestimating the actual noise levels that
would be present when classes are in session and the heating/
ventilation systems would be activated. Reverberation times
ranged from 0.2 to 1.27 sec. Although reverberation times were
acceptable in many of the classrooms, 41% of the classrooms
tested had reverberation times longer than the ANSI recom-
mended 0.6 sec RT60. These findings are in agreement with the
results of older studies showing higher noise levels and longer
than optimal reverberation times in classrooms (Kodaras 1960;
Sanders 1965; McCroskey & Devens 1975; Bradley 1986).
Although guidelines regarding acceptable acoustic condi-
tions relevant to designing classrooms are helpful in promoting
better classroom acoustics, the standard does not address
concerns about the noise generated in occupied classrooms.
Substantial noise levels (primarily competing speech) can be
generated by students. Estimates of the noise levels in active
classrooms vary considerably. A survey of the literature on
noise levels in occupied classrooms (Picard & Bradley 2001)
revealed multiple reports of levels between 50 and 60 dBA,
with the highest noise levels (65 to 75 dBA) occurring in the
classrooms of the youngest children. Several recent reports
have focused on the variations in levels during activities within
a classroom. For example, Shield and Dockrell (2004) reported
1Department of Otolaryngology, New York University School of Medi-
cine, New York, New York; and2Department of Speech Communication
Arts and Sciences, Brooklyn College, CUNY, Brooklyn, New York.
0196/0202/10/3103-0336/0 • Ear & Hearing • Copyright © 2010 by Lippincott Williams & Wilkins • Printed in the U.S.A.
336

Page 2

a difference of 20 dB between the quietest and noisiest
classroom activities. A recent study by Larsen and Blair (2008)
is of particular interest because noise levels were measured
during classroom activity in four 4th grade classrooms that
were in compliance with the ANSI classroom standard, that is,
the reverberation times were all ?0.6 sec and the background
noise level in the unoccupied rooms were ?35 dBA. The
measurements revealed that signal to noise ratio (SNR) varied
with the classroom activity and the student’s location in the
classroom. With the teacher “lecturing,” SNRs varied from ?9
dB in the front of the classroom to ?6 dB in the center and ?3
dB in the back of the classroom. Thus, even under conditions
in which children were listening quietly, the SNR was poorer
than the ?15 dB recommended in the ANSI standard. Depend-
ing on the activity, the noise generated by classroom activities
resulted in increases in level of up to 33 dB higher than the
unoccupied noise level. During activities in which the teacher
was carrying out one-on-one work or children were working on
projects, SNRs were lesser than 0 dB. This study is in
agreement with a number of other studies reporting SNRs
during classroom activities ranging from negative SNRs to
?10 dB (data are summarized by Crandell & Smaldino 2000
and Picard & Bradley 2001).
In view of the relatively poor SNRs and the wide range of
reverberant conditions in current elementary school class-
rooms, it is important to understand how the combination of
noise and reverberation will affect speech recognition perfor-
mance of children as a function of age. Yet, relatively few
studies have examined the effects of noise and reverberation on
the speech recognition of children for a range of conditions
typical of classroom environments. Early studies demonstrated
that reverberation, noise, and their combination degraded
speech recognition performance. However, limitations of the
studies included monaural testing, failure to include a range of
reverberation times typical of classroom environments, and
failure to consider possible developmental effects relevant to
the perception of degraded speech. Finitzo-Hieber and Tillman
(1978) demonstrated that recognition performance in a group
of children aged 8 to 12 yrs decreased with increases in noise
level (?12 to 0 dB SNR) and with increases in reverberation
(anechoic, 0.4 and 1.2 sec reverberation). They also determined
that the combined effect of reverberation and noise was greater
than would be predicted based on decrements in performance
due to either degradation alone. The reverberation times
assessed include a short RT60(0.4 sec) that would be consid-
ered optimal and a longer time (1.2 sec) typical of a room
known to be poor for speech communication. However, the
range of reverberation times between 0.4 and 1.2 sec is more
typical of classroom environments and thus results are infor-
mative for classrooms with reverberation times lower than the
ANSI-recommended standard, and for classrooms clearly hav-
ing poor acoustics (1.2 sec). In addition, because testing was
performed monaurally, the results may overestimate reverber-
ation and noise effects. Yacullo and Hawkins (1987) also
focused on monaural speech recognition of 8- and 9-yr-old
children (recognition of words in sentences) under reverberant
(anechoic and RT60? 0.8 sec) and multitalker babble noise
(SNR ? ?2 and ?6 dB). The limited set of data revealed
significant decreases in performance with reverberation and
with increasing noise, as well as a significant interaction effect,
but again the reverberation time exceeds the ANSI recommen-
dation and focuses on poorer SNRs than the recommended.
Furthermore, the data obtained at the poorer SNR may have
resulted in floor effects for some of the subjects.
A more comprehensive evaluation of reverberation time
was carried out by Bradley (1986) who focused on speech
recognition performance of 12- to 13-yr-olds (grades 7 and 8)
in 10 test classrooms. Reverberation times in the 10 classrooms
varied between 0.39 and 1.2 sec, and background noise levels
(as measured during or after the test) varied between 39 and 45
dBA. The speech level from the loud speaker was varied during
testing to obtain SNRs of ?12 to ?15 dB. Thus, this study
represents the performance of students in quiet classrooms in
which the speech level is varied. The background noise is
primarily from heating/ventilation systems, small movements
of the children, and noises that are generated outside the
classroom. These test conditions do not necessarily represent
the background noises in an active classroom in which com-
peting speech signals or group activities are taking place.
Results of the study revealed that word recognition scores
decreased with increasing reverberation time and increased
with increasing SNR. Importantly, as reverberation time in-
creased, the SNR required for equivalent speech recognition
performance also increased. On the basis of the data from this
study, Bradley recommended that to optimize speech recogni-
tion performance, reverberation times of 0.4 to 0.5 sec and a
background noise level of 35 dBA would be appropriate for a
classroom of 300 m3, but a lower background noise level of 30
dBA would be more appropriate for younger children.
Several studies have addressed developmental aspects of the
recognition of reverberant speech. Neuman and Hochberg
(1983) demonstrated that speech recognition in quiet (0, 0.4,
and 0.6 sec reverberation) decreased with increasing reverber-
ation, and performance under reverberant conditions improved
with age (children aged 5 to 13 yrs and young adults). In
addition, binaural versus monaural performance was com-
pared; binaural speech recognition of reverberant speech was
significantly higher, and binaural benefit was much larger for
the youngest children than for older children or adults. This
study was performed in quiet and did not include a reverber-
ation time exceeding the current ANSI recommendation for
classrooms. Johnson (2000) measured the effect of multitalker
babble noise (quiet, ?13 dB SNR), reverberation (anechoic
and RT60? 1.3 sec), and their combination on the monaural
nonsense syllable identification of children ranging in age from
6 to 15 yrs and adults. Results were in agreement with
outcomes of other studies demonstrating that (a) speech rec-
ognition performance was degraded by noise or by reverbera-
tion alone, (b) the combined effect of noise and reverberation
was greater than that would be predicted by either degradation
alone, and (c) performance improved with age. The data
revealed a differential developmental pattern of performance in
the children, depending on the degradation. Although adult-like
consonant-identification performance was obtained by the 14-
to 15-yr-old children for the singly degraded speech stimuli,
consonant-identification performance by the 14- to 15-yr-old
children was significantly poorer than the adults for the
combined condition of reverberation and noise. However,
results do not allow conclusions about how children of differ-
ent ages will perform in noise in classrooms with reverberation
times in the range recommended in classroom standards
because (a) a single reverberation time much longer than
NEUMAN ET AL. / EAR & HEARING, VOL. 31, NO. 3, 336–344
337

Page 3

recommended for classrooms was used, (b) a single noise level
more favorable than would be found in many classrooms was
evaluated, and (c) testing was monaural.
Bradley and Sato (2008) assessed speech recognition per-
formance (monosyllabic closed-set rhyme test) of groups of
children aged 6, 8, and 11 yrs in 41 classrooms, most of which
had reverberation times in compliance with ANSI recommen-
dations. The results of the study confirmed a developmental
effect for listening to speech in noise. The younger children
required better SNRs than older children to obtain similar
scores. Analysis of the data revealed that to obtain a mean
score of 95%, the 6-yr-olds required an SNR 4 dB better than
the 8-yr-olds and 7 dB better than 11-yr-olds. However, the
data revealed large differences among individual listeners in
each age group. Modeling of the data revealed that although an
SNR of 15 dB would allow 80% of the 11-yr-olds to obtain
near-perfect speech recognition performance on the monosyl-
labic word test, 8-yr-olds would require an 18-dB SNR and
6-yr- olds would require ?20-dB SNR.
Thus, previous research has revealed a developmental effect
with regard to the recognition of reverberant speech in noise
(Bradley & Sato 2008) under reverberant conditions that would
be considered acceptable according to current ANSI guidelines
and for a reverberation time longer than that recommended for
classrooms (Johnson 2000). The purpose of the present study
was to determine how the speech recognition performance of
children and adults with normal hearing is affected in noisy,
reverberant environments (multitalker competing noise) for a
range of reverberation times typical of modern elementary
school classrooms. The Bamford-Kowal-Bench Speech in
Noise Test (BKB-SIN, Version 1.03) Audio CD (Etymotic
Research, Elk Grove Village, IL, 2005) was chosen as the test
material because words in sentences are most likely to reveal
reverberant effects and because normative data for speech in
noise performance is available for children and adults for the
premixed speech and noise stimuli. Test stimuli were generated
with both speech and noise co-located, because this would
simplify testing and could be considered to appropriately typify
listening in noise in the classroom, because spatial release from
masking is greatly reduced in reverberant conditions (Plomp
1976). Reverberation times of 0.3, 0.6, and 0.8 sec were
included to represent a short reverberation time, the ANSI-
recommended reverberation time, and a longer than recom-
mended reverberation time. This test material allows the
identification of the SNR required for 50% performance
(SNR-50), and in this study, it is used to compare performance
among children in different age groups. This approach has been
accepted as a means of avoiding ceiling and floor effects that
are problematic when testing at fixed SNRs. It also allows for
a comparison of an individual’s performance with normative
data obtained from well-defined groups of listeners with
normal hearing. The procedure used in this study derived
thresholds from data obtained at multiple SNRs. This also
made it possible to determine the SNR required for a high level
of performance (95% recognition), as well as the slope of the
performance-intensity (PI) function for each reverberation
time. The results of this study provide information about how
the required SNR changes with reverberation in elementary
school children as a function of age and how children’s
performance compares with that of adults. The data should be
helpful in determining whether the acoustic requirements for
classrooms should be age dependent.
MATERIALS AND METHODS
Reverberant test materials were created by convolving
binaural room impulse responses (BRIRs) with recordings of
anechoic test materials. This approach is commonly used to
create “virtual” test environments (Koehnke & Besing 1996;
Cameron et al. 2006). The impulse responses were recorded in
a single classroom. This classroom had an average reverbera-
tion time of 0.8 sec, thus exceeding ANSI 2002 recommenda-
tions. The reverberation time was modified to create two
additional reverberant classroom conditions by hanging ab-
sorptive acoustic panels on the upper portions of the walls.
Measurement of Noise Characteristics and Reverberation
Time
The room was rectangular in shape with a volume of 227.5
m3(10.06 m ? 6.65 m ? 3.4 m). The front wall, back wall, and
ceiling were sheet rock. The side walls were cement block. One
wall had five large windows extending from waist height to the
ceiling. The opposite wall had two doors providing access to
the hallway. The room is currently not being used as a
classroom and was relatively empty. The back of the room had
built in furniture. Two cabinets with glass doors were at either
side at the front of the room. A small area rug covered the
center of the floor (?50% of the area).
A Tannoy CPA5 loudspeaker was used to generate the
stimuli for measurements of the room acoustics and for
obtaining the impulse responses for creation of the test mate-
rials. The loudspeaker was placed in front of the classroom (1.2
m from the front wall) and equidistant from the side walls. The
center of the loudspeaker was at a height of 1.5 m. This
location represents the position and height of a teacher standing
in front of the room. The measurement/recording location was
at a distance of 5.5 m from the loudspeaker and represents the
position of a student sitting toward the back of the classroom.
This position was beyond the critical distance (i.e., the distance
at which the intensity of the direct sound equals that of the
reflected sound) and thus was in the reverberant field.
Background noise level measurements were obtained in the
unoccupied classroom using a Larson-Davis 800 B Sound
Level Meter (SLM) and 826-10 preamplifier with1⁄2 inch 2559
microphone. Calibration was checked using a Larson Davis
Model CA250 calibrator. The noise level at the measurement
locationwas32dBA.Thisbackgroundnoiselevelisconsideredto
be acceptable according to the ANSI guidelines and was suffi-
ciently low to allow measurement of reverberation time and
impulse response recordings for use in stimulus preparation.
Reverberation measurements were obtained using maxi-
mum length sequence (MLS) stimuli (length, 2.73 sec), 48 kHz
sampling rate, and 16 bit quantization. Stimuli were generated
using a SONY VAIO UX280P Micro PC with full duplex input
and output capabilities. The output of the computer was sent to
a Crown D75 amplifier and signals were played through the
front loudspeaker. The microphone was mounted on a stand at
the student’s location. The output of the microphone was sent
to the SLM using the Larson Davis 800-EC extension cable.
The analog output of the SLM was sent to the input of the
microphone of the computer and the impulse response was
NEUMAN ET AL. / EAR & HEARING, VOL. 31, NO. 3, 336–344
338

Page 4

derived by cross-correlating the original MLS with the re-
corded MLS. Impulse to noise ratios were greater than 45 dB
at all frequencies used for calculation of RT60values. Rever-
beration times in octave bands were derived from the early
decay time of the impulse response.
To modify the reverberation time, All Noise Control
ANC600 absorptive panels were hung from hooks mounted on
the walls close to the ceiling. The average reverberation time
(average of 500, 1000, and 2000 Hz) of the classroom before
modification was 0.84 sec. With nine panels hung, the average
reverberation time was reduced to 0.59 sec. This approximates
the ANSI-recommended reverberation time of 0.6 sec. With 41
panels, a reverberation time of 0.31 sec was obtained. This
short reverberation time complies with the American Speech
Language Hearing Association (1995) position statement rec-
ommendation and is also typical of the lowest reverberation
time that might be found in a classroom (Yang & Bradley
2009).
The measured reverberation times in octave bands for each
of the reverberant conditions are given in Table 1.
Recording of BRIRs
The Knowles Electronics Manikin for Auditory Research
(KEMAR) was placed at the recording location at a height
approximating that of a child seated at a desk (ears were 0.8 m
above the floor, as specified in ANSI S12.60–2002). BRIR
recordings were obtained for each reverberation condition.
MLS stimuli were presented from the front loudspeaker. Stereo
recordings of the MLS sequences were recorded using Ety-
motic Research ER11 microphones placed at the medial
opening of Bruel and Kjaer DB 100 couplers (KEMAR
eardrum locations). The diffuse field inverse filter (DFI) setting
was used on the Etymotic Research ER 11 preamplifier. The
output of the KEMAR left ear microphone was sent to the left
channel, and the right ear microphone to the right channel of
the Sony Vaio input.
The impulse responses were derived by cross-correlating
the KEMAR microphone outputs with the original MLS.
Before creating the test materials, the impulse responses were
truncated. Anything before the first primary impulse was
eliminated. Further truncation resulted in an impulse response
containing only the information about reflections that were
above the noise floor.
Creation of Reverberant Test Materials
Test materials • The truncated BRIRs for each of the class-
room reverberation conditions were convolved with the cali-
bration tracks and the .wav files from the BKB-SIN test
materials using standardized linear convolution algorithms in
MATLAB?. The BKB-SIN test was developed to be suitable
for testing children and persons with cochlear implants. Nor-
mative data on the original recordings are available for children
aged 5 yrs and older (Etymotic Research 2005). The test
consists of 18 equivalent list pairs. Short, simple sentences (5
to 7 words, first grade language level) are presented against
four-talker babble. The test is simple to administer because the
changes in background noise level are prerecorded. There are
three key words per sentence, except for the first sentence
which has four key words. The level of background noise is
increased by 3 dB with each sentence. The listener repeats the
sentence, and key words are scored as correct or incorrect. SNR
of ?21 dB to ?6 dB SNR are assessed. The SNR yielding 50%
correct performance (SNR-50) is calculated for each list pair
using a formula provided by the developers of the test material.
Test recordings • In this study, six BKB-SIN list pairs were
used for testing to avoid repetition of test sentences or lists
within subjects. Two list pairs were administered per reverber-
ant condition. An additional list pair was used for practice. For
ease of data collection, nine test CDs were created, each
containing a practice list (list pair 1) and a pseudorandomiza-
tion of list pairs 2 to 7. Each of the nine subjects per age group
was tested with a different CD. Thus, within each age group,
each list pair appeared three times per test condition.
Within each test session, each reverberation condition was
assessed once in the first half of the session, and then again in
the second half of the session. The order of testing the different
reverberation conditions was counterbalanced. There were
three orders of presentation for the RT condition: low (L, 0.3
sec)–mid (M, 0.6 sec)–high (H, 0.8 sec), M–H–L, and H–L–M.
If a subject received the L–M–H order in the first half of the
session, the order of conditions was reversed in the second half
(e.g., L–M–H became H–M–L).
Participants
Sixty-three children aged 6 to 12 yrs and nine adults served
as listeners. Children were grouped according to their chrono-
logical age, with nine children per year. All participants had
normal hearing as established through audiometric screening
(20 dB HL at 500, 1000, 2000, and 4000 Hz). Children were
typically developing monolingual speakers of American En-
glish and had no history of recurrent middle ear disorders or
insertion of ventilating tubes for treatment of recurrent otitis
media with effusion. None of the children received special
education services (e.g., for a learning disability, attention
deficit/hyperactivity disorder, cognitive disability, specific lan-
guage impairment, etc.) or support services in the schools or
through a private practitioner (e.g., speech-language therapy,
reading/academic tutoring). Further, only children with age
appropriate performance on the Peabody Picture Vocabulary
Test-III (PPVT III, Dunn & Dunn 1997) and without articula-
tion problems on the Goldman Fristoe 2 Test of Articulation
(Goldman & Fristoe 2000) were included in this study. Nine
young adults, aged 20 to 32 yrs, who were native speakers of
American English with normal hearing, served as control
subjects.
Data Collection Techniques
Data were collected in a single test session in most cases.
Audiometric testing, the PPVT III, and the Goldman Fristoe 2
were administered first to determine eligibility. Data were then
collected using the reverberant test materials. Hearing screen-
TABLE 1. RT60(sec) at octave band frequencies
125
Hz
250
Hz
500
Hz
1000
Hz
2000
Hz
4000
Hz
8000
Hz
Average
RT60
0.84
0.57
0.31
0.59
0.64
0.38
0.92
0.49
0.40
0.89
0.48
0.29
0.82
0.60
0.31
0.80
0.64
0.33
0.68
0.65
0.45
0.39
0.45
0.34
RT, reverberation time.
NEUMAN ET AL. / EAR & HEARING, VOL. 31, NO. 3, 336–344
339

Page 5

ing of children was carried out using a Grason Stadler GSI 17
or 61 Clinical Audiometer. Adults were tested using the GSI 61
audiometer. The sentence recognition threshold testing using
the BKB-SIN reverberant test materials was administered using
a portable CD player and Etymotic Research ER6i Isolator
earphones. The volume control of the CD player was set such
that the output of the calibration signal associated with the
reverberant recordings was equal to 71 dB SPL in a Bruel and
Kjaer DB100 ear simulator mounted on a Larson Davis1⁄2 inch
microphone (2559) and a Larson Davis 800 B Precision
Integrating Sound Level Meter. This level was used to simulate
a level of 70 dB SPL in the soundfield, a level that is in the
range of recommended test levels for the BKB-SIN (Etymotic
Research 2005) and would also be consistent with the average
voice level of a teacher in a noisy classroom (Pearsons et al.
1976).
Data Analysis
SNR-50 was calculated for each participant in each rever-
berant condition using the formula supplied with the original
test materials. Because two list pairs were administered per test
condition, the SNR-50 was calculated for each list pair and
then the two scores were averaged. Data were analyzed using
a mixed model repeated measures analysis of variance
(ANOVA) with reverberation time as the within factor and age
group as the between factor.
In addition, reverberant SNR loss was determined for
each age group. SNR loss is the deviation of a listener or
group of listeners from the normative SNR-50 for a group of
adults with normal hearing (Etymotic Research 2005, p. 2).
In this case, the SNR loss reveals the increase in SNR
required for 50% performance resulting from reverberation
(in comparison with the SNR required for adults for the test
material in noise alone).
A PI function was derived for each participant using the
percent correct score at each SNR level for each reverberant
condition. To determine the slope at the 50% intercept, a third-
order polynomial was fit to the PI function for each subject for
each reverberant condition. The average slope was then deter-
mined. Slope data were also analyzed using a mixed model
repeated measures ANOVA. Finally, the SNR required for 95%
performance was determined from the average third-order poly-
nomials for each age group and was compared across age groups.
RESULTS
Mean SNR-50s as a function of reverberation and of age are
displayed in Figure 1A. Inspection of the figure reveals that
SNR-50 increases with increase in the reverberation time. The
SNR required for 50% performance decreases with increasing
age for each test condition. Normative data for the anechoic
condition taken from the BKB-SIN users manual (Etymotic
Research 2005, p. 22) are also shown for comparison. Data
presented in Figure 1B are discussed in detail in the Follow-Up
Study section below.
The repeated measures ANOVA on the SNR-50 data
revealed that SNR-50 changed significantly as a function of
Fig. 1. (A, left) Mean SNR-50 (and standard error) as a function of age for each reverberant condition. “x” values are means reported in the BKB-SIN manual
(Etymotic Research 2005). (B, right) Mean SNR-50 (and standard error) in a group of adults under anechoic, pseudoanechoic, and 0.3-sec reverberation
conditions. BKB-SIN, Bamford-Kowal-Bench Speech in Noise; RT, reverberation time; SNR, signal to noise ratio.
NEUMAN ET AL. / EAR & HEARING, VOL. 31, NO. 3, 336–344
340

Page 6

reverberation (F ? 40.83, degrees of freedom [df] ? 2, 128,
p ? 0.001) and as a function of age (F ? 22.6, df ? 7, 64, p ?
0.001). The interaction between reverberation and age was not
significant (F ? 0.62, df ? 14, 128. p ? 0.05). Post hoc testing
with Tukey’s Honestly Significant Differences (HSD) test
revealed that each increase in reverberation time resulted in a
significant increase in SNR-50 (p ? 0.01). Mean SNR-50
values were 2.7, 3.4, and 4.4 dB for RTs of 0.3, 0.6, and 0.8
sec, respectively. Recall that lower SNR-50 values are consis-
tent with better performance. The Tukey’s HSD post hoc
analysis results for the factor of age are given in Table 2. Age
groups whose means do not differ significantly are indicated by
cells with asterisks in a given column. The SNR-50 for adults
was significantly lower than for any of the children. The
younger children (6-, 7-, and 8-yr-olds) had significantly higher
(poorer) SNR-50s than older children (10–12-yr-olds).
The mean slopes of the PI functions at the 50% intercept
were calculated. A repeated measures ANOVA revealed that
the slope changed significantly as a function of reverberation
(F ? 17.05, df ? 2,128, p ? 0.001) and age (F ? 8.67, df ?
7, 64; p ? 0.001). The slope of the PI function decreased
significantly with each increase in reverberation (p ? 0.05).
The mean slope data and Tukey’s HSD post hoc analysis
results for the factor of reverberation are presented in Table 3.
The mean slope of the function was significantly steeper for the
adults than for the children. Mean slopes and results of the
Tukey’s HSD post hoc test are listed in Table 4. The interaction
of reverberation and age was not significant.
The mean SNR loss with reverberation was calculated by
subtracting the mean reverberant score at 50% for each age
group from the SNR-50 for normal-hearing listeners under
anechoic conditions. SNR loss is shown in Figure 2. Under
anechoic conditions, with this test material, the SNR-50 for
adults is ?2.5 dB. In reverberant environments, the SNR loss
decreases with increasing age and decreasing reverberation.
The youngest children show larger SNR losses, whereas adults
show a relatively small SNR loss.
Figure 3 illustrates the average third-order polynomial fits
to the PI function for each reverberant condition for (a) the
youngest age group (6 yrs), median age group (9 yrs), and
oldest age group (12 yrs) and (b) adults. The adults have the
steepest functions and the 6-yr-old children have the shallowest
functions. Clearly, for a given combination of reverberation
time and SNR, the youngest children have the poorest scores.
Scores increase with increasing age. For the test material used
in this study, adults require approximately 8 dB SNR for the
0.8-sec condition to obtain a speech recognition score of
approximately 95%, but 15 dB SNR would be required for the
9- and 12-yr-olds. The 6-yr-olds would require an SNR
exceeding 15 dB SNR to reach 95% performance. For the 0.6
sec condition, 15 dB SNR would yield 95% performance for
even the 6-yr-olds for the test material used in the study.
TABLE 2. Results of Tukey’s HSD test on mean SNR-50 scores
as a function of age
Age, yrs
SNR-50
Mean Group 1 Group 2Group 3Group 4
Adult
11
12
10
9
8
7
6
?0.09
2.16
2.23
3.05
4.01
5.14
5.18
5.94
****
****
****
****
**** ****
****
****
****
****
****
Homogeneous groups (alpha ? 0.05) appear in columns and are indicated by asterisks.
HSD, Honestly Significant Differences; SNR, signal to noise ratio.
TABLE 3. Results of Tukey’s HSD test on mean slope as a
function of reverberation time
RT
Slope (%/dB) at the
50% interceptGroup 1 Group 2Group 3
Low
Mid
High
7.8
7.3
6.7
****
****
****
Homogeneous groups (alpha ? 0.05) appear in columns and are indicated by asterisks.
HSD, Honestly Significant Differences; RT, reverberation time.
TABLE 4. Results of Tukey’s HSD test on mean slope as a
function of age
Age, yrs
Slope (%/dB) at the
50% intercept Group 1Group 2
6
7
8
9
10
11
12
Adult
6.7
7.4
7.0
6.8
6.6
7.4
7.5
8.9
****
****
****
****
****
****
****
****
Homogeneous groups (alpha ? 0.05) appear in columns and are indicated by asterisks.
HSD, Honestly Significant Differences.
Fig. 2. SNR loss as a function of age for three reverberant conditions (re.
norms for adults under anechoic conditions). RT, reverberation time; SNR,
signal to noise ratio.
NEUMAN ET AL. / EAR & HEARING, VOL. 31, NO. 3, 336–344
341

Page 7

Follow-Up Study
Having noted that the SNR-50 for the 0.3 sec reverberant
threshold was 2 to 3 dB higher than the normative data reported
for an anechoic condition for all age groups (see Fig. 1A), a
small follow-up experiment was carried out with a group of
normal-hearing adults to verify that the higher threshold for the
0.3-sec reverberant condition was caused by the introduction of
reverberation and was not an artifact introduced by the record-
ing process.
A set of “pseudoanechoic” stimuli were created by truncat-
ing the binaural room impulse response (0.3-sec condition)
before the first reflection. This truncated BRIR was then
convolved with the original BKB-SIN .wav files. Using a
procedure identical to that used in the main study, a new group
of seven adults with normal hearing were tested using the
original BKB-SIN materials (anechoic), the pseudoanechoic
stimuli, and the 0.3-sec stimuli. Mean scores for the three test
conditions are illustrated in Figure 1B. Mean SNR-50 scores
for the three conditions were ?2.7 dB for the anechoic
condition, ?3.1 dB for the pseudo-anechoic, and ?1.1 dB for
the 0.3-sec condition. The mean SNR-50 measured with the
original BKB-SIN materials (anechoic condition) is in close
agreement with the norms supplied by the manufacturer (?2.5
dB, SD 0.8 dB).
A repeated measures ANOVA revealed that mean SNR-50
scores differed significantly as a function of reverberation time
(F ? 10.36, df ? 2,12, p ? 0.01). Tukey’s HSD post hoc
testing revealed that the mean scores for the anechoic and
pseudoanechoic conditions did not differ significantly, but the
SNR-50 for the 0.3-sec reverberation time was significantly
higher than either “anechoic” condition. These results confirm
that the elevated threshold for the 0.3-sec reverberant condition
was a result of the reverberation rather than an artifact
introduced during the generation of test stimuli.
DISCUSSION
The conditions tested in this study represent the listening
situations of a student sitting in the reverberant field when both
the speech and the noise are equally affected by reverberation,
because both the speech and noise were presented from the
same loudspeaker at 0° azimuth. For each reverberant condi-
tion, speech recognition performance was measured using the
BKB-SIN test. Results revealed that under reverberant condi-
tions, younger children had higher SNR-50s (required better
SNRs at threshold) than older children. The 12-yr-old children
required better SNR than adults. Because children older than
12 yrs were not tested, it is not clear at what age adult-like
performance is achieved. These results are in agreement with
findings of previous studies showing a developmental aspect of
speech recognition performance in noise and reverberation, and
extend findings to a range of reverberant conditions typical of
classrooms.
The SNR-50 increased significantly with increasing rever-
beration. The lack of an interaction between reverberation and
age in the ANOVA points to a similarity in the relative
elevation of threshold for each reverberation condition for the
children and the adults when reverberation is added to noise.
On average, there was a similar increase in threshold (mean 1.7
dB) for each age group between the shortest and longest
reverberation conditions. This shift in threshold compares
favorably with the predicted change in SNR for 50% perfor-
mance in reverberation using the regression equation derived
by Bradley (1986, Eq. [1]).
Although testing under anechoic conditions was not carried
out on the children in this study, a comparison of the expected
performance for noise alone (taken from the normative data
provided with the test material) reveals that higher thresholds
are required by all age groups under reverberant conditions
than would be expected for noise alone. This conclusion is
bolstered by the follow-up study confirming the same finding
in adults who were tested under anechoic (original BKB-SIN
materials) and pseudoanechoic materials (BKB-SIN recordings
convolved with an impulse response truncated before the first
reflection). The pattern of decrease in SNR required for 50%
performance with increasing age roughly parallels the pattern
of performance in anechoic conditions evident in the BKB-SIN
norms. On average, for each age group, the 0.3-sec reverber-
ation elevates the threshold by approximately 2 dB, the 0.6-sec
condition by 3 dB, and the 0.8 sec by 4 dB.
Although threshold for adults approximated 0 dB even in
the 0.8-sec condition, the children required positive SNRs to
obtain 50% performance in all conditions tested. Recall that the
0.8-sec condition represents a classroom whose reverberation
time exceeds the recommended ANSI reverberation time. In
this reverberant condition, the average threshold for the 6-yr-
old children was 6.9 dB SNR, whereas the threshold for the
12-yr-old children was 3.1 dB SNR. The 0.6-sec condition
represents the ANSI-recommended reverberation time. In this
condition, the threshold for the 6-yr-olds was 5.9 dB SNR and
for the 12-yr-olds was 2 dB SNR. In the 0.3-sec condition, the
thresholds improved by a small amount; 6-yr-olds required 5.1
dB SNR and the 12-yr-olds required 1.7 dB. Adult-like
performance was not obtained by the oldest children (age, 12
yr) included in the study. Although use of the SNR-50 allows
comparisons among listeners and across age groups at thresh-
Fig. 3. Average third-order polynomial fit to performance-intensity func-
tions for the three reverberant conditions for 6-, 9-, and 12-yr-old children
and for adults. RT, reverberation time; SNR, signal to noise ratio.
NEUMAN ET AL. / EAR & HEARING, VOL. 31, NO. 3, 336–344
342

Page 8

old, it is important to know more about the SNRs required for
good performance in noisy classroom environments. An exam-
ination of the PI functions for the various age groups included
in this study reveals that the 15-dB SNR recommended in the
ANSI standard would seem to give good speech recognition
performance in all three reverberant conditions for all but the
6-yr-olds in the 0.8-sec reverberant condition. On average, the
6-yr-olds would require a higher SNR to obtain 95% recogni-
tion. It is likely that if the classroom had an even longer
reverberation time that even better SNRs would be required.
Because of the paucity of information regarding the com-
bined effects of reverberation and noise on speech recognition
performance using sentence material, it is difficult to compare
our findings with those of earlier studies. Yacullo and Hawkins
(1987) tested 8- to 10-yr-old children using sentence material
and a 12-talker babble. Testing was carried out monaurally
under anechoic conditions and in a small classroom (90 m3)
with a mean reverberation time of 0.81 sec at SNRs of ?2 and
?6 dB. In the anechoic condition, the mean scores were 50.5%
at 2 dB SNR and 85.4% at 6 dB SNR. In contrast, the SNR-50
under anechoic conditions for 7- to 10-yr-old children would be
expected to be 0.8 dB (Etymotic Research 2005). Thus, the
SNR yielding 50% performance in the study by Yacullo and
Hawkins is approximately 1.2 dB higher than the expected
SNR-50 for the BKB SIN anechoic condition. In the reverber-
ant condition (RT60? 0.8 sec), Yacullo and Hawkins reported
mean speech recognition scores of 36.8% at ?6 dB and 16.9%
at ?2 dB. In the current study, the mean speech recognition
scores in the 0.8-sec condition at 6 dB SNR were 52, 62, and
67% for 8-, 9-, and 10-yr-olds, respectively. The average scores
at 2 dB were 27, 39, and 45% for 8-, 9-, and 10-yr-olds,
respectively. The poorer performance in the study by Yacullo
and Hawkins may be due to several factors. Different speech
and noise materials were used in the two studies. As pointed
out earlier, testing in our study was binaural, whereas Yacullo
and Hawkins tested monaurally. It is known that binaural
listening is beneficial in reverberant conditions (Moncur &
Dirks 1967). Finally, the classroom used by Yacullo and
Hawkins (90 m3) was much smaller than the classroom in this
study (228 m3). Galster (2007) recently showed that speech
recognition performance will differ in rooms with nominally
the same reverberation time, but differing in volume. Specifi-
cally, speech recognition performance is poorer in rooms with
smaller volumes.
Bradley and Sato (2008) assessed speech recognition per-
formance in classrooms with a mean reverberation time of 0.4
sec. They derived best-fit regression functions from the PI
functions obtained with a monosyllabic closed-set test for first,
third, and sixth graders. The SNR required for 95% perfor-
mance was 15.5 dB for the 6-yr-olds, 12.5 dB for the 8-yr-olds,
and 8.5 dB for the 11-yr-olds. The SNRs for 95% performance
determined from the PI functions in this study are remarkably
similar, considering the differences in test materials. For the
0.6-sec condition, SNRs required for 95% performance were
estimated from the PI functions as 15, 12, and 9 dB for the 6-,
8-, and 11-yr-olds, respectively. For the 0.3-sec condition,
SNRs required for 95% performance were 15, 11, and 9 dB.
The slope of the PI function became slightly shallower as
reverberation increased. This is likely due to changes in both
the speech signal and the noise caused by reverberation. In the
test conditions in this study, both speech and noise were
affected equally by reverberation. In the less reverberant
condition, the gaps in the noise still allow “glimpses” of the
speech signal. However, as reverberation increases, the noise
becomes more steady state thus reducing the release from
masking available under anechoic conditions (see George et al.
2008). This reduction in the modulation of the four-talker
babble used in this study was confirmed through inspection of
the waveforms of the noise in the anechoic and the three
reverberant test conditions.
Guidelines for interpreting SNR loss in adults are provided
in the BKB SIN manual (Etymotic Research 2005). SNR loss
of 0 to 3 dB is considered within normal limits, 3 to 7 dB a mild
loss, 7 to 15 dB a moderate loss, and ?15 dB a severe loss.
Using these guidelines, it seems that on average the 6-, 7-, and
8-yr-olds exhibit a “moderate” SNR loss under all reverberant
conditions, the 9- to 12-yr-olds exhibit a “mild” SNR loss, and
the adults are still within the normal range in all three
reverberant conditions tested.
The results of this study confirm developmental changes in
speech recognition performance in noise in a range of rever-
berant environments typifying a short, ANSI recommended,
and longer than recommended reverberation time for a location
in the reverberant field in a room typical of many elementary
school classrooms. Children’s SNR-50 values decrease with
increasing age. The oldest group in this study (age, 12 yr) had
thresholds significantly higher than adults. The literature on
classroom acoustics shows that occupied classrooms typically
have SNRs ?10 dB, with the classrooms of the youngest
children having SNRs between 0 and 5 dB (or even lower).
SNRs in the range of 5 to 7 dB would result in approximately
50% performance for the youngest children and less than
optimal performance for older children, whereas adults would
be expected to obtain maximal performance at these SNRs.
Improving the SNR will increase performance, but larger
increases in SNR will be required for children than adults
because the PI function is shallower for children than adults. In
the reverberant conditions tested in this study, on average,
SNRs of 15 dB could be anticipated to yield speech recognition
scores of 95% for children aged 6 yrs and older in classrooms
with reverberation times in compliance with the ANSI stan-
dard. In the longer reverberation time included in this study, the
youngest group required an SNR exceeding 15 dB. Results
support the importance of attention to classroom acoustics and
emphasize the need for maximizing SNR in classrooms,
especially in classrooms designed for early childhood grades.
The data collected in this study describe performance of
normal-hearing children under conditions of noise and rever-
beration. The results of this study raise concerns that even
higher SNRs might be required for children expected to have
greater difficulty recognizing degraded speech. The test
materials and protocol developed for this study can be used
in future studies on populations whose speech identification
performance may be particularly susceptible to degradations
to the speech signal (e.g., children or adults with hearing
impairment, language disorders, and second-language learn-
ers). Normative data collected in this study will serve as a
baseline against which children and adults from other
populations can be compared.
NEUMAN ET AL. / EAR & HEARING, VOL. 31, NO. 3, 336–344
343

Page 9

ACKNOWLEDGMENTS
The authors thank Frank Iglehart and the Clarke School for the Deaf for
providing access to the classroom and the absorptive panels used during the
recording process. They are grateful to Mario Svirsky, Matthew Fitzgerald,
and Elad Sagi for helpful discussions.
This work was supported by the National Institute on Disability and
Rehabilitation Research. The contents of this paper were developed under
a grant from the Department of Education, NIDRR grant number
H133E030006. However, those contents do not necessarily represent the
policy of the Department of Education, and you should not assume
endorsement by the Federal Government. Etymotic Research provided
ER6i earphones and ear tips for testing.
Portions of this study were presented in a poster at the 2009 meeting of the
American Academy of Audiology.
Address for correspondence: Arlene C. Neuman, PhD, NYU School of
Medicine, Department of Otolaryngology, 550 First Avenue, NBV 5E5,
New York, NY 10016. E-mail: arlene.neuman@nyumc.org.
Received October 14, 2009; accepted December 23, 2009.
REFERENCES
American National Standards Institute. (2002). Acoustical Performance
Criteria, Design Requirements and Guidelines for Schools. ANSI
12.60–2002. New York: ANSI.
American Speech Language Hearing Association. (1995). Position State-
ment and Guidelines for Acoustics in Educational Settings. ASHA, 37
(suppl. 14), 15.
Australian/New Zealand Standard. (2000). Acoustics-Recommended De-
sign Sound Levels and Reverberation Times for Building Interiors.
AS/NZS 2107: Sydney and Wellington: Sydney, Australia and Well-
ington, New Zealand: Standards Australia International Ltd. And
Standards New Zealand.
Bradley, J. S. (1986). Speech intelligibility studies in classrooms. J Acoust
Soc Am, 80, 846–854.
Bradley, J. S., & Sato, H. (2008). The intelligibility of speech in elementary
school classrooms. J Acoust Soc Am, 123, 2078–2086.
Cameron, S., Dillon, H., & Newall, P. (2006). Development and evaluation
of the listening in spatialized noise test. Ear Hear, 27, 30–42.
Crandell, C. C., & Smaldino, J. J. (2000). Classroom acoustics for children
with normal hearing and with hearing impairment. Lang Speech Hear
Serv Sch, 31, 362–370.
Department for Education and Skills. (2003). Building Bulletin 93:
Acoustic design for schools. London: The Stationary Office. Retrieved
August 20, 2009 from http://www.teachernet.gov.uk/management/
resourcesfinanceandbuilding/schoolbuildings/environ/acoustics/.
Dunn, L. M., & Dunn, L. M. (1997). Peabody Picture Vocabulary Test III.
Circle Pines, MN: American Guidance Service.
Etymotic Research. (2005). Bamford-Kowal-Bench Speech in Noise Test
(Version 1.03) Audio CD. Elk Grove Village, IL: Etymotic Research.
Finitzo-Hieber, T., & Tillman, T. W. (1978). Room acoustics effects on
monosyllabic word discrimination ability for normal and hearing-
impaired children. J Speech Hear Res, 21, 440–458.
Galster, J. A. (2007). The Effect of Room Volume on Speech Recognition in
Enclosures with Similar Mean Reverberation Time. Doctoral Disserta-
tion, Vanderbilt University, Nashville, Tennessee.
George, E., Festen, J., & Houtgast, T. (2008). The combined effects of
reverberation and nonstationary noise on sentence intelligibility.
J Acoust Soc Am, 124, 1269–1277.
Goldman, R., & Fristoe, M. (2000). Goldman-Fristoe Test of Articulation
(2nd ed.). Circle Pines, MN: American Guidance Service.
Johnson, C. E. (2000). Children’s phoneme identification in reverberation
and noise. J Speech Lang Hear Res, 43, 144–157.
Knecht, H. A., Nelson, P. B., Whitelaw, G. M., et al. (2002). Background
noise levels and reverberation times in unoccupied classrooms: predic-
tions and measurements. Am J Audiol, 11, 65–71.
Kodaras, M. J. (1960). Reverberation times of typical elementary school
classrooms. Noise Control, 6, 17–19.
Koehnke, J., & Besing, J. M. (1996). A procedure for testing speech
intelligibility in a virtual listening environment. Ear Hear, 17, 211–217.
Larsen, J. B., & Blair, J. C. (2008). The effect of classroom amplification
on the signal-to-noise ratio in classrooms while class is in session. Lang
Speech Hear Serv Sch, 39, 451–460.
McCroskey, R. L., & Devens, J. S. (1975). Acoustic characteristics of
public school classrooms constructed between 1890 and 1960.
NOISEXPO Proc, 101–103.
Moncur, J. P., & Dirks, D. (1967). Binaural and monaural speech
intelligibility in reverberation. J Speech Hear Res, 10, 186–195.
Neuman, A. C., & Hochberg, I. (1983). Children’s perception of speech in
reverberation. J Acoust Soc Am, 73, 2145–2149.
Pearsons, K. S., Bennett, R. L., & Fidell, S. (1976). Speech levels in
various noise environments (Report No. EPA-600/1–77-025), Washing-
ton DC: U.S. Environmental Protection Agency.
Picard, M., & Bradley, J. S. (2001). Revisiting speech interference in
classrooms. Audiology, 40, 221–244.
Plomp, R. (1976). Binaural and monaural speech intelligibility of con-
nected discourse in reverberation as a function of azimuth of a single
competing sound source (speech or noise). Acustica, 34, 200–211.
Sanders, D. (1965). Noise conditions in normal school classrooms. Except
Child, 31, 344–353.
Shield, B., & Dockrell, J. E. (2004). External and internal noise surveys of
London primary schools. J Acoust Soc Am, 115, 730–738.
Yacullo, W. S., & Hawkins, D. B. (1987). Speech recognition in noise and
reverberation by school-age children. Audiology, 26, 235–246.
Yang, W., & Bradley, J. S. (2009). Effects of room acoustics on the
intelligibility of speech in classrooms for young children. J Acoust Soc
Am, 125, 922–933.
NEUMAN ET AL. / EAR & HEARING, VOL. 31, NO. 3, 336–344
344