Key Words
audiology - binaural advantage - directional advantage - gold standard - pure-tone
thresholds - single-sided deafness - spatial advantage - speech recognition in noise
ability
INTRODUCTION
In the field of audiology, pure-tone threshold testing has been called the “gold standard”
for auditory assessment ([Sindhusake et al, 2001]; [Shargorodsky et al, 2010]). The degree of hearing loss has been described in terms of pure-tone average (PTA)
or pure-tone thresholds. According to [Goodman (1965)], normal hearing corresponds to a range of PTAs for thresholds at 500, 1000, and
2000 Hz. On the other hand, [Jerger and Jerger (1980)] described normal hearing based on individual pure-tone thresholds that fall in the
range of 0–20 dB HL for 250–8000 Hz (re: ANSI-69). [Steinberg et al (1940)] reported that individuals with pure-tone thresholds of 25 dB HL (at frequencies
up to 1760 Hz) will be aware of a hearing impairment for the perception of speech
under conditions of public address, such as in a church, theater, or around the conference
or dinner table.
The World Health Organization (WHO) and American Medical Association (AMA) Ratings
of Hearing Impairment and Single-Sided Deafness (SSD)
The WHO states that a PTA0.5, 1.0, 2.0, 4.0 kHz ≤ 25 dB HL for the better ear represents no impairment ([Mathers et al, 2000]). According to this classification scheme, a patient with normal pure-tone thresholds
for the better ear and a profound hearing loss for the poorer ear (PTA0.5, 1.0, 2.0, 4.0 kHz ≥ 81 dB HL) would have “no or very slight hearing problems.” [Olusanya et al (2014)] wrote that this definition excludes individuals with mild or unilateral hearing
impairment. The WHO’s classification of hearing disregards any functional impairment
of speech perception in background noise for unilateral hearing loss or SSD.
The WHO published the International Classification of Functioning, Disability and
Health (ICF). This document recognizes contextual factors such as environmental noise
on hearing impairment. The ICF lists each function of an individual and defines disability
as a decrement of each functioning domain. According to [WHO (2010)], the ICF is impractical for the determination of disability in daily practice; therefore,
they developed the WHO Disability Assessment Schedule to address this need and provide
a standardized method to measure health and disability across cultures. In the section
on cognition, there is one question regarding speech recognition in noise ability.
[Danermark et al (2013)] commented that the ICF and the ICF Core Sets were developed as a way to counteract
the fragmented approach of specialization found in modern healthcare. Currently, however,
there is no formal audiological test battery associated with the ICF Core Sets.
The AMA’s Guides to the Evaluation of Permanent Impairment ([AMA, 2008]) states that if the PTA0.5, 1.0, 2.0, 3.0 kHz is ≤25 dB HL, there is no impairment present for the ability to hear “everyday sounds
under everyday listening conditions.” If the PTA0.5, 1.0, 2.0, 3.0 kHz is >91.7 dB HL, the binaural hearing impairment is rated at 100% because this represents
the loss of the ability to hear everyday speech. Moreover, a patient with SSD and
a PTA0.5, 1.0, 2.0, 3.0 kHz ≤ 25 dB HL for the better ear would receive a binaural hearing impairment rating
of 16.7%. According to the AMA, for every dB greater than a PTA of 25 dB HL, a monaural
hearing impairment of 1.5% is assigned. When only one ear exhibits a hearing impairment,
0% impairment is assigned for the unimpaired ear and the following formula is used:
binaural hearing impairment = [5(percent hearing impairment for the better ear) +
(percent hearing impairment in the poorer ear)]/6. In the case of SSD, the monaural
hearing impairment is 0% for the better ear and 100% for the poorer ear. The binaural
hearing impairment according to the formula is [5(0%) + (100%)]/6 = 16.7%. However,
the ratings of binaural hearing impairment for the perception of speech in everyday
listening conditions are not supported by the literature for simulated or actual single-sided
hearing deficits.
Early Discussions on Monaural Versus Binaural Hearing
Vern O. Knudsen was the president of the Acoustical Society of America from 1933 to
1935. In his lecture at the 10th anniversary meeting, he reported an investigation
of binaural advantage ([Knudsen, 1939]). Eight participants with “normal hearing” were tested, first with both ears open
and then with one ear wearing a “special type of stopper” which provided an attenuation
of about 35 dB. The average speech recognition score was 72% for the binaural condition
and 69% for the monaural condition. It was noted that several participants were aware
of an increase in listening difficulty during the monaural condition. The author also
noted, “On other occasions I have engaged in group conversation with one of my ears
closed, and have found it difficult to shift auditory attention from one speaker to
another; it is especially trying when two or more persons speak at the same time.”
[Koenig (1950)] used a “binaural telephone system” with a separate microphone, amplifier, and receiver
for each ear of the study participants. For an alternate condition, the signals from
only one microphone were delivered to both ears. No systematic data collection was
reported for this study. The author noted that the binaural system provided the ability
to squelch reverberation and background noise as compared to the system with a common
microphone routed to each ear. [Haskins and Hardy (1960)] reported critical observations of patient preferences for two versus one hearing
aid for bilateral hearing losses. The authors wrote that, “much work must be devoted
to the development of better indicators of success in terms of stereophonic hearing:
and especially of ability to handle informational material in the presence of competing
messages.”
Self-Report of Hearing Difficulties for Patients with Unilateral Hearing Losses and
SSD
Even though the WHO considers patients with unilateral hearing impairment or SSD to
have “no impairment,” there is evidence that these types of hearing losses result
in noticeable hearing difficulties. [Dwyer et al (2014)] used the Speech, Spatial, and Qualities of Hearing scale (SSQ) subscales ([Gatehouse and Noble, 2004]; [Gatehouse and Akeroyd, 2006]) to determine the self-perception of hearing difficulties for participants with
normal hearing and unilateral hearing loss. The 21 participants in the normal-hearing
group had unaided thresholds <30 dB HL for both ears. The 30 members of the unilateral
hearing loss group had a mean PTA (0.25–6.0 kHz) of 13.0 dB HL for the better ear
and a severe-to-profound hearing loss for the poorer ear. The authors did not report
the pure-tone thresholds or PTA for the poorer ears in this study. According to the
SSQ responses, the unilateral hearing loss group reported significantly poorer speech
recognition in background noise ability than the normal-hearing group (p < 0.001).
[Douglas et al (2007)] used the SSQ to document the self-report of hearing abilities for control and SSD
groups. There were 127 participants in the control group with an age range from 50
to 80 years (mean = 52 years, SD = 13.2 years). The mean hearing threshold for the
better ear was 16.4 dB HL (SD = 12.6 dB). No “significant hearing asymmetry” was found
for the control-group participants. Forty-four patients with profound unilateral hearing
loss after an acoustic neuroma surgery were in the SSD group. The age range was from
20 to 74 years (mean = 51.3 years, SD = 13). The average hearing threshold for the
better ear was 16.4 dB HL (SD = 13.2) for the patient group. The authors did not report
the pure-tone thresholds or PTA for the poorer ears in this study. The SSQ ratings
of speech recognition in noise ability for the SSD group were significantly poorer
than those in the control group (p < 0.05).
Quantification of Speech Recognition in Noise Deficits for Simulated and Actual Asymmetrical
Conductive Hearing Losses
[Carhart (1965)] evaluated the binaural advantage for 16 “normal hearers.” The ages of the participants
were not given. Target monosyllabic words (phonetically balanced items, Normative
Update Test 2) were presented at 0° and 32 dB HL from one soundfield loudspeaker.
A second soundfield loudspeaker positioned at 90° was used for the presentation of
competing sentences. The level of the competing sentences was varied to achieve signal-to-noise
ratios (SNRs) of −24, −18, −12, and −6 dB. The SNR represents the difference in levels
between the signal (speech) and the noise (competing sentences). For example, a −24
dB SNR indicates that the signal was 24 dB below the level of the noise. The more
negative the SNR, the more difficult the listening task. A more positive SNR represents
a more favorable listening condition. Testing was also conducted for a quiet condition.
Each loudspeaker was six feet from the participant. The monaural listening condition
was achieved by occluding the ear closest to the competing sentences with an earplug
(shadowed ear condition). Mean discrimination scores (percent correct) were obtained
for each condition. While the binaural advantage was only 1.2% for the quiet condition,
it ranged from 3.3% to 8.7% across the noise conditions. From these data, the authors
determined that the binaural advantage provided an overall improvement in SNR of 2.5
dB.
[Persson et al (2001)] measured speech recognition in noise ability for binaural and monaural listening
conditions. Sixteen individuals with pure-tone thresholds ≤20 dB HL from 250 to 8000
Hz participated in the study. The mean age was 28 years. Testing was conducted in
an anechoic chamber. Speech stimuli were delivered via a loudspeaker at 0° for all
conditions. Noncoherent speech-weighted noise was presented bilaterally from loudspeakers
positioned at ±45° or ±90°. All soundfield loudspeakers were 1.25 m from the center
of each participant’s head. The noise was low-pass filtered with a cutoff frequency
of 1 kHz. A “hearing protector” was used for one ear to simulate a monaural listening
condition (simulated conductive hearing loss). Earplug types varied across participants.
The criterion for sufficient attenuation was ≥25 dB when both ears were plugged. When
this criterion was met, an earplug was removed from one of the ears for the monaural
condition.
Two types of speech in noise tests were administered. One test used a recording of
a male speaker reciting 50 phonetically balanced Swedish words presented at 65 dBC
while the noise was presented at 60 dBC (5 dB SNR). Each word was preceded by the
carrier phrase, “Now you hear…” A percent correct score was determined for the noise
±45° and ±90° conditions. The second test was a just-follow-conversation test. The
target speech was a recording of a female speaker reading text from a novel. The level
of the noise was fixed at 60 dBC. Each participant was instructed to adjust the volume
control so that they could just follow the context if they really concentrate on listening.
The mean SNR was calculated for each participant. For all test conditions, performances
for the binaural listening condition were significantly better than for the monaural
listening condition (p < 0.05). With minimum attenuation of 25 dB provided by the earplugs, speech presented
at 65 dBC would have reached the plugged ears at ∼40 dBC. Thus, there may have been
some contribution from the plugged ears during the monaural test conditions. The range
of earplug attenuation across participants was not reported. Nonetheless, this study
demonstrates the binaural advantage for a simulated unilateral hearing loss, even
though the ear with the earplug may have contributed to the “monaural” speech recognition
performances.
[Byun et al (2015)] determined the effect of surgical correction of unilateral congenital atresia on
26 children from 10 to 16 years of age. The participants were tested presurgery and
12 mo postsurgery. The self-perception of hearing ability was evaluated using the
Korean version of the SSQ. The ability to recognize speech in speech-spectrum noise
was measured using the Korean version of the Hearing in Noise Test (HINT; [Moon et al, 2008]) in a soundfield. The mean PTA(0.5, 1.0, 2.0, 3.0 kHz) improved from 63.9 to 39.4 dB 1 year postsurgery (p < 0.001). A comparison of mean speech recognition in noise performances revealed
improvements for all HINT conditions 12 mo postsurgery. Speech was always presented
at 0°. For the test condition where the noise was presented toward the atretic ear,
the HINT threshold improved 1.5 dB postsurgery (p = 0.014). For the Noise Front condition, a nonsignificant improvement of 0.8 dB was
found postsurgery. For the condition where the noise was directed toward the nonatretic
ear, the HINT threshold improved 1.7 dB postsurgery (p = 0.0005). The HINT Noise Composite Score improved by 1.0 dB postsurgery (p = 0.045). SSQ scores improved postsurgery for 82.6% of the participants. A significant
improvement for mean SSQ scores was found 12 mo postsurgery (p < 0.05).
Binaural Advantage for Speech Recognition in Noise Ability in a Simulated Soundfield
[Arsenault and Punch (1999)] determined the binaural advantage for speech recognition in noise ability for 10
participants with pure-tone thresholds ≤20 dB HL for 250–6000 Hz. The speech stimuli
were nonsense syllables (consonant–vowel or vowel–consonant) from the City University
of New York Nonsense Syllable Test ([Resnick et al, 1975]). Speech was presented at 0° and fixed at 65 dBC. Cafeteria noise was presented
at 270° (noise left condition). The level of the cafeteria noise was adjusted for
five SNR conditions; −8, −4, 0, 4, and 8 dB SNR. Recordings of the stimuli were made
using a Knowles Electronics Mannequin for Auditory Research (KEMAR) for each of the
SNR conditions. The speech and noise recordings were presented to 10 participants
with normal pure-tone thresholds. The ages of the participants were not presented
in this study. A mean binaural advantage of 4.9 dB was found for the binaural noise
left versus monaural (shadowed ear) performances. A mean binaural advantage of 12.3
dB was found for the binaural noise left versus monaural (unshadowed ear) performances.
The Effects of the Spatial Separation of Speech and Noise Signals on Binaural and
Monaural (Simulated Conductive Hearing Loss) Speech Recognition in Noise Ability
[Dubno et al (2008)] evaluated the benefit of the spatial separation of speech and noise stimuli and
the binaural advantage. The authors examined the differences between spatial separation
of the target speech and speech-shaped noise both with and without the head shadow
effect. The HINT ([Nilsson et al, 1994]; [Vermiglio, 2008]) was used with a nonstandard protocol to measure speech recognition in noise ability.
The speech-shaped noise was presented at 62 dB SPL. The level of the speech was adaptively
varied based on the participant’s response. A 3-dB step-size was used for first four
reversals and a 2-dB step-size was used for the remaining eight reversals. Fifteen
to 17 sentences were presented to obtain each HINT threshold. The threshold was defined
as the average level of the last six reversals. Each data point was the average of
two thresholds. HINT sentences were presented at 0° and speech-shaped noise at 0°,
90°, or ±90° in the soundfield. The results showed that spatial separation of the
speech and noise provided a greater benefit with the head shadow effect (4.8 dB SNR)
than without the head shadow effect (1.6 dB SNR) (p < 0.0001).
In the binaural and monaural listening conditions for the second experiment, HINT
sentences (speech) were presented at 0° and the speech-shaped masker was presented
at 90° in the soundfield. For the monaural listening condition, the ear closest to
the noise was plugged. This could also be called the “shadowed” monaural condition
where the head blocks the noise as it reaches the unplugged ear. The authors found
that even with an unfavorable SNR, thresholds for binaural listening were significantly
better compared with the monaural listening condition (1.6 dB; p < 0.0001). According to the authors, “with the speech emanating from the front and
the noise to one side, thresholds measured with the near ear plugged improved significantly
with the addition of a second [unshadowed] ear with an unfavorable SNR. Thus, the
advantage of interaural difference cues provided by a second ear outweighed that ear’s
poorer SNR.” This study shows the importance of conducting speech recognition in noise
ability where the head shadow effects may influence performance. This is in contrast
to pure-tone threshold testing where there are no head-shadow effects.
Dubno and colleagues used an E-A-R foam earplug (Aearo Company, Indianapolis, IN).
According to the authors, the earplug was inserted deeply into the participant’s ear
canal and maximum attenuation was confirmed by the participant. The authors further
stated that the articulation index (AI) values predicted minimal contribution of the
plugged ear to speech recognition in noise ability, “as estimated from participants’
thresholds, the levels of the speech and masker, and the manufacturer’s specification
for attenuation across frequency provided by the earplug.” The noise-reduction rating
for the E-A-R foam earplug was 29 dB. The highest average level of the speech at threshold
in this study was ∼61 dB SPL. This means that at threshold the sentences would reach
the plugged ear at ∼32 dB SPL. Because all of the participants had normal hearing,
the sentences would have reached the plugged ear above threshold. Again, the “deaf”
ear in this study would have heard the speech stimuli delivered “monaurally.” This
may have influenced the binaural advantage found in this study. Furthermore, whereas
the authors used the AI to predict the contribution of the plugged ear, [Vermiglio et al (2012)] demonstrated that the AI is not a strong predictor of speech recognition in noise
ability.
Summary of Literature Review
Pure-tone thresholds have been used as the “gold standard” for the assessment of the
ability to hear ([Shargorodsky et al, 2010]). The WHO has stated that as long as the better hearing ear has a PTA(0.5, 1.0, 2.0, 4.0 kHz) ≤ 25 dB HL, the individual will have no or very slight hearing problems ([Mathers et al, 2000]). However, this conclusion has not been borne out in self-report studies of patients
with unilateral hearing loss ([Dwyer et al, 2014]; [Byun et al, 2015]) or SSD ([Douglas et al, 2007]). According to the SSQ results, the patient groups in these studies reported significantly
poorer speech recognition in noise ability than the control groups. Furthermore, measurable
speech recognition in noise deficits have been found for participants with a simulated
unilateral conductive hearing loss ([Persson et al, 2001]; [Dubno et al, 2008]) and unilateral congenital atresia ([Byun et al, 2015]). Deficits have also been found for simulated SSD in a virtual soundfield environment
([Arsenault and Punch, 1999]).
Binaural advantage for speech recognition in noise ability has been demonstrated within
atretic patients pre- and postsurgery ([Gray et al, 2009]; [Byun et al, 2015]), and between control participants and patients with unilateral hearing losses ([Ruscetta et al, 2005]). [Dubno et al (2008)] investigated binaural advantage within participants between binaural and monaural
(plugged) conditions. However, since the monaural condition included an earplug for
the nontest ear, the extent of the binaural advantage may have been limited because
of the potential audibility of the speech stimuli for the plugged ear. [Arsenault and Punch (1999)] measured the binaural advantage for the ability to recognize nonsense syllables
in cafeteria noise. Testing was conducted binaurally and monaurally under headphones
in a simulated soundfield environment via KEMAR recordings. This approach decreases
the risk of audibility of the speech stimulus by the nontest ear when compared with
studies that used an earplug for the monaural listening conditions. Their methods
included testing with the spatial separation of the speech and noise signals.
Purpose
The overall goal of the present study was to evaluate the WHO and AMA ratings of hearing
impairment for SSD. The purpose of the study was to demonstrate binaural advantage
and conversely the impact of simulated SSD for speech recognition in noise ability.
Similar to the study by [Arsenault and Punch (1999)], participants with normal pure-tone thresholds were evaluated binaurally and monaurally.
However, instead of listening to nonsense syllables in cafeteria noise, the participants
listened to HINT sentences in steady-state speech-shaped noise in a simulated soundfield
environment. All stimuli were presented under headphones. This method provides less
interference from crossover to the nontest ear than previous studies that used earplugs.
The listening conditions included the spatial separation of the speech and noise stimuli.
Both binaural and spatial advantage measures were determined for the participants.
The spatial advantage (or the improvement in speech recognition in noise ability with
the spatial separation of the speech and noise stimuli) was determined for the binaural
and monaural conditions. The following research questions were addressed:
-
What is the binaural advantage for speech recognition in noise ability within participants?
-
What is the relationship between monaural and binaural HINT thresholds?
-
What is the spatial advantage for the binaural condition?
-
What is the spatial advantage for the monaural condition?
METHODS
Permission to conduct this research study was obtained from the East Carolina University
Institutional Review Board. Participation criteria included normal pure-tone thresholds
(≤25 dB HL for 500–8000 Hz) and clear ear canals bilaterally. All participants were
native speakers of American English. Twenty-eight participants were initially tested
for this study. Data from three participants were omitted because of pure-tone thresholds
poorer than 25 dB HL. The average age was 20.8 years (SD = 1.27). The age range was
from 19 to 25 years. Data from 24 females and one male were included in this study.
The American English HINT ([Nilsson et al, 1994]; [Vermiglio, 2008]) was used to measure the ability to recognize speech in steady-state noise. Sentences
were presented in “speech-shaped” noise at a fixed level of 65 dBA. Each HINT condition
was conducted using a single list of 20 sentences. Testing was conducted in a simulated
or “virtual” soundfield environment using KEMAR head-related transfer-functions. All
stimuli were prerecorded. The sentences were presented at 0° for each test condition.
The noise was presented at 0°, 90°, and 270° for the Noise Front, Noise Right, and
Noise Left conditions, respectively. All test conditions were randomized. Telephonic
TDH-50P headphones were used to deliver the stimuli. Headphone testing allows for
the evaluation of monaural (or simulated SSD) and binaural hearing ability under conditions
where the speech and noise are spatially separated.
The simulated binaural and monaural noise conditions delivered under headphones are
illustrated in [Figures 1]–[3]. The Noise Front condition was measured binaurally and monaurally. The monaural
Noise Front thresholds were measured for the left and right ears. The Noise Left and
Noise Right thresholds were measured for the ear closest to the noise source (“unshadowed
ear”). The monaural Noise Left thresholds were measured for the left ear and the monaural
Noise Right thresholds were measured for the right ear.
Fig. 1 Illustration of the binaural and monaural Noise Front conditions simulated under
headphones.
Fig. 2 Illustration of the binaural and monaural Noise Left conditions simulated under headphones.
The “unshadowed” ear is used for the monaural condition.
Fig. 3 Illustration of the binaural and monaural Noise Right conditions simulated under
headphones. The “unshadowed” ear is used for the monaural condition.
The HINT uses an adaptive protocol where the level of the sentence presentation varies
based on the response of the participant. The participant’s task is to listen to and
repeat the sentence heard in the presence of the speech-shaped noise. If the participant
correctly repeats the sentence, the level of the speech for the following sentence
is decreased. If the participant incorrectly repeats the sentence, the level of the
speech for the following sentence is increased. A 4-dB step-size is used for the first
four sentence presentations. A 2-dB step-size is used for the remaining sentences.
The HINT threshold is the SNR where a participant recognizes 50% of the sentences.
The “variability” was also determined. This is the standard deviation (SD) of the
SNRs used for each test run. It is a measure of the stability or consistency of the
participant’s responses. The binaural advantage for each HINT condition was determined
by subtracting the binaural from the monaural threshold for each noise condition.
The spatial advantage was determined by subtracting the Noise Side threshold from
the Noise Front threshold for the binaural and monaural conditions. The HINT test
was administered using custom software provided by the House Ear Institute in Los
Angeles, CA.
Statistical analyses were conducted using the JMP Pro (V.12) software. The matched
pairs t-test was conducted to determine if there were statistically significant differences
between the binaural and monaural conditions. The Pearson correlation coefficient
was calculated between binaural and monaural test results.
RESULTS
The descriptive statistics for the HINT thresholds are presented in [Table 1]. The more negative the threshold in dB SNR, the better the speech recognition in
noise performance. The average binaural Noise Front threshold was –1.76 dB SNR. This
indicates that when the speech signal is 1.76 dB below the level of the noise, the
participants on average recognize 50% of the sentences. The average binaural thresholds
for the Noise Left and Noise Right conditions were –8.58 and –8.38 dB SNR, respectively.
The average monaural thresholds for the Noise Front (left ear) and Noise Front (right
ear) conditions were –0.41 and –0.70 dB SNR, respectively. The average monaural thresholds
for the Noise Left and Noise Right conditions were 3.00 and 2.53 dB SNR, respectively.
The descriptive statistics for the variability of each threshold run are presented
in [Table 2]. The average mean variability across all HINT conditions was 1.96 dB. This indicates
that the participant responses were consistent across the test conditions.
Table 1
Descriptive Statistics for All HINT Thresholds (dB SNR)
|
HINT Noise Front Binaural
|
HINT Noise Front Monaural (Left ear)
|
HINT Noise Front Monaural (Right ear)
|
HINT Noise Left Binaural
|
HINT Noise Left Monaural (Left ear)
|
HINT Noise Right Binaural
|
HINT Noise Right Monaural (Right ear)
|
|
Mean
|
−1.76
|
−0.41
|
−0.70
|
−8.58
|
3.00
|
−8.38
|
2.53
|
|
SD
|
0.92
|
1.21
|
1.21
|
1.97
|
1.51
|
1.95
|
1.30
|
|
n
|
25
|
25
|
25
|
25
|
25
|
25
|
25
|
|
Maximum
|
1
|
1.7
|
1.9
|
−0.7
|
6.5
|
−0.6
|
5.2
|
|
Minimum
|
−3.7
|
−3.2
|
−3.1
|
−11
|
0.7
|
−10.7
|
0.1
|
|
Range
|
4.7
|
4.9
|
5.0
|
10.3
|
5.8
|
10.1
|
5.1
|
Table 2
Descriptive Statistics of the Variability for All HINT Thresholds
|
HINT Noise Front Binaural
|
HINT Noise Front Monaural (Left ear)
|
HINT Noise Front Monaural (Right ear)
|
HINT Noise Left Binaural
|
HINT Noise Left Monaural (Left ear)
|
HINT Noise Right Binaural
|
HINT Noise Right Monaural (Right ear)
|
|
Mean
|
1.82
|
1.74
|
1.94
|
2.06
|
2.04
|
2.10
|
2.00
|
|
SD
|
0.27
|
0.31
|
0.44
|
0.45
|
0.51
|
0.50
|
0.38
|
|
n
|
25
|
25
|
25
|
25
|
25
|
25
|
25
|
|
Maximum
|
2.6
|
2.4
|
3
|
3.2
|
3.6
|
3
|
2.9
|
|
Minimum
|
1.4
|
1.2
|
1.3
|
1.6
|
1.2
|
1.5
|
0.4
|
|
Range
|
1.2
|
1.2
|
1.7
|
1.6
|
2.4
|
1.5
|
1.5
|
Notes: Variability = standard deviation for each HINT threshold search.
The binaural advantage represents the improvement in speech recognition ability for
two ears versus one. [Figure 4] illustrates the binaural advantage for the Noise Front conditions. Most of the 25
participants demonstrated a binaural advantage. However, a monaural advantage was
found for two (8%) of the participants. [Figure 5] shows a binaural advantage for all 25 participants (100%) for the Noise Side conditions.
The descriptive statistics for the binaural advantage are presented in [Table 3]. This was determined by subtracting the binaural from the monaural thresholds for
each HINT condition. A positive value indicates a binaural advantage. Negative values
indicate a monaural advantage. [Table 4] shows that significant binaural advantages were found for the Noise Front and Noise
Side conditions with Bonferroni correction (p < 0.01). On average, binaural speech recognition in noise performances were significantly
better than the monaural performances. The average binaural advantage for the Noise
Front condition was 1.2 and 11.25 dB for the Noise Side conditions.
Fig. 4 Binaural advantage for the Noise Front condition (AS = left ear, AD = right ear).
Fig. 5 Binaural advantage for the Noise Side conditions (AS = left ear, AD = right ear).
Table 3
Descriptive Statistics for the Binaural Advantage (dB)
|
Binaural Advantage Noise Front (Re: Left ear)
|
Binaural Advantage Noise Front (Re: Right ear)
|
Binaural Advantage Noise Left
|
Binaural Advantage Noise Right
|
|
Mean
|
1.35
|
1.06
|
11.58
|
10.92
|
|
SD
|
1.25
|
1.27
|
1.90
|
1.92
|
|
n
|
25
|
25
|
25
|
25
|
|
Maximum
|
3.3
|
3.8
|
14.1
|
14.3
|
|
Minimum
|
−1.9
|
−1.9
|
4.9
|
5.4
|
|
Range
|
5.2
|
5.7
|
9.2
|
8.9
|
Table 4
Binaural Advantage = Average Monaural Minus Binaural Thresholds for each HINT Condition
(Matched Pairs Analysis in JMP)
|
Noise Front (AS)
|
Noise Front (AD)
|
Average Noise Front
|
Noise Left
|
Noise Right
|
Average Noise Side
|
|
Binaural advantage
|
1.35 dB
|
1.06 dB
|
1.21 dB
|
11.58 dB
|
10.92 dB
|
11.25 dB
|
|
p-value with Bonferroni correction
|
<0.0006
|
0.0024
|
<0.0006
|
<0.0006
|
<0.0006
|
<0.0006
|
|
Estimated maximum intelligibility change
|
13.44%
|
10.57%
|
12.01%
|
85.33%
|
82.85%
|
84.09%
|
Notes: AS = left ear; AD = right ear.
According to [Nilsson et al (1994)], a 1-dB change in SNR is equivalent to a 10% change in speech recognition ability.
Freed (personal communication) created a feature in a previous version of the HINT
system ([HEI, 2007]) that was used to estimate the maximum change in intelligibility of a given HINT
threshold in reference to the mean normal performance. Freed’s method was used to
estimate the maximum change in intelligibility ([Table 4]) for the monaural as compared to the binaural thresholds (or predicted maximum binaural
improvement in percent correct). For the Noise Front listening condition, the maximum
improvement in speech recognition ability for two ears versus one is 13.44% in reference
to the monaural left-ear condition and 10.57% in reference to the monaural right-ear
condition. The average binaural advantage for the Noise Front conditions was 12.01%.
By contrast, the maximum improvement in speech-recognition ability for the Noise Side
conditions was much greater: 85.33% for the Noise Left condition and 82.85% for the
Noise Right condition. The average maximum improvement across the Noise Side conditions
was 84.09%.
Recall that the spatial advantage represents the improvement in thresholds when the
noise is spatially separated from the speech. The spatial advantage is illustrated
in [Figure 6] for the binaural conditions and [Figure 7] for the monaural conditions. For the binaural conditions, all participants with
the exception of one demonstrated a spatial advantage. All of the participants demonstrated
a spatial advantage for the monaural conditions. The spatial advantages are presented
in [Table 5]. This was determined by subtracting the Noise Side from the Noise Front thresholds
for the binaural and monaural conditions. The average spatial advantage was 6.72 and
−3.32 dB for the binaural and monaural conditions, respectively. This indicates that
while a significant improvement in HINT performance was found when the speech and
noise were spatially separated for the binaural condition, a significant spatial deficit
was found for the monaural (unshadowed) condition (p < 0.01). In other words, when referencing the unshadowed ear (monaural condition),
it is easier to recognize speech in noise when the stimuli are delivered from 0° than
when the noise is spatially separated from the speech. The estimated maximum intelligibility
improvement for the binaural conditions was 60.73% in reference to the Noise Left
condition and 59.33% in reference to the Noise Right condition. There was an estimated
maximum intelligibility change of 60.03% across the binaural conditions. For the monaural
conditions, there was a significant spatial advantage: −33.09% in reference to the
Noise Left condition and −31.44% in reference to the Noise Right condition. Across
these two conditions, there was an average estimated maximum intelligibility change
of −32.27%. This indicates that for the monaural conditions it was much easier to
listen to speech in noise when both stimuli are presented from 0° than when the speech
and noise are spatially separated. This may be considered intuitive because the noise
is directed toward the test (unshadowed) ear for the monaural conditions.
Fig. 6 Spatial advantage for the binaural conditions.
Fig. 7 Spatial advantage for the monaural conditions.
Table 5
Spatial Advantage = Noise Front Minus Noise Side Thresholds for Each for the Binaural
and Monaural Conditions (Matched Pairs Analysis in JMP)
|
Binaural NF - NL
|
Binaural NF - NR
|
Average Binaural
|
Monaural (AS) NF - NL
|
Monaural (AD) NF - NR
|
Average Monaural
|
|
Spatial advantage
|
6.82 dB
|
6.62 dB
|
6.72 dB
|
−3.41 dB
|
−3.23 dB
|
−3.32 dB
|
|
p-value with Bonferroni correction
|
<0.0006
|
<0.0006
|
<0.0006
|
<0.0006
|
<0.0006
|
<0.0006
|
|
Estimated maximum intelligibility change
|
60.73%
|
59.33%
|
60.03%
|
−33.09%
|
−31.44%
|
−32.27%
|
Notes: NF = Noise Front threshold; NR = Noise Right threshold; NL = Noise Left threshold.
The correlation coefficients for the binaural versus monaural conditions and p-values are presented in [Table 6]. No significant correlations were found between the binaural Noise Front and Noise
Side thresholds. A relatively strong positive correlation was found between the binaural
Noise Left and Noise Right performances (r = 0.6372, p < 0.05). A moderate correlation was found between binaural Noise Left and monaural
Noise Left thresholds (r = 0.4264, p < 0.05) and between the binaural Noise Left and monaural Noise Right thresholds (r = 0.4560, p < 0.05). No significant correlations were found between the binaural Noise Right
and monaural Noise Front or Noise Side thresholds. A significant positive correlation
was found for monaural Noise Front (left versus right ear) thresholds (r = 0.5719, p < 0.05). However, no significant correlation was found between the monaural Noise
Side conditions. Significant positive correlations were also found between the monaural
Noise Front (left ear) versus the monaural Noise Left conditions (r = 0.5451, p < 0.05) and between the monaural Noise Front (left ear) versus the monaural Noise
Right conditions (r = 0.4546, p < 0.05). While no significant correlation was found between the monaural Noise Front
(right ear) and the monaural noise Left thresholds, a significant correlation was
found between the Noise Front (right ear) and the monaural Noise Right thresholds
(r = 0.4334, p < 0.05).
Table 6
Correlation Coefficients for the Binaural vs. Monaural Conditions (p-values in Parentheses)
|
Binaural Noise Left
|
Binaural Noise Right
|
Monaural Noise Front (AS)
|
Monaural Noise Front (AD)
|
Monaural Noise Left (AS)
|
Monaural Noise Right (AD)
|
|
Binaural noise front
|
0.0288 (0.8914)
|
0.0425 (0.8400)
|
0.3395 (0.0969)
|
0.3103 (0.1311)
|
−0.1421 (0.4979)
|
0.3067 (0.1359)
|
|
Binaural noise left
|
|
0.6372 (0.0006)
|
0.1880 (0.3681)
|
0.0696 (0.7409)
|
0.4264 (0.0335)
|
0.4560 (0.0220)
|
|
Binaural noise right
|
|
|
0.1551 (0.4590)
|
0.1823 (0.3831)
|
0.1964 (0.3467)
|
0.3598 (0.0773)
|
|
Monaural noise front (AS)
|
|
|
|
0.5719 (0.0028)
|
0.5451 (0.0048)
|
0.4546 (0.0224)
|
|
Monaural noise front (AD)
|
|
|
|
|
0.3697 (0.0689)
|
0.4334 (0.0304)
|
|
Monaural noise left (AS)
|
|
|
|
|
|
0.2526 (0.2232)
|
Notes: AS = left ear; AD = right ear. Significant correlations at the p < 0.5 level are presented in bold font.
DISCUSSION
A significant binaural advantage was found for all of the HINT conditions (p < 0.01). The average binaural advantage for the Noise Front condition was 1.21 dB.
This corresponds to an estimated maximum change in intelligibility of 12.01%. The
average binaural advantage for the Noise Side conditions was 11.25 dB. This corresponds
to an estimated maximum change in intelligibility of 84.09%. In other words, in reference
to the unshadowed ear for the Noise Side condition, an individual on average will
detect up to 84.09% more of the conversation with two ears than when listening with
only one ear. This is in stark contrast to the degree of hearing impairment for SSD
as determined by the WHO and AMA methods based on PTA. Recall that [Arsenault and Punch (1999)] found a binaural advantage for the recognition of nonsense syllables in cafeteria
noise of 12.3 dB for the “Noise Left” condition. This is similar to the 11.58 dB binaural
advantage for the present study for the recognition of sentences in steady-state speech-shaped
noise for the Noise Left condition.
A significant spatial advantage was found for the binaural and monaural conditions
(p < 0.01). The average binaural spatial advantage was 6.72 dB. This corresponds to
an estimated maximum change in intelligibility of 60.03%. However, the average spatial
advantage for the monaural HINT condition was −3.32 dB, which corresponds to an estimated
maximum change in intelligibility of −32.27%. In other words, for the monaural condition
the participants performed poorer when the speech and noise were spatially separated
than when the stimuli were presented from the same location. Spatial advantage is
not reflected in the results from the PTA methods used by the WHO and AMA.
The Monaural Test Condition and Simulated SSD
The results demonstrated the binaural advantage and conversely the effect of simulated
SSD in a virtual soundfield environment presented under headphones. It should be noted
that although no signal was presented to the simulated deaf ear during the monaural
conditions, it was possible for stimuli delivered to the monaural test ear to reach
the simulated deaf ear. Recall that TDH-50P headphones were used for the delivery
of all stimuli for the present study. According to [Blackwell et al (1991)], the TDH-50P has a mean interaural attenuation of 51.25 dB for 250 Hz, 57.08 dB
for 500 Hz, 60.00 dB for 1000 Hz, 60.00 dB for 2000 Hz, and 64.17 dB for 4000 Hz.
Across the frequency range of 250–4000 Hz, there is an average attenuation of 58.5
dB for the TDH-50P headphones. The highest average HINT threshold was 3 dB SNR for
the monaural Noise Left condition. This means that across test conditions the highest
level of the sentences at threshold was 68 dBA. This speech level minus the average
interaural attenuation for the TDH 50P headphones of 58.5 dB, results in an average
speech level of 9.5 dBA delivered to the “deaf” ear. According to the unpublished
norms at East Carolina University, the HINT threshold in quiet is 25.42 dBA. Therefore,
the speech stimulus presented to the hearing ear in the monaural condition would reach
the simulated deaf ear via bone conduction at an inaudible level. Overall, the monaural
conditions in this study appear to be a reasonable simulation of SSD. This is in contrast
to studies that have used an earplug for the monaural listening condition where the
nontest ear may have heard the target speech delivered to the test ear ([Persson et al, 2001]; [Dubno et al, 2008]).
SSD and the Rating of Hearing Impairment
Recall that according to the WHO ([Mathers et al, 2000]), a better ear PTA(0.5, 1.0, 2.0, 4.0 kHz) ≤ 25 dB HL represents “no impairment.” This implies that patients with SSD would
have “no or very slight hearing problems.” The results of the present study and previous
studies, however, do not support this classification when the ability to recognize
speech in background noise is a factor and especially in conditions where the speech
and noise are spatially separated. Even though the WHO has published the ICF, which
recognizes contextual factors such as environmental noise on hearing impairment, speech
recognition in noise ability is not a factor for the WHO’s classification of hearing
impairment. The AMA Guides ([AMA, 2008]) have adopted the terminology and conceptual framework of disablement as put forth
by the WHO’s ICF. The ICF framework is intended to describe and measure health and
disability.
According to the AMA Guides, hearing loss is evaluated using pure-tone thresholds. This is considered an “objective”
measure of hearing. However, a pure-tone threshold test relies on the subjective impression
of audibility from the patient. The AMA Guides states that binaural hearing impairment is determined through the evaluation of pure-tone
thresholds for both ears. However, the term “binaural” means the perception of sound
with both ears. The combination of monaural pure-tone thresholds is not equivalent
to hearing with both ears together. The rationale behind the rating of hearing impairment
is related to the ability to hear “everyday speech.”
According to the Hearing Handicap Guide published by the American Academy of Otolaryngology and the American Council of Otolaryngology
([AAO-ACO, 1979]), a handicap or impairment is “a medical condition that affects one’s personal efficiency
in the activities of daily living.” The AAO-ACO wrote that the basis for the calculation
of hearing handicap should be modified to reflect the understanding of speech, not
only in a quiet environment but also in the presence of some noise. At that time,
they noted that there was no standardized test for this assessment; therefore, they
recommended a determination of hearing impairment based on the decibel sum of pure-tone
threshold levels for 500, 1000, 2000, and 3000 Hz. According to their rating scheme,
pure-tone thresholds of 25 dB HL from 500 to 3000 Hz would represent 0% hearing impairment.
For a deaf ear, the patient would have a hearing impairment of 100%. The binaural
impairment is equal to five times the percent impairment for the better ear plus one
times the percent impairment for the poorer ear. This amount is then divided by six
to determine the percent of handicap. For the patient with SSD in this example, the
percent handicap is equal to 16.7%. This is the same method for the determination
of hearing impairment described in the AMA Guides.
[AAO-ACO (1979)] referenced recommendations for the improvement of the evaluation of hearing impairment
from the Council on Physical Medicine and Rehabilitation ([CPMR, 1955]). The CPMR noted, “…the most valid way to measure the ability to hear speech correctly
is to use words or sentences.” They further commented that, “hearing and recognizing
the spoken word is more than receiving a set of independent signals; it is a dynamic
process in which time is a factor and in which there are complicated interactions.
It seems logical to use speech as the material in a test of the ability to hear speech.”
The authors recommended that once the test materials are developed, the methods for
calculating hearing disability should be verified in terms of sentence intelligibility.
The [AAO-ACO (1979)] PTA method for the determination of hearing handicap was based in part on a study
conducted by [Kryter et al (1962)]. The authors evaluated the relationship between speech recognition in noise ability
and pure-tone thresholds for 114 adult male soldiers with various audiometric configurations.
Speech recognition ability was conducted in quiet and in noise with a spectrum shape
similar to the long-term idealized spectrum of speech. Testing was conducted using
words and sentences. The speech stimulus was delivered at 65 dB [sic]. Speech-recognition
ability was determined for a range of SNRs. The results demonstrated a strong relationship
(R = 0.806) between a PTA(1.0, 2.0, 3.0 kHz) and speech recognition in noise ability. The authors concluded that this PTA would
be a reasonably valid method for the determination to understand everyday speech.
There are two problems with this conclusion. First, the authors omitted data from
participants with normal pure-tone thresholds, and second, the strong positive relationship
between PTA(1.0, 2.0, 3.0 kHz) and speech recognition in noise results was most likely driven by the audibility
of the test materials.
Not all investigations of the relationship between PTA and speech recognition in noise
performances have revealed strong correlations. [Vermiglio et al (2012)] reported no significant relationships between PTA (0.5–2.0, 3.0–6.0, or 0.5–6.0
kHz) and thresholds for the HINT Noise Front condition. [Wilson et al (2007)] reported weak but significant correlations between PTA (0.5–2.0 kHz) and Bamford-Kowal-Bench
Speech-In-Noise test results (r = 0.292, p < 0.05). On the other hand, [Tschopp and Züst (1994)] reported a relatively strong relationship between PTA (0.5–4 kHz) and the German
Speech-in-Noise test (r = 0.740, p < 0.05). The strength of the relationships between PTA and speech recognition in
noise ability may be affected by the audibility of the stimulus ([Vermiglio, 2007]). This in turn is dependent on the level of the stimuli and the configuration of
hearing sensitivity. Furthermore, listeners with normal pure-tone thresholds may exhibit
a range of speech recognition in noise performances that match those of hearing-impaired
participants ([Dickson et al, 1946]; [Vermiglio et al, 2012]). The ability to recognize speech in noise should be measured directly and not inferred
from the PTA.
Measures of Speech Recognition in Noise Ability
The measurement of the ability to recognize speech in noise has been investigated
over the past 80 years. [Fletcher (1929)] demonstrated how performance improved as the SNR improved. [Fry (1942)] suggested that pure-tone thresholds would not represent a complete assessment of
the ability to hear. He recommended the development of a speech-recognition-in-noise
test for Royal Air Force candidates. The development of this “efficiency test” was
presented by [Dickson et al (1946)]. The test protocol included listening to recorded words and sentences in the presence
of aircraft noise. The authors noted “…it was difficult to predict from a subject’s
audiogram whether he was likely to do well or badly in an articulation test in a noise
field.” They also stated that speech recognition in noise scores from a “sample of
subjects with ‘normal’ audiograms may scatter just as widely as those from a group
of subjects with varying degrees of hearing loss.” The authors recommended a two-part
protocol for hearing evaluations. The first part was the efficiency test. If the candidate
failed this test, no further evaluations were conducted and the candidate was rejected.
If the candidate passed the efficiency test, then pure-tone threshold testing was
conducted to rule out any clinical reason for rejecting the candidate in spite of
passing the efficiency test. It is apparent from the protocol that the ability to
recognize speech in background noise was considered more important than the ability
to detect pure-tone stimuli. Even though speech recognition in noise testing has been
conducted since the 1920s, it has not found widespread use in audiology clinics today.
[ASHA (2015)] noted in a survey of 1811 audiologists that only 30% routinely conduct a validation
of outcomes using speech recognition in noise testing.
Dr. Shelly Chadha, the medical officer for the WHO Program for Prevention of Deafness
and Hearing Loss, reported that the WHO defines disabling hearing loss as “hearing
loss that is >40 dB” in the better hearing ear for adults ([Fabry and Clark, 2017]). She also stated that hearing problems and the resources available to address these
problems vary across the world. She noted that, “The public health perspective requires
developing a model of service and care that may ensure best practice but also allow
customization to meet regional or country-specific issues.” Although speech recognition
in noise testing may not be available for all areas of the world, when it is available
it would be preferable to rate hearing impairment based on speech-in-noise test results
than PTA.
Summary and Conclusion
According to the WHO and AMA PTA methods, the degree of hearing impairment for an
individual with SSD is none or “slight.” This is contrary to self-report data from
patients with SSD ([Douglas et al, 2007]). Studies that have investigated the binaural advantage where the nontest ear was
plugged during the monaural condition ([Carhart, 1965]; [Persson et al, 2001]; [Dubno et al, 2008]) may have produced results that were confounded by the audibility of the stimuli
for the nontest ear. This may have resulted in an underestimation of the binaural
advantage. An alternative to measuring binaural benefit is to obtain monaural and
binaural speech recognition in noise thresholds under headphones in a virtual soundfield
environment ([Arsenault and Punch, 1999]). This minimizes the risk of the nontest ear hearing the speech stimuli for the
monaural test conditions.
It was noted by the [CPMR (1955)] that speech would be a better test stimulus than pure-tone thresholds when determining
the degree of hearing impairment. Furthermore, the [AAO-ACO (1979)] made this very important point, “The limit of normal for the hearing of speech may
reasonably be assumed for the present to be comparable to that for the hearing of
pure-tones in the speech-frequency range, but it should be defined more precisely
in the future when more measurements and better validation become available.” Today
we have the instruments for a better validation of hearing impairment than PTA. The
use of speech recognition in noise testing allows clinicians, researchers, and physicians
to quantify hearing impairment not solely on pure-tone threshold results, but on the
ability to perceive speech in background noise.
It is clear from the results of the present and previous studies that the PTA method
used by the WHO and AMA for the determination of hearing impairment underestimates
the effect of SSD on the perception of speech in the presence of background noise.
The WHO and AMA criteria for the determination of hearing impairment should be updated
to include speech recognition in noise testing with and without the spatial separation
of the speech and noise stimuli. In this way, actual, as opposed to inferred perceptions
of speech in noisy environments may be determined. This will provide a much-needed
improvement in the ratings of hearing impairment.
Abbreviations
AAO-ACO:
American Academy of Otolaryngology and the American Council of Otolaryngology
AI:
articulation index
AMA:
American Medical Association
CPMR:
Council on Physical Medicine and Rehabilitation
HINT:
Hearing in Noise Test
ICF:
International Classification of Functioning, Disability and Health
KEMAR:
Knowles Electronics Mannequin for Auditory Research
PTA:
pure-tone average
SD:
standard deviation
SNR:
signal-to-noise ratio
SSD:
single-sided deafness
SSQ:
Speech, Spatial, and Qualities of Hearing scale
WHO:
World Health Organization
