speech understanding in quiet and noise, with and without hearing aids

10
Mathias Ha ¨ llgren* Birgitta Larsby* Bjo ¨ rn Lyxell $ Stig Arlinger* *Division of Technical Audiology, Department of Neuroscience and Locomotion $ Department of Behavioural Sciences, Linko ¨ping University, Linko ¨ ping, Sweden Original Article International Journal of Audiology 2005; 44:574 /583 Speech understanding in quiet and noise, with and without hearing aids Comprensio ´ n del lenguaje en silencio y con ruido, con y sin auxiliares auditivos Abstract Speech recognition and cognitive functions important for speech understanding were evaluated by objective mea- sures and by scores of perceived effort, with and without hearing aids. The tests were performed in silence, and with background conditions of speech spectrum random noise and ordinary speech. One young and one elderly group of twelve hearing-impaired subjects each participated. Hear- ing aid use improved speech recognition in silence (7 dB) and in the condition with speech as background (2.5 dB S/N), but did not change the perceived effort scores. In the cognitive tests no hearing aid benefit was seen in objective measures, while there was an effect of hearing aid use in scores of perceived effort, subjects reported less effort. There were no age effects on hearing aid benefit. In conclusion, hearing aid use may result in reduced effort in listening tasks that is not associated with improvement in objective scores. Sumario El reconocimiento del lenguaje y las funciones cognitivas importantes para la comprensio ´n del lenguaje se valoraron por medio de mediciones objetivas y con puntuaciones sobre el esfuerzo percibido, con y sin auxiliares auditivos. Las pruebas se aplicaron en silencio y con ruido aleatorio de fondo en el espectro del lenguaje y con lenguaje ordinario. Participaron un grupo de jo ´venes y uno de adultos mayores, de 12 personas hipoacu ´sicas cada uno. El auxiliar auditivo mejoro ´ el reconocimiento del lenguaje en silencio (7 dB) y en la condicio ´n de lenguaje de fondo (2.5 dB S/N) pero no cambio ´ las puntuaciones de esfuerzo percibidas. En las pruebas cognitivas no se apreciaron beneficios con el auxiliar auditivo con medidas objetivas, mientras que si hubo un efecto del uso del auxiliar auditivo en las puntuaciones de esfuerzo percibido, al reportar los sujetos un menor esfuerzo. No existieron efectos de la edad en el beneficio del auxiliar auditivo. En conclusio ´n, el uso del auxiliar auditivo puede resultar en un esfuerzo reducido para las tareas de atencio ´n que no se asocian con la mejorı ´a de las puntuaciones objetivas. In many situations in today’s society we are exposed to avariety of sounds which make communication more difficult. Listening in noise is related to a high degree of perceived effort (Larsby et al, 2005) and is often also associated with reduced speech understanding. It is well known that hearing-impaired persons suffer from problems with speech understanding in noise (e.g. van Rooij & Plomp, 1990). A hearing impairment is composed of two factors: 1) an attenuation factor whose effect is similar to reducing the overall level of both speech and noise, and 2) a distortion factor which means that hearing-impaired listeners need a higher speech-to-noise ratio to reach the same degree of speech recognition as normal-hearing persons (Hagerman 1984, Plomp 1986). Amplification through hearing aids compensates for the attenuation factor, but the distortion factor is harder to deal with. Hearing aid use improves speech perception in quiet conditions mainly due to increased audibility. In noise, there are reports of benefit (Alcantara et al, 2003; Haskell et al, 2002; Larson et al, 2002; Shanks et al, 2002) as well as of no benefit (e.g. Gustafsson & Arlinger, 1994) of hearing aid use in speech processing tasks. If a hearing aid benefit is seen it is often significantly smaller than in silence (Cord et al, 2000). Hearing aid benefit expressed as change in signal-to-noise ratio for a specified performance is used in many studies. However, changes in perceived effort of listening have rarely been studied. Speech understanding, among other things, depends on cognitive abilities (Lyxell et al, 2003; Pichora-Fuller, 2003; Lunner, 2003; Gatehouse et al, 2003) such as working memory capacity, speed in verbal information processing and phonolo- gical skills. In noise, there are higher demands of cognitive compensatory mechanisms to restore and interpret the limited and distorted sensory signal. For example, the use of visual information provides better speech understanding, but is cogni- tively demanding and requires extra resources (Larsby et al, 2005; Ha ¨llgren et al, 2001). In order to study these functions of speech processing in both auditory and audiovisual modalities a cognitive test battery has previously been developed (SVIPS- Speech and Visual Information Processing System; Ha ¨llgren et al, 2001). This test battery includes tests of different cognitive skills in the dimensions of both accuracy and speed of performance. Previous studies from our group using the SVIPS test battery have shown that performance in noise is generally worse and processing of information is more time-consuming than in quiet (Larsby et al, 2005; Ha ¨llgren et al, 2001). In addition, we found that different noises have differential effects on speech processing (Larsby et al, 2005). It was concluded that interfering noise has negative effects on level of accuracy and speed of processing. More specifically, hearing-impaired subjects have more problems in noise with temporal variations and some ISSN 1499-2027 print/ISSN 1708-8186 online DOI: 10.1080/14992020500190011 # 2005 British Society of Audiology, International Society of Audiology, and Nordic Audiological Society Received: May 10, 2004 Accepted: Mathias Ha ¨llgren Department of Neuroscience and Locomotion, Division of Technical Audiology, University Hospital, S-581 85 Linko ¨ ping, Sweden E-mail [email protected]

Upload: liu-se

Post on 30-Nov-2023

0 views

Category:

Documents


0 download

TRANSCRIPT

Mathias Hallgren*Birgitta Larsby*Bjorn Lyxell$

Stig Arlinger*

*Division of Technical Audiology,Department of Neuroscience andLocomotion$Department of Behavioural Sciences,Linkoping University, Linkoping,Sweden

Original Article

International Journal of Audiology 2005; 44:574�/583

Speech understanding in quiet and noise, with

and without hearing aids

Comprension del lenguaje en silencio y con ruido, con ysin auxiliares auditivos

AbstractSpeech recognition and cognitive functions important forspeech understanding were evaluated by objective mea-sures and by scores of perceived effort, with and withouthearing aids. The tests were performed in silence, and withbackground conditions of speech spectrum random noiseand ordinary speech. One young and one elderly group oftwelve hearing-impaired subjects each participated. Hear-ing aid use improved speech recognition in silence (7 dB)and in the condition with speech as background (2.5 dBS/N), but did not change the perceived effort scores. Inthe cognitive tests no hearing aid benefit was seen inobjective measures, while there was an effect of hearingaid use in scores of perceived effort, subjects reported lesseffort. There were no age effects on hearing aid benefit. Inconclusion, hearing aid use may result in reduced effort inlistening tasks that is not associated with improvement inobjective scores.

SumarioEl reconocimiento del lenguaje y las funciones cognitivasimportantes para la comprension del lenguaje sevaloraron por medio de mediciones objetivas y conpuntuaciones sobre el esfuerzo percibido, con y sinauxiliares auditivos. Las pruebas se aplicaron en silencioy con ruido aleatorio de fondo en el espectro del lenguajey con lenguaje ordinario. Participaron un grupo dejovenes y uno de adultos mayores, de 12 personashipoacusicas cada uno. El auxiliar auditivo mejoro elreconocimiento del lenguaje en silencio (7 dB) y en lacondicion de lenguaje de fondo (2.5 dB S/N) pero nocambio las puntuaciones de esfuerzo percibidas. En laspruebas cognitivas no se apreciaron beneficios con elauxiliar auditivo con medidas objetivas, mientras que sihubo un efecto del uso del auxiliar auditivo en laspuntuaciones de esfuerzo percibido, al reportar lossujetos un menor esfuerzo. No existieron efectos de laedad en el beneficio del auxiliar auditivo. En conclusion,el uso del auxiliar auditivo puede resultar en un esfuerzoreducido para las tareas de atencion que no se asociancon la mejorıa de las puntuaciones objetivas.

In many situations in today’s society we are exposed to a variety

of sounds which make communication more difficult. Listening

in noise is related to a high degree of perceived effort (Larsby et

al, 2005) and is often also associated with reduced speech

understanding. It is well known that hearing-impaired persons

suffer from problems with speech understanding in noise (e.g.

van Rooij & Plomp, 1990). A hearing impairment is composed

of two factors: 1) an attenuation factor whose effect is similar to

reducing the overall level of both speech and noise, and 2) a

distortion factor which means that hearing-impaired listeners

need a higher speech-to-noise ratio to reach the same degree of

speech recognition as normal-hearing persons (Hagerman 1984,

Plomp 1986). Amplification through hearing aids compensates

for the attenuation factor, but the distortion factor is harder to

deal with. Hearing aid use improves speech perception in quiet

conditions mainly due to increased audibility. In noise, there are

reports of benefit (Alcantara et al, 2003; Haskell et al, 2002;

Larson et al, 2002; Shanks et al, 2002) as well as of no benefit

(e.g. Gustafsson & Arlinger, 1994) of hearing aid use in speech

processing tasks. If a hearing aid benefit is seen it is often

significantly smaller than in silence (Cord et al, 2000). Hearing

aid benefit expressed as change in signal-to-noise ratio for a

specified performance is used in many studies. However, changes

in perceived effort of listening have rarely been studied.

Speech understanding, among other things, depends on

cognitive abilities (Lyxell et al, 2003; Pichora-Fuller, 2003;

Lunner, 2003; Gatehouse et al, 2003) such as working memory

capacity, speed in verbal information processing and phonolo-

gical skills. In noise, there are higher demands of cognitive

compensatory mechanisms to restore and interpret the limited

and distorted sensory signal. For example, the use of visual

information provides better speech understanding, but is cogni-

tively demanding and requires extra resources (Larsby et al,

2005; Hallgren et al, 2001). In order to study these functions of

speech processing in both auditory and audiovisual modalities a

cognitive test battery has previously been developed (SVIPS-

Speech and Visual Information Processing System; Hallgren et

al, 2001). This test battery includes tests of different cognitive

skills in the dimensions of both accuracy and speed of

performance. Previous studies from our group using the SVIPS

test battery have shown that performance in noise is generally

worse and processing of information is more time-consuming

than in quiet (Larsby et al, 2005; Hallgren et al, 2001). In

addition, we found that different noises have differential effects

on speech processing (Larsby et al, 2005). It was concluded that

interfering noise has negative effects on level of accuracy and

speed of processing. More specifically, hearing-impaired subjects

have more problems in noise with temporal variations and some

ISSN 1499-2027 print/ISSN 1708-8186 onlineDOI: 10.1080/14992020500190011# 2005 British Society of Audiology, InternationalSociety of Audiology, and Nordic Audiological Society

Received:May 10, 2004Accepted:

Mathias HallgrenDepartment of Neuroscience and Locomotion, Division of Technical Audiology,University Hospital, S-581 85 Linkoping, SwedenE-mail [email protected]

individuals are more distracted by noise with meaningful

content. Previous studies have shown that competing speech

with meaningful content is more difficult to ignore than noise

without meaning (Tun et al, 2002; Tun & Wingfield, 1999). Since

speech processing in everyday life is often disturbed by other

persons speaking, it is important to study how the cognitive load

on the listener’s speech processing depends on different acoustic

and linguistic characteristics in the disturbing background. In

the present study, we compare speech understanding in slightly

modulated speech spectrum noise (Hagerman, 1982) and in

ordinary speech (one speaker); see Figure 1.

Whether or not hearing aid amplification facilitates perfor-

mance at cognitive tasks, in silence as well as in noise, is a

question for debate. In silence, amplification increases audibility

which increases sensory driven bottom-up processing and

reduces the need for cognitively demanding top-down processing.

In noise, both the speech signal and the noise are amplified,

which may make the signal more audible. However, at the same

time the masking and the distraction from the noise increases,

leading to higher demands on cognitive processing. Both speech

understanding and the degree of distraction depend on the type

of interfering noise. Gatehouse et al (2003) found that listeners

with greater degrees of cognitive ability exhibited better speech

recognition, which was most evident in noise with temporal

variations. Davis (2003), however, showed that listeners with

better cognitive functions benefited less from amplification in

noise.

It has been shown that the negative effects of noise lead to an

increased degree of listening effort (Larsby et al, 2005). An

interesting issue is how hearing aid amplification changes the

degree of perceived effort. One possibility is that the increased

effort in noise is reduced or eliminated by hearing aid

amplification. Another possibility is that hearing aid amplifica-

tion leads to a larger cognitive challenge, and thus to an

increased degree of perceived effort.

An increased degree of perceived effort with a non-affected

performance in speech recognition tasks might, in the long term,

have consequences for speech understanding. The goal for

hearing aid fitting must both be to optimise speech under-

standing and to minimise the degree of perceived effort.

The fact that elderly, hearing-impaired persons have addi-

tional problems in noisy environments has been well documen-

ted (Gustafsson & Arlinger, 1994; CHABA, 1988). Cognitive

abilities decrease with increasing age and the elderly face more

problems in speech understanding tasks (Larsby et al, 2005;

Hallgren et al, 2001; Ronnberg, 1990). How the decreased

cognitive ability influences whether the elderly benefit from

amplification and the increased audibility provided by hearing

instruments is an interesting question.

The objective of the present investigation is to study the effect

of hearing aids on perceived effort, speech perception, and

cognitively demanding speech understanding tasks, in silence as

well as in noisy environments. Specifically, we investigate how

the benefit from hearing aid amplification is related to:

�/ the background condition of silence or noise

�/ the meaningfulness of the interfering noise

�/ the age of the hearing-impaired subject

�/ the signal presentation modality (auditory/audiovisual)

Methods

SubjectsAll subjects were hearing-impaired with sensorineural hearing

losses of mild-to-moderate degree. They were all experienced

hearing aid users (at least nine months) and used bilaterally

fitted hearing aids (Table 1) with their ordinary settings. The

majority of the subjects wore Oticon Digifocus II hearing aids

(20 of 24), where the signal processing is two-band automatic

gain control with a short time constant in the low-frequency

band and a long time constant in the high-frequency band. Two

age groups, each including twelve subjects, participated in the

study, a young group aged 25�/45 years (mean�/36.8, SD�/7.1)

and an elderly group aged 65�/80 years (mean�/71.8, SD�/3.8).

The subjects were paid a small amount (approximately t 10) for

their participation.

Figure 1. Waveforms of the two noise distracters.

Table 1. Hearing aids of the participating subjects

Kind of HA Number

Oticon Adapto 1

Oticon Digifocus II 20

Resound Danalogic 163D 1

Siemens Cosmea Top S 1

Widex Senso 1

Speech understanding in quiet and noise,with and without hearing aids

Hallgren/Larsby/Lyxell/Arlinger 575

Background conditionAll tests, described below, were performed in quiet and in the

presence of two background noises, resulting in the following

background conditions:

1. Hagerman noise (HN). The Hagerman noise is a computer-

generated noise using digital samples of 5 word sentences

from the Hagerman test (Hagerman, 1982). The noise is

slightly amplitude modulated to simulate the amplitude

variations in speech babble (Hagerman, 1982); see Figure 1.

2. Speech. The voice of a female speaker reading a continuous

story from the novel ‘Nils Holgerssons underbara resa

genom Sverige’ (The Wonderful Adventures of Nils) by

Selma Lagerlof; see Figure 1.

3. ‘No-noise’.

Hagerman speech testSpeech recognition was measured with the Hagerman speech test

(Hagerman, 1982). This speech material consists of eleven lists

with ten sentences in each. Each sentence contains five low-

redundancy words. All sentences have the same structure: name

�/ verb �/ number �/ adjective �/ noun. The same 50 words appear

in all the lists but in different combinations. Between each

sentence there is a pause, long enough for the subject to repeat

the words that he recognised. The session started with a training

list and the S/N was initially large. Then the speech signal was

adjusted adaptively in order to reach 40% correct responses

(Hagerman & Kinnefors, 1995). Two more lists were presented

and the mean value of the S/N in these sentences was used as the

outcome measure.

SVIPSA cognitive test battery was used for assessment of speech and

visual information processing skills (SVIPS�/Speech and Visual

Information Processing System). The SVIPS tests, where both

number of correct answers and reaction times were measured,

are described below.

. Semantic decision making. The subject’s task was to decide

whether a word belonged to a certain pre-defined semantic

category or did not (compare to Shoben, 1982). Four trials

were used with 32 items, of which sixteen items belonged to a

semantic category and sixteen items did not. The four

categories used were ‘colors’, ‘occupations’, ‘diseases’, and

‘parts of the body’.

. Lexical decision making. The subject’s task was to judge

whether a combination of three letters was a real word or a

non-word. Thirty items were used in the test, fifteen being real

words and fifteen not. The real words used in the present test

were all familiar Swedish words according to Allen (1970).

. Name matching. The subject’s task was to judge whether two

presented letters were the same (e.g., A�/A) or not (A�/B). Ten

pairs of letters were presented.

The tests were presented in two different modalities:

. In an auditory version, the stimuli were presented acoustically

via a loudspeaker.

. In an audiovisual version, the stimuli were presented acous-

tically via a loudspeaker and the face of the reader was

simultaneously visible on a computer screen.

Rating perceived effort during listeningWe asked the subject to rate the degree of perceived effort during

performance of the tasks in all background conditions in the

different modalities of presentation of the signal. We used a

version of Borg’s CR-10 scale to score the degree of perceived

effort (Borg, 1990). This scale involves a combination of ratio

and category scaling where verbal expressions and numbers are

used congruently to determine values on a ratio scale ranging

from 0 (none at all) to 10 (extremely great). If the experience was

greater than 10 the subject was allowed to use larger numbers

than 10.

Word recognition testTo obtain an indication that the different words in the SVIPS

test battery were actually heard by the subject, a test list

comprising 25 real words from the SVIPS tests was put together.

This was presented to the subject at the stimulus and noise level

used in the SVIPS test (see below) as a last task in each

background condition.

Test procedureThe subject was seated in a sound-isolated chamber. The

acoustic signal in the tests was presented by a loudspeaker at a

distance of one meter in front of the subject. The visual

information in SVIPS was presented via a computer monitor,

showing the reader’s face and shoulders. The subject pressed

predefined response buttons, one for ‘yes’ and one for ‘no’.

Without hearing aids the sound level of the stimuli was 75 dB

SPL (equivalent level) as standard, but was adjusted to a higher

level if practice stimuli were not correctly repeated and clearly

heard. With hearing aids the stimuli were always presented at 75

dB SPL. A signal to noise ratio of �/10 dB was used in the

SVIPS tests. For more technical information see Hallgren et al

(2001). The same noise level as used in the SVIPS tests was also

used in the Hagerman test.

The tests were performed in two sessions (1.5 hours each) with

at least one week in between. Half of the subjects performed the

tests with hearing aids in the first session and without hearing

aids in the second session and vice versa for the other half of the

group. In both sessions all tests were performed three times, with

the different background conditions. The order of background

conditions was balanced across the subjects. The Hagerman

speech test was always performed first, followed by the SVIPS

battery, while the order of auditory and audiovisual modalities

was balanced across subjects.

After completing the Hagerman speech test and the SVIPS

battery, respectively, the subject was asked to rate the degree of

perceived effort by giving a numerical value on the Borg Scale

corresponding to the perceived effort during the listening tasks.

As the last task in each background condition, the subject had

to repeat the words in the word recognition test list.

Statistical analysisPure-tone hearing threshold levels were measured for the

frequency range 0.125�/8.000 kHz and entered into an ANOVA

to evaluate differences between subject groups. Further ANO-

VAs were carried out with test results in the Hagerman speech

test (dB SPL or signal-to-noise ratio) and SVIPS tests (accuracy

or reaction time) as dependent variables. When ceiling effects

were present, as in accuracy measurements in the SVIPS tests, an

576 International Journal of Audiology, Volume 44 Number 10

arcsine-transformation was carried out to avoid dependence

between means and standard deviations. The transformed values

were used in the ANOVA and the raw scores were used in the

figures. ANOVAs were also carried out with the scores from

rating of perceived effort. Finally, an ANOVA was performed

with the number of correct words identified in the word

recognition test as dependent variable. A p-value of 0.05 was

considered statistically significant throughout.

Results

Hearing thresholdsThe hearing threshold levels for the two groups are presented in

Figure 2. A three-way ANOVA was performed, with age group

(young, elderly) as between-group factor, audiometric frequency

(0.125, 0.25, 0.5, 1, 2, 3, 4, 6, and 8 kHz) and ear (left, right) as

within-subject factors and hearing threshold as dependent

variable. There were significant main effects of frequency (pB/

0.001) and ear (p�/0.03), but not of age. The interaction between

age-group and frequency was significant (pB/0.001). In the low-

frequency range the hearing threshold levels for the young

subjects were poorer than for the elderly subjects, while the

reverse was true in the high-frequency range. The interaction

between age group, frequency and ear was significant (p�/0.04).

Planned comparisons showed that the young subjects did not

differ between ears, neither in low (0.125�/1 kHz) nor in high

frequencies (4�/8 kHz), whereas the elderly differed between ears

in the high-frequency range only.

Hagerman speech test

OBJECTIVE MEASUREMENTS

Two ANOVA?s were performed to study the results of the

Hagerman speech test. In the no-noise condition, a two-way

ANOVA was performed with age group as between-group factor,

hearing aid use (yes, no) as within-subject factor and sound

pressure level for 40% correct word recognition as dependent

variable. There was a significant main effect of hearing aid use

(pB/0.001). With hearing aids, the sound pressure level needed

to obtain 40% correct word recognition was 7 dB lower than

without aids. No age effect was found. In the two noises, a three-

way ANOVA was performed with age group as between-group

factor, hearing aid usage (yes, no) and noise condition (Hager-

man noise, Speech) as within-subject factors and S/N as

dependent variable. There was a main effect of hearing aid use

(p�/0.008). The improvement in S/N with hearing aids was 1.6

dB (averaged over noise conditions). There was also a main

effect of noise condition (p�/0.02); the S/N was �/3.8 dB in

Hagerman noise and �/5.8 dB in Speech (averaged over hearing

aid use). No age effect was found. A significant interaction was

found between hearing aid use and noise condition (p�/0.006);

see Figure 3. The hearing aid benefit was less in Hagerman noise

than in Speech.

PERCEIVED EFFORT

The perceived effort values at the 40% correct response level in

the Hagerman speech test were entered in a three-way ANOVA,

with age-group as between-group factor and hearing aid use

(yes, no) and background condition (No-noise, Hagerman noise,

Speech) as within-subject factors. There was a significant main

effect of background condition (p�/0.042). The perceived effort

was 6.1 in silence, 6.6 in Hagerman noise and 7.3 in Speech.

There were no significant main effects of hearing aid use or age.

No significant interactions were found.

SVIPSTwo four-way ANOVAs were performed for each cognitive test

with age group as between-group factor; modality (audiovisual,

Figure 2. Mean audiograms9/one standard deviation for the two groups for the right and left ears.

Speech understanding in quiet and noise,with and without hearing aids

Hallgren/Larsby/Lyxell/Arlinger 577

auditory), hearing aid use (yes, no), background condition

(No-noise, Hagerman noise, Speech) as within-subject factors;

and percent correct answers or reaction time as dependent

variables.

ACCURACY

Significant effects of age were found where the elderly showed

poorer results than the young subjects in the lexical and semantic

tests; see Figure 4a. There were significant main effects of

modality in all tests. Performance in the audiovisual modality

was superior to that in the auditory; see Figure 4b. There were

also significant main effects of background condition in all tests;

see Figure 4c. No main effects of hearing aid use were found.

REACTION TIME

Significant effects of modality were found in all tests. Processing

in the audiovisual modality took longer than in the auditory

modality; see Figure 5a. Significant effects of background

condition were seen in the lexical and semantic test; see Figure

5b. No main effects of hearing aid use or age were found. A

significant interaction was found between modality and back-

ground condition in the semantic test (p�/0.048). The fact that

the audiovisual modality led to longer reaction times was most

obvious in the no-noise condition and less pronounced in the

noisy background situations. Another interaction was found

between age group and background condition in the name-

matching test (p�/0.002). The young group had shorter reaction

times in the background condition of Speech than in the

Hagerman noise, while the reverse was true for the elderly group.

PERCEIVED EFFORT

The ratings of perceived effort in the SVIPS battery were

analysed in a four-way ANOVA. Age group was used as

between-group factor and modality (audiovisual, auditory),

hearing aid use (yes, no) and background condition (No-noise,

Hagerman noise, Speech) as within-subject factors. There were

significant main effects of modality (pB/0.001), hearing aid use

(pB/0.001) and background condition (pB/0.001); see Figure 6.

The perceived effort was rated 1.5 units higher in the auditory

than in the audiovisual modality, and 0.7 units lower with than

without hearing aids. In the different background conditions, the

perceived effort was rated highest in Speech (5.4), lower in

Hagerman noise (4.4) and lowest in silence (2.9). There

was no significant main effect of age. A significant interaction

between age group and modality (p�/0.036) showed that the

elderly perceived less effort than the young group in the

audiovisual modality but comparable effort in the auditory

modality. There was also a significant interaction between

background condition and modality (p�/0.001). The relatively

small difference between the auditory and audiovisual modality

in silence increased in noise and was largest in Speech. A

significant interaction between background condition and hear-

ing aid use (p�/0.019) showed that the subjective benefit of

hearing aids was high in silence and decreased in background

noise; see Figure 7. Finally, an interaction between age group,

modality, and background condition (p�/0.002), showed that

the effect on perceived effort due to the visual contribution in the

different background conditions was different for the two age-

groups; see Figure 8.

Word recognition testThe results from the speech test were analysed in a three-way

ANOVA. Age group was used as between-subject factor, hearing

aid use (yes, no) and background condition (No-noise, Hager-

man noise, Speech) as within-subject factors and the number of

correctly heard items as dependent variable. There was a

significant main effect of noise (pB/0.001). In silence 96.5% of

the test words were identified, in Hagerman noise the corre-

sponding value was 92% and in Speech 91.8%. There was also a

significant main effect of hearing aid use (p�/0.022). 94.6% of

the words were correctly repeated with hearing aids and 92.3%

without hearing aids. There was no significant main effect of age.

A significant interaction between hearing aid use and back-

ground condition (p�/0.018) showed that the benefit of hearing

aid use was different in the various background conditions; see

Figure 9.

Discussion

Speech understanding comprises many processes, from word

recognition to higher order abilities, and the final result in

understanding of content depends on how these interact. Many

of these processes are strenuous, and even if performance on

speech understanding is good, it might be at the cost of a higher

degree of effort. The obvious improvement of speech recognition

with hearing aids in quiet listening environments is not at all

obvious in noise. The purpose of the present study was to

investigate the effect of hearing aid use on word recognition and

cognitive functions important for speech understanding, in

silence as well as in two different noises with different cognitive

involvements. The hearing-impaired subject is deprived of the

complete auditory input signal and is, to a higher degree than

normal-hearing subjects, dependent on compensatory mechan-

isms to understand the message (Larsby et al, 2005).

Figure 3. Hearing thresholds (S/N) in the Hagerman noise andin background speech, with and without hearing aids, in theHagerman speech test.

578 International Journal of Audiology, Volume 44 Number 10

Peripheral hearingWhile there was no main effect of age group in HTL, there were

differences between the groups at different frequencies. The

elderly subjects mainly comprised subjects with presbyacusis

whereas the young group had sensorineural hearing impairments

of various etiology. Thus, the HTLs for the elderly subjects were

poorer than for the younger subjects in the high-frequency range

and better in the low-frequency range (Figure 2). This is a

confounding factor that cannot be ignored.

Hagerman speech testIn the Hagerman speech test there was a benefit of hearing aid

use without background noise. The benefit was on average

relatively small (7 dB SPL), which might be explained by the fact

that the hearing losses in the lower frequency range were

moderate (see Figure 2). With speech as the background, there

was also a hearing aid benefit (2.5 dB S/N), while there was no

hearing aid benefit in the Hagerman noise. Lunner (2003) did

not find a hearing aid benefit in the Hagerman test with the

original Hagerman noise either. There is a difference in hearing

aid benefit between the background condition with relatively

large temporal variations (Speech) and the background condi-

tion with small temporal variations (Hagerman noise); see

Figure 3. This finding agrees with those of Alcantara et al

(2003) and Gatehouse et al (2003), who both showed that

hearing aid benefit in speech-shaped noise with amplitude

Figure 4. Significant main effects of age (a), modality (b), and background condition (c) in the correct answers parameter in thedifferent cognitive tests. P-values shown represent the degree of statistical significance according to the ANOVA.

Speech understanding in quiet and noise,with and without hearing aids

Hallgren/Larsby/Lyxell/Arlinger 579

modulations was superior to that in noise without such

modulations. Hearing aid amplification increases the audibility

of the speech target signal. In the Hagerman noise the masking

effect of the noise is relatively constant, while with speech as a

masker the masking effect varies over time; see Figure 1. In the

short periods of silence or low level in the speech masker, the

subject can utilise the gaps to hear the amplified target signal.

This agrees with the findings of previous studies (Gustafsson &

Arlinger, 1994; Hygge et al, 1992; Festen & Plomp, 1990;

Duquesnoy, 1983).

For the scores of perceived effort, there was no effect of

hearing aid in the Hagerman test. This is probably due to the

adaptive process of signal-to-noise adjustment to reach 40%

correct responses. Despite the adaptive procedure, there was a

main effect of background condition, where speech as a back-

ground was more demanding than Hagerman noise and the least

demanding condition was the one without noise.

SVIPSThe effects of age, modality and background condition in the

SVIPS battery agree with results obtained in previous studies

from our group (Larsby et al, 2005; Hallgren et al, 2001). For the

parameter of interest in this study, hearing aid use, no main

effect was seen in the ANOVA. Word recognition in the SVIPS

tests was good even without hearing aids, since the subjects were

allowed to adjust the level for optimum speech recognition. The

results of the word recognition test (Figure 9) show that over

90% of the items in the SVIPS battery were correctly repeated,

even in noise. In the SVIPS battery, two major mechanisms

interact, word recognition, which to a high degree reflects

Figure 5. Significant main effects of modality (a) and back-ground condition (b) on reaction time in the different cognitivetests. P-values shown represent the degree of statistical signifi-cance according to the ANOVA.

Figure 6. Significant main effects of modality (a), hearing aiduse (b), and background condition (c) on perceived effort in thecognitive tests.

580 International Journal of Audiology, Volume 44 Number 10

peripheral hearing, and decision making, which depends more

on cognitive demanding top-down driven speech understanding

processes. Incomplete word recognition complicates decision

making, which can be compensated for by redundancy in the test

situation. In the semantic test, it is known that the target item

belongs to a predefined category, which facilitates the decision-

making. In the lexical test, where the task is to judge whether the

item is a real word or a non-word, the non-words have very low

redundancy and are hard to identify if recognition is incomplete.

In this case, a hearing aid benefit was actually seen in the

background condition without noise (planned comparison, p�/

0.018). This effect was only seen in the auditory modality, where

there was no visual contribution. The name matching test, where

the task is to decide if two letters are the same or not, is also a

test with low redundancy items that is easy to perform with

visual support but very difficult without visual support (Larsby

et al, 2005; Girin et al, 2001). In the auditory modality there was

indeed a hearing aid benefit in the background condition

without noise in the reaction time parameter (planned compar-

ison, p�/0.017). Taken together, despite the relatively easy word

recognition in the SVIPS tests, there was a hearing aid benefit in

tasks with low redundancy test items in the auditory modality

without background noise. In noise, the positive effect of

amplification on the target signal did not occur. To recognize

low redundancy stimuli one has to identify every speech sound in

the test word/pair of letters and any kind of noise is likely to

disturb this.

For the scores of perceived effort, there was a significant main

effect of hearing aid use. The subjects experienced less effort with

hearing aids than without. The significant interaction between

hearing aid use and background condition verifies that the

benefit from hearing aid use is largest in the no-noise condition,

decreases in the Hagerman noise, and is smallest in the condition

with speech as background (see Figure 7). Planned comparisons

showed a significant benefit from hearing aids in the no-noise

condition (pB/0.001) and in the Hagerman noise (p�/0.008), but

not in the background condition with speech.

In the Hagerman noise, audibility increases somewhat with

hearing aid amplification, as seen in the word recognition test

(Figure 9), but no hearing aid effect is seen in performance in

the SVIPS tests. It is likely that the increased audibility with

hearing aids makes the subjects more confident and thus the

decision-making less strenuous. The fact that we did not see a

hearing aid benefit in the background condition with speech

could either have been due to the lack of improved audibility

or to a larger effect of distraction caused by the temporal

pattern or the meaningfulness of the masker. Previous studies

(Larsby et al, 2005; Tun et al, 2002; Tun & Wingfield, 1999)

have shown that competing noise with meaningful content is

more difficult to ignore than noise without meaning. It is also

Figure 7. Perceived effort in the different background condi-tions, with and without hearing aids, in the cognitive tests.

Figure 8. Interaction between background condition, modality, and age group in the perceived effort score in the cognitive tests.

Speech understanding in quiet and noise,with and without hearing aids

Hallgren/Larsby/Lyxell/Arlinger 581

likely that listening in the gaps requires a higher degree of

demanding top-down processing, such as semantic and lexical

decision-making. This is in line with Gatehouse et al (2003),

who showed that the ability to capitalise on the temporal dips

in the noises is greater for listeners with better cognitive

function.

For the subjective scores from the SVIPS test battery, the

perceived effort was rated for all the tests overall (for each

background condition and modality). The specific analysis of

low redundancy items made in the objective measurement is

therefore not possible. However, there was a main effect of

hearing aid use. Hearing aid use thus results in less perceived

effort, despite unaffected speech understanding. The high degree

of perceived effort without hearing aid use would presumably at

some stage negatively affect the degree of fatigue and the ability

to concentrate which would, in turn, affect speech understand-

ing.

General discussionThe benefit from hearing aid amplification can be evaluated

along many different dimensions (for a review, see Humes,

1999). The most common is to record improvement in speech

recognition in quiet and in noise. The benefit for an individual in

everyday life depends on the audibility of the speech signal, but

it is also influenced by other factors. Critical is the person’s

ability to extract the limited and distorted speech signal from

environmental noise and at the same time ignore irrelevant

distracting information. This ability depends on sensory auto-

matic bottom-up driven processing as well as on higher order

top-down driven processing. When the signal-to-noise ratio

becomes unfavourable and the processing goes from being easy

and automatic to being difficult and cognitively demanding,

then the degree of perceived effort is likely to increase (Pichora-

Fuller, 2003). A measure of perceived effort has, in this study

and in Larsby et al (2005), been shown to be a valuable tool to

complement the objective measure of speech recognition. Also

Humes (1999) pointed out the importance of including a

measure of subjective listening effort. This dual approach gives

a more complete picture of the person’s ability to make use of

amplification in difficult listening situations. For example, for

the background of speech, there was a significant hearing aid

effect in the Hagerman test, when measuring the S/N for 40%

correct word recognition. This effect was not seen in the

subjective scores, neither in the Hagerman nor in the SVIPS

test. For the background condition of Hagerman noise, no effect

of hearing aid use was found in the objective S/N in the

Hagerman test, but a significant effect was found in the

subjective scores in the SVIPS test.

All hearing aid benefit measurements depend on methodolo-

gical circumstances and the outcome measure used. The signal-

to-noise ratio, for example, is dependent on level and presenta-

tion mode of noise and speech stimuli, the test material used,

and the hearing aid settings. The different outcome measures in

the SVIPS test are also dependent on the test material and

presentation level of test stimuli and background noise. In the

present study, the SVIPS tests were performed at an S/N of

�/10 dB, which clearly allows the hearing-impaired subjects

to hear most of the test items and also to utilize

redundant information. In this case we see no hearing aid

benefit in noise. In the Hagerman speech test, however, where

we measure around the 40% threshold we see a hearing aid

benefit in noise, which is most evident in the background

condition of speech. In daily life we sometimes, especially in

background noise, listen at threshold, but most of the time we

listen above threshold where persons with moderate hearing

impairments can function quite well, if they make use of context

and visual support (Larsby, 2005; Hallgren, 2001). It is

noteworthy, however, that despite no hearing aid benefit in

performance there was a lower degree of perceived effort with

hearing aids, a fact whose importance cannot be overestimated.

It is a mission for the hearing aid professional to make the

hearing-impaired person aware of the consequences that active

listening with hearing aids has on both performance and on

perceived effort.

Administering the SVIPS with a less favourable S/N ratio than

�/10 dB is a further and necessary step to understand the

mechanisms underlying the speech understanding process in

noise and the benefit achieved from hearing aids. The outcome

of SVIPS performed at different S/N ratios gives a more

complete picture of a person’s ability to both recognise and

understand a spoken message.

In conclusion, the most benefit derived from amplification

with hearing aids was shown in the background condition

without noise. The hearing threshold for Speech was 7 dB SPL

lower in the Hagerman speech test and performance was

increased in cognitive tasks with low redundancy items. In noise,

the only effect of hearing aid use was a lower hearing threshold

for Speech in the Hagerman speech test (2.5 dB S/N).

Despite minor benefits of hearing aid amplification in the

objective measures, especially in the cognitive tests, significantly

less effort was perceived when hearing aids were used. This

underlines the importance of considering perceived effort as a

dimension when evaluating hearing aid benefit, in further

research as well as in clinical practice.

Figure 9. Correct answers (%) in the different backgroundconditions, with and without hearing aids, in the word recogni-tion test.

582 International Journal of Audiology, Volume 44 Number 10

Acknowledgements

Thanks are due to The Swedish Council for Working Life and

Social Research (FAS) and to The Swedish Association of Hard

of Hearing People (HRF) for generous support. Thanks are also

due to Erica Billermark for assistance during test performance.

A special thanks goes to all our subjects who willingly

participated in the study.

References

Alcantara, J.I., Moore, B.C., Kuhnel, V. & Launer, S. 2003. Evaluation ofthe noise reduction system in a commercial digital hearing aid. Int JAudiol , 42, 34�/42.

Allen, S. 1970. Frequency dictionary of present-day Swedish (in Swedish:Nusvensk frekvensbok) . Stockholm: Almquist & Wiksell.

Borg, G. 1990. Psychophysical scaling with applications in physical workand the perception of exertion. Scand J Work Environ Health , 16,55�/58.

CHABA (Committee on Hearing, Bioacoustics and Biomechanics) 1988.Speech understanding and aging. J Acoust Soc Am , 83, 859�/895.

Cord, M.T., Leek, M.R. & Walden, B.E. 2000. Speech recognition abilityin noise and its relationship to perceived hearing aid benefit. J AmAcad Audiol , 11, 475�/483.

Davis, A. 2003. Population study of the ability to benefit fromamplification and the provision of a hearing aid in 55�/74-year-oldfirst-time hearing aid users. Int J Audiol , 42, 2S39�/2S52.

Duquesnoy, A. J. 1983. Effect of a single interfering noise or speechsource upon the binaural sentence intelligibility of aged persons.J Acoust Soc Am , 74, 739�/743.

Festen, J. M. & Plomp, R. 1990. Effects on fluctuating noise andinterfering speech on the speech-reception threshold for impairednormal hearing. J Acoust Soc Am , 88, 1725�/1736.

Gatehouse, S., Naylor, G. & Elberling, C. 2003. Benefits from hearingaids in relation to the interaction between the user and theenvironment. Int J Audiol , 42, S77�/S85.

Girin, L., Schwartz, J-L. & Feng, G. 2001. Audio-visual enhancement ofspeech in noise. J Acoust Soc Am , 109, 3007�/3020.

Gustafsson, H. A. & Arlinger, S. D. 1994. Masking of speech byamplitude-modulated noise. J Acoust Soc Am , 95, 518�/529.

Hagerman, B. 1982. Sentences for testing speech intelligibility in noise.Scand Audiol , 11, 79�/87.

Hagerman, B. 1984. Clinical measurements of speech reception tresholdin noise. Scand Audiol , 13, 57�/63.

Hagerman, B. & Kinnefors, C. 1995. Efficient adaptive methods formeasuring speech reception threshold in quiet and in noise. ScandAudiol , 24, 71�/77.

Haskell, G.B., Noffsinger, D., Larson, V.D., Williams, D.W., Dobie, R.A.& Rogers, J.L. 2002. Subjective measures of hearing aid benefit in theNIDCD/VA clinical trial. Ear Hear, 23, 301�/307.

Humes, L.E. 1999. Dimensions of hearing aid outcome. J Am AcadAudiol , 10, 26�/39.

Hygge, S., Ronnberg, J., Larsby, B. & Arlinger, S. 1992. Normal-hearingand hearing-impaired subjects’ ability to just follow conversation incompeting speech, reversed speech, and noise backgrounds. J SpeechHear Res, 35, 208�/215.

Hallgren, M., Larsby, B., Lyxell, B. & Arlinger, S. 2001. Evaluation of acognitive test battery in young and elderly normal-hearing andhearing-impaired subjects. J Am Acad Audiol , 12, 357�/370.

Larsby, B., Hallgren, M., Lyxell, B. & Arlinger, S. 2005. Cognitiveperformance and perceived effort in speech processing tasks: effectsof different noise backgrounds in normals and in hearing-impairedsubjects. Int J Audiol , 44, 131�/143.

Larson, V.D., Williams, D.W., Henderson, W.G., Luethke, L.E., Beck,L.B., Noffsinger, D., et al. 2002. A multi-center, double blind clinicaltrial comparing benefit from three commonly used hearing aidcircuits. Ear Hear, 23, 269�/276.

Lunner, T. 2003. Cognitive function in relation to hearing aid use. Int JAudiol , 42, S49�/S58.

Lyxell, B., Andersson, U., Borg, E. & Ohlsson, I-S. 2003. Working-memory capacity and phonological processing in deafened adultsand individuals with a severe hearing impairment. Int J Audiol , 42,S86�/S89.

Pichora-Fuller, M.K. 2003. Processing speed and timing in aging adults:psychoacoustics, speech perception, and comprehension. Int JAudiol , 42, S59�/S67.

Plomp, R. 1986. A signal-to-noise ratio model for the speech receptionthreshold of the hearing-impaired. J Speech Hear Res , 29, 146�/154.

Ronnberg, J. 1990. Cognitive and communicative function: The effect ofchronological age. Eur J Cogn Psychol , 2, 253�/275.

Shanks, J.E., Wilson, R.H., Larson, V. & Williams, D. 2002. Speechrecognition performance of patients with sensorineural hearing lossunder unaided and aided conditions using linear and compressionhearing aids. Ear Hear, 23, 280�/290.

Shoben, E. 1982. Semantic and lexical decisions. In C.R. Puff (ed.),Handbook of research methods in human memory and cognition . NewYork: Academic Press.

Tun, P. A., O’Kane, G. & Wingfield, A. 2002. Distraction by competingspeech in young and older adult listeners. Psychol Aging , 17, 453�/

467.Tun, P. A. & Wingfield, A. 1999. One voice too many: Adult age

differences in language processing with different types of distractingsounds. J Gerontol , 54B, 317�/327.

van Rooij, J. C. G. M. & Plomp, R. 1990. Auditive and cognitive factorsin speech perception by elderly listeners. II. Multivariate analysis.J Acoust Soc Am , 88, 2611�/2624.

Speech understanding in quiet and noise,with and without hearing aids

Hallgren/Larsby/Lyxell/Arlinger 583