Past studies of speech-on-speech masking in young adults (YA) indicate that the intelligibility of masked speech can improve if the target and masker speech are in different languages. Current work investigated whether a linguistic masking release is obtained in older adults (OA) with age-typical hearing abilities. Participants listened to English sentences in the presence of two-talker Spanish or English maskers. A similar masking release with Spanish-language maskers was obtained for OA and YA listeners, despite greater accuracy by YA listeners. In speech-on-speech masking, older listeners can thus improve speech intelligibility by utilizing nonenergetic linguistic differences between the target and masker speech.

Intelligibility of speech masked by competing speech from other talkers is a major challenge for older listeners due to both energetic and informational masking effects (Pichora-Fuller and Souza, 2003; Helfer and Freyman, 2008; Mattys et al., 2009). Energetic masking negatively affects the audibility of acoustic cues in the target speech at the level of the auditory periphery due to spectral and temporal overlap between the target and masker. Informational masking, defined as a threshold elevation which cannot be explained by energetic masking alone, affects speech recognition at more central processing levels. Informational masking may result from similarity of the signal and the masker in terms of their acoustic properties, or in terms of linguistic information contained in phonemic, prosodic, syntactic, and semantic structure (Mattys et al., 2009; Brouwer et al., 2012). Previous research indicates that speech comprehension in children and young adults with normal hearing improved when the target and masker speech were presented in different languages, as compared to when the target and the masker were in the same language (Rhebergen et al., 2005; Calandruccio et al., 2013; Calandruccio et al., 2016). However, a masking release induced by a language mismatch between the target and masker may be influenced by peripheral hearing acuity and more central auditory factors, including linguistic and cognitive processing, which decline with age. The purpose of this study was to examine the effect of masker language on speech intelligibility in older listeners with hearing acuity typical of their age.

Several studies with young adults have examined the conditions that lead to a linguistic masking release (LMR): a release from masking when the target and masker speech are in different languages, as compared to when the target and the masker are in the same language. For instance, Dutch target speech was more intelligible for native Dutch listeners when masked by Swedish-language maskers than by Dutch-language maskers (Rhebergen et al., 2005). A similar effect was found when English target speech was presented in a two-talker babble masker in Mandarin, compared to when the masker was in English. The amount of release was further shown to be influenced by the degree of linguistic similarity between the target and masker languages. Calandruccio et al. (2013), using additional target-masker language pairs, found that the more linguistically distant the masker language was from the target language, the greater the amount of masking release. In that study, an English target with a Mandarin masker produced greater release than an English target with a Dutch masker. Thus, the improvement in performance with different language maskers may arise from several levels of linguistic similarity across target and masker language pairs.

In addition to linguistic similarity between the target and the masker, previous studies also indicated that LMR can be affected by energetic factors. Specifically, a significant masking release was found only at the least favorable signal-to-noise ratios (SNRs) tested across studies—generally between 0 to −5 dB (Calandruccio et al., 2010; Calandruccio et al., 2013; Brouwer et al., 2012). Additionally, the masking release was only significant with a two-talker speech-babble masker. When the number of talkers in the babble was increased to six, there was no significant effect of changing the masker language (Van Engen and Bradlow, 2007). The lack of a masking release associated with the greater number of talkers may result from less language-specific information contained in the multi-talker babble due to the greater spectral and temporal density of the masker (Brouwer et al., 2012). With a less detectable linguistic structure in the six-talker masker, the ability to segregate the target and the masker based on their linguistic incongruency may decrease. On the other hand, in addition to better audibility of the signal afforded by the dips in the masker envelope of a two-talker masker, listeners seem to take advantage of the differences between the linguistic structure of the target and the maskers. Specifically, in difficult listening conditions, i.e., an SNR of −5 dB, performance with a two-talker English babble masker was superior to that with a spectrally matched and modulated noise masker (Calandruccio et al., 2010). As overall intelligibility decreased in lower SNR conditions, the role of masker linguistic content increased.

Furthermore, LMR can be obtained even with bilingual listeners highly familiar with both the target and the masker languages (Calandruccio and Zhou, 2014). In this case, however, the LMR of bilingual listeners is attenuated compared to that of monolingual listeners (Calandruccio et al., 2013; Van Engen, 2010; Brouwer et al., 2012). Similarly, a release from masking was found when the target and masker speech are in the same language but differ in phonological and prosodic structure. Calandruccio et al. (2010) used different degrees of nonnative accented speech as maskers of a target signal spoken in the same language as the masker, i.e., English. The researchers examined the effect of Mandarin-accented masking speech that varied in terms of accentedness and intelligibility. At low SNRs of −3 and −5 dB, masking release increased as the overall intelligibility decreased. Overall, past work suggests that increasing linguistic similarity between the target and the masker can decrease the magnitude of LMR, while neither energetic nor linguistic differences alone can fully account for the effect.

Age-related decline in speech understanding has been attributed to multiple factors pertaining to changes in the auditory periphery, central auditory system, and general cognitive processes (Humes and Dubno, 2010). Typically, older individuals have increasing difficulty understating speech in noise, especially when masking noise is also speech. Even when contributions of energetic masking are controlled, older listeners demonstrate greater difficulties with speech in noise than younger adults (Humes and Dubno, 2010). In addition, changes in speech rate, speaker clarity, or accentedness have a more negative effect on speech perception by older than younger adults (Versfeld and Dreschler, 2002).

It has been suggested that an age-related decline in audition along with amodal aspects of attention, working memory, inhibition, and a general slowing of cognitive processing can contribute to greater informational masking in older listeners (Pichora-Fuller and Souza, 2003; Rajan and Cainer, 2008; Humes and Dubno, 2010). Nevertheless, older adults are able to effectively utilize syntactic and semantic contextual cues to maintain a high level of speech perception accuracy (Pichora-Fuller and Souza, 2003). Specifically, Tun et al. (2002) demonstrated that in recall and recognition of spoken English words, older adults improved more than young adults when the language of the masker was changed from English to Dutch. This finding suggests that older adults are able take advantage of the language-mismatch between the target and the masker speech to improve memory for words. However, it is unknown if a similar LMR effect in older adults can be obtained for speech intelligibility, which is less demanding of memory and cognitive processing than recall and recognition memory tasks.

The objective of the present study was to determine the effect of masker language on the intelligibility of target speech in older adults. It was expected that, similar to younger listeners, older adults with age appropriate hearing abilities including a mild hearing loss, would demonstrate a LMR due to a linguistic mismatch between the target and masker languages. Based on prior findings of greater interference of informational maskers with the performance of older adults (Tun et al., 2002; Pichora-Fuller and Souza, 2003; Shinn-Cunningham and Best, 2008), older adults could be expected to show a greater LMR than younger listeners. This expectation is consistent with previous research in young and older adults, reviewed above, which demonstrated a release from masking when the target and the masker were in different languages.

Sixteen older adults (OA), 11 females and 5 males, ages 61 to 78 years [mean = 68.63 years, standard deviation (SD) = 6.08] participated with hearing sensitivity ranging from normal to a mild hearing loss. Their hearing sensitivity was assessed using a four-frequency pure-tone average (PTA) of 500, 1000, 2000, and 4000 Hz with results tabulated for the better-hearing ear. Across subjects, PTA ranged between 12.5 dB hearing level (HL) and 41.25 dB HL (mean = 24.61 dB HL, SD = 7.88). Hearing-in-noise abilities in the older group, as measured with quick-sin, a standard clinical test of speech-in-noise perception, ranged between an SNR Loss of 0 to 8 dB SNR loss (mean = 1.97 dB, SD = 1.8). SNR Loss represents the estimated SNR value in dB that allows for 50% correct intelligibility, normalized to the performance of young normal-hearing adults. Eleven young normal-hearing adults (YA), 8 females and 3 males, ages 22 to 26 years old (mean = 24.36 years, SD = 1.03) served as controls. Subjects in the YA group passed a hearing screening at 20 dB HL for 500, 1000, 2000, and 4000 Hz pure tones. All but one study participants were native English speakers. One participant in the OA group was not a native speaker of English who identified herself as fluent in English. Although several subjects in both groups had some familiarity with Spanish, on average their reported fluency level was low.

Target sentences were recordings of Bamford-Kowal-Bench (BKB) sentence lists by a female speaker. Each list contains 16 simple meaningful sentences with three to four keywords in each sentence. To enable greater comparisons with previous work that investigated LMR, the masking speech was two-talker babble masker in English and in Spanish previously used by Calandruccio et al. (2014). The English masker was composed of the “Jake and the Beanstalk” story and the Spanish masker was composed of the Spanish translation of the same story “Juan y Los Frijoles.” To minimize spectral and temporal differences between the two maskers, the same bilingual talkers were used for both English and Spanish maskers. Furthermore, the long-term average speech spectrum (LTASS) of the maskers was normalized to reduce spectral differences. The target signal was always presented at 75 dB sound pressure level (SPL) while the level of the background babble was varied. Four different SNRs were used for each group to create a psychometric function. For the younger group, SNRs of 2.5, 0, −2.5, and −5 dB were used. For the older group, SNRs of 5, 2.5, 0, and −2.5 dB were used to accommodate a different performance range and to avoid ceiling and floor effects. For the older group, easier SNR conditions were created to obtain performance roughly comparable to that of the young group. Altogether, there were three common SNRs between the two groups, −2.5, 0, and 2.5 dB.

For each subject, eight lists were presented. Four of the lists were presented with the English masker and four were presented with the Spanish masker. Prior to testing, the participants were instructed that they will hear English sentences presented in background babble which will be either in English or Spanish. The order of masker language was randomized across list presentations. Within each list the four different SNR conditions changed for each consecutive sentence starting from the easiest to the hardest and then repeating in this order until the end of the list was reached. This was done to make the SNR presentation order more predictable for older adults, minimizing their discomfort during the presentation at more challenging SNRs. Subjects were instructed to repeat what they heard after listening to each sentence and guess when they were not sure. Verbal responses were scored manually by the tester based on keywords correct.

Overall, consistent with expectations, older adults demonstrated lower intelligibility scores than younger listeners in all comparable SNR conditions (Fig. 1). Averaged across English and Spanish masker conditions, OA intelligibility ranged from a mean of 47.4% (SD = 18.2) at an SNR of −2.5 dB to 94.5% (SD = 4.5) at an SNR of 2.5 dB. In the same conditions, YA intelligibility ranged from the mean of 65.1% (SD = 19.5) to 92.4% (SD = 9.5). Both YA and OA subjects demonstrated LMR. The average amount of masking release across the three shared SNR conditions was 13.45% for OA and 15.17% for YA subjects. For individual YA listeners, it ranged between 0% and 74.27%, while for OA listeners it ranged between 0% and 56.27%. In both groups, the magnitude of the release increased as SNR decreased. This is similar to the findings of previous research with young-adult listeners (Van Engen and Bradlow, 2007).

Fig. 1.

Mean performance in RAU of the younger (circle) and older (square) listener groups as a function of the different SNR levels. There was a significant effect of the masker's language between the English masker (filled icons) and the Spanish masker (open icons) and a significant effect of SNR.

Fig. 1.

Mean performance in RAU of the younger (circle) and older (square) listener groups as a function of the different SNR levels. There was a significant effect of the masker's language between the English masker (filled icons) and the Spanish masker (open icons) and a significant effect of SNR.

Close modal

To evaluate statistical significance of these effects intelligibility scores were converted into rationalized arcsine units (RAU) to normalize the error variance (Fig. 1). A split-plot analysis of variance (ANOVA) was conducted to compare the performance of the YA and the OA groups on two factors: masker language and SNR levels that were common across groups. The analysis revealed a significant effect of masker language [F(1, 25) = 47.53, p < 0.001] and SNR [F(2,50) = 299.99, p < 0.001]. Additionally, there was a significant interaction between the listener's group and SNR [F(2,50) = 5.22, p = 0.009]. This was due to variation in the performance of the two groups at different SNR levels. Specifically, follow up independent sample t-tests revealed that there was no significant differences between the two groups for the 2.5-dB SNR [t(25) = 1.6, p = 0.12], while the performance of the two groups differed significantly with SNRs of 0 and −2.5 dB [t(25) = 3.5, p = 0.002 and t(25) = 3.8, p = 0.001]. Neither the two-way interaction between masker language and group nor the three-way interaction between masker language, SNR, and group were significant. The absence of these significant interactions indicates that the magnitude of LMR did not differ between the younger and older listeners.

As an alternative metric of masking release, LMR magnitude was compared between the two groups at fixed performance levels. This was done to determine whether lack of group differences in the amount of masking release observed at specific SNR levels was influenced by the differences in overall intelligibility that distinguished the younger from the older subject group. Based on individual psychometric functions, a masking release was measured in terms of the change in SNR needed to maintain a fixed performance level when the masker language was changed from English to Spanish. Speech masking release in terms of SNR delta is shown for both subject groups in Fig. 2 at four levels of percent correct. On average, YA listeners exhibited a slightly larger masking release than OA, with neither group showing an effect of performance level in the range of 70% to 85% correct. Despite the apparent trend for a greater amount of speech masking release by younger listeners, none of the between group comparisons of SNR difference was statistically significant at any of the four fixed performance levels (p > 0.05, range 0.07–0.12).

Fig. 2.

Mean masking release in terms of delta signal-to-noise ratio (SNR) at four performance levels with subject group indicated by bar color. Error bars represent 1 SD.

Fig. 2.

Mean masking release in terms of delta signal-to-noise ratio (SNR) at four performance levels with subject group indicated by bar color. Error bars represent 1 SD.

Close modal

Finally, a repeated measure of covariance (ANCOVA) was conducted to examine potential contributions of age and hearing sensitivity in the OA listeners to LMR. Age, hearing sensitivity (as measured with PTA) and hearing in noise ability (as measured with the quick-sin) were included as covariates. The ANCOVA test showed no significant involvement of these control variables on the effects of masker language, SNR, or their interaction (p > 0.05). Thus variation in the age and hearing levels in the older group did not significantly contribute to LMR.

Older adults with either normal hearing or a mild hearing loss showed a significant masking release due to a linguistic mismatch between the target and masker in a speech-on-speech masking task. The masking release for both younger and older adult listeners was inversely related to SNR (Fig. 1). When assessed as the SNR delta needed to maintain a fixed performance level when the masker language was changed from English to Spanish, the average speech masking release was roughly 1.0 to 1.5 dB for OA listeners and 2 to 2.5 for YA listeners, with little effect of performance level in the range of 70% to 85% correct. (Fig. 2). However, these differences in the amount of release were not significant between the groups. Thus, results demonstrate that older listeners can benefit from a linguistic mismatch between the signal and the masker in a similar way to younger listeners. However, it is important to note that the subjects in our older group had relatively good hearing sensitivity and speech-in-noise ability.

Consistent with past research, OA listeners performed more poorly than young adults at speech recognition in the presence of a speech masker. However, the extent of the speech masking release was not significantly different for the two groups when based on either evaluation at a common SNR or a common performance level. With aging, increased susceptibility to informational masking due to diminishing cognitive resources is associated with greater reliance on nonenergetic contextual cues (Pichora-Fuller and Souza, 2003). In that case, a greater masking release for the older than younger listeners could be expected when the language of the target is different from the language of the masker. However, this expectation was not confirmed in the present study, with YA listeners consistently demonstrating slightly, though not significantly, greater LMR than OA listeners. Presumably, other factors related to the perception of masked speech in older adults such as higher pure tone thresholds, dip-listening ability, distortion in auditory processing, cognitive status might have interfered with OA listeners' ability to benefit from masker language mismatch. For instance, deficits in temporal fine-structure processing in older adults (Sheft et al., 2012) might have interfered with segregation of the target and masker talkers, reducing the availability of language specific acoustic cues that support LMR. Unable to segregate the speakers' voices as effectively as younger listeners, older listeners may not have utilized language-dependent processing as effectively in further segregation of the signal and the masker speech.

The absence of a significant difference in LMR for the YA and OA groups contrasts with the results of Tun et al. (2002) who reported a greater improvement by older adults in recall and recognition of spoken words. The difference between the findings of the present study and those of Tun et al. suggests that the amount of LMR can be further modulated during memory encoding and access stages, that require greater cognitive involvement than the more immediate repetition of words during the intelligibility task. It is thus possible that older adults may benefit from LMR even more in speech perception tasks that place greater demands on memory and cognitive resources.

The findings of the present study indicate several directions for future work that can clarify the mechanisms of LMR. First, the hearing sensitivity and speech-in-noise abilities of the older adults studied were only mildly impaired. We do not know how older adults with greater degrees of hearing loss would perform with linguistically mismatched targets and maskers. It is possible that with a greater degree of hearing loss, hearing impaired listeners would be less able to benefit from LMR. Additionally, cognitive and linguistic abilities were neither measured nor controlled. It is possible, that aspects of cognitive and linguistic abilities may modulate the amount of benefit that can be derived from a linguistic mismatch between the target and masker languages. Finally, though maskers were matched in terms of LTASS and showed comparable characteristics of their average temporal envelopes, conditions do not allow for specific determination of the acoustic basis of the speech masking release.

Hearing in noise is often a challenging task for older adults regardless of the extent of hearing loss. The present findings demonstrate that under adverse listening conditions in which target speech audibility is compromised by energetic masking, both older and younger adult listeners can effectively exploit nonenergetic linguistic differences between the target and masker to increase speech intelligibility. Future work manipulating the informational and energetic aspects of masking in the linguistic-masking paradigm may help isolate detrimental components of speech masking for older adults, and also indicate specific cues that may benefit their speech recognition in noise.

We would like to thank Dr. Alyssa Herrera for her assistance with participant testing and Dr. Lauren Calandruccio for help with stimulus selection and productive discussions.

1.
Brouwer
,
S.
,
Van Engen
,
K. J.
,
Calandruccio
,
L.
, and
Bradlow
,
A. R.
(
2012
). “
Linguistic contributions to speech-on-speech masking for native and non-native listeners: Language familiarity and semantic content
,”
J. Acoust. Soc. Am.
131
(
2
),
1449
1464
.
2.
Calandruccio
,
L.
,
Bradlow
,
A. R.
, and
Dhar
,
S.
(
2014
). “
Speech-on-speech masking with variable access to the linguistic content of the masker speech for native and nonnative english speakers
,”
J. Am. Acad. Audiol.
25
(
4
),
355
366
.
3.
Calandruccio
,
L.
,
Brouwer
,
S.
,
Van Engen
,
K. J.
,
Dhar
,
S.
, and
Bradlow
,
A. R.
(
2013
). “
Masking release due to linguistic and phonetic dissimilarity between the target and masker speech
,”
Am. J. Audiol.
22
(
1
),
157
164
.
4.
Calandruccio
,
L.
,
Dhar
,
S.
, and
Bradlow
,
A. R.
(
2010
). “
Speech-on-speech masking with variable access to the linguistic content of the masker speech
,”
J. Acoust. Soc. Am.
128
(
2
),
860
869
.
5.
Calandruccio
,
L.
,
Leibold
,
L. J.
, and
Buss
,
E.
(
2016
). “
Linguistic masking release in school-age children and adults
,”
Am. J. Audiol.
25
,
34
40
.
6.
Calandruccio
,
L.
, and
Zhou
,
H.
(
2014
). “
Increase in speech recognition due to linguistic mismatch between target and masker speech: Monolingual and simultaneous bilingual performance
,”
J. Speech Lang. Hear. Res.
57
(
3
),
1089
1097
.
7.
Helfer
,
K. S.
, and
Freyman
,
R. L.
(
2008
). “
Aging and speech-on-speech masking
,”
Ear Hear.
29
(
1
),
87
98
.
8.
Humes
,
L. E.
, and
Dubno
,
J. R.
(
2010
). “
Factors affecting speech understanding in older adults
,” in
The Aging Auditory System: Perceptual Characterization and Neural Bases of Presbycusis, Springer Handbook of Auditory Research
, edited by
S.
Gordon-Salant
,
R. D.
Frisina
,
A.
Popper
, and
D.
Fay
(
Springer
,
Berlin
), Chap. 8, pp.
211
258
.
9.
Mattys
,
S. L.
,
Brooks
,
J.
, and
Cooke
,
M.
(
2009
). “
Recognizing speech under a processing load: Dissociating energetic from informational factors
,”
Cogn. Psychol.
59
(
3
),
203
243
.
10.
Pichora-Fuller
,
M. K.
, and
Souza
,
P. E.
(
2003
). “
Effects of aging on auditory processing of speech
,”
Int. J. Audiol.
42
(S
2
),
11
16
.
11.
Rajan
,
R.
, and
Cainer
,
K. E.
(
2008
). “
Ageing without hearing loss or cognitive impairment causes a decrease in speech intelligibility only in informational maskers
,”
Neuroscience
154
(
2
),
784
795
.
12.
Rhebergen
,
K. S.
,
Versfeld
,
N. J.
, and
Dreschler
,
W. A.
(
2005
). “
Release from informational masking by time reversal of native and non-native interfering speech
,”
J. Acoust. Soc. Am.
118
(
3
),
1274
1277
.
13.
Sheft
,
S.
,
Shafiro
,
V.
,
Lorenzi
,
C.
,
McMullen
,
R.
, and
Farrell
,
C.
(
2012
). “
Effects of age and hearing loss on the relationship between discrimination of stochastic frequency modulation and speech perception
,”
Ear Hear.
33
(
6
),
709
720
.
14.
Shinn-Cunningham
,
B. G.
, and
Best
,
V.
(
2008
). “
Selective attention in normal and impaired hearing
,”
Trends Amplif.
12
,
283
899
.
15.
Tun
,
P. A.
,
O'Kane
,
G.
, and
Wingfield
,
A.
(
2002
). “
Distraction by competing speech in young and older adult listeners
,”
Psychol. Aging
17
(
3
),
453
467
.
16.
Van Engen
,
K. J.
, and
Bradlow
,
A. R.
(
2007
). “
Sentence recognition in native- and foreign-language multi-talker background noise
,”
J. Acoust. Soc. Am.
121
(
1
),
519
526
.
17.
Van Engen
,
Kristin J.
(
2010
). “
Similarity and familiarity: Second language sentence recognition in first- and second-language multi-talker babble
,”
Speech Commun.
52
,
943
953
.
18.
Versfeld
,
N. J.
, and
Dreschler
W. A.
(
2002
). “
The relationship between the intelligibility of time- compressed speech and speech in noise in young and elderly listeners
,”
J. Acoust. Soc. Am.
111
,
401
408
.