This study followed up Wang, Shu, Zhang, Liu, and Zhang [(2013). J. Acoust. Soc. Am. 34(1), EL91–EL97] to investigate factors influencing older listeners' Mandarin speech recognition in quiet vs single-talker interference. Listening condition significantly interacted with F0 contours but not with semantic context, revealing that natural F0 contours provided benefit in the interference condition whereas semantic context contributed similarly to both conditions. Furthermore, the significant interaction between semantic context and F0 contours demonstrated the importance of semantic context when F0 was flattened. Together, findings from the two studies indicate that aging differentially affects tonal language speakers' dependence on F0 contours and semantic context for speech perception in suboptimal conditions.

In speech recognition, smooth and rapid integration of perceptual and cognitive processes is required to extract and use the information contained in spoken language. The operation of the integrated system may deteriorate with age due to parallel or disparate declines in perceptual and cognitive domains. Although older adults' speech recognition in quiet is generally quite good, differences between older and young listeners tend to emerge under suboptimal listening conditions. For example, it is well documented that older listeners often experience difficulty recognizing and understanding speech spoken at a fast rate or when presented concurrently with background noise/interfering speech (Gordon-Salant and Fitzgibbons, 1993; Helfer and Freyman, 2008; Anderson et al., 2013). However, there are controversies on the explanations for the age-related difficulty with different classes of explanations respectively locating the difficulty in perceptual or cognitive declines associated with normal aging. There is evidence that older adults still perform more poorly on speech recognition tasks in adverse listening conditions than young listeners even when they have normal hearing ability, indicating that cognitive factors may play an important role beyond auditory/perceptual factors (Tun et al., 2002; Hasher et al., 2008; Anderson Gosselin and Gagné, 2011).

Studies on speech recognition in quiet and adverse listening conditions have shown inconsistent results regarding how aging affects the use of linguistic context. Evidence from priming studies in which target words were preceded by a meaning-related/unrelated word or sentence reported larger priming effects for older adults than young adults (see meta-analysis by Laver and Burke, 1993). Further evidence comes from some studies on predictability from sentence context which showed that older listeners benefited more from meaningful context than young listeners in recognizing sentence-final target words of noise-vocoded sentences or normal sentences in the presence of multi-talker babble (Pichora-Fuller et al., 1995; Frisina and Frisina, 1997; Sheldon et al., 2008). By contrast, Wingfield et al. (1994) reported that during recognizing words at different points in a sentence, young and old listeners showed similar improvement when there was context preceding the target words, whereas older listeners could not use context following the targets as effectively as their younger counterparts. Dubno et al. (2000) also showed that both older and young adults derive equal context benefit for recognizing sentence-final words in the presence of masking broadband noise. These results suggest that age difference in contextual benefit may differ across listening conditions and tasks, in which speech recognition relies on different cognitive processes such as processing speed, working memory, inhibition, and selective attention. For example, speech recognition in broadband noise requires inhibition of noise distraction while recognition of speech in multi-talker babbles requires selective attention to the targets in addition to distraction inhibition.

While English has been the primary target language in most studies, studies on other languages with typologically distinct acoustic and prosodic properties are needed to test the generalizability of previous findings. In a tonal language like Mandarin Chinese, fundamental frequency (F0) is a primary cue for distinguishing lexical meanings (Wang, 1973). This contrasts with non-tonal languages such as English in which variation in F0 is mainly employed for prosodic and pragmatic purposes such as emphasis, sentence modality (declarative vs interrogative), and emotion (Cutler et al., 1997). Although tones have a lexical status in Chinese, flattening F0 patterns seldom cause lexical ambiguity in Mandarin because most of the modern Chinese words consist of two syllables, which is completely different from ancient Chinese in which most words consisted of only one syllable. In our previous study (Wang et al., 2013), Mandarin sentences with flat F0 patterns were as intelligible to young listeners as sentences with natural F0 patterns when presented in quiet. However, when sentences were presented to young listeners in adverse listening conditions such as in a background of speech-shaped noise or against a competing talker, flat F0 patterns led to substantially lower intelligibility compared to natural F0 patterns (Wang et al., 2013; Chen et al., 2014). These results suggest that the interaction between lower-level F0 information and higher-level semantic information on speech recognition by young listeners is modulated by listening condition. Specifically, semantic information compensates for the distorted lexical tone contours in quiet, while natural F0 contours are important for the recognition of Mandarin sentences in adverse listening conditions. It has been shown that there is an age-related decline in the ability to use F0 information to separate target message from maskers and to interpret emotional valence of speech (Mitchell et al., 2011; Lee, 2013). A recent study revealed that Mandarin-speaking older listeners scored significantly lower than young listeners in tone recognition (Yang et al., 2015), which might be attributed to reduced neural representation of F0 information in speech for the older listeners (Anderson et al., 2013). Given these changes in F0 perception with age, the question arises whether older listeners have a deficit in using F0 information in conjunction with higher-level semantic context to assist sentence recognition in quiet and adverse listening conditions.

The present study aimed to investigate Mandarin speech recognition by older adults in quiet and single-talker interference conditions. Considering the interaction of semantic context, F0 contours, and listening condition on Mandarin speech recognition by young adults (Wang et al., 2013), we were particularly interested in exploring how these factors separately or interactively affected older adults' performance, which allows a better understanding of how aging affects perceptual and cognitive abilities associated with top-down and bottom-up processing for speech recognition.

Thirty-eight (20 males) older adults were recruited from three communities in Beijing [age range, 55–65; mean, 61.2; standard deviation (SD), 3.1]. All participants were native speakers of Mandarin Chinese and passed a hearing screening at 25 dB hearing level for octave frequencies between 125 and 4000 Hz. No participants had a history of neurological, psychiatric, or neuropsychological problems. Written informed consent was obtained from all participants. The study was approved by the Institutional Review Board of the National Key Laboratory of Cognitive Neuroscience and Learning at Beijing Normal University. A mixed-design was adopted with F0 contours (normal vs flat) as a between-subject factor (19 participants in each condition) while semantic context (normal sentence vs word list sentence) and listening condition [quiet, signal-to-noise ratio (SNR) = +5 dB] as within-subject factors. Presentation order of the four within-subject conditions was randomized.

The materials were used in our previous study on young adults (Wang et al., 2013). The normal sentences were 20 simple declarative Chinese sentences with a variety of topics that are familiar to both young and older listeners. Each sentence was composed of 5–9 words (4–5 content words plus 0–4 functional words) with a simple syntactic structure. Words from the entire pool of the normal sentences were pseudo-randomly selected to form the word list sentences matched in number of content and functional words with the normal sentences. Word list sentences were syntactically correct but semantically meaningless at the whole sentence level. That is, recognition of both word list and normal sentence recruits syntactic and semantic processing, but only normal sentence provides the meaningful semantic context of “who does what to whom” beyond random lexical semantics. The normal sentences and word list sentences were read by a male native speaker of Chinese. Manipulation of F0 was done using Praat (Institute of Phonetic Sciences, University of Amsterdam; downloadable at www.praat.org). A flat F0 contour was created for each sentence at the sentence's mean F0 and the resulting monotonous sentence was resynthesized using the PSOLA method (Fig. 1). Consonant-misplaced sentences constructed by replacing the onset consonant of each syllable in the normal sentences with another consonant were used as masker stimuli to minimize the effects of informational masking because consonant-misplaced sentences were not meaningful at both lexical and sentential levels (Xu et al., 2013). The masker sentences were read by a female native Mandarin speaker, which was to enable separation of the target message from the interfering speech. The listeners were instructed to listen to the male speaker for the content rather than the identity of the target speaker throughout (Scott et al., 2004). Each target sentence was combined with interfering speech at the SNR level of +5 dB with the target and masker sentences fixed at 70 and 65 dB sound pressure level (SPL), respectively. The +5 dB SNR and SPL of the target and masker sentences were the same as our previous study (Wang et al., 2013). However, the −5 dB SNR condition in our previous study was not tested here because a pilot study with older listeners showed a floor effect in recognition scores for several conditions (e.g., normal sentence with flat F0 contours, word list sentence with natural or flat F0 contours).

Fig. 1.

(Color online) Acoustic features of sample speech stimuli. Broad-band spectrograms (SPG: 0–5 kHz), intensity envelopes (INT: 50–100 dB), and fundamental frequency contours (F0: 50–250 Hz) are displayed for (A) normal sentence and its pitch-flattened counterpart; (B) word list sentence and its pitch-flattened counterpart. Note: translation of the example sentences. Normal sentence: The old man often comes to the park for a walk. Word list sentence: The telephone received the enemy's passengers.

Fig. 1.

(Color online) Acoustic features of sample speech stimuli. Broad-band spectrograms (SPG: 0–5 kHz), intensity envelopes (INT: 50–100 dB), and fundamental frequency contours (F0: 50–250 Hz) are displayed for (A) normal sentence and its pitch-flattened counterpart; (B) word list sentence and its pitch-flattened counterpart. Note: translation of the example sentences. Normal sentence: The old man often comes to the park for a walk. Word list sentence: The telephone received the enemy's passengers.

Close modal

We followed similar experimental procedures in our previous study (Wang et al., 2013), except that listeners in the current study were instructed to verbally repeat rather than write down the whole target normal/word list sentences in order to reduce the cognitive burden on the older listeners. Their responses were recorded and scored by W.J. and checked by an independent auditor blind to the experiment. Listeners were tested individually in a sound-attenuated booth with ambient noise level below 15 dB(A). The stimuli were presented via loudspeakers (Edifier R18, Edifier Technology Co. Ltd., Beijing, China) and the sound level of the stimuli was calibrated to 65 dB SPL at the subject's head. With semantic context and listening condition as within-subject factors, each listener was presented with a total of 40 trials: 10 normal sentences and 10 word list sentences in each listening condition. Practice sentences were provided before the experiment for all the conditions.

Speech recognition accuracy was determined by a keyword-correct count where each content word provided a potential key word (Scott et al., 2004; Wang et al., 2013). Percent-correct score was obtained based on the number of correct responses (see Fig. 2) and converted to rationalized arcsine unit (RAU) for further analysis (Studebaker, 1985). A 2 × 2 × 2 repeated measures analysis of variance (ANOVA) was carried out, with semantic context and listening condition as the within-subject factors and F0 contours as the between-subject factor. Results showed that the three main effects were all highly significant [sentence context: F(1, 36) = 172.91, mean square error (MSE) = 62.374, p < 0.001, partial η2 = 0.828; F0 contours: F(1, 36) = 31.95, MSE = 611.568, p < 0.001, partial η2 = 0.47; listening condition: F(1, 36) = 155.56, MSE = 64.317, p < 0.001, partial η2 = 0.812], Thus, there is clear evidence that speech recognition in older listeners was compromised by the lack of sentential semantic context and natural F0 contours, as well as by the presence of interfering speech.

Fig. 2.

Mean word-report scores (RAU) and standard deviation for all conditions in the older and young groups. * Significant at p < 0.00625 with the α level set at 0.05. Error bars represent SD across subjects. Note: NS, normal sentence; WS, word list sentence, NF, normal F0 contours, FF, flat F0 contours. Results of the young group are based on Wang et al. (2013).

Fig. 2.

Mean word-report scores (RAU) and standard deviation for all conditions in the older and young groups. * Significant at p < 0.00625 with the α level set at 0.05. Error bars represent SD across subjects. Note: NS, normal sentence; WS, word list sentence, NF, normal F0 contours, FF, flat F0 contours. Results of the young group are based on Wang et al. (2013).

Close modal

ANOVA tests further revealed significant interaction effects between F0 contours and listening condition [F(1, 36) = 17.47, MSE = 64.317, p < 0.001, partial η2 = 0.327] and between F0 contours and semantic context [F(1, 36) = 7.66, MSE = 62.374, p = 0.009, partial η2 = 0.175]. These interactions showed that Mandarin speech recognition accuracy disproportionately decreased for flat-contour sentences versus natural-contour sentences when the listening condition changed from quiet to the interfering background, and when semantic context changed from normal sentence to word list sentence. However, there was no significant interaction between semantic context and listening condition [F(1, 36) = 0.17, MSE = 82.515, p = 0.682, partial η2 = 0.005], revealing that speech recognition accuracy decreased to a similar degree for normal sentence versus word list sentence when listening condition changed from quiet to the interfering background (Fig. 3). The 3-way interaction effect between semantic context, listening condition and F0 contours was not significant [F(1, 36) = 0.34, MSE = 82.515, p = 0.566, partial η2 = 0.009].

Fig. 3.

Word-report scores (RAU) sorted by the interaction effects. (A) Significant interaction between F0 contours and listening condition (p < 0.001), (B) significant interaction between F0 contours and semantic context (p = 0.009), (C) non-significant interaction between listening condition and semantic context (p = 0.682). Error bars represent SD across subjects.

Fig. 3.

Word-report scores (RAU) sorted by the interaction effects. (A) Significant interaction between F0 contours and listening condition (p < 0.001), (B) significant interaction between F0 contours and semantic context (p = 0.009), (C) non-significant interaction between listening condition and semantic context (p = 0.682). Error bars represent SD across subjects.

Close modal

In order to explore the possible age difference in use of semantic context and F0 contours during speech recognition in quiet and single-talker interference conditions, eight Independent-sample T-tests were carried with Bonferroni correction (p ≤ 0.00625 with the α level set at 0.05) for multiple comparisons based on the data of the present and our previous study (Wang et al., 2013) (Fig. 2). The data from the two studies could not merge together to carry out ANOVAs because they differ in the design, with listening condition as a between-subject in the previous study and a within-subject factor in the current study. The results showed that older listeners performed worse than young listeners in recognizing normal sentence with natural (t(42) = 2.878, p = 0.006) or flat F0 contours (t(41) = 2.897, p = 0.006) in the single-talker interference condition. The two groups performed equally well in recognizing normal sentence with natural or flat F0 contours in quiet, and also did not differ for word list sentences irrespective of listening condition and F0 contour manipulation. These results indicated that recognition of normal sentences rather than word list sentences against interfering speech could be impaired in older listeners when compared with young adult listeners.

The present study investigated the use of sentence context and F0 contours by older listeners during Mandarin speech recognition in quiet and single-talker interference conditions. The significant main effects of semantic context and F0 contours demonstrated that older listeners could use both cues to assist speech recognition. Furthermore, the significant interaction between F0 contours and listening condition revealed that natural F0 contours provided extra benefit in the interference condition, whereas the absence of interaction effect between semantic context and listening condition showed that semantic context contributed to speech recognition to a similar degree in both the quiet and interfering backgrounds.

In a tonal language like Mandarin Chinese, F0 contours play an important role in speech recognition, especially when speech is presented without contextual information (e.g., isolated syllables) or in adverse listening conditions (e.g., in the presence of N-talker babbles) (Patel et al., 2010; Wang et al., 2013; Yang et al., 2015). The significant main effect of F0 contours and interaction effects between F0 contours and listening condition/semantic context indicate that older listeners have the ability to use F0 information to assist speech recognition, especially in the presence of interfering speech or absence of sentential semantic context. Flatting F0 contours alters tonal and prosodic information at the same time, thus it is not possible to isolate the effects of altered tonal versus prosodic feature for the present study. This issue needs further investigation, e.g., by using speech software to make stimuli with altered prosodic but intact tonal information or vice versa. However, the current finding that there was significant difference in recognition accuracy between word list sentence with and without natural F0 patterns seems to indicate that word-level F0 patterns (i.e., lexical tones) could be used by older listeners during Mandarin speech recognition. This finding is further confirmed by the results of direct comparisons between young and older listeners, which revealed no significant age-related difference in recognizing word list sentence with natural F0 patterns. Our results appear to be inconsistent with those of some previous studies that have revealed a decline in the ability of older listeners to use F0 information in speech recognition (Mitchell et al., 2011; Lee, 2013; Yang et al., 2015). The discrepancies might be related to differences in subjects' characteristics and language materials in the current and previous studies. For example, our participants had an age range of 55–65 and thus were relatively younger than those of previous studies. Further research that includes more refined age groups of older listeners and stimuli with smaller F0 differences is needed to better understand whether and how F0 information is used by older listeners of various ages during speech recognition.

There is controversy as to how semantic context is used by older listeners in speech recognition under adverse listening conditions. In the current study, the significant main effect of semantic context and absence of interaction effect between semantic context and listening condition jointly revealed that sentential context contributed to speech recognition by older listeners, but to a similar degree in both the quiet and interfering backgrounds. The lack of interaction between semantic context and listening condition in older adults differs from the result of our previous study with young listeners (Wang et al., 2013), in which recognition of the word list sentence degraded to a larger extent than a normal sentence when the listening condition changed from quiet to interfering backgrounds. The lack of extra semantic context effect in older listeners' speech recognition against interfering speech is consistent with the results of direct comparisons between young and older listeners, which revealed significant age-related differences in recognizing normal sentences as well as no significant between-group differences in recognizing word list sentences in the single-talker interference condition. However, caution is necessary in directly comparing and interpreting the different interaction effects across studies that did not use identical experimental protocols. Our finding that older listeners had decreased ability to use semantic context to assist speech perception against interfering speech is inconsistent with the results of some previous studies which have shown that older listeners benefit more from semantic context than young listeners in adverse listening conditions (Laver and Burke, 1993; Pichora-Fuller et al., 1995; Frisina and Frisina, 1997; Sheldon et al., 2008). The inconsistency might be partly attributable to the difference in masker stimuli. Specifically, broadband noise or multi-talker babble was used in the previous studies. In the current study, however, we used single-talker babble as masker stimuli with marked gender difference between the target voice and the interfering voice. Interestingly, in another study in which single-talker babble was used as masker, greater semantic priming for older adults than young adults was only observed when context was presented in quiet; in the single-talker babble condition, however, the facilitation effect for context was only observed in young adults but absent in older listeners (Aydelott et al., 2010). Collectively, these results seem to indicate that older listeners' ability to use semantic context in adverse listening conditions is affected by the type of masker stimuli. It is noteworthy that the maskers used in the present study are consonant-misplaced sentences that are syntactically anomalous and semantically uninterpretable, and thus have energetic plus minimal effect of informational masking on the targets. Our findings may indicate that such single-talker interference would readily distract older adults' attention from integration of semantic context irrespective of whether the information provided by maskers is relevant to the targets. Different types of maskers (e.g., broadband noise, speech-shaped noise, time-reversed speech, one- and multi-talker babble) might have different effects on speech recognition because they differ to a great degree in masking targets and distracting listeners (energetic, informational, or both). A detailed discussion of such effects is beyond the scope of this report because only single-talker interfering speech was included in our study. It merits further investigation how the auditory and cognitive processes are affected by characteristics of maskers as a function of age and listening condition.

This research was supported by grants from the Program for New Century Excellent Talents in University (Grant No. NCET–13–0691) and the Science Foundation of Beijing Language and Culture University (Fundamental Research Funds for the Central Universities) (Grant No. 16WT02) to L.J.Z., and in part from the Natural Science Foundation of China (Grant No. 81461130018) to H.S. Y.Z. was additionally supported by the Grand Challenges Exploratory Research Grant Award from the University of Minnesota.

2.
Anderson
,
S.
,
White-Schwoch
,
T.
,
Parbery-Clark
,
A.
, and
Kraus
N.
(
2013
). “
A dynamic auditory-cognitive system supports speech-in-noise perception in older adults
,”
Hear. Res.
300
(
3
),
18
32
.
1.
Anderson Gosselin
,
P.
, and
Gagné
,
J. P.
(
2011
). “
Older adults expend more listening effort than young adults recognizing speech in noise
,”
J. Speech Lang. Hear. Res.
54
(
3
),
944
958
.
3.
Aydelott
,
J.
,
Leech
,
R.
, and
Crinion
,
J.
(
2010
). “
Normal adult aging and the contextual influences affecting speech and meaningful sound perception
,”
Trends Amplif.
14
(
4
),
218
232
.
4.
Chen
,
F.
,
Wong
,
L. L.
, and
Hu
,
Y.
(
2014
). “
Effects of lexical tone contour on Mandarin sentence intelligibility
,”
J. Speech Lang. Hear. Res.
57
(
1
),
338
345
.
5.
Cutler
,
A.
,
Dahan
,
D.
, and
van Donselaar
,
W.
(
1997
). “
Prosody in the comprehension of spoken language: A literature review
,”
Lang. Speech
40
(
Pt. 2
),
141
201
.
6.
Dubno
,
J. R.
,
Ahlstrom
,
J. B.
, and
Horwitz
,
A. R.
(
2000
). “
Use of context by young and aged adults with normal hearing
,”
J. Acoust. Soc. Am.
107
,
538
546
.
7.
Frisina
,
D. R.
, and
Frisina
,
R. D.
(
1997
). “
Speech recognition in noise and presbycusis: Relations to possible neural mechanisms
,”
Hear. Res.
106
,
95
104 (1997)
.
8.
Gordon-Salant
,
S.
, and
Fitzgibbons
,
P.
(
1993
). “
Temporal factors and speech recognition performance in young and elderly listeners
,”
J. Speech Hear. Res.
36
,
1276
1285
(1993).
9.
Hasher
,
L.
,
Lustig
,
C.
, and
Zacks
,
R.
(
2008
). “
Inhibitory mechanisms and the control of attention
,” in
Variation in Working Memory
, edited by
A.
Conway
,
C.
Jarrold
,
M.
Kane
,
A.
Miyake
, and
J.
Towse
(
Oxford University Press
,
London
), pp.
227
249
.
10.
Helfer
K. S.
, and
Freyman
R. L.
(
2008
). “
Aging and speech-on-speech masking
,”
Ear Hear.
29
(
1
),
87
98
.
12.
Laver
,
G. D.
, and
Burke
,
D. M.
(
1993
). “
Why do semantic priming effects increase in old age? A meta-analysis
,”
Psychol. Aging
8
(
1
),
34
43
.
13.
Lee
,
J. H.
(
2013
). “
Age-related deficits in the processing of fundamental frequency differences for the intelligibility of competing voices
,”
Korean J. Audiol.
17
(
1
),
1
8
.
15.
Mitchell
,
R. L.
,
Kingston
,
R. A.
, and
Barbosa Bouças
,
S. L.
(
2011
). “
The specificity of age-related decline in interpretation of emotion cues from prosody
,”
Psychol. Aging
26
(
2
),
406
414
.
16.
Patel
,
A. D.
,
Xu
,
Y.
, and
Wang
,
B.
(
2010
). “
The role of F0 variation in the intelligibility of Mandarin sentences
,” in
Proceedings of Speech Prosody
,
Chicago, IL
.
17.
Pichora-Fuller
,
M. K.
,
Schneider
,
B. A.
, and
Daneman
M.
(
1995
). “
How young and old adults listen to and remember speech in noise
,”
J. Acoust. Soc. Am.
97
(
1
),
593
608
.
18.
Scott
,
S. K.
,
Rosen
,
S.
,
Wickham
,
L.
, and
Wise
,
R. J.
(
2004
). “
A positron emission tomography study of the neural basis of informational and energetic masking effects in speech perception
,”
J. Acoust. Soc. Am.
115
(
2
),
813
821
.
19.
Sheldon
,
S.
,
Pichora-Fuller
,
M. K.
, and
Schneider
,
B. A.
(
2008
). “
Priming and sentence context support listening to noise-vocoded speech by younger and older adults
,”
J. Acoust. Soc. Am.
123
(
1
),
489
499
.
20.
Studebaker
,
G. A.
(
1985
). “
A ‘rationalized’ arcsine transform
,”
J. Speech Hear. Res.
28
(
3
),
455
462
.
21.
Tun
,
P. A.
,
O'Kane
,
G.
, and
Wingfield
,
A.
(
2002
). “
Distraction by competing speech in younger and older listeners
,”
Psychol. Aging
17
,
453
467
.
23.
Wang
,
J.
,
Shu
,
H.
,
Zhang
,
L.
,
Liu
,
Z.
, and
Zhang
,
Y.
(
2013
). “
The roles of fundamental frequency contours and sentence context in Mandarin Chinese speech intelligibility
,”
J. Acoust. Soc. Am.
34
(
1
),
EL91
EL97
.
22.
Wang
,
W. S. Y.
(
1973
). “
The Chinese language
,”
Sci. Am.
228
,
50
60
.
24.
Wingfield
,
A.
,
Alexander
,
A. H.
, and
Cavigelli
,
S.
(
1994
). “
Does memory constrain utilization of top-down information in spoken word recognition? Evidence from normal aging
,”
Lang. Speech
37
,
221
235
.
25.
www.praat.org (Last viewed March 23, 2017).
26.
Xu
,
G.
,
Zhang
,
L.
,
Shu
,
H.
,
Wang
,
X.
, and
Li
,
P.
(
2013
). “
Access to lexical meaning in pitch-flattened Chinese sentences: An fMRI study
,”
Neuropsychologia
51
(
3
),
550
556
.
27.
Yang
,
X.
,
Wang
,
Y.
,
Xu
,
L.
,
Zhang
,
H.
,
Xu
,
C.
, and
Liu
,
C.
(
2015
). “
Aging effect on Mandarin Chinese vowel and tone identification
,”
J. Acoust. Soc. Am.
138
(
4
),
EL411
EL416
.