Content uploaded by Wentao Gu
Author content
All content in this area was uploaded by Wentao Gu on Mar 09, 2021
Content may be subject to copyright.
Prosodic Characteristics of Genuine and Mock (Im)polite Mandarin Utterances
Chengwei Xu1 and Wentao Gu1,2
1School of Foreign Languages and Cultures, Nanjing Normal University, China
2School of Chinese Language and Literature, Nanjing Normal University, China
13815877927@163.com, wtgu@njnu.edu.cn
Abstract
As specialized social affects in speech communication, mock
politeness and mock impoliteness are usually characterized by
unique prosodic patterns that conflict with the literal meanings.
To give a quantitative analysis of prosodic characteristics, a
context-elicited discourse completion task was conducted to
collect genuine and mock (im)polite Mandarin utterances in
both imperative and interrogative modes. Results revealed that
prosodic features played roles in a complex way. Mock polite
speech showed a higher maximum F0 and intensity, a wider
range as well as a higher variability of F0 and intensity, a
lower HNR, and a higher jitter than genuine polite speech,
whereas mock impolite speech showed a lower
mean/maximum F0 and intensity, a narrower range as well as a
lower variability of F0 and intensity, a slower speech rate, a
higher HNR, and lower jitter, shimmer and H1-H2 than
genuine impolite speech. In the perceptual experiment, the
lower identification rates on mock (im)politeness indicated
that perceptual judgement was influenced by literal meanings.
Politeness ratings further showed that mock (im)polite speech
was less (im)polite than genuine (im)polite speech, suggesting
a good correspondence between prosodic manifestations and
perceived politeness. Moreover, interrogatives sounded more
polite than imperatives, also verifying the Tact Maxim
principle for politeness.
Index Terms: (im)politeness, genuine/mock, prosody,
Mandarin speech
1. Introduction
(Im)politeness is an important topic in pragmatics study. Most
research to date, however, has focused on genuine politeness
and impoliteness, yet many issues related to (im)politeness do
not straightforwardly fit these two opposing extremes.
Culpeper et al. [1] proposed that interpersonal interactions
involved mixed messages, that is, the messages that contain
features mixed with polite and impolite interpretations.
Among them, mock politeness and mock impoliteness are two
social attitudes that have been widely discussed.
Two important principles of (im)politeness mismatch (i.e.,
Irony Principle and Banter Principle) proposed by Leech [2]
have been used as the basis for the conceptualization of mock
(im)politeness. The Irony Principle underlines impoliteness to
listeners, while the Banter Principle emphasizes politeness in
which untrue and impolite utterances are superficial in nature.
Leech’s two principles were later integrated into Culpeper’s
model of impoliteness [3, 4], where he defined mock
politeness as “an impoliteness understanding that does not
match the surface form or semantics of the utterance or the
symbolic meaning of the behavior” and mock impoliteness as
“the recontextualisation of impoliteness in socially opposite
contexts creates socially opposite effects, namely affectionate,
intimate bonds amongst individuals and the identity of that
group.” A more recent work on the metapragmatics of mock
politeness conducted by Taylor [5] gave another definition, i.e.,
“mock politeness occurs when there is an (im)politeness
mismatch leading to an implicature of impoliteness.” What is
more, the only phraseology that can potentially predict mock
polite behaviors is over-politeness [6], which is considered as
a mismatch between the speaker’s intention and self-
presentation, including instances where the speaker hides
insincerity for manipulating the listener [7].
Thus, the present study employed over-politeness as a
manifestation of mock politeness. Since the Chinese word ᛘ
Nin2 (an honorific form of ‘you’) is a good indicator of
politeness, mock politeness will be indicated when Nin2 is
used in a close relationship. Also, the mismatch between a
close relationship and an impolite expression is a good
indicator of mock impoliteness.
While most previous studies on the relationship between
speech prosody and (im)politeness were based on qualitative
and impressionistic observations on fundamental frequency
(F0) contours [4, 8-10], recent works have focused more on
quantitative analysis of prosodic parameters including F0,
duration, and intensity, in a combination with more qualitative
variables such as pragmatic expressions of (im)politeness
categories [11, 12]. A study of Catalan [11] revealed a higher
F0, a narrower F0 range, and a lower maximum intensity in
mock impolite speech than in genuine impolite speech,
whereas a study of German and Polish [12] showed that mock
impolite speech had a lower intensity and a larger F0
variability than genuine impolite speech. Voice quality
features are also closely related to politeness marking. For
example, speakers of Korean used a breathier voice in
informal/impolite speech than in formal/polite speech, for
significant differences in H1-H2, HNR, jitter, and shimmer
were found between the two types of speech [13].
In particular, China is conventionally known as “a nation
of etiquette,” where being polite to others is a traditional virtue.
Therefore, how Chinese people differentiate genuine and
mock (im)polite speech acoustically and perceptually deserves
particular investigation. This study looked into prosodic
manifestations and perceptual attributes of Mandarin speech
conveying four types of attitudes relevant to (im)politeness,
i.e., genuine politeness (GP), mock politeness (MP), genuine
impoliteness (GI), and mock impoliteness (MI).
Also, according to the Tact Maxim, the first of the six
principles for politeness [2], interrogative (as an indirect
speech act) is generally more polite than imperative (as a
direct speech act) when requesting others. Therefore, we also
included sentence mode (interrogative vs. imperative) as
another control factor in the present study.
Copyright © 2020 ISCA
INTERSPEECH 2020
October 25–29, 2020, Shanghai, China
http://dx.doi.org/10.21437/Interspeech.2020-32314153
2. Speech data
2.1. Corpus design
To keep a balance between naturalness and controllability,
elicited speech was used in this study. An oral version of the
Discourse Completion Task, originated from pragmatics study
and later widely used in prosodic research, was adopted as the
method of elicitation. For a comparison of genuine and mock
politeness, we designed a set of target sentences expressing
request associated with literally polite wording. For a
comparison of genuine and mock impoliteness, we designed a
set of target sentences expressing criticism associated with
literally impolite wording. For each target sentence, two
scenarios were designed to elicit genuine and mock attitudes,
respectively. The target sentence was placed at the end of each
scenario. Here are a set of example scenarios, each of which
were used twice, with either an interrogative or an imperative
sentence (shown as 1 and 2, respectively) embedded at the end:
x Genuine politeness:
You took a bus to school. Your seat was by the window,
and you were going to get off. You said to the passenger next
to you politely:
(1) “哫✖ᛘ㜭ᥚал?” (Could you move over, please?)
(2) “哫✖ᛘᥚал!” (Please move over!)
x Mock politeness:
Your boy/girlfriend and you took a bus to work together.
Your seat was by the window. Your boy/girlfriend was
playing his/her phones all the time without any interaction
with you and even ignored your words. You were ready to get
off, but he/she was still unaware of it. So, you said to him/her:
(1) “哫✖ᛘ㜭ᥚал?” (Could you move over, please?)
(2) “哫✖ᛘᥚал!” (Please move over!)
x Genuine impoliteness:
You helped your son with his homework in the evening.
When it came to a math problem which you had told him at
least three times, he still could not make it. You said to him:
(1) “㜭䮯⛩ᗳ?” (Can you pay more attention?)
(2) “䮯⛩ᗳ!” (Pay more attention!)
x Mock impoliteness:
You went to class with your roommate in the morning.
Your roommate was a very careless person. When he/she
entered the classroom, he/she suddenly realized that he/she
forgot to bring his/her schoolbag. You said to him/her:
(1) “㜭䮯⛩ᗳ?” (Can you pay more attention?)
(2) “䮯⛩ᗳ!” (Pay more attention!)
2.2. Data recording
Twenty-four native speakers of Mandarin (12M, 12F) were
recruited for speech recording. They were graduate students at
the age of 22-29 (mean = 23.9, SD = 1.7), with a language
proficiency of Mandarin at 2A level, which is generally good
enough. None of them had any experience of performing and
vocal expression. They were reasonably remunerated for their
participation.
After each speaker was familiarized with the materials and
could express the intended attitudes appropriately, speech
recording was conducted in a sound-proof booth, with a
cardioids microphone Neumann U87Ai placed about 15cm in
front of the mouth. The speaker was asked to read the context
first and then speak aloud the target sentence. Speech signals
were recorded with a sampling rate of 44,100Hz at a 16-bit
precision. Eighty utterances (20 target sentences × 2 sentence
modes × 2 attitudes) were collected for each speaker. Thus,
there were 1920 target utterances altogether.
3. Acoustic analysis
For each target utterance, we measured five fundamental
frequency (F0) parameters, five intensity parameters, five
voice quality parameters, and speech rate. Five F0/intensity
parameters included the mean, standard deviation, max, min,
and range of F0/intensity, whereas five voice quality
parameters were jitter, shimmer, Harmonics-to-Noise Ratio
(HNR), H1-H2, and H1-A1. Among them, H1-H2 and H1-A1
were extracted using VoiceSauce, while others were extracted
using Praat. All F0 parameters were measured in semitone (st)
with a reference of 50 Hz, and all intensity parameters were
transformed to z-scores.
Statistical comparisons were then conducted in each pair
of attitudes, one between genuine vs. mock politeness, and the
other between genuine vs. mock impoliteness. In each pair, all
16 parameters were analyzed with linear mixed-effects models
(LMMs) using the Jamovi program. Attitude type (genuine vs.
mock) and sentence mode (imperative vs. interrogative) were
fixed effects, while sentence was a random effect.
Results of statistical comparison on four sets of parameters
are shown in Tables 1-4, where asterisk indicates a significant
difference between GP and MP, while cross indicates a
significant difference between GI and MI. It should be noted
that there was no comparison between (GP, MP) and (GI, MI).
Table 1. Statistical comparison of F0 parameters.
F0 (st)
GP
MP
GI
MI
F0_Mean
22.61
23.11
23.54†
22.23†
F0_std
1.62*
1.78*
2.21†
1.97†
F0 Min
17.04
17.06
16.06
15.73
F0_Max
26.76*
27.71*
29.31†
27.58†
F0_Range
11.04*
13.22*
17.49†
13.92†
Table 2. Statistical comparison of intensity parameters.
Intensity (dB)
GP
MP
GI
MI
Intensity_Mean
58.16*
62.05*
64.15†
60.06†
Intensity_std
14.12*
16.27*
17.08†
15.53†
Intensity Min
23.80
23.36
23.71†
23.52†
Intensity_Max
67.30*
72.09*
74.78†
69.68†
Intensity_Range
43.50*
48.72*
51.08†
46.16†
Table 3. Statistical comparison of speech rate.
GP
MP
GI
MI
Speech rate (syl/s)
5.41
5.55
5.81†
5.18†
Table 4. Statistical comparison of voice quality parameters.
Voice quality
GP
MP
GI
MI
Jitter
0.02*
0.02*
0.02†
0.02†
Shimmer
0.09
0.09
0.11†
0.10†
HNR
13.08*
12.53*
11.20†
12.60†
H1-H2
6.16
5.62
5.96†
5.60†
H1-A1
3.96
3.82
6.64
6.56
4154
3.1. Fundamental frequency
In a comparison of GP and MP, there was no significant main
effect on F0_Mean (p = 0.081) and F0_Min (p = 0.834).
However, F0_std was by 0.16st significantly higher (p =
0.010), F0_Max was by 0.95st higher (p = 0.020), and
F0_Range was by 2.18st higher (p = 0.013) in MP than in GP.
There were significant interaction effects of
Attitude×Mode on F0_Mean (β = -0.29, SE = 0.14, t = -2.08, p
= 0.038), F0_std (β = -0.15, SE = 0.08, t = -1.96, p=0.050),
F0_Max (β = -0.85, SE = 0.24, t = -3.50, p < 0.001) and
F0_Range (β = -1.85, SE = 0.67, t = -2.76, p = 0.006). Simple
effect analysis showed that only for interrogatives, F0_Mean
was by 0.73st significantly higher (p = 0.032), F0_std was by
0.23st higher (p =0.001), F0_Max was by 1.47st higher (p =
0.002), and F0_Range was by 3.25st higher (p = 0.001) in MP
than in GP.
In a comparison of GI and MI, F0_Mean was by 1.31st
significantly lower (p < 0.001), F0_std was by 0.24st lower (p
=0.003), F0_Max was by 1.73st lower (p < 0.001), and
F0_Range was by 3.57st narrower (p < 0.001) in MI. There
was no significant main effect on F0_Min (p = 0.326).
There was a significant interaction effect of
Attitude×Mode on F0_Mean (β = -0.59, SE = 0.19, t = -3.15, p
= 0.002). For interrogatives MI was by 1.09st lower than GI (p
= 0.001), while for imperatives MI was by 1.69st lower than
GI (p < 0.001). There was a significant interaction effect on
F0_std (β = 0.44, SE = 0.09, t = 4.73, p < 0.001), and only for
interrogatives MI was by 0.46st lower than GI (p < 0.001).
There was a significant interaction effect on F0_Min (β = -2.41,
SE = 0.47, t = -5.09, p < 0.001), and only for imperatives MI
was by 1.61st lower than GI (p < 0.001). Also, there was a
significant interaction effect on F0_Range (β = 2.86, SE =
0.74, t = 3.87, p <0.001). For interrogatives MI was by 5.19st
lower than GI (p < 0.001), while for imperatives MI was by
2.33st lower than GI (p = 0.020). No significant interaction
effect was found on F0_Max.
3.2. Intensity
In a comparison of GP and MP, Intensity_Mean was by
3.89dB significantly higher (p < 0.001), Intensity_std was by
2.15dB higher (p < 0.001), Intensity_Max was by 4.79dB
higher (p < 0.001), and Intensity_Range was by 5.22dB higher
(p < 0.001) in MP. There was no significant main effect on
Intensity_Min (p = 0.893).
There was also a significant interaction effect of
Attitude×Mode on Intensity_Mean (β = -0.81, SE = 0.29, t = -
2.78, p = 0.005). For interrogatives MP was by 4.28dB higher
than GP (p < 0.001), while for imperatives MP was by 3.47dB
higher than GP (p < 0.001). There was a significant interaction
effect on Intensity_std (β = -1.46, SE = 0.19, t = -7.79, p <
0.001). For interrogatives MP was by 2.71dB higher than GP
(p < 0.001), while for imperatives MP was by 1.23dB higher
than GP (p = 0.001). There was a significant interaction effect
on Intensity_Max (β = -1.55, SE = 0.35, t = -4.40, p < 0.001).
For interrogatives MP was by 5.55dB higher than GP (p <
0.001), while for imperatives MP was by 3.99dB higher than
GP (p < 0.001). Also, a significant interaction effect was
found on Intensity_Range (β = -2.48, SE = 0.38, t = -6.55, p <
0.001). For interrogatives MP was by 6.08dB higher than GP
(p < 0.001), while for imperatives MP was by 3.57dB higher
than GP (p < 0.001).
In a comparison of GI and MI, Intensity_Mean was by
4.08dB significantly lower (p < 0.001), Intensity_std was
1.55dB lower (p < 0.001), Intensity_Min was by 0.19dB lower
(p < 0.001), Intensity_Max was by 5.10dB lower (p < 0.001),
and Intensity_Range was by 4.92dB lower (p < 0.001) in MI.
There was also an interaction effect of Attitude×Mode on
Intensity_Min (β = -0.99, SE = 0.30, t = -4.23, p < 0.001).
Only for imperatives MI was by 1.01dB lower than GI (p <
0.001). However, no significant interaction effect was found
on Intensity_Mean (p = 0.642), Intensity_std (p = 0.493),
Intensity_Max (p = 0.522), and Intensity_Range (p = 0.178).
3.3. Speech rate
In a comparison of GP and MP, no significant main effect was
found on speech rate (p = 0.178). However, there was a
significant interaction effect of Attitude×Mode (β = -0.26, SE
= 0.07, t = -3.74, p < 0.001). Only for interrogatives MP was
by 0.27syl/s faster than GP (p = 0.018).
In a comparison of GI and MI, speech rate was by
0.64syl/s significantly slower in MI (p < 0.001). There was
also a significant interaction effect of Attitude×Mode (β = -
0.31, SE = 0.10, t = -3.09, p = 0.002). For interrogatives MI
was by 0.27syl/s slower than GI (p < 0.001), while for
imperatives MI was by 0.79syl/s slower than GI (p < 0.001).
3.4. Voice quality
In a comparison of GP and MP, jitter was by 0.00083
significantly higher (p < 0.001), and HNR was by 0.55 lower
(p = 0.003) in MP. There was also a significant interaction
effect of Attitude×Mode on HNR (β = 0.49, SE = 0.21, t =2.29,
p = 0.020). Only for interrogatives MP was by 0.82 lower than
MP (p < 0.001). No significant main or interaction effect was
found on shimmer, H1-H2, and H1-A1.
In a comparison of GI and MI, jitter was by 0.00083
significantly lower (p < 0.001), shimmer was by 0.011 lower
(p < 0.001), HNR was by 1.40 higher (p < 0.001), and H1-H2
was by 0.36 lower (p = 0.035) in MI. There was no significant
main effect of Attitude on H1-A1 (p = 0.816). Also, there was
no significant interaction effect of Attitude×Mode on any of
these parameters.
4. Perceptual experiment
4.1. Method
The perceptual experiment consisted of two tasks, i.e., an
identification task and a politeness rating task. Twelve native
listeners of Mandarin at similar ages (mean = 24.7; SD = 1.0)
participated in the identification task, and other 21 native
listeners at similar ages (mean = 24.4; SD = 2.7) participated
in the politeness rating task. None of them had a reported
history of auditory or cognitive disorders.
After a perceptual validation test conducted via the online
platform www.wjx.cn by four native listeners who were all
graduate students in linguistics, altogether 320 utterances from
speech recordings were selected as stimuli, which were
partitioned into four lists of 80 utterances using a full Latin-
square counterbalancing procedure, with four attitude types
rotating across the lists. Each participant was assigned
randomly one of the four lists for perceptual test.
In the identification task, for each utterance a participant
was asked to choose one out of the four given attitudes. In the
politeness rating task, for each utterance a participant was
asked to give a score of politeness on a 7-point Likert scale,
where 7 represents highly polite and 1 indicates highly
impolite.
4155
4.2. Results
As shown in Table 5, the ordering of the identification rates
among four attitudes is GP > GI > MP > MI. Genuine
(im)politeness is much better identified than the mock ones. In
addition, the rate of identification in GP is higher for
interrogatives (97.50%) than for imperatives (90.83%); in MP
and GI, however, the rates of identification are higher for
imperatives than for interrogatives.
As shown in Table 6, the ordering of the average scores of
politeness among the four attitudes is GP > MP > MI > GI. In
addition, interrogatives consistently have higher average
scores of politeness than imperatives in all four attitudes.
Table 5. Rates of identification (%).
Attitude
Rate (%)
Sentence mode
Imperative
Interrogative
GP
94.17
90.83
97.50
MP
64.58
69.17
60.00
GI
82.71
87.50
77.92
MI
56.87
57.50
56.25
Table 6. Scores of politeness rating.
Attitude
Score
Sentence mode
Imperative
Interrogative
GP
6.02
5.72
6.32
MP
4.72
4.26
5.19
GI
2.04
1.80
2.28
MI
2.72
2.64
2.80
5. Discussion
By acoustic analysis, we examined prosodic manifestations of
genuine and mock (im)polite Mandarin speech. Since mock
politeness is deemed to be less polite than genuine politeness
while mock impoliteness is deemed to be more polite than
genuine impoliteness, as was also verified in the perceptual
experiment, we can compare our findings with previous
studies on (im)polite speech.
In the first place, the max, range, and SD of F
0 were
significantly higher in MP than in GP, whereas the mean, max,
range, and SD of F0 was significantly lower in MI than in GI.
There were also significant interaction effects between attitude
type and sentence mode. For example, in a comparison of GP
and MP, only for interrogatives there was a significant
difference between MP and GP.
As revealed in previous studies, a stereotype of (im)polite
speech is that polite speech tends to have a higher pitch than
impolite speech [14-16] (though there is not always an
acoustic evidence; instead, speech rate may play a more
important role [17]). The results on F0 here seem to contradict
with this stereotype. However, we interpret the results in the
way that in expressing mock politeness speakers tend to raise
F0 to signal an exaggerated ‘politeness,’ whereas in expressing
mock impoliteness they tend to lower F0 to signal an
exaggerated ‘impoliteness’ – in both case an exaggeration may
be an indicator of mock expression.
It was also found that MP had a higher intensity than GP,
while MI had a lower intensity than GI. In a sense this
coincides with the reports in previous studies [11-13, 18-21].
For speech rate, MI was significantly slower than GI, while
only for interrogatives MP was significantly faster than GP
(for imperatives there was no significant difference). This
basically coincides with the finding in [17].
A higher jitter and a lower HNR were found in MP than in
GP. This coincides with the finding on Catalan that the speech
with a higher politeness had lower jitter and shimmer [21].
Also, lower jitter, shimmer and H1-H2, as well as a higher
HNR were found in MI than in GI, suggesting a more stable
and less breathy voice in MI. This coincides with the finding
of a breathier voice in the informal/impolite speech of Korean
[13].
After acoustic analysis, the perceptual experiment further
tested the rates of identification of attitude type and the scores
of politeness. The rate of identification was lower in MP and
MI than in GP and GI, suggesting that perceptual judgement
was influenced by literal meanings. In addition, in GP the rate
of identification was higher for interrogatives than for
imperatives, which coincides with the Tact Maxim principle
that interrogatives tend to be more polite than imperatives. In
contrast, the results were reverse in MP and GI, which
however still coincides with the Tact Maxim principle, for MP
and GI are intrinsically associated with impoliteness.
The ordering of the scores of politeness, GP > MP > MI >
GI, coincides with the expectation that mock (im)politeness is
less (im)polite than the genuine ones, and the attitudes named
with politeness, whether genuine or mock, sound more polite
than the attitudes named with impoliteness. In addition,
interrogatives sounded consistently more polite than
imperatives, again verifying the Tact Maxim principle.
6. Conclusions
This study conducted both acoustic analysis and perceptual
experiment on genuine and mock (im)polite Mandarin speech.
In acoustic analysis, a number of prosodic features, including
F0, intensity, speech rate and voice quality, were found to play
significant roles in distinguishing genuine and mock (im)polite
Mandarin speech. In perceptual experiment, the lower
identification rates on mock (im)politeness indicated that
perceptual judgement was influenced by literal meanings,
while politeness ratings showed that mock (im)polite speech
was perceptually less (im)polite than the genuine ones, based
on which we further found a good correspondence between
prosodic manifestations and perceived politeness. In addition,
interrogatives were found to be perceptually more polite than
imperatives, which verified the Tact Maxim principle for
politeness.
Future work needs to take into account facial expression,
gesture and other physical signals to find a combinatorial
multimodal strategy in expressing genuine and mock
(im)politeness.
7. Acknowledgments
This work is supported jointly by the National Social Science
Fund of China (13&ZD189), and the project for Jiangsu
Higher Institutions’ Excellent Innovative Team for Philosophy
and Social Sciences (2017STD006).
8. References
[1] J. Culpeper and M. Terkourafi, “Pragmatic approaches
(im)politeness,” In: Culpeper, J., Haugh, M. & Kádár, D. Z.
(eds.), The Palgrave Handbook of Linguistic (Im)politeness.
London: Palgrave Macmillan, pp. 11–39, 2017.
4156
[2] G. N. Leech, Principles of Pragmatics. London: Longman, 1973.
[3] J. Culpeper, “Towards an anatomy of impoliteness,” Journal of
Pragmatics 25(3), pp. 349–368, 1996.
[4] J. Culpeper, Impoliteness: Using Language to Cause Offence,
Cambridge: Cambridge University Press, 2011.
[5] C. Taylor, Mock Politeness in English and Italian, Amsterdam:
John Benjamins Publishing Company, 2016.
[6] C. Taylor, “Beyond Sarcasm: The metalanguage and structures
of mock politeness,” Journal of Pragmatics 87, pp. 127–141,
2015.
[7] A. Paternoster, “Inappropriate inspectors: Impoliteness and over-
politeness in Ian Rankin’s and Andrea Camilleri’s crime series,”
Language and Literature 21(3), pp. 311–324, 2012.
[8] A. Wichmann, “The attitudinal effects of prosody, and how they
relate to emotion,” Proc. ISCA Workshop on Speech and
Emotion, pp.143–148, Newcastle, Northern Ireland, 2000.
[9] E. Devís and F. Cantero, “The intonation of mitigating
politeness in Catalan,” Journal of Politeness Research 10(1), pp.
127–149, 2014.
[10] L. Astruc, M. M. Vanrell and P. Prieto, “Cost of the action and
social distance affect the selection of question intonation in
Catalan,” In: Armstrong, M., Henriksen, N., Vanrell, M. (eds.),
Interdisciplinary Approaches to Intonational Grammar in Ibero-
Romance Intonation. Amsterdam: John Benjamins, pp. 91–114,
2016.
[11] S. McKinnon and P. Prieto, “The role of prosody and gesture in
the perception of mock impoliteness,” Journal of Politeness
Research 10(2), pp.185–219, 2014.
[12] B. Andreeva, B. Silvia and B. William, “Prosodic cues of
genuine and mock impoliteness in German and Polish,” Proc.
Speech Prosody, pp. 999–1003, Boston, MA, 2016.
[13] B. Winter and S. Grawunder, “The phonetic profile of Korean
formal and informal speech registers,” Journal of Phonetics 40
(6), pp. 808–815, 2012.
[14] E. Ofuka, J. D. Mckeown, M. G. Waterman and P. J. Roach,
“Prosodic cues for rated politeness in Japanese speech,” Speech
Communication 32(3), pp. 199–217, 2000.
[15] Y. Ohara, “Finding one’s voice in Japanese: A study of the pitch
levels of L2 users,” In: Pavlenko, A. Brackledge, A. Piller, I. and
Teutsch-Dwye, M. (eds.), Multilingualism, Second Language
Learning, and Gender. New York: Mounton de Gruyter, pp.
231–254, 2001.
[16] M. Ito, “Politeness and voice quality – The alternative method to
measure aspiration noise,” Proc. Speech Prosody, pp. 213-216,
Nara, Japan, 2004.
[17] W. Gu, T. Zhang, and H. Fujisaki, “Prosodic analysis and
perception of Mandarin utterances conveying attitudes,” Proc.
INTERSPEECH, pp. 1069–1072, Florence, Italy, 2011.
[18] J. J. Ohala, “An ethological perspective on common cross-
language utilization of F0 of voice,” Phonetica 41(1), pp. 1–16,
1984.
[19] L. Brown, B. Winter, K. Idemaru and S. Grawunder, “Phonetics
and politeness: perceiving Korean honorific and non-honorific
speech through phonetic cues,” Journal of Pragmatics 66(1):
45–60, 2014.
[20] E. Sherr-Ziarko, “Prosodic properties of formality in
conversational Japanese,” Journal of the International Phonetic
Association 23(4), pp. 1–22, 2018.
[21] I. Hübscher, J. Borràs-Comes & P. Prieto, “Prosodic mitigation
characterizes Catalan formal speech: the frequency code
reassessed,” Journal of Phonetics 65, pp.145–159, 2017.
4157