Content uploaded by Mariapaola D'Imperio
Author content
All content in this area was uploaded by Mariapaola D'Imperio on Mar 21, 2018
Content may be subject to copyright.
The expression of politeness and pitch height in Russian imperatives
Aleksandra Chikulaeva1 & Mariapaola D’Imperio1
1 Aix Marseille Université, CNRS, LPL UMR 7309, 13100, Aix-en-Provence, France
aleksandra.chikulaeva@etu.univ-amu.fr, mariapaola.dimperio@lpl-aix.fr
Abstract
Based on the theory biological codes [1], the Frequency code
[2] claims that pitch height is a universal correlate of
politeness. Other frameworks, while taking a pragmatic
approach, [3], [4] claim that high pitch can be employed in
both polite and impolite contours and argue for the
importance of socio-pragmatic variables in the expression of
politeness. Work on Russian prosody suggests though that
the degree of politeness decreases with higher f0 of falling
contours in imperatives [5], [6], [7]. The present study
investigates the relationship between f0 height, pitch accent
type and conveyed attitude in Russian imperatives when
social distance (power relationship) is manipulated. A
discourse completion task, in which both speakers’ power
and attitude were manipulated, was carried out to test our
hypotheses. Our results show that higher f0 values are found
for both rising and falling polite imperatives, except for
downstepped pitch accents. Moreover, speakers’ social
power did not show a significant effect. Our findings
underline the need to take into account pitch accent type and
speech act to predict fundamental frequency values in polite
contexts.
Index Terms: Prosody, intonation, Russian, Frequency
code, paralinguistic meaning, politeness, social distance.
1. Introduction
The metaphor of the Frequency code [1] is based on an
analogy between larynx size and body size in the animal
kingdom, whereby higher pitched sounds are emitted by
smaller creatures. Gussenhoven [2] later proposed a
paralinguistic, “affective” interpretation of the Frequency
code, in which high pitch serves to express social meanings,
such as “politeness”, “friendliness” and “submissiveness”.
Crosslinguistically, high correlations between “perceived
politeness” and mean f0 were found for rising contours in
English and Dutch [8].
The theory of pragmatic politeness [9] underlines the
fact that social variables, such as social distance, power
imbalance and degree of imposition, may affect the choice
of politeness strategy and ways of softening face threat.
Recent production and perception experiments in prosody
[10], [11], [12] have also revealed the influence of social
distance on fundamental frequency contour and choice of
pitch accent in vocatives and questions in Catalan. Hence,
social distance can interact with politeness expression in
complex ways and affect pitch values and/or pitch accent
and contour composition.
Previous studies testing the Frequency code have mainly
concentrated on rising contours (yes/no questions, polite
requests, vocatives), while other types of speech acts,
especially face threatening acts, have received little
attention. For example, imperative acts, allowing both
falling and rising contours, have a stronger illocutionary
aspect and may be employed with different degrees of
politeness. Still, prosodic ways of mitigating or aggravating
face threat in imperative sentences have never been
systematically studied before.
In Russian, it has been suggested that speaker’s attitude
and/or relative social power [5] determines the choice of
pitch contour direction (rise vs. fall), suggesting that,
following [3], [13], imperatives with a rising contour (or
prosodically open) are employed when a subordinate
addresses a superior, while falling contours would express
dominance. The use of rising contours in imperatives has
also been connected with friendliness [14], [5].
However, other body of research points to controversial
evidence as to the relationship between pitch height and
politeness in Russian imperatives. For example, [6] reports
that imperatives only show falling patterns, hence
confirming the Frequency code by claiming that higher pitch
peaks would soften imperativeness, while ordering
imperatives would tend to show lower falls. However, the
corpus of Russian intonation [7] lists examples in which
higher falls in imperatives are marked as being less polite.
Given that the role of the Frequency code in Russian
imperatives is controversial, i.e. whether pitch height is
affected by politeness in falling contours, we hypothesized
that Russian imperatives allow both falls and rises, and that
the choice of contour is made by the speaker also according
to relative social power.
Specifically, we hypothesized that speakers would
employ rising contours when addressing an interlocutor with
higher power, while usi ng falls when addressing a speaker
with lower power, as claimed by [5]. According to evidence
provided by Odé [15], [7] and results of a pilot study [16],
we also tested a modified version of the Frequency code.
Specifically, we hypothesized that in falling imperative
contours higher pitch would be used to convey a less polite
attitude and/or higher social distance. In other words, we
expected nuclear pitch accent type (rising or falling) to
differently interact with the Frequency code, so that higher
pitch would correlate with polite attitude only in rising pitch
accents and not in falling ones.
2. Method
2.1. Hypotheses
In this study, the following hypotheses were tested: 1) In
high power contexts, speakers would produce a falling H+L*
pitch accent in directive acts containing imperatives, while
in low power context they would employ a rising L+H*
configuration; 2) High f0 peaks would correlate positively
with degree of politeness in rising pitch accents, and
negatively in falling ones.
2.2. Participants
11 native speakers of Russian (all females, mean age=21.5,
sd=3.9) took part in a production experiment. All the
participants had spent up to 6 months in France, were all L1
Russian speakers and had spent most of their lives in Russia.
All of them were informed about the experimental procedure
and data anonymity in advance and signed a participation
consent.
2.3. Corpus and procedure
Participants were invited to perform a discourse completion
task - DCT [17] containing 18 situations in which they had
to roleplay characters with either high or low social power
while employing either a polite or impolite attitude for each
of these roles. Target phrases contained 9 items built from a
quadrisyllabic target word (a verb in imperative form, 2nd
person, plural) with stress on the third syllable, plus an object
pronoun. Each target phrase appeared in all the four
conditions, i.e. high and low power, polite and impolite
attitude, resulting in 36 productions per session.
Table 1: Examples of contexts used in the study and
target phrase.
High-power context
“You are a museum director. During the
reconstruction, you find out that the workers left an
antique statue lying on the floor. You are calling one of
them and saying…
Polozhite ih!
“Put it down!”
Low power context
“You work in a shop selling crystal-made
decorations. Once you see that one of the visitors
ignored the warning sign and is holding a fragile
crystal figure in his hands, you are telling him to …
Polozhite ih!
“Put it down!”
Contexts and target sentences were presented in the form
of a powerpoint presentation and in randomized order.
Participants were not limited in time when reading the
contexts. The experiment was piloted by Perceval software
[18]. Also, we relied on speakers’ awareness of the
conscious nature of politeness and impoliteness [19]. Hence,
our participants were asked to utter each target phrase either
in the most polite or in the most impolite way. This
instruction was given before each target phrase.
Each participant repeated the session three times with a
pause of 2-5 minutes. Our total data consist of 108
productions by participant, for a total of 1188 stimuli, though
only the 2nd and 3rd sessions were submitted to the analysis
1
We excluded final rising contours because the status of high
boundary tones in Russian remains undetermined; [15] claims that
given that recordings from the first sessions had to be
excluded because of technical problems during recording.
As a result, we had 8 repetitions of the target item by
speaker. The total duration of the experiment did not exceed
30 minutes.
2.4. Annotation and measurements
Only utterances with low L-L% boundary tones were
submitted to the analysis (663 sound files)
1
. The final set of
utterances was analyzed in Praat [20]. Each soundfile was
annotated with the use of textgrids. Annotations included a
manually syllabified tier, stress location, a phonetic
transcription [22] and a point tier with pitch accents and
boundary tones (see Figure 1). The tonal transcription
combined general ToBI conventions [23] with Russian-
specific ToRI guidelines [7], [15]. As for the acoustic
measures, in order to estimate peak height, we automatically
extracted f0 maxima corresponding to the H targets in the
nuclear pitch accents.
3. Results
3.1. Descriptive analysis
According to our predictions we expected that half of the
elicited contours (the polite renditions) would be produced
with a rising nuclear pitch accent, while the other half would
show a fall. Unexpectedly, our data yielded a much larger
variability. First, only one pitch accent from those listed in
Table 2 appears to match a pattern previously described in
ToRI for imperative utterances. This is the H+L* accent
(Figure 1), which despite claims of being typical for
“imperatives and commands” [7], appeared only in one fifth
of our data and mainly in the polite condition with low
power. Two of the least frequently occurring accent types
were excluded from the overall analysis (H+H* and an L*,
representing less than 3% of the overall data).
Table 2: List of nuclear pitch accents with frequency of
occurrence frequency according to power context and
politeness.
Pitch
accent
Total
number
Politeness condition
H*
215
(34%)
Polite: 58
Impolite: 157
High power: 34
Low power: 24
High power: 83
Low power: 74
L+H*
234
(37%)
Polite: 150
Impolite: 84
High power: 78
Low power: 72
High power: 42
Low power: 42
H+!H*
56 (9%)
Polite: 17
Impolite: 39
High power: 7
Low power: 10
High power: 16
Low power: 23
H+L*
158
(19%)
Polite: 113
Impolite: 45
High power: 46
Low power: 67
High power: 22
Low power: 23
High Boundary Tones (HBTs) do not occur in Russian, while [24]
mentions HBTs as a part of Russian intonational grammar. Due to
this indeterminacy, we excluded HBT data from the analysis.
Figure 1: Pitch contour, spectrogram and annotation of two
renditions of Zamenite ih “Change!” with polite attitude
(top), and impolite attitude (bottom), both with high power.
3.2. Relationship between power context and pitch accent
The data was analyzed in R studio [25] with the packages
lme4 [26] and lmerTest [27]. An informal inspection of
Figure 2, showing pitch accent type by power condition,
indicates high similarity between the accent distribution for
both power contexts.
As for the statistical analysis, we fitted a logistic
regression (glm) to test if pitch accent type could predict
power context. The model had Pitch Accent (PA) as a four-
leveled explanatory variable (each level corresponded to one
of the accents listed in Table 2). Results show that only the
presence of the falling H+L* nuclear accent was marginally
significant for low power contexts (α = .05, β = 0.4, z = 1.8,
p = .07). A post-hoc Wald Test further confirmed that that
the PA variable did not significantly predict the results (F =
1.31, p= 0.3).
3.3. Politeness effect
We then tested whether PA type could predict
politeness/impoliteness (Table 2). We hence fitted a logistic
regression with Politeness (2 levels) as dependent variable
and PA as explanatory variable. Our data showed that
productions with a rising H* accent were mostly likely to
occur in the impolite condition (β=-0.9, z=-6.2, p<.001),
while H+L* and a L+H* were more likely to occur in the
polite condition (H+L*: β=1.9, z=7.3, p<.001; L+H* : β=1.5,
z=7.3, p<.001), while the occurrence of the H+!H*
downstepped pitch accent did not reach significance in any
of the conditions (β=0.09, z=0.27, p=0.7). An additional
Wald Test confirmed that PA type was a significant
predictor of the politeness condition (F = 17.17, p<0.001).
Figure 2: Distribution of the annotated pitch accents
according to power contexts, (high power context is
marked in red, low power in blue).
3.4. F0 height analysis
We then went on to test if f0 peak height for the H target
within each nuclear PA would be affected by politeness
across power conditions.
As it can be noted in Figure 3, overall it appears that
polite productions were associated to higher f0 values for the
target H tone within the nuclear PA. We hence fitted a
generalized linear mixed effects model (GLMM) with
Speaker and Target Word as random effects, and Politeness,
Power and PA type as fixed effects. The retained model
included all fixed effects, sum coded, and all interactions.
The Politeness and Power factors included two levels each,
while the PA factor included 4 levels (one for each of the
accents in Table 2). The model included 412 data points.
The model showed that impolite productions were
produced with lower f0 peaks than polite ones (β= -17.5, t=-
7.72, p<.001). H peaks in rising accents appeared to be
higher than the grand mean, though the effect was not
significant (H*: β=5.57, t=1.5, p=.12; L+H* : β=8.7, t=1.3,
p=.21). A further test of the interaction between f0 height
and politeness using sum coding revealed, though, that the
effect of impoliteness was negative for H* (β=-10.3, t=-2.9,
p=.001) while positive for H+!H* (β=12.17, t=2.4, p=.01),
which is in line with our predictions (politeness correlates
with high pitch peaks in rising accents and not in falling
ones). Testing interaction terms with other accents revealed
a trend for higher f0 peaks in polite L+H* rises and polite
H+L* falls, though the effect was not significant (L+H*:
β=8.3, t=1.8, p=.06; H+L* (β=.19, t=.5, p=.9).
Different from the Politeness factor, the main effect of
Power did not reach significance (β = -2.9, t = -1.3, p = 0.2).
Also, no interaction was found between Power and
Politeness (β = 0.9, t = 0.4, p= 0.7). Given that Power did not
interact with any of the dependent or independent variables,
we omitted it from our final retained model.
Figure 3: F0 height for each pitch accent type by power
context (high power=left panel; low power=right panel)
and politeness (green boxes=polite; red boxes=impolite).
4. Discussion
In the present study, we explored the influence of the social
variable of power on intonation realizations of Russian
imperatives with either polite or impolite intended attitude,
and tested the Frequency code [1], [2] for both rising and
falling pitch accent types.
The influence of power imbalance on intonation
contours was found to be significant in a number of previous
studies for other languages [3], [4], [10]. Our results showed,
though, that power context did not predict the distribution of
falling and rising pitch accents in our data. What is more, the
only pitch accent showing a quasi-significant trend showed
a pattern going in the opposite direction relative to our
prediction: the falling H+L*, instead of being associated
with high power contexts, was more likely to appear in the
low power one. Similarly, a study on Catalan [11] did not
find an effect of power in vocatives, which was accounted
for by the relative unimportance of this social parameter in
Catalan culture.
In our study, we can only speculate that the lack of this
effect might be due to limitations in the speaker sample
composition (only young female speakers), which was
conditioned by limited access to Russian speakers in Aix-en-
Provence. It is possible that all the subjects, when belonging
to the same age and sex group, followed a culturally specific
stereotype suggesting that young people, and especially
young women, should always come across as polite (see
similar sociolinguistic behavior for Korean [28]). If this
stereotype has a real effect on speaker’s productions, it
would mean that our participants could not fully adopt a
required dominant behavior and produce differential power
patterns which we tried to induce through the experimental
task.
The Frequency code was then tested by comparing f0
peak height of H target tones in polite and impolite
imperative contours (for each pitch accent type),
hypothesizing that the relationship between fundamental
frequency and politeness should take into account tonal
direction within the pitch accent. This is because the
Frequency code, while originally associating higher f0 with
politeness and friendliness, has different outcomes in rising
and falling accents in some languages, such as Korean [28],
in which lower pitch appears to be associated with politeness
instead of impoliteness. However, while for languages such
as Korean power imbalance appears to affect intonation, this
is still controversial for Russian. In fact, our data did not find
an effect of power imbalance. On the other hand, the
Frequency code was somewhat confirmed by the f0 peak
data, showing higher values for polite utterances, except for
the downstepped H+!H* accent. This pattern might be due
to a specific application of the Effort code by which a higher
f0 peak in the downstopped accent would be a proxy for a
greater pitch excursion, hence being employed to convey
insistence and a lower degree of politeness. However, it is
also possible that different ‘codes’ might be applied in the
same communicative acts. Our current data do not allow us
to choose between these two possible accounts.
In addition, the experiment revealed an interaction
between PA type and politeness, with clearest effects for the
H* and H+!H* accents (most frequent for impolite attitude)
and L+H* (most frequent for polite attitude). The
explanation of this distribution possibly lies in the scope of
peak tonal alignment, which seems to be employed by
Russian listeners when discriminating questions from
assertions [29]. Future tests manipulating peak timing in
Russian imperatives might reveal if there is a “window”
related to meanings of insistence and impatience, which may
convey impoliteness within the same speech act. Finally, our
findings appear to be in line with [3] for English and [11] for
Catalan, showing that downstep (in either the nuclear accent
or the boundary tone) was correlated with extreme finality
and dominance, whereas a rising L+H* accent appeared to
convey politeness addressed to a superior, or to an equal
interlocutor. Future wok will have to better address
language-specific uses of the Frequency code and its
interaction with social distance and intonation contour type.
5. Conclusion
A production study on pitch accent type and f0 height in
Russian imperatives showed that neither social power nor
politeness have an effect on choice of tonal direction within
nuclear pitch accents. Despite previous findings on Russian
intonation, our data show that different nuclear pitch accents
can be employed in imperatives, though rising L+H* accents
appear to be most frequent for polite utterances. This result
suggests that the intonation of imperative speech acts in
Russian is not influence by social distance. As for pitch
height, our data show that it can be modulated by politeness,
in line with the Frequency code. In fact, all pitch accents to
the exception of the downstepped H+!H* showed higher f0
peaks for the polite condition than for the impolite one.
Moreover, a variety of tonal patterns produced by our
speakers had not been yet described in the literature,
underlying the need for providing a more in-depth
description of Russian intonational patterns in the future.
6. References
[1] Ohala, John J., (1984). “An ethological perspective on
common cross-language utilization of f0 of voice”. Phonetica
41, 1–16.
[2] Gussenhoven, C. (2002). “Intonation and interpretation:
phonetics and phonology”. In B. Bel & I. Marlien (Eds.),
Proceedings of Speech Prosody (pp. 47 – 57). Aix-en-
Provence : Université de Provence.
[3] Culpeper, J. and Bousfield, D., and Wichmann, A. (2003).
“Impoliteness revisited: With special reference to dynamic and
prosodic aspects”. Journal of Pragmatics 35, 1545-1579.
[4] Culpeper, Jonathan (2011). “Politeness and impoliteness”. In:
Karin Aijmer and Gisle Andersen (eds.) Sociopragmatics,
Volume 5 of Handbooks of Pragmatics edited by Wolfram
Bublitz, Andreas H. Jucker and Klaus P. Schneider. Berlin:
Mouton de Gruyter, 391-436.
[5] Kodzasov, S. V. (2002). “Symbolism of Russian Intonation
[Simvolika russkoi intonazii]”, Papers of the International
seminar “Dialogue’2002” [Trudi mezhdunarodnogo seminara
Dialog’2002]. vol.1: Theoretical problems, [teoreticheskie
problemi]. Moscow.
[6] Nikolaev, A. I., Artimiev, I. T. (2011). “Intonational transfer
of communicative sentence-types in the context of artificial
multilingualism”. Journal of North-Eastern Federal
University, Yakutsk [ВЕСТНИК СВФУ], 2011, vol. 8, № 4.
[7] Odé, C. (2008). Transcription of Russian intonation, ToRI, a
free interactive research tool and learning module,
http://www.fon.hum.uva.nl/tori/index.html.
[8] Chen, A., Gussenhoven, C., Rietveld, T., (2004). “Language-
specificity in the perception of paralinguistic intonational
meaning”. Language and Speech 4, 311–350.
[9] Brown, P., Levinson, S.C. (1987). Politeness. Some universals
in language use. Cambridge University Press, New York.
[10] Astruc, L. - Vanrell, M.M. - Prieto, P. (2016). “Cost of the
action and social distance affect the selection of question
intonation in Catalan”. In M. E. Armstrong, N. Henriksen, and
M. M. Vanrell (eds.), Intonational grammar in Ibero-
Romance. Approaches across linguistic subfields.
Amsterdam: John Benjamins, pp. 93-114.
[11] Borràs-Comes, J., Sichel-Bazin, R., and Prieto, P. (2015).
“Vocative intonation preferences are sensitive to politeness
factors”. Language and Speech 58(1), pp. 68-83.
[12] Nadeu, M., and Prieto, P. (2011). “Pitch range, gestural
information, and perceived politeness”, Catalan. Journal of
Pragmatics, vol. 43, (3), February 2011, pp. 841–854.
[13] Wichmann, A. (2004). “The intonation of please-requests: A
corpus-based study”. Journal of Pragmatics 36, 1521-1549.
[14] Hirst, D., Di Cristo, A. (1998). Intonation Systems: A Survey
of Twenty Languages. Cambridge University Press.
[15] Odé, C. (2008). “Transcription of Russian intonation, ToRI,
an interactive research tool and a learning module on the
Internet”, Dutch Contributions to the Fourteenth International
Congress of Slavists, Ohrid: Linguistics (SSGL 34).
Amsterdam – New York: Rodopi, 2008, 431-449.
[16] Chikulaeva, A. (2017). The expression of politeness and pitch
height in Russian. Unpublished Master’s Thesis, Aix
Marseille University, Aix-en-Provence, France.
[17] Blum-Kulka, S., House, J., & G. Kasper. (1989).
“Investigating cross-cultural pragmatics: An introductory
overview”. In Blum-Kulka, S., House, J., & G. Kasper (Eds.),
Cross-cultural pragmatics: Requests and apologies (pp. 1-34).
Norwood, NJ: Ablex, p. 13-14.
[18] André, C., Chio, A., Cavé, C., Teston, B. (2003) PERCEVAL:
PERCeption EVALution Auditive et Visuelle (computer
program), retrieved from http: //www.lpl.univ-
aix.fr/~lpl/dev/perceval.
[19] Aubergé, V., Gestalt, A., (2002) “Morphology of Prosody
Directed by Functions”, Speech Prosody 2002 Proc., 151-154,
Aix-en-Provence, France.
[20] Boersma, P., Weenink, D. 2010. Praat: doing phonetics by
computer [Computer program]. Version 5.0, retrieved from
http://www.praat.org/.
[21] Paschen, L. (2016). “Boundary tones indicate turn allocation
in Russian”. Proceedings of ConSOLE 23, Leiden, pp. 493-
508.
[22] IPA Chart, http://www.internationalphoneticassociation.org,
available under a Creative Commons Attribution-Sharealike
3.0 Unported License. Copyright © 2015 International
Phonetic Association.
[23] Beckman, M. & G. A. Elam. 1997. Guidelines for ToBI
Labelling. Version 3. Unpublished ms., Ohio State University.
[24] Paschen, L. (2016). “Boundary tones indicate turn allocation
in Russian”. Proceedings of ConSOLE 23, Leiden, pp. 493-
508.
[25] RStudio Team (2015). RStudio: Integrated Development for
R. RStudio, Inc., Boston, MA URL http://www.rstudio.com/.
[26] Bates, D., Maechler, M., Bolker, B., Walker, S. 2014. lme4:
Linear mixed-effects models using Eigen and S4. R package
version 1.1-7.
[27] Kuznetsova, A., Brockhoff, P. B. and Christensen, R. H.
(2016). lmerTest: Tests in Linear Mixed Effects Models. R
package version 2.0-33. https://CRAN.R-
project.org/package=lmerTest.
[28] Brown, L., Winter, B., Idemaru, K., & Grawunder, S. (2014).
“Phonetics and politeness: Perceiving Korean honorific and
non-honorific speech through phonetic cues”. Journal of
Pragmatics, vol. 66, February 2014, pp. 45 – 60.
[29] Rathcke, T. (2006). “Relevance of f0 Peak Shape and
Alignment for the Perception of a Functional Contrast in
Russian”. In: Proceedings of the 3rd Conference on Speech
Prosody, 2-5 May 2006, Dresden, Germany.
