Content uploaded by Olga Dmitrieva
Author content
All content in this area was uploaded by Olga Dmitrieva on Jan 31, 2016
Content may be subject to copyright.
LANGUAGE ATTITUDES AND LISTENER-ORIENTED PROPERTIES IN
NON-NATIVE SPEECH
Olga Dmitrieva, Wai Ling, Law, Mengxi Lin, Yuanyuan Wang, Jenna Conklin, Ashley Kentner
Purdue University
{odmitrie, wlaw, lin211, wang861, jconkli, akentner}@purdue.edu
ABSTRACT
The study examines a number of acoustic properties
of non-native speech directed to a native speaker, a
non-native speaker with a shared first language
background, and a non-native speaker with a different
first language. Results demonstrate that the
interlocutor condition interacts with the language
attitudes factor: Participants with more positive
attitudes towards their second language (English)
differ along several acoustic dimensions from
participants with more positive attitudes towards their
first language (Mandarin), especially when
interacting with native speakers of English. Expanded
vowel space, higher articulation rate, and increased
pitch adopted by English-oriented participants in
interactions with native speakers of English may be
indicative of their greater positive emotional
involvement in the interaction.
.
Keywords: non-native speech, language attitudes,
listener-oriented, vowel space, rate of speech
1. INTRODUCTION
It has been known for some time that speakers can
adjust the acoustic characteristics of their speech to
accommodate the communicative needs of the
listeners. The speaking style directed at increasing
speech intelligibility, dubbed ‘clear speech’, has been
shown to be adopted by native speakers in specific
communicative settings, for example in the presence
of noise or when addressing hearing-impaired
listeners [11]. Clear speech is typically characterized
by a decrease in the rate of speech, higher pitch and
expanded pitch range, an increase in vowel duration,
and expanded vowel space. Other populations of
listeners who elicit similar adaptations in native
speech include foreigners, infants and young
children, and even pets [5], [12], [14].
Less is known about listener-oriented speaking
style adaptations that may occur in non-native (L2)
speech. Perceptual studies demonstrate that clear
speech produced by proficient L2 speakers leads to
intelligibility benefits comparable to those produced
by native speakers’ clear speech [13]. There is also
some evidence that different pairings of native and
non-native interlocutors may result in changes in
speaking style, as assessed via the differences in
resulting intelligibility and degree of phonetic
convergence [7], [8], [13], [15]. These findings
suggest that not only non-native speakers are able to
modify their speech ‘at will’, they may do so in the
absence of explicit instructions (e.g. to speak more
clearly) but in spontaneous response to the change of
listener and their perceived communicative needs.
The present study investigates the possible effects
that the change of the interlocutor characteristics in
terms of native language background may have on the
acoustic properties of non-native speech. More
specifically, we are testing the hypothesis that non-
native speakers may choose to speak more ‘clearly’
to a particular group of listeners; those that the
speakers expect to experience the greatest
intelligibility-related difficulty with their accented
speech. Listeners with whom the speakers don’t share
a common native language and those who have less
exposure to non-native speech may belong to this
category.
However, listener-oriented adaptations in non-
native speech may also be modulated by speakers’ L2
proficiency levels and attitudes towards their second
language. In particular, participants with a more
positive attitude to their second language and a
greater motivation to be perceived as a proficient
speaker by native listeners may choose to speak more
clearly when addressing the native speakers of their
L2.
We address these questions by examining the
acoustic characteristics of non-native speech
addressed to native and non-native listeners (those
with the same and different L1 backgrounds) in light
of speakers’ language attitudes.
2. METHODS
2.1. Participants
Thirteen participants (5 women, 8 men) took part in
the study to date. All were native speakers of
Mandarin from the same dialectal area: Northern
regions of mainland China (north of Yangtze River).
Participants were recruited on the campus of a major
Midwestern university and received payment for their
participation. All participants completed a post-test
questionnaire, adapted from [3] and [10], with
detailed questions concerning their proficiency in
second language (English, self-rated), amount of first
and second language use (in hours per week), quality
of linguistic interactions (with native vs. non-native
speakers of the language), and language attitudes
(how much importance they attach to being perceived
as a proficient/authentic speaker of their native vs.
non-native language). Only language attitude results
are discussed in this paper.
Three confederates (all women) served as the
conversation partners in the experimental sessions.
The first confederate was a native speaker of
Mandarin (a non-native speaker with the same L1
background as the participants); the second
confederate was a native speaker of the Midwestern
dialect of American English (a native speaker of the
participants’ L2); the third confederate was a native
speaker of Russian (a non-native speaker with a
different L1 background). Both non-native
confederates learned English as a second language in
adolescence/adulthood and spoke noticeably
accented English. The participants were also notified
of the confederates’ native language backgrounds
during the introduction part prior to the experimental
sessions. All confederates are authors on this paper.
2.2. Materials
Three different versions of the map, similar to those
found in the HCRC Map Task Corpus [1], were
created for the experiment. Each map contained the
same 13 labeled landmarks, which were arranged in a
different order and connected with a different route
on each map. Both the participant and the confederate
were given a copy of the same map per interaction,
however the confederate’s map did not have the route.
2.3. Procedure
After participants had given informed consent, they
were instructed to complete the map task three times.
For each task, participants were instructed to explain
the route on their map to the task partner (one of the
three confederates) such that the partner could
replicate the route on their map. Participants were
informed that their task partner does not have the
route drawn on their map. Confederates were
presented as fellow participants in order to allow for
the most natural interaction possible and avoid any
formality that may have been induced by speaking
knowingly with an experimenter. During the map
task, participants were seated across the table from
the confederate, in the sound-attenuated booth. A
custom divider did not allow the task partners to see
each other’s maps but did not interfere with visual
contact. The order of interactions with three
confederates was counterbalanced across
participants. Each interaction lasted for about 10
minutes, the entire experiment lasting between 30 and
40 minutes. Both the participants’ and confederates’
voices were recorded digitally to separate channels.
2.4. Measurements
The participants’ recordings were manually
annotated for the syllable and stressed vowel
boundaries in the target words (map landmarks). The
values of the first two formant frequencies (F1 and
F2) at midpoint of the vowel were collected. Formant
values were examined for outliers and corrected
manually where necessary.
Average pitch per syllable was also obtained using
an autocorrelation pitch tracking algorithm. Outlying
values due to pitch tracking errors were removed
from the analysis. All the annotations and
measurements were done in Praat [4].
Vowel space between the point vowels [i] [æ] [u]
and [ɑ] was calculated by adding the areas of the two
triangles, that between vowels [i] [æ] and [u] and that
between [u] [æ] and [ɑ]. The areas of the triangles
were found using the formula in (1), where x
corresponds to the F1 value, y corresponds to the F2
value, and A, B, and C stand for the three point
vowels.
(1)
|
2)(()()(
|ByAyCxAyCyBxCyByAx
Aabc
Articulation rate was calculated by dividing the
number of syllables in each participant’s response
(estimated as a number of vocalic segments) by the
participant’s phonation time (total response time
minus silence time).
Attitude ratio was calculated based on
participants’ responses to the Language Attitudes part
of the questionnaire. Participants rated on the 6-point
scale (0-strongly disagree, 6-strongly agree)
statements such as “I identify with an
English/Mandarin-speaking culture” and “I want
others to think I am a native/proficient speaker of
English/Mandarin”. The totals of points in English-
related statements and Mandarin-related statements
were obtained and a Mandarin/English attitudes ratio
(AR) was calculated. The ratio of 1 indicated that the
participants valued the authenticity of their Mandarin
and English-speaking identities equally. A ratio lower
than 1 indicated a greater value of the English-
speaking identity, while a ratio greater than 1
indicated a greater value of the Mandarin-speaking
identity.
2.5. Analysis
Participants were divided into two groups based on
the attitudes ratio: Group 1 contained seven
participants whose AR was higher than 1 (Mandarin-
oriented); Group 2 contained six participants whose
AR was 1 or lower (English-oriented). Acoustic
parameters were checked for interactions of the AR
grouping variable (English-oriented vs. Mandarin-
oriented) with the interlocutor’s native language
variable (English, Mandarin, and Russian) in a series
of repeated measures ANOVAs. Analyses with
significant interactions were followed up by repeated
measures ANOVAs within each AR group.
3. RESULTS
3.1. Vowel space
There was a significant interaction between the AR
variable and the Interlocutor’s Language (IL) in the
analysis of vowel space: F(2,22)=5.907, p<0.01,
which indicated that English-oriented and Mandarin-
oriented groups demonstrated different vowel space
patterns across the interlocutor conditions. Figure 1
shows that the two groups diverged in terms of vowel
space in the native English-speaking condition.
Figure 1: Vowel space in English, Mandarin, and
Russian Interlocutor Conditions for English-
oriented and Mandarin-oriented groups.
Follow-up within-group analyses showed a
significant effect of IL within the English-oriented
group: F(2,10)=5.507, p<0.05. In this group of
participants, vowel space was more expanded when
they were addressing a native English-speaking
interlocutor compared to interactions with non-native
listeners. Post hoc pairwise comparisons (Bonferroni)
showed a strong trend for a significant difference in
terms of vowel space between English and Russian
conditions (p=0.074).
3.2. Articulation rate
There was a significant interaction between the AR
factor and the IL factor in the analysis of articulation
rate: F(2,22)=5.631, p<0.05. Figure 2 shows that
participants in English-oriented and Mandarin-
oriented groups used different articulation rates when
addressing native English-speaking participants.
Figure 2: Articulation rate in English, Mandarin,
and Russian Interlocutor Conditions for English-
oriented and Mandarin-oriented groups.
Within-group analyses demonstrated a significant
effect of Interlocutor Language within the Mandarin-
oriented group: F(2,12)=4.001, p<0.05. These
participants spoke slower when addressing the native
English-speaking interlocutor compared to
interactions with non-native interlocutors. While the
effect of Interlocutor Language did not reach
significance within the English-oriented group, the
quantitative tendency was opposite to that of the
Mandarin-oriented group, similarly to the pattern of
vowel space results.
3.3. Pitch
The analysis of mean f0 showed a significant
interaction between the AR factor and the IL factor:
F(2,22)=5.512, p<0.05. Figure 3 shows that English-
oriented and Mandarin-oriented groups of
participants adopted different mean levels of pitch
across different interlocutor conditions. In particular,
English-oriented participants spoke with a higher f0
when addressing English and Russian-speaking
interlocutors, while Mandarin-oriented participants
spoke with a higher f0 when addressing Mandarin
participants.
Figure 3: Mean f0 in English, Mandarin, and
Russian Interlocutor Conditions for English-
oriented and Mandarin-oriented groups.
Within-group comparisons demonstrated a near-
significant effect of Interlocutor Language within the
English-oriented group: F(2,10)=4.103, p=0.05.
4. DISCUSSION
The results demonstrated that listener-oriented
adaptation in the non-native speech of the participants
were strongly influenced by their language attitudes.
English-oriented and Mandarin-oriented groups of
participants made difference acoustic adjustments in
their speech across the three interlocutor conditions.
Especially prominent is the quantitative tendency to
treat the native English-speaking group differently
from the two non-native groups. English-oriented
speakers used a more hyperarticulated (expanded)
vowel space when addressing native English
listeners, while for Mandarin-oriented speakers the
tendency was in the opposite direction. This finding
is consistent with the prediction that non-native
speakers who value their L2 identity and strive to be
perceived as authentic/proficient L2 speakers will
choose to speak more clearly to native English
speakers.
However, the articulation rate results are pointing
in a different direction. In this analysis, English-
oriented group of speakers spoke faster when
addressing native English listeners, while Mandarin-
oriented speakers spoke slower to native English
speakers. Clear speech is typically characterized with
a slower speech rate and it is difficult to reconcile this
result with the clear speech pattern. However, it is
plausible that rate of speech varies in English- and
Mandarin-oriented groups across interlocutor
condition as function of speakers’ emotional
involvement during the interaction. Research shows
that rate of speech increases as speakers take a
stronger stance in the conversation [6]. English-
oriented speakers may be showing a greater degree of
emotional involvement in the interactions with native
speakers by increasing articulation rate.
Mean f0 results are also largely consistent with
this interpretation: English-oriented speakers’ speech
was on average higher-pitched when addressing
English and Russian listeners than the speech of
Mandarin-oriented group, while Mandarin-oriented
speakers adopted higher pitch when addressing
Mandarin listeners. Higher pitch has also been shown
to correlate with greater emotional engagement,
stance taking, and positive affect in speech [2]. [9].
Thus, the present results demonstrate that groups
of non-native speaker adopt different listener-
oriented strategies depending on the value they attach
to their first and second language. The findings are
most consistent with the interpretation that speakers
who value their L2 identity are more positively
emotionally involved in the interactions with native
speakers, which is manifested in hyperarticulated
vowel space, faster articulation rate, and higher pitch.
Speakers who value their L1 identities demonstrate
nearly opposite acoustic patterns in interactions with
native and non-native listeners.
These results expand our understanding of
listener-oriented properties of non-native speech
beyond the clear speech settings, where participants
are explicitly instructed to modify their speech. They
show an interaction of the listener characteristics
(such as that of a potential ‘judge’ of speakers’
authenticity) and speaker characteristics (such as
attitudes to one’s first and second language) in a more
spontaneous conversational environment.
5. REFERENCES
[1] Anderson, A., Bader, M., Bard, E., Boyle, E., Doherty,
G.M., Garrod, S., Isard, S., Kowtko, J., McAllister, J.,
Miller, J., Sotillo, C., Thompson, H.S. Weinert, R.
1991. The HCRC Map Task Corpus. Language and
Speech 34, 351-366.
[2] Belyk, M., Brown, S. (2014). The acoustic correlates of
valence depend on emotion family. Journal of
Voice, 28(4), 523-e9.
[3] Birdsong, D., Gertken, L.M., Amengual, M. 2012.
Bilingual Language Profile: An Easy-to-Use
Instrument to Assess Bilingualism. COERLL,
University of Texas at Austin. Web. 20 Jan.
[4] Boersma, P., Weenink, D. 2001. Praat, a system for
doing phonetics by computer. Glot International
5:9/10, 341-345.
[5] Burnham, D., Kitamura, C., Vollmer-Conna, U. 2002.
What's new, pussycat? On talking to babies and
animals. Science, 296(5572), 1435-1435.
[6] Freeman, V., Wright, R., Levow, G.-A., Luan, Y.,
Chan, J., Tran, T., Zayats, V., Antoniak, M., Ostendorf,
M. 2014. Phonetic correlates of stance-taking. The
Journal of the Acoustical Society of America, 136(4),
2175-2175.
[7] Hayes-Harb, R., Smith, B. L., Bent, T., Bradlow, A. R.
2008. The interlanguage speech intelligibility benefit
for native speakers of Mandarin: Production and
perception of English word-final voicing
contrasts. Journal of phonetics, 36(4), 664-679.
[8] Kim, M., Horton, W. S., Bradlow, A. R. 2011. Phonetic
convergence in spontaneous conversations as a
function of interlocutor language distance. Laboratory
phonology, 2(1), 125-156.
[9] Laukka, P., Neiberg, D., Forsell, M., Karlsson, I.,
Elenius, K. 2011. Expression of affect in spontaneous
speech: Acoustic correlates and automatic detection of
irritation and resignation. Computer Speech &
Language, 25(1), 84-104.
[10] Marian, V., Blumenfeld, H. K., Kaushanskaya, M.
2007. The language experience and proficiency
questionnaire (LEAP-Q): Assessing language profiles
in bilinguals and multilinguals. Journal of Speech
Language and Hearing Research, 50(4), 940-967.
[11] Picheny, M.A., Durlach, N.I., Braida, L.D. 1986.
Speaking Clearly for the Hard of Hearing II Acoustic
Characteristics of Clear and Conversational Speech.
JSHR 29, 434-436.
[12] Scarborough, R., Brenier, J., Zhao, Y., Hall-Lew, L.,
and Dmitrieva, O. 2007. An acoustic study of real and
imagined foreigner-directed speech. Proceedings of the
15th International Congress of Phonetic Sciences,
2165–2168.
[13] Smiljanić, R., Bradlow, A. R. 2011. Bidirectional
clear speech perception benefit for native and high
proficiency non-native talkers and listeners:
Intelligibility and accentedness). The Journal of the
Acoustical Society of America, 130(6), 4020-4031.
[14] Uther, M., Knoll, M.A., Burnham, D. 2007. Do you
speak E-NG-L-I-SH? A comparison of foreigner- and
infant-directed speech. Speech Communication 49, 2-7.
[15] Van Engen, K. J., Baese-Berk, M., Baker, R. E., Choi,
A., Kim, M., Bradlow, A. R. 2010. The Wildcat Corpus
of native-and foreign-accented English:
Communicative efficiency across conversational dyads
with varying language alignment profiles. Language
and speech, 53(4), 510-540.
