1. INTRODUCTION
⌅ In the last decades, while there has been a growing body of work on the acquisition of non-native Spanish segments (i.e., Chen, 2007Chen, Y. (2007a). A comparison of Spanish produced by Chinese L2 learners and native speakers-an acoustic phonetics approach [Doctoral dissertation]. Department of Philosophy, University of Illinois at Urbana-Champaign.
; Cobb & Simonet, 2015Cobb, K., & Simonet, M. (2015). Adult second language learning of Spanish vowels. Hispania, 47-60.
; Liu, 2019Liu, Z. (2019). Análisis de las obstruyentes en chino y en español como L3: estudio acústico y perceptivo para la categorización de errores [Tesis doctoral]. Departamento de Filología Española, Universitat Autònoma de Barcelona.
; Morrison, 2003Morrison, G. S. (2003). Perception and production of Spanish vowels by English speakers. In Proceedings of the 15th international congress of phonetic sciences (pp. 1533-1536). Barcelona, Spain.
), stress (i.e., Chen, 2007bChen, Y. (2007b). From tone to accent: the tonal transfer strategy for Chinese L2 learners of Spanish. In Proceedings of 16th International Congress of Phonetic Sciences (pp. 6-10). Saarbrücken, Germany.
; Cortés Moreno, 2005Cortés
Moreno, M. (2005). Análisis experimental del aprendizaje de la
acentuación y la entonación españolas por parte de hablantes nativos de
chino. Phonica, 1, 1-25.
; Kim, 2015Kim,
J.-Y. (2015). Perception and production of Spanish lexical stress by
Spanish heritage speakers and English L2 learners of Spanish. In Proceedings of the 6th Conference on Laboratory Approaches to Romance Phonology (pp. 106-128). Cascadilla, Somerville, MA.
; Kimura, Sensui, & Takasawa, 2015), prominence (i.e., Kim, 2016Kim, J. Y. (2016). The perception and production of prominence in Spanish by heritage speakers and L2 learners [Doctoral dissertation]. Department of Spanish and Portuguese, University of Illinois at Urbana-Champaign.
; Van Maastricht, Krahmer, & Swerts, 2016Van
Maastricht, L., Krahmer, E., & Swerts, M. (2016). Prominence
Patterns in a Second Language: Intonational Transfer From Dutch to
Spanish and Vice Versa. Language Learning, 66(1), 124-158.
), and intonation contours (i.e., Gabriel & Kireva, 2014Gabriel,
C., & Kireva, E. (2014). Prosodic transfer in learner and contact
varieties: Speech rhythm and intonation of Buenos Aires Spanish and L2
Castilian Spanish produced by Italian native speakers. Studies in Second Language Acquisition, 36(2), 257-281.
; Henriksen, Geeslin, & Willis, 2010Henriksen,
N. C., Geeslin, K. L., & Willis, E. W. (2010). The development of
L2 Spanish intonation during a study abroad immersion program in León,
Spain: Global contours and final boundary movements. Studies in Hispanic and Lusophone Linguistics, 3(1), 113-162.
; Silva & Barbosa, 2017Silva,
C. C., & Barbosa, P. A. (2017). The contribution of prosody to
foreign accent: A study of Spanish as a foreign language. Loquens, 4(2), e041-e041.
; Trimble, 2013Trimble,
J. C. (2013). Perceiving Intonational Cues in a Foreign Language :
Perception of Sentence Type in Two Dialects of Spanish. In C. Howe, S.
E. Blackwell, & M. L. Quesada (Eds.), 15th Hispanic Linguistics Symposioum (pp. 78-92). Athens, USA.
; Yuan et al., 2019Yuan,
C., González-Fuente, S., Baills, F., & Prieto, P. (2019). Observing
pitch gestures favors the learning of spanish intonation by mandarin
speakers. Studies in Second Language Acquisition, 41(1), 5-32.
),
little is known about the acoustic-phonetic realization of pitch and
temporal patterns in L2 Spanish, particularly in environments of
language contact between tone and non-tone languages such as Chinese and
Spanish. Therefore, the goal of the present study is to fill in the gap
by examining cross-linguistic differences of pitch and temporal
profiles between first- (L1) and second-language (L2) speakers of
Peninsular Spanish.
Pitch profiles consist of the oscillations of
fundamental frequency (F0) and are claimed to have quasi-universal and
language-specific characteristics in human communication (Chen, Gussenhoven, & Rietveld, 2004Chen,
A., Gussenhoven, C., & Rietveld, T. (2004). Language-specificity in
the perception of paralinguistic intonational meaning. Language and Speech, 47(4), 311-349.
; Gussenhoven & Chen, 2000Gussenhoven, C., & Chen, A. (2000). Universal and language-specific effects in the perception of question intonation. In 6th International Conference on Spoken Language Processing (pp. 91-94). Beijing, China.
).
The generalizability in the use of pitch to convey certain
paralinguistic meanings is often explained with biologically determined
codes. For example, the frequency code proposes that high pitch is
related to a small larynx and often serves as a marker of uncertainty,
whilst low pitch is associated with a larger organ of production and is
used to signal assertiveness (Gussenhoven, 2002Gussenhoven, C. (2002). Intonation and Interpretation : Phonetics and Phonology. In International Conference on Speech Prosody 2002 (pp. 47-57). Aix-en-Provence, France.
; Ohala, 1983Ohala, J. J. (1983). Cross-language use of pitch: an ethological view. Phonetica, 40(1), 1-18.
).
However, despite this commonality, it is broadly recognized that
language communities differ from each other in the specific phonetic
implementation of pitch patterns, such as register and range. For
instance, by combining the linguistic and the long-term distributional
(LTD) measures, Mennen et al. (2012)Mennen,
Ineke, Schaeffler, F., & Docherty, G. (2012). Cross-language
differences in fundamental frequency range: A comparison of English and
German. Journal of the Acoustical Society of America, 131(3), 2249-2260.
found that English female speakers had a significantly higher F0
register and a larger F0 span than their German counterparts. Similar
cross-linguistic differences in pitch profiles have also been observed
for Polish vs. English (Majewski et al., 1972Majewski, W., Hollien, H., & Zalewski, J. (1972). Speaking fundamental frequency of Polish adult males. Phonetica, 25(2), 119-125.
), Russian vs. German (Nebert, 2013Nebert, A. U. (2013). Der Tonhöhenumfang der deutschen und russischen Sprechstimme: Vergleichende Untersuchung zur Sprechstimmlage (Hallesche Schriften zur Sprechwissenschaft und Phonetik 46). Frankfurt Am Main: Lang.
), Mandarin vs. English (Keating & Kuo, 2012Keating, P., & Kuo, G. (2012). Comparison of speaking fundamental frequency in English and Mandarin. Journal of the Acoustical Society of America, 132(2), 1050-1060.
), Mandarin vs. Japanese (Shi et al., 2014Shi,
S., Zhang, J., & Xie, Y. (2014). Cross-language comparison of F0
range in speakers of native Chinese, native Japanese and Chinese L2 of
Japanese: Preliminary results of a corpus-based analysis. In Proceedings of the 9th International Symposium on Chinese Spoken Language Processing (pp. 241-244). Singapore.
), Slavic and Germanic languages (Andreeva et al., 2014Andreeva,
B., Demenko, G., Wolska, M., Möbius, B., Zimmerer, F., Jügler, J.,
Oleskowicz-Popiel, M. & Trouvain, J. (2014). Comparison of pitch
range and pitch variation in Slavic and Germanic languages. In Proceedings to the 7th Speech Prosody Conference (pp. 776-780). Dublin, Ireland.
), and many others (see Mennen et al., 2012 Mennen,
Ineke, Schaeffler, F., & Docherty, G. (2012). Cross-language
differences in fundamental frequency range: A comparison of English and
German. Journal of the Acoustical Society of America, 131(3), 2249-2260.
and Ordin & Ineke Mennen., 2017Ordin, M., & Ineke Mennen. (2017). Cross-Linguistic Differences in Bilinguals’ Fundamental Frequency Ranges. Journal of Speech, Language, and Hearing Research, 60(6), 1493-1506.
for a review). Apart from the influence of the L1 prosodic system and
some physiological factors such as vocal tract length, gender, and age,
the language-specific pitch properties are possibly more closely linked
to some social-cultural attributes. Unmistakable evidence for this is
that Japanese speakers, particularly women, have a higher F0 register
and F0 span than native speakers of Chinese (Shi et al., 2014Shi,
S., Zhang, J., & Xie, Y. (2014). Cross-language comparison of F0
range in speakers of native Chinese, native Japanese and Chinese L2 of
Japanese: Preliminary results of a corpus-based analysis. In Proceedings of the 9th International Symposium on Chinese Spoken Language Processing (pp. 241-244). Singapore.
), Dutch (Van Bezooijen, 1995Van Bezooijen, R. (1995). Sociocultural aspects of pitch differences between Japanese and Dutch women. Language and Speech, 38(3), 253-265.
), American English, and Spanish (Hanley et al., 1966Hanley, T. D., Snidecor, J. C., & Ringel, R. L. (1966). Some acoustic differences among languages. Phonetica, 14(2), 97-107.
).
The preference for high pitches shown by Japanese women is explained in
the context of their relative powerlessness in social status and the
gender roles they are expected to play according to cultural
conventions.
Furthermore, since the speech of a foreign language
often entails some degree of interaction, the cross-language
differences between the first and the second language can be expected to
impact the target speech patterns. Studies have shown that most L2
segmental and suprasegmental errors could be attributed to a prosodic
transfer from the L1 system into the phonetic and phonological knowledge
of the L2 (Graham & Post, 2018Graham, C., & Post, B. (2018). Second language acquisition of intonation: Peak alignment in American English. Journal of Phonetics, 66, 1-14.
; Ineke Mennen, 2015Mennen, Ineke. (2015). Beyond segments: towards an L2 intonation learning theory ( LILt ). In Prosody and language in contact (pp. 171-188). Berlin, Heidelberg: Springer.
).
However, importantly, several studies have found that some deviated use
of pitch is common in L2 speech, revealing itself as a consistent
development trajectory during the L2 speech-learning process. For
example, the results in previous literature (i.e., Busà & Urbani, 2011Busà, M. G., & Urbani, M. (2011). A Cross Linguistic Analysis of Pitch Range in English L1 and L2. In XVII International Congress of Phonetic Sciences (pp. 380-383). Hong Kong, China.
; Chen, 1972Chen, G. T. (1972). A comparative study of pitch range of native speakers of Midwestern English and Mandarin Chinese: An acoustic study [Doctoral dissertation]. University of Wisconsin.
; Mennen, Schaeffler, & Dickie, 2014Mennen,
Ineke, Schaeffler, F., & Dickie, C. (2014). Second language
acquisition of pitch range in german learners of english. Studies in Second Language Acquisition, 36(2), 303-329.
; Shi et al., 2014Shi,
S., Zhang, J., & Xie, Y. (2014). Cross-language comparison of F0
range in speakers of native Chinese, native Japanese and Chinese L2 of
Japanese: Preliminary results of a corpus-based analysis. In Proceedings of the 9th International Symposium on Chinese Spoken Language Processing (pp. 241-244). Singapore.
; Ullakonoja, 2007Ullakonoja, R. (2007). Comparison of pitch range in Finnish (L1) and Russian (L2). In Proceedings of the 16th International Congress of Phonetic Sciences (pp. 1701-1704). Saarbrücken, Germany.
; Yuan et al., 2018Yuan,
J., Dong, Q., Wu, F., Luan, H., Yang, X., Lin, H., & Liu, Y.
(2018). Pitch characteristics of L2 English speech by Chinese speakers: A
large-scale study. In Proceedings of the Annual Conference of the International Speech Communication Association (pp. 2593-2597). Hyderabad.
)
suggest that foreign speakers, regardless of their L1-L2 backgrounds,
are often characterized by a narrower F0 range and less variable pitch
when producing the L2 speech on the utterance level. In contrast, on the
phonemic level, Chinese L2 speakers were reported to have a wider pitch
span and smaller F0 fluctuations than native English speakers, mostly
due to the negative attachment of L1 lexical tones to stressed syllables
in the L2 (Ding et al., 2016Ding,
H., Hoffmann, R., & Hirst, D. (2016). Prosodic transfer: A
comparison study of f0 patterns in L2 english by chinese speakers. In 8th International Conference on Speech Prosody (pp. 756-760). Boston, US.
; J. Yuan et al., 2018Yuan,
J., Dong, Q., Wu, F., Luan, H., Yang, X., Lin, H., & Liu, Y.
(2018). Pitch characteristics of L2 English speech by Chinese speakers: A
large-scale study. In Proceedings of the Annual Conference of the International Speech Communication Association (pp. 2593-2597). Hyderabad.
).
The difficulty of accurately implementing the target pitch profiles has
been mainly correlated with the L2 learners’ lack of confidence and
insecurity when speaking a foreign language (Ding et al., 2016Ding,
H., Hoffmann, R., & Hirst, D. (2016). Prosodic transfer: A
comparison study of f0 patterns in L2 english by chinese speakers. In 8th International Conference on Speech Prosody (pp. 756-760). Boston, US.
; Shi et al., 2014Shi,
S., Zhang, J., & Xie, Y. (2014). Cross-language comparison of F0
range in speakers of native Chinese, native Japanese and Chinese L2 of
Japanese: Preliminary results of a corpus-based analysis. In Proceedings of the 9th International Symposium on Chinese Spoken Language Processing (pp. 241-244). Singapore.
; Yuan et al., 2018Yuan,
J., Dong, Q., Wu, F., Luan, H., Yang, X., Lin, H., & Liu, Y.
(2018). Pitch characteristics of L2 English speech by Chinese speakers: A
large-scale study. In Proceedings of the Annual Conference of the International Speech Communication Association (pp. 2593-2597). Hyderabad.
),
and not merely due to the language specificities and the different
socio-cultural identities. Another plausible factor that may constrain
the pitch variance is the learners’ increased cognitive efforts in
producing segments and stress (Zimmerer et al., 2014Zimmerer,
F., Jügler, J., Andreeva, B., Möbius, B., & Trouvain, J. (2014).
Too cautious to vary more? A comparison of pitch variation in native and
non-native productions of French and German speakers. In Ni. Campbell,
D. Gibbon, & D. Hirst (Eds.), Proceedings to the 7th Speech Prosody Conference (pp. 1037-1041). Dublin, Ireland.
).
Nevertheless, fortunately, studies showed that, with the aid of speech
technology or with developing their proficiency in L2, learners were
able to fine-tune the production of the L2 pitch and finally approach
native-like pitch patterns (Hincks & Edlund, 2009Hincks, R., & Edlund, J. (2009). Using speech technology to promote increased pitch variation in oral presentations. In International Workshop on Speech and Language Technology in Education (pp. 1-4). Wroxall, UK.
; Ullakonoja, 2007Ullakonoja, R. (2007). Comparison of pitch range in Finnish (L1) and Russian (L2). In Proceedings of the 16th International Congress of Phonetic Sciences (pp. 1701-1704). Saarbrücken, Germany.
).
On the other hand, L2 speech is also found to be characterized by a decrease in oral fluency (Peters, 2019Peters, J. (2019). Fluency and speaking fundamental frequency in bilingual speakers of High and Low German. In Proceedings of the 19th International Congress of Phonetic Sciences (pp. 1-5). Melbourne, Australia.
). The differences in fluency between the L1 and the L2 are frequently measured by various temporal metrics. For example, Ding et al. (2016)Ding,
H., Hoffmann, R., & Hirst, D. (2016). Prosodic transfer: A
comparison study of f0 patterns in L2 english by chinese speakers. In 8th International Conference on Speech Prosody (pp. 756-760). Boston, US.
showed that, in comparison with native English speakers, Chinese
learners tend to have a lower speech rate and articulation rate in their
L2 English. Lee and Sidtis (2017)Lee,
B., & Sidtis, D. V. L. (2017). The bilingual voice: Vocal
characteristics when speaking two languages across speech tasks. Speech, Language and Hearing, 20(3), 174-185.
and Peters (2019)Peters, J. (2019). Fluency and speaking fundamental frequency in bilingual speakers of High and Low German. In Proceedings of the 19th International Congress of Phonetic Sciences (pp. 1-5). Melbourne, Australia.
made similar observations. The decrease in speech fluency in the
non-native language has been explained with reference to the same
psychological and cognitive factors as L2 pitch compression-cautiousness
and increased cognitive efforts when speaking a foreign language.
However, unlike the two variables of speech rate and articulation rate,
the temporal assumption of pitch change rate is controversial,
especially when it is examined in a stress language such as English
compared to a tone language like Chinese. For instance, Yuan et al. (2018)Yuan,
J., Dong, Q., Wu, F., Luan, H., Yang, X., Lin, H., & Liu, Y.
(2018). Pitch characteristics of L2 English speech by Chinese speakers: A
large-scale study. In Proceedings of the Annual Conference of the International Speech Communication Association (pp. 2593-2597). Hyderabad.
reported a faster pitch change rate for L1 English speakers than for L2 Chinese learners, while in Ding et al. (2016)Ding,
H., Hoffmann, R., & Hirst, D. (2016). Prosodic transfer: A
comparison study of f0 patterns in L2 english by chinese speakers. In 8th International Conference on Speech Prosody (pp. 756-760). Boston, US.
, there was no significant difference between the two language groups with regards to the speed of pitch changes.
Despite the large body of cross-linguistic analyses of pitch and
temporal differences, it is somewhat difficult to compare the results of
these findings. This is partly because the F0 estimation methods and
the fluency measures used for evaluating the pitch and temporal
properties differed across studies. Another aspect is that the distinct
discourse conditions designed to elicit the speech may also cause
inconsistent results. For instance, Yuan and Liberman (2014)Yuan, J., & Liberman, M. (2014). F0 declination in English and Mandarin broadcast news speech. Speech Communication, 65, 67-74.
reported that Chinese native speakers have a wider pitch range and
greater F0 fluctuations in broadcast news speech than native English
speakers. However, regarding prose passages (Keating & Kuo, 2012Keating, P., & Kuo, G. (2012). Comparison of speaking fundamental frequency in English and Mandarin. Journal of the Acoustical Society of America, 132(2), 1050-1060.
), there was no significant difference in pitch range on the utterance level between Chinese and English speech.
Given the inconsistency of prior results and the typological differences between Chinese and Spanish, it is of great importance to examine the pitch and temporal characteristics in the CH-ES language pair, which has received little attention in the prosodic field to date. Of particular interest to us is to investigate (1) whether the pitch and temporal profiles produced by L2 Chinese learners are highly dependent on their L1 properties or if they support the L2 general trend hypothesis, (2) whether speakers’ pitch and temporal implementations are influenced by the gender and the level of proficiency in Spanish, and finally (3) whether the production of L2 pitch and temporal features reflects different levels of difficulty depending on question type and stress position. For these purposes, we extend the previous studies by accounting for proficiency level, gender, question type, and stress position, which allows us to examine the interaction between proficiency and other fixed factors concerning various pitch and temporal metrics.
2. METHODOLOGY
⌅2.1. Participants
⌅The participants of this study included: 5 female native speakers of Peninsular Spanish and 32 learners of Spanish (26 females and 6 males) whose first language is Mandarin Chinese. The ages of Chinese learners ranged from 21 to 31 (mean age: 24.09; SD = 2.53), while those of L1 Spanish speakers ranged from 18 to 24, with a mean age of 23.2 years (SD = 4.87). All subjects were divided into three language groups according to their proficiency level in Spanish: intermediate (B1-B2 level), advanced (C1-C2 level), and native. The Spanish proficiency of most Chinese speakers was judged using the information from their most recent official language qualification DELE (Diploma of Spanish as a Foreign Language). Chinese learners who did not have this certificate (approximately 15%) were asked to self-evaluate their L2 proficiency based on the Spanish language courses they had completed. The criteria for the six levels of European language proficiency were explained to those participants to help them to reach a reliable self-assessment.
Although the age of acquisition and the length of exposure to the target language are reported to influence L2 speech (Cadierno et al., 2020Cadierno,
T., Hansen, M., Lauridsen, J. T., Eskildsen, S. W., Fenyvesi, K.,
Jensen, S. H., & aus der Wieschen, M. V. (2020). Does younger mean
better? Age of onset, learning rate and shortterm L2 proficiency in
young Danish learners of English. Vigo International Journal of Applied Linguistics, 17, 57-86.
; Kharkhurin, 2008Kharkhurin,
A. V. (2008). The effect of linguistic proficiency, age of second
language acquisition, and length of exposure to a new cultural
environment on bilinguals’ divergent thinking. Bilingualism: Language and Cognition, 11(2), 225-243.
; Pfenninger & Singleton, 2016Pfenninger,
S. E., & Singleton, D. (2016). Age of onset, socio-affect and
cross-linguistic influence: a long-term classroom study. Vigo
International Journal of Applied Linguistics, 13, 147-179.
),
we did not control for these variables, as this would have
significantly reduced the number of L2 Chinese participants. However,
most of the Chinese learners in this study acquired Spanish in adulthood
(mean age: 18.81; SD = 2.08). Only one subject reported starting
to learn Spanish at 12 years of age. All the Chinese participants were
in an immersion situation at the time of recording. Although the length
of their stay in Spain had varied, the average exposure time of L2
advanced learners (mean length: 22.80 months; SD = 18.02) was generally longer than that of L2 intermediate speakers (mean length: 19.13 months; SD = 9.51).
2.2. Task and materials
⌅ The corpus was elicited by utilizing the DCT (Discourse Completion Task) technique (Billmyer & Varghese, 2000Billmyer,
K., & Varghese, M. (2000). Investigating instrument-based pragmatic
variability: Effects of enhancing discourse completion tests. Applied Linguistics, 21(4), 517-552.
; Félix-Brasdefer, 2010Félix-Brasdefer, J. C. (2010). Data collection methods in speech act performance. Speech Act Performance: Theoretical, Empirical and Methodological Issues, 26(41), 69-82.
).
Specifically, we designed 15 brief dialogues structured as situational
contexts to elicit five question types with different functional
meanings in Spanish, namely, information-seeking yes-no question (‘YN’), information-seeking wh-question (‘WH’), disjunctive question (‘DJ’), confirmation-seeking yes-no question (‘CYN’), and confirmation-seeking tag question (‘TAG’). The conversational interaction was initiated by an
interlocutor with whom the participant was familiar so that
politeness-related effects (e.g., power, and social distance) could be
minimized (Borràs-Comes, Sichel-Bazin, & Prieto, 2015Borràs-Comes,
J., Sichel-Bazin, R., & Prieto, P. (2015). Vocative intonation
preferences are sensitive to politeness factors. Language and Speech, 58(1), 68-83.
; Roseano et al., 2015Roseano,
P., Fernández Planas, A. M., Elvira-García, W., Massó, R. C., &
Celdrán, E. M. (2015). La entonación de las preguntas parciales en
catalán. Revista Española de Lingüística Aplicada, 28(2), 511-554.
). A sample context for eliciting the disjunctive question is as follows:
-
Interlocutor: Has invitado a un buen amigo a tu piso para una cena. Después de acabar los platos principales, le preguntas si quiere tarta o helado de postre. (You have invited a good friend to your apartment for dinner. After finishing the main courses, you ask her if she wants cake or ice cream for dessert.)
-
Participant: ¿Quieres tarta o helado? (Do you want cake or ice cream?)
Each of the five question types varied in the nuclear stress position
(two positions: penultimate syllable stress-paroxytone; final syllable
stress-oxytone). To facilitate L2 speakers’ comprehension during the
task, all test items consisted of words with high frequency for L1 and
L2 Spanish speakers (Tanaka & Terada, 2011Tanaka-Ishii, K., & Terada, H. (2011). Word familiarity and frequency. Studia Linguistica, 65(1), 96-116.
).
The recordings took place in a soundproof room with a head-mounted
microphone. Speech files were digitalized at a sampling rate of 44.1 kHz
and with a quantization precision of 16 bits. Each utterance was saved
separately and annotated to a TextGrid object in Praat (Boersma & Weenink, 2020Boersma, P., & Weenink, D. (2020). Praat: doing phonetics by computer [Computer program]. Version 5.3.82. http://www.praat.org/
).
2.3. Data extraction
⌅ For the purposes of this paper, two types of measurements were
conducted: (a) pitch and (b) temporal measures. In order to extract the
pitch information from the utterances, firstly, the ESPS algorithm (‘get
F0’) (Talkin, 1995Talkin, D. (1995). A robust algorithm for pitch tracking (RAPT). In W. B. Klejin & K. K. Paliwal (Eds.), Speech coding and synthesis (pp. 495-518). Elsevier Science B.V.
) was automatically conducted in Praat with the pitch floor and ceiling set to 70 Hz and 600 Hz, respectively.
A time step of 10 ms was used for the computation of F0. After the
automatic extraction, the raw F0 data were corrected manually, unvoicing
those pitch points with octave jumps or measurement errors, such as
false voicing in silent fragments, creaky voice, and laryngealization.
The linear results in Hz were then transformed into the near-logarithmic
scale (ERB-rate), which is one of the best psycho-acoustic measures for
modeling the intonational equivalence between men and women, and for
capturing the F0 differences across languages (Nolan, 2003Nolan, F. (2003). Intonational equivalence: an experimental evaluation of pitch scales. In Proceedings of the 15th international congress of phonetic sciences (Vol. 771, pp. 2-5). Barcelona, Spain.
).
In specific, pitch characteristics in this study were evaluated by
means of the three F0 variables: (1) 80% pitch span on the utterance
level (the 90th and 10th percentile span), (2) absolute span on the
syllable level (the 100th percentile span), and (3) pitch dynamism
quotient (abbreviated as PDQ). The PDQ metric was included as a
normalization of the F0 variation data since it can minimize the effects
caused by gender and different group size. The PDQ value gives an
account of the pitch variability in the utterance, and it is calculated
by dividing the standard deviation by the F0 mean. In general, the
previous literature indicates that the higher the PDQ, the more variable
the speech (Shi, Zhang, & Xie, 2014Shi,
S., Zhang, J., & Xie, Y. (2014). Cross-language comparison of F0
range in speakers of native Chinese, native Japanese and Chinese L2 of
Japanese: Preliminary results of a corpus-based analysis. In Proceedings of the 9th International Symposium on Chinese Spoken Language Processing (pp. 241-244). Singapore.
; Wang & Qian, 2018; Zimmerer et al., 2014Zimmerer,
F., Jügler, J., Andreeva, B., Möbius, B., & Trouvain, J. (2014).
Too cautious to vary more? A comparison of pitch variation in native and
non-native productions of French and German speakers. In Ni. Campbell,
D. Gibbon, & D. Hirst (Eds.), Proceedings to the 7th Speech Prosody Conference (pp. 1037-1041). Dublin, Ireland.
).
Further, considering the temporal traits, three variables were examined
between L1 and L2 speech: (1) pitch change rate (the average of the
absolute pitch differences in every 10-ms interval), (2) speech rate
(number of syllables / total duration of the utterance), and (3)
articulation rate (number of syllables / (total duration-internal
pauses). The minimum pause length calculated for fluency judgments was
set to 0.05 s instead of the larger values of 0.25 s adopted in the
study of Peters (2019)Peters, J. (2019). Fluency and speaking fundamental frequency in bilingual speakers of High and Low German. In Proceedings of the 19th International Congress of Phonetic Sciences (pp. 1-5). Melbourne, Australia.
.
The underlying reason is that the speech materials used in our
experiment were single utterances with an average syllable number of
5.8-unlike the passages in Peters (2019)Peters, J. (2019). Fluency and speaking fundamental frequency in bilingual speakers of High and Low German. In Proceedings of the 19th International Congress of Phonetic Sciences (pp. 1-5). Melbourne, Australia.
that frequently required the use of long pauses as a linguistic cue for narrative segmentation (Oliveira, 2002Oliveira, M. (2002). The role of pause occurrence and pause duration in the signaling of narrative structure. In International Conference for Natural Language Processing in Portugal (pp. 43-51). Springer.
).
2.4. Statistical analysis
⌅ The data analysis was conducted in the R environment (R Core Team, 2020R Core Team (2020). R: A language and environment for statistical computing [Computer software manual]. Vienna, Austria.
). A linear mixed-effects analysis was carried out using the lmerTest package for R (Kuznetsova et al., 2017Kuznetsova, A., Brockhoff, P. B., & Christensen, R. H. B. (2017). lmerTest package: tests in linear mixed effects models. Journal of Statistical Software, 82(13).
).
The six pitch and temporal parameters (80% span on the utterance level,
PDQ, 100% span on the syllable level, pitch change rate, speech rate,
and articulation rate) were entered into the model successively as
dependent variables, with Proficiency Level in Spanish (intermediate < advanced < native), Gender (female vs. male), Question Type (i.e., YN, WH, DJ, CYN, and TAG), Stress Type (Oxytone vs. Paroxytone), and their possible interactions as fixed
effects. Participants were included as random effects with all possible
random intercepts. The significance of the main effects was tested using
the ANOVA function. P-values were fitted by eliminating the non-significant effects of the initial model and calculated with Satterthwaite’s method (Kuznetsova et al., 2017Kuznetsova, A., Brockhoff, P. B., & Christensen, R. H. B. (2017). lmerTest package: tests in linear mixed effects models. Journal of Statistical Software, 82(13).
). The post-hoc analysis was performed using the single-step function of the multcomp package (Hothorn et al., 2016Hothorn,
T., Bretz, F., Westfall, P., Heiberger, R. M., Schuetzenmeister, A.,
Scheibe, S., & Hothorn, M. T. (2016). Package ‘multcomp.’ Simultaneous Inference in General Parametric Models. Project for Statistical Computing, Vienna, Austria.
) supported by the emmeans algorithm (Lenth et al., 2019Lenth, R., Singmann, H., & Love, J. (2019). Emmeans: Estimated marginal means, aka least-squares means. R Package Version 1.3.4.
).
3. RESULTS
⌅The following two sections present the results of the three pitch variables measured on the utterance (80 % F0 span, and PDQ) and syllable level (100 % F0 span), and the results of the three temporal parameters (pitch change rate, speech rate, and articulation rate).
3.1. Pitch results
⌅ First, we considered the differences in the use of pitch across the
three language groups. The analysis of variance indicated that Proficiency Level was not a significant factor for the three pitch variables (see Table 1). However, Figures 1, 2, and 3 indicate that Chinese intermediate (hereafter CI) and advanced learners
(hereafter CA) tend to produce a less variable pitch and narrower span
on the utterance and syllable levels compared to L1 Spanish speakers
(hereafter SN). These findings generally are consistent with previous
studies that reported a reduced pitch for non-native speakers (Busà & Urbani, 2011Busà, M. G., & Urbani, M. (2011). A Cross Linguistic Analysis of Pitch Range in English L1 and L2. In XVII International Congress of Phonetic Sciences (pp. 380-383). Hong Kong, China.
; Mennen, Schaeffler, & Docherty, 2007Mennen,
I, Schaeffler, F., & Docherty, G. (2007). Pitching it differently: a
comparison of the pitch ranges of German and English speakers. In 16th International Congress of Phonetic Sciences (pp. 1769-1772). Saarbrücken, Germany.
; Shi et al., 2014Shi,
S., Zhang, J., & Xie, Y. (2014). Cross-language comparison of F0
range in speakers of native Chinese, native Japanese and Chinese L2 of
Japanese: Preliminary results of a corpus-based analysis. In Proceedings of the 9th International Symposium on Chinese Spoken Language Processing (pp. 241-244). Singapore.
; Yuan et al., 2018Yuan,
J., Dong, Q., Wu, F., Luan, H., Yang, X., Lin, H., & Liu, Y.
(2018). Pitch characteristics of L2 English speech by Chinese speakers: A
large-scale study. In Proceedings of the Annual Conference of the International Speech Communication Association (pp. 2593-2597). Hyderabad.
; Zimmerer et al., 2014Zimmerer,
F., Jügler, J., Andreeva, B., Möbius, B., & Trouvain, J. (2014).
Too cautious to vary more? A comparison of pitch variation in native and
non-native productions of French and German speakers. In Ni. Campbell,
D. Gibbon, & D. Hirst (Eds.), Proceedings to the 7th Speech Prosody Conference (pp. 1037-1041). Dublin, Ireland.
), suggesting that there may be a universal trend of pitch range compression in L2 speech. Additionally, the results in Figures 1, 2, and 3 indicated that, in comparison with the lower proficiency group (i.e.,
the CI group), highly proficient learners of the CA group were closer to
SN speakers in the implementation of the F0 pitch, although this trend
was not strong enough to be statistically significant (see Table 2).
| 80 % utterance span | PDQ | 100 % syllable span | |
|---|---|---|---|
| Proficiency | 2.99. | 2.80. | 2.53. |
| QuestionType | 10.53*** | 8.99*** | 22.26*** |
| Gender | 0.00 | 8.76** | 1.33 |
| StressType | 3.12. | 4.00* | 0.42 |
| Proficiency*QuestionType | 9.98*** | 8.42*** | 8.58*** |
| Proficiency* StressType | 3.54* | 0.22 | 0.37 |
| 80 % utterance span | 100 % syllable span | PDQ | |
|---|---|---|---|
| CI-CA | t = −1.802, p = 0.179 | t = −1.783, p = 0.185 | t = −1.791, p = 0.182 |
| SN-CA | t = 1.029, p = 0.559 | t = 0.766, p = 0.723 | t = 0.932, p = 0.620 |
| SN-CI | t = 2.289, p = 0.067 | t = 2.029, p = 0.116 | t = 2.190, p = 0.083 |
Next, as with the Question Type factor, it is apparent in Table 1 that there is a significant main effect on the three pitch variables. In contrast, the factors Gender and Stress Type were found to be significant only for the variable of PDQ. In particular, our results indicated that female speakers (mean PDQ: 0.175) had significantly more F0 variability than males (mean PDQ: 0.127) in speech [t(70) = 2.14, p < 0.05). We also observed a significant effect of Stress Type on the variable of PDQ. Specifically, it is noteworthy in Figure 2 (see the right panel) that participants of the three language groups consistently had a more variable pitch in questions with a paroxytone than those with an oxytone in the final word.
As with the 80% utterance span, Figure 1 shows that the two Chinese groups had a wider pitch span in questions
ending with a paroxytone word, but this tendency was statistically
significant only for the CA group [t(539) = 3.07, p <
0.01]. Regarding the SN group, we did not find a statistically
significant difference in realizing the pitch between the two stress
types [t(539) = 0.04, p = 0.76], although SN speakers were
more likely to compress the F0 span in questions ending with a
paroxytone word (see the right panel of Figure 1).
The pitch performance exhibited by the CI and CA groups may be because
the paroxytone is the most frequent and unmarked stress pattern in
Spanish and, therefore, the most familiar one for L2 speakers (Defior & Serrano, 2017Defior, S., & Serrano, F. (2017). Learning to Read Spanish. In L. Verhoeven & C. Perfetti (Eds.), Learning to Read across Languages and Writing Systems (pp. 243-269). Cambridge University Press.
; Roca, 2019Roca,
I. (2019). Spanish Word Stress: an updated multidimensional account. In
R. Goedemans, J. Heinz & H. van der Hulst (Eds.), The Study of Word Stress and Accent: Theories, Methods and Data (pp. 256-292). Cambridge University Press.
).
This means that Chinese learners may experience the least cognitive
difficulties when producing such stressed words in Spanish, which allows
more planning time to fine-tune the corresponding pitch profiles in a
native-like way. In contrast, it is unclear why SN speakers had an
opposite trend for implementing the F0 span between the two stress
types. Since we only had five Spanish subjects in this work, future
investigations with a larger sample size are needed to validate this
finding.
The results of the linear mixed model also revealed a strong interaction effect between Proficiency Level and Question Type on the three pitch variables (see Table 1). The post-hoc analysis indicated that the pitch performance of CI and CA learners was highly dependent on the question type in which they were engaged. More precisely, we found that, in comparison with the SN group, the CI and CA group had a particularly narrower span and less pitch variability in DJ [e.g. 80% span: CI-SN: t(2) = −4.04, p < 0.001; CA-SN: t(2) = −3.79, p < 0.001] and YN questions [e.g. PDQ: CI-SN: (t(2) = −3.35, p < 0.01); CA-SN: (t(2) = −2.47, p < 0.05)]. By contrast, in WH questions, it is noteworthy that the two Chinese groups had a higher PDQ and a wider pitch span on both utterance and syllable levels than the SN group (see Figures 1, 2, and 3). This finding can be explained by the overproduction of WH questions by Chinese learners. Specifically, we notice that some L2 learners, irrespective of their level of proficiency, tend to produce a high-rising nuclear pitch accent or a final rising boundary tone in WH questions. Although the final rising contour can also be used in WH questions, it is not frequently found in the L1 native speech (i.e., all the SN speakers in our study produced the WH questions with a final-falling pitch movement) since the interrogative particles in Spanish (e.g., qué, dónde, quién, cuál) are clear enough for signaling this type of question.
3.2. Temporal results
⌅ The main effects of the linear mixed models fitted for the three temporal variables are shown in Table 3. For ease of exposition, we discuss these results by referring to Figures 4 and 5,
which display the specific temporal values produced by the three
language groups in the five question types. First, considering
individual effects, the output in Table 3 revealed that there was a significant main effect of Proficiency and Question Type on the outcome variables of pitch change rate, speech rate, and articulation rate. By contrast, Stress Type and Gender were insignificant factors for the three temporal variables. Moreover, the pairwise comparisons of Proficiency Level showed that, in comparison with the SN group, the two Chinese groups had a significantly lower pitch change rate [CI-SN: t(2) = −4.71, p < 0.001; CA-SN: t(2) = −3.75, p < 0.01], speech rate [CI-SN: t(2) = −5.71, p < 0.001; CA-SN: t(2) = −5.62, p < 0.001], and articulation rate [CI-SN: t(2) = −5.58, p < 0.001; CA-SN: t(2) = −5.44, p < 0.001] in their speech. These findings corroborate previous
studies that reported a reduced oral fluency for L2 speakers in the
non-native language (Ding et al., 2016Ding,
H., Hoffmann, R., & Hirst, D. (2016). Prosodic transfer: A
comparison study of f0 patterns in L2 english by chinese speakers. In 8th International Conference on Speech Prosody (pp. 756-760). Boston, US.
; Peters, 2019Peters, J. (2019). Fluency and speaking fundamental frequency in bilingual speakers of High and Low German. In Proceedings of the 19th International Congress of Phonetic Sciences (pp. 1-5). Melbourne, Australia.
).
Nevertheless, unlike our previous findings-which showed that
high-proficiency Chinese learners achieved a target-like pitch
performance-(see Section 3.1), we did not observe any significant
improvement in speech rate and articulation rate between the CI and CA
groups.
| Pitch change rate | Speech rate | Articulation rate | |
|---|---|---|---|
| Proficiency | 11.23*** | 18.75*** | 17.75*** |
| QuestionType | 14.95*** | 10.56*** | 4.35** |
| Gender | 0.03 | 2.71 | 3.26. |
| StressType | 0.01 | 0.22 | 0.00 |
| Proficiency*QuestionType | 11.46*** | 3.80*** | 2.64** |
Further, the results in Table 3 indicated a strong interaction between Proficiency and Question Type on the three temporal variables. Particularly, as shown in Figure 5, SN speakers had higher values of pitch change rate than the CI and CA learners in all questions except for WH questions. As discussed above, the faster pitch change in L2 WH questions may be attributed to the fact that most Chinese learners excessively varied their F0 contours by producing either a high pitch accent or a final rising boundary in the nuclear position. In addition, although each question type was realized with a specific temporal value, the two Chinese groups were consistently lower than the SN speakers regarding the speech and articulation rates (see Figure 5). Finally, it is interesting that the results of speech rate and articulation rate were similar in this work. This is perhaps because the speech stimuli used in this work consisted of short utterances produced with low frequency and short pauses.
4. DISCUSSION
⌅The aim of the present study was to investigate the L2 production of Spanish questions by Chinese speakers with regards to pitch and temporal characteristics and to explore the factors that may contribute to the pitch and temporal deviations in L2 speech. Six pitch and temporal metrics of L1 and L2 Spanish speakers were examined and compared using a linear mixed-effects analysis. The findings of our study are discussed below.
First, our results confirm that there are indeed some
cross-linguistic differences between Spanish L1 and L2 regarding pitch
performance. The evidence in support of this is that the L2 Spanish in
this study was produced with a narrower span (on both utterance and
syllable levels) and less variable pitch than that of L1 native
speakers. This supports previous studies that reported a pitch range
compression effect for L2 speakers with typologically different L1
backgrounds (e.g., Busà & Urbani, 2011Busà, M. G., & Urbani, M. (2011). A Cross Linguistic Analysis of Pitch Range in English L1 and L2. In XVII International Congress of Phonetic Sciences (pp. 380-383). Hong Kong, China.
; Liu, 2005Liu, Y. H. (2005). La entonación del español hablado por taiwaneses. Biblioteca Phonica, 2. www.ub.es/lfa
; Mennen et al., 2007Mennen,
I, Schaeffler, F., & Docherty, G. (2007). Pitching it differently: a
comparison of the pitch ranges of German and English speakers. In 16th International Congress of Phonetic Sciences (pp. 1769-1772). Saarbrücken, Germany.
, 2012Mennen,
Ineke, Schaeffler, F., & Docherty, G. (2012). Cross-language
differences in fundamental frequency range: A comparison of English and
German. Journal of the Acoustical Society of America, 131(3), 2249-2260.
, 2014Mennen,
Ineke, Schaeffler, F., & Dickie, C. (2014). Second language
acquisition of pitch range in german learners of english. Studies in Second Language Acquisition, 36(2), 303-329.
; Peters, 2019Peters, J. (2019). Fluency and speaking fundamental frequency in bilingual speakers of High and Low German. In Proceedings of the 19th International Congress of Phonetic Sciences (pp. 1-5). Melbourne, Australia.
; Shi et al., 2014Shi,
S., Zhang, J., & Xie, Y. (2014). Cross-language comparison of F0
range in speakers of native Chinese, native Japanese and Chinese L2 of
Japanese: Preliminary results of a corpus-based analysis. In Proceedings of the 9th International Symposium on Chinese Spoken Language Processing (pp. 241-244). Singapore.
; Ullakonoja, 2007Ullakonoja, R. (2007). Comparison of pitch range in Finnish (L1) and Russian (L2). In Proceedings of the 16th International Congress of Phonetic Sciences (pp. 1701-1704). Saarbrücken, Germany.
; Urbani, 2012Urbani,
M. (2012). Pitch range in L1/L2 English. An analysis of f0 using LTD
and linguistic measures. In M. G. Busà & S. Antonio (Eds.), Methodological perspectives on L2 prosody: Papers from ML2P 2012 (pp. 79-83).
; Yuan et al., 2018Yuan,
J., Dong, Q., Wu, F., Luan, H., Yang, X., Lin, H., & Liu, Y.
(2018). Pitch characteristics of L2 English speech by Chinese speakers: A
large-scale study. In Proceedings of the Annual Conference of the International Speech Communication Association (pp. 2593-2597). Hyderabad.
; Zimmerer et al., 2014Zimmerer,
F., Jügler, J., Andreeva, B., Möbius, B., & Trouvain, J. (2014).
Too cautious to vary more? A comparison of pitch variation in native and
non-native productions of French and German speakers. In Ni. Campbell,
D. Gibbon, & D. Hirst (Eds.), Proceedings to the 7th Speech Prosody Conference (pp. 1037-1041). Dublin, Ireland.
).
The consistency of the findings for L2 pitch and temporal production
suggests that non-native learners may have universal developmental
pathways for acquiring specific aspects of L2 speech, independent of the
specificity of their L1 system. We cannot provide a definitive
explanation for this quasi-universal effect in L2 speech. However,
rather than being shaped by the L1 phonetic system, the compressed pitch
patterns in L2 have previously been attributed to the lack of
confidence and insecurity of L2 learners when speaking a non-native
language (Peters, 2019Peters, J. (2019). Fluency and speaking fundamental frequency in bilingual speakers of High and Low German. In Proceedings of the 19th International Congress of Phonetic Sciences (pp. 1-5). Melbourne, Australia.
; Volín, Poesová, & Weingartová, 2015Volín,
J., Poesová, K., & Weingartová, L. (2015). Speech melody properties
in English, Czech and Czech English: Reference and interference. Research in Language, 13(1), 107-123.
; Yuan et al., 2018Yuan,
J., Dong, Q., Wu, F., Luan, H., Yang, X., Lin, H., & Liu, Y.
(2018). Pitch characteristics of L2 English speech by Chinese speakers: A
large-scale study. In Proceedings of the Annual Conference of the International Speech Communication Association (pp. 2593-2597). Hyderabad.
).
Additionally, the increased cognitive efforts in producing the
non-native segmental or suprasegmental features (i.e., vowels and
consonants, stress, and prominence) are also plausible factors that may
lead to a lower pitch variability in L2 utterances. For instance, Zimmerer et al. (2014)Zimmerer,
F., Jügler, J., Andreeva, B., Möbius, B., & Trouvain, J. (2014).
Too cautious to vary more? A comparison of pitch variation in native and
non-native productions of French and German speakers. In Ni. Campbell,
D. Gibbon, & D. Hirst (Eds.), Proceedings to the 7th Speech Prosody Conference (pp. 1037-1041). Dublin, Ireland.
pointed out that L2 learners can frequently overlook the variation of
F0 pitch in a native-like way because they are too focused on the
correct production of words and stress in the non-native language.
Another noteworthy point in the pitch results is the F0 span at the
syllable level. As a typical tone language, Chinese makes use of the F0
information for encoding lexical tone meanings (Yuan, 2011Yuan, J. (2011). Perception of intonation in Mandarin Chinese. Journal of the Acoustical Society of America, 130(6), 4063-4069.
).
Therefore, it is expected that Chinese learners would show greater F0
variations on the syllable level because of L1 tonal transfer. However,
unlike Ding et al. (2016)Ding,
H., Hoffmann, R., & Hirst, D. (2016). Prosodic transfer: A
comparison study of f0 patterns in L2 english by chinese speakers. In 8th International Conference on Speech Prosody (pp. 756-760). Boston, US.
,
we did not find a wider pitch span on the syllable level for Chinese
learners of Spanish. This seems to imply that the production of the L2
syllable span was not necessarily affected by the learners’ long-term
experience with a tone language. The discrepancy between the results
could be justified by the distinct language pairs examined in the
experiment: In Ding et al. (2016)Ding,
H., Hoffmann, R., & Hirst, D. (2016). Prosodic transfer: A
comparison study of f0 patterns in L2 english by chinese speakers. In 8th International Conference on Speech Prosody (pp. 756-760). Boston, US.
,
English was the Chinese learners’ L2, whereas in our study, it was
Spanish. Future studies regarding the pitch range differences between
English and Spanish at the syllable level would help us elucidate
whether this is the primary cause of the discordances found. On the
other side, based on our observed data, another possible explanation for
the reduced syllable span in L2 Spanish might be that Chinese learners
were too cautious to vary the pitch due to a lack of intonational skills
and language experience, thereby exhibiting a flat F0 contour without
many fluctuations until they reached the great F0 changes in the nuclear
location. Further investigations of L2 phonetic performance are
required to test this hypothesis, considering the position sensitivity
of pitch changes in the utterance.
Further, although the factor proficiency statistically failed to reach significance in the three pitch variables, the results seem to suggest that Chinese learners of L2 Spanish can progressively fine-tune their production of F0 values and approach a target-like pitch pattern with increasing proficiency in their L2. Moreover, results of the three pitch parameters revealed a strong interaction between proficiency level and question type, illustrating that the L2 learning of pitch implementation details is susceptible to pragmatically different question types. For instance, we found that Chinese intermediate and advanced learners consistently had a reduced pitch span and lower PDQ in all utterances except for WH questions. As is clear from the above discussion, the opposite performance of Chinese speakers on WH-questions can account for their overproduction of a high pitch accent or a final-rising boundary in the nuclear position. Or, in a more general way, it can be attributed to the fact that learners were unfamiliar with the target intonation contours of WH-questions due to the typological distance between the L1 and the L2. Thus, most would simply assume that Spanish WH-questions are produced with a high pitch in the utterance-final location based on their knowledge of the typical use of the F0 cue.
As with other question types (e.g., the information-seeking yes-no question and the disjunctive question), we found that most F0 targets
in the utterance-final position could be accurately achieved by Chinese
learners, while those in the prenuclear position were deviated and
produced with a less variable contour. In this regard, our findings
suggest that the compressed pitch in L2, rather than being solely
determined by psychological and cognitive factors (i.e., uncertainty,
cautiousness, and increased efforts when speaking the L2), is also
constrained by the learners’ knowledge of the target intonation
categories. Overall, the different pitch performance of the L2 speakers
in the five question types gives support to previous findings which
proposed a scaffolding from the phonological to phonetic dimensions (Cortés Moreno, 2004Cortés Moreno, M. (2004). Análisis acústico de la producción de la entonación española por parte de sinohablantes. Estudios de Fonética Experimental, 13, 80-110.
; Yuan et al., 2019Yuan,
C., González-Fuente, S., Baills, F., & Prieto, P. (2019). Observing
pitch gestures favors the learning of spanish intonation by mandarin
speakers. Studies in Second Language Acquisition, 41(1), 5-32.
),
suggesting that there is a hierarchy of difficulties in implementing
the L2 pitch patterns depending on the prosodic similarities and
dissimilarities between the first and the target language.
Considering the gender effect, our study revealed that men and women
differ significantly only in the variable of PDQ. Congruent with
previous works (Ordin & Ineke Mennen., 2017Ordin, M., & Ineke Mennen. (2017). Cross-Linguistic Differences in Bilinguals’ Fundamental Frequency Ranges. Journal of Speech, Language, and Hearing Research, 60(6), 1493-1506.
),
female speakers in our study varied their F0 contours more frequently
than male speakers. The gender differences in pitch variability are more
closely linked to the speakers’ willingness to express emotions in
communication rather than physiological factors. Research has shown that
humans express a range of emotions by readily modulating their F0
pitch, and female speakers tend to express most emotions more frequently
than males in speech-except for pride and power (Brebner, 2003Brebner, J. (2003). Gender and emotions. Personality and Individual Differences, 34(3), 387-394.
; Pisanski et al., 2020Pisanski,
K., Raine, J., & Reby, D. (2020). Individual differences in human
voice pitch are preserved from speech to screams, roars and pain cries. Royal Society Open Science, 7(2), 191642.
).
In this sense, the greater pitch variance observed in the data of
female speakers could be attributed to their greater emotional
involvement in speech than male participants. Nevertheless, because the
number of male and female speakers differed strongly in this task, this
research needs to be replicated with a well-balanced design to
consolidate the results presented here.
Further interesting findings related to pitch are that the F0 variation was highly modulated by the stress type, whereby all speakers produced more variable pitch in questions with a final-paroxytone word than in those with a final-oxytone word. Similarly, for the 80 % F0 span, Chinese learners (particularly those of the advanced group) showed a significantly wider pitch span in questions ending with a paroxytone word. We speculate that this could be related to the relative cognitive efforts required to process the two stress types for L2 learners. Since the paroxytone is the most frequent and unmarked stress pattern in Spanish (hence the most familiar one for L2 learners), Chinese speakers may show fewer difficulties when producing it in questions and have more planning time, allowing them to better approach a target-like pitch profile. Although L1 Spanish speakers had a reduced pitch span in sentences with a final-paroxytone word, this effect did not reach statistical significance, and their average pitch span was still higher than that of Chinese learners with such stimuli. So far, we have no clear explanation for the behaviour of Spanish speakers. Since there were only five native subjects in the control group, future investigations with a larger sample size are required to test whether there is a difference of pitch span for L1 Spanish speakers in questions ending with different stress patterns.
Regarding the temporal characteristics, our
study revealed significantly lower pitch change rate, speech rate, and
articulation rate in L2 Spanish. These results are consistent with
previous studies that reported a similar reduction of oral fluency (Ding et al., 2016Ding,
H., Hoffmann, R., & Hirst, D. (2016). Prosodic transfer: A
comparison study of f0 patterns in L2 english by chinese speakers. In 8th International Conference on Speech Prosody (pp. 756-760). Boston, US.
; Peters, 2019Peters, J. (2019). Fluency and speaking fundamental frequency in bilingual speakers of High and Low German. In Proceedings of the 19th International Congress of Phonetic Sciences (pp. 1-5). Melbourne, Australia.
) and slower pitch rises and falls in L2 speech than L1 speech (Yuan et al., 2018Yuan,
J., Dong, Q., Wu, F., Luan, H., Yang, X., Lin, H., & Liu, Y.
(2018). Pitch characteristics of L2 English speech by Chinese speakers: A
large-scale study. In Proceedings of the Annual Conference of the International Speech Communication Association (pp. 2593-2597). Hyderabad.
).
Moreover, it has been noted that although Chinese is a lexical tone
language with F0 peaks or valleys in every syllable, the speed of F0
changes is not significantly faster than in stress languages such as
English (Xu & Sun, 2002Xu, Y., & Sun, X. (2002). Maximum speed of pitch change and how it may relate to speech. Journal of the Acoustical Society of America, 111(3), 1399-1413.
).
If this is the case, we speculate that there is no negative transfer of
L1 Chinese in terms of the pitch change rate in this study. The lower
values of Chinese L2 learners on the three temporal metrics might also
be attributed to their increased cognitive efforts in producing the
segments or their lack of experience in the target speech.
Additionally, the interaction effects found for the three temporal
variables indicate that the proficiency effect was strongly modulated by
question type. Whereas the speech rate and articulation rate were lower
in all question types for L2, the average pitch change rate showed an
exception for the WH questions in which the F0 directions varied more
frequently in L2 than in L1. Since there is no indication that the L2
deviation on WH questions was caused by the systematic differences
between the two languages, we speculate that the higher values of pitch
change rate and F0 span in WH questions reflected overproduction by
Chinese speakers due to a lack of target intonational knowledge.
Finally, the main effect of proficiency seems to suggest a trend of
pitch improvement with learners’ increasing L2 proficiency. In
particular, our study replicates previous findings (i.e., Ullakonoja, 2007Ullakonoja, R. (2007). Comparison of pitch range in Finnish (L1) and Russian (L2). In Proceedings of the 16th International Congress of Phonetic Sciences (pp. 1701-1704). Saarbrücken, Germany.
; Yuan et al., 2018Yuan,
J., Dong, Q., Wu, F., Luan, H., Yang, X., Lin, H., & Liu, Y.
(2018). Pitch characteristics of L2 English speech by Chinese speakers: A
large-scale study. In Proceedings of the Annual Conference of the International Speech Communication Association (pp. 2593-2597). Hyderabad.
; Zimmerer et al., 2014Zimmerer,
F., Jügler, J., Andreeva, B., Möbius, B., & Trouvain, J. (2014).
Too cautious to vary more? A comparison of pitch variation in native and
non-native productions of French and German speakers. In Ni. Campbell,
D. Gibbon, & D. Hirst (Eds.), Proceedings to the 7th Speech Prosody Conference (pp. 1037-1041). Dublin, Ireland.
)
that highly proficient learners were closer to L1 native speakers in
the realization of pitch change rate, pitch span on the utterance and
syllable level, and pitch variability. Further, as suggested by
neurobehavioral research, the advantage of high-proficiency speakers in
the L2 can be attributed to their enhanced ability to use higher-level
cognition (i.e., attention) to process non-native speech components (Archila-Suerte et al., 2012Archila-Suerte,
P., Zevin, J., Bunta, F., & Hernandez, A. E. (2012). Age of
acquisition and proficiency in a second language independently influence
the perception of non-native speech. Bilingualism, 15(1), 190.
, 2015Archila-Suerte,
P., Zevin, J., & Hernandez, A. E. (2015). The effect of age of
acquisition, socioeducational status, and proficiency on the neural
processing of second language speech sounds. Brain and Language, 141, 35-49.
).
5. CONCLUSION
⌅The study presented here was intended to explore the pitch and temporal characteristics of native and Chinese L2 speakers of Spanish. Using six different metrics, we examined the pitch and temporal implementation in five question types of Peninsular Spanish and obtained several important findings regarding the cross-linguistic differences in the speech. First, congruent with previous literature on L2 speech, the results of this study suggest that Chinese speakers of L2 Spanish deviate from L1 native speakers mainly in the compression of pitch span (both on the utterance and syllable levels) and pitch variability, and the strong reduction of pitch change rate, speech rate, and articulation rate. Second, these pitch and temporal deviations in L2 speech are attributed to psychological-cognitive factors and the learners’ lack of knowledge and intonation skills in the target language rather than physiological factors or the L1 effect.
From the pedagogical perspective, our findings hold important implications for understanding the cross-linguistic differences between L1 and the speech, underlining the importance of preparing special training methods with varied materials and contexts to reduce learners’ foreign accents and improve their phonetic knowledge of the L2. Further research on native Chinese and native Spanish will be conducted to explore more cross-linguistic differences that may account for the L2 speech deviations. It is also interesting to consider how pitch span and pitch variability are realized depending on the syntactic and phonological positions of the phrase and in which locations L2 learners mostly deviate from the L1 native speakers.