International Journal of Speech-Language Pathology, 2011; 13(6): 526–535
Proficiency in English sentence stress production by Cantonese speakers who speak English as a second language (ESL)
Int J Speech Lang Pathol Downloaded from informahealthcare.com by University of Hong Kong Libraries on 11/30/11 For personal use only.
MANWA L. NG1 & YANG CHEN2 1
University of Hong Kong, Hong Kong, SAR China, and 2Duquesne University, Pittsburgh, PA, USA
Abstract The present study examined English sentence stress produced by native Cantonese speakers who were speaking English as a second language (ESL). Cantonese ESL speakers’ proficiency in English stress production as perceived by English-speaking listeners was also studied. Acoustical parameters associated with sentence stress including fundamental frequency (F0), vowel duration, and intensity were measured from the English sentences produced by 40 Cantonese ESL speakers. Data were compared with those obtained from 40 native speakers of American English. The speech samples were also judged by eight naı¨ve listeners who were native speakers of American English for placement, degree, and naturalness of stress. Results showed that Cantonese ESL speakers were able to use F0, vowel duration, and intensity to differentiate sentence stress patterns. Yet, both female and male Cantonese ESL speakers exhibited consistently higher F0 in stressed words than English speakers. Overall, Cantonese ESL speakers were found to be proficient in using duration and intensity to signal sentence stress, in a way comparable with English speakers. In addition, F0 and intensity were found to correlate closely with perceptual judgement and the degree of stress with the naturalness of stress.
Keywords: Bilingualism, cross-linguistic, acoustics, stress.
Introduction Errors in English produced by foreign speakers are readily noticeable by native ears, and such errors contribute to the perceived accent in L2 English. Accent originates from deviations in articulation (segmental errors) and prosody (suprasegmental errors) in L2 English production. The more the phonetic deviation exists, the greater the amount of accent is perceived (Flege, 1980; Flege & Davidian, 1985; Major, 2001; Ng, Chen, & Sadaka, 2008; Tarone, 1980). Accent present in L2 English may adversely affect the effectiveness of communication of foreign language (L2) speakers. Among the many aspects of suprasegmental features, English stress has been considered a major difficulty in speaking English as a second language (ESL). Previous studies have reported errors in English stress produced by L2 English speakers, as compared to native speakers of English. For example, it has been observed by many ESL teachers that English produced with distorted stress pattern by Cantonese ESL speakers are commonly perceived (Chan & Li, 2000; Thompson & Gaddes, 2005). Perceptually, all syllables within a sentence appear to be produced with equal or similar amount of stress, with little or no
difference between stressed and unstressed syllables within a sentence. This might be related to the fact that Cantonese is a syllable-timed language, while English is a stress-timed language. In fact, mastery of English stress production has been regarded as a strong indicator of proficiency of spoken English as an L2 (Cheng, 1968). In English, the important linguistic information carried by lexical stress and sentence stress has been discussed (e.g., Raphael, Borden, & Harris, 2007). Lexical stress refers to the pattern of stress within the boundary of a word. For example, the placement of lexical stress of the word subject determines if it is a noun or a verb (i.e., PERmit vs perMIT). Sentence stress acts as a pointer indicating the important information within the sentence. For example, the more important word of the phrase ‘‘This is my green book’’ changes from ‘‘this’’ to ‘‘my’’ in sentences ‘‘THIS is my green book’’ and ‘‘This is MY green book’’. Previous studies consistently reported that stressed and unstressed syllables within a sentence could be differentiated by fundamental frequency (F0), vowel duration, and vowel intensity (Chen, Robb, Gilbert, & Lerman, 2001; Cooper & Sorensen, 1981;
Correspondence: Manwa L. Ng, University of Hong Kong, 5/F Prince Philip Dental Hospital, 34 Hospital Road, Sai Ying Pun, Hong Kong, PR China. Tel: þ852-28590582. Fax: þ852-25590060. E-mail:
[email protected] ISSN 1754-9507 print/ISSN 1754-9515 online ª 2011 The Speech Pathology Association of Australia Limited Published by Informa UK, Ltd. DOI: 10.3109/17549507.2011.580783
Int J Speech Lang Pathol Downloaded from informahealthcare.com by University of Hong Kong Libraries on 11/30/11 For personal use only.
Proficiency in English sentence stress production Cooper, Eady, & Mueller, 1985; Fry, 1955; Raphael et al., 2007). Stressed syllables are associated with higher F0, and longer and louder vowel segment (e.g., Fry, 1955; Raphael et al., 2007). It follows that proficiency in L2 English sentence stress production can be quantitatively described by using F0, and duration and intensity of the vocalic segment. In comparing the English sentences produced by adult Mandarin ESL speakers and native AmericanEnglish speakers, Chen et al. (2001) found that, similar to native English speakers, Mandarin ESL speakers exhibited higher average F0, longer vowel duration, and greater vowel intensity in stressed than unstressed words. However, comparing between Mandarin ESL and English speakers, Mandarin speakers produced the stressed English words with significantly higher F0, and male Mandarin speakers stressed the words with shorter vowel duration than English speakers (Chen et al., 2001). Deviation in F0, vowel duration, and intensity in the English sentences produced by Mandarin ESL speakers was believed to contribute to the accent perceived. It was suggested that such deviation in F0, vowel duration, and intensity is the result of the interference of Mandarin (L1) on the production of English (L2) (Chen, 2006; Flege, 1980). According to Chen et al. (2001), English sentence intonation indicated by F0 values was due to the fact that Mandarin is a tone language; English intonation was affected by the intrinsic lexical tones of Mandarin words. In addition, the shorter vowel duration in the stressed English words might be related to the syllable-time nature of Mandarin. Mandarin syllables occupied relatively constant duration across a sentence. When Mandarin ESL speakers spoke English, intensity appeared to be the most comparable with English speakers, although Mandarin speakers tended to produce stressed and unstressed words at a greater intensity than English speakers (Chen et al., 2001). These findings are generally supported by other studies of ESL speakers of a tone language (Cheng, 1968; Chun, 1982), in which difficulty in acquiring a native-like stress pattern of English has been reported (Chun, 1982). Despite the several studies of how L1 Mandarin influences production of L2 English sentence stress production, only one study provided objective measurements of L2 English sentence stress, and discussion was made in relation to L1 Mandarin. Also, no perceptual experiment has been carried out to examine how L2 English stress patterns are perceived by native English speakers. Although English sentence stress has been found to be greatly associated with proficiency of speaking English as a second language, how English sentence stress is produced by Cantonese ESL speakers is not known. It is hypothesized that, with the influence of L1, Cantonese ESL speakers may demonstrate deviations in producing L2 English sentence stress. To better understand how native Cantonese ESL speakers produce English sentence stress, the present
527
study served as an extension of Chen et al.’s study, and perceptually and acoustically examined English sentence stress produced by Cantonese ESL speakers. An array of acoustical measures was used, and comparison was made between native Cantonese ESL speakers and English speakers. Cantonese is a tone language in which six contrastive lexical tones are identified: the high level, high rising, mid level, low falling, low rising, and low level tones. That is, the same syllable may have different meanings when produced at different lexical tones. For example, the syllable /si/ means poem, history, to try, time, market, and to serve if produced at the aforementioned tones, respectively. The present study quantitatively measured the deviations in L2 English sentence stress produced by adult Cantonese ESL speakers based on F0, vowel duration, and intensity. The array of acoustical measurements was then correlated with the perceived accuracy in placement, amount, and naturalness of stress in L2 English sentences perceived by native English-speaking listeners. Based on previous studies on Mandarin ESL speakers, and the marked difference in the language structures between English and Cantonese (stressedtime vs syllable-timed), it is predicted that Cantonese ESL speakers produce English sentence stress with greater accuracy in intensity, and least in duration.
Method Participants Speakers. Forty Cantonese speakers (20 males and 20 females) who were speaking English as a second language (ESL) participated in the study. The selection criteria for the Cantonese ESL speakers included: (1) the speakers were native Hong Kong Cantonese adult speakers with ages between 19–27 years (mean age ¼ 23 years) who also spoke English as a second language; (2) Cantonese was their first language and they had learned English for over 10 years; (3) they demonstrated a high level of proficiency in listening and speaking English; and (4) they had obtained at least a grade C in the HKCEE (Hong Kong Certificate of Education Examination) or a D in the HKAL (Hong Kong Advanced Level Education) or an A in the speaking section in the University Entrance Test for Speaker of Other Languages. According to NARIC (The National Recognition Information Centre for the United Kingdom), this was equivalent to a grade B or above in GCSE (General Certificate of Secondary Education) or GCE (General Certificate of Education) Advanced Level exam (HKEAA, 2008a), or *6.0 or above in an IELTS (International English Language Testing System) exam (HKEAA, 2008b), or a grade C or above in the IGSCE (International General Certificate of Secondary Education) exam (HKEAA, 2008c). A second group of speakers included 40 native American-English monolingual adult speakers
Int J Speech Lang Pathol Downloaded from informahealthcare.com by University of Hong Kong Libraries on 11/30/11 For personal use only.
528
M. L. Ng and Y. Chen
(20 males and 20 females). The native English speakers were selected from North America and judged as speaking the Mid-Atlantic variation of General American English by a practicing speechlanguage pathologist. The recruitment criteria for the English speakers included: (1) they were between 21–32 years old (mean age¼25 years); (2) they grew up in the Inland North region of the US, (3) they consistently demonstrated /A, O/ distinction in their spoken English, (4) their English lacked the chain shift of vowels commonly identified in the Southern US, and (5) they consistently demonstrated prominent rhoticized /r/ in their spontaneous speech. All Cantonese and English speakers were physically healthy and free of speech, language, or hearing problems. Listeners. Ten native American English adult speakers served as listeners in the present study. They were selected from the Northeast region of the US and judged as speaking the General American English by local speech-language pathologists practicing in North America. The English-speaking listeners were aged from 20–32 years old (average age¼24 years) and had no prior training of or exposure to Cantonese. The listeners had no reported history of hearing problems and were literate with at least college-level education. Speech materials The speech material used in the study was similar to that used by Chen et al. (2001). Sentence stress pattern was evaluated from participant’s production of the sentence ‘‘I bought a cat there’’. This sentence was produced in four distinct manners by placing the primary stress on one of four different words (stressed word is in italicized capital letters): (1) I bought a cat there; (2) I BOUGHT a cat there; (3) I bought a CAT there; and (4) I bought a cat THERE. Each sentence was produced three times and the order at which they were produced was randomized. A total of 12 English sentences were eventually produced by each participant, giving a grant total of 960 (80 speakers64 sentences63 trials) sentences being produced by all speakers. Audio recordings All recording took place in a sound-treated booth. Prior to the actual recording, the speakers were allowed to practice the speech materials as often as they needed. Only sentences that were fluently produced with no misarticulation and with the primary stress correctly placed were used. Investigators who were proficient in English monitored the entire recording procedure. The speakers were instructed to produce the sentences at a comfortable conversational loudness level and a normal speaking rate. Speech samples were recorded by using a
high-quality microphone (Shure, SM 58) and a preamplification system (M-Audio, MobilePre USB). During the recording, the microphone was placed *10 cm away from the speaker’s mouth. To calibrate for intensity, 10 data points of each of the three pure tones of different intensities (40 dBSPL, 50 dBSPL, and 60 dBSPL) were recorded, based on which a calibration equation was calculated using linear regression and it was used to calculate the actual intensity. The acoustic signals were digitized at 20 kHz and 16 bits/sample by using Praat and then stored in a computer for later analyses. Acoustic analysis During the acoustic analysis, the waveform of each recorded sentence was displayed and cycle-to-cycle F0, intensity, and vowel duration values were measured from the vocalic nucleus of each stressed and unstressed words. Due to the great variability of vowels (Fry, 1955), all measurements were made only from the vowel nucleus of the stressed targets. To measure vowel duration, cursors were manually placed at the first and the last visible glottal cycles, and vowel duration value was given by the time interval between the two markers. To trace F0 and intensity of the vowel, F0 and intensity measurements were obtained from three different locations (beginning, mid-point, and termination) of each vowel segment. To minimize the possible coarticulatory influence of the pre- and post-vocalic consonants on the vowel, the beginning of the vowel portion was defined as the first 50 ms of periodic activity following the third pulse. Similarly, vowel termination was defined as the last 50 ms of the periodic activity preceding the last three glottal pulses. The actual intensity level was calculated by using the intensity amplitudes and the calibration equation previously obtained. Two measurements for evaluating sentence stress were used: (1) average sentence stress and (2) acrosssentence stress. For average sentence stress, the average F0, intensity, and vowel duration values of the different stressed words were measured. However, stress is a relative concept with the stressed syllable more prominent than other syllables (Laver, 1994). Different stressed syllables may not demonstrate equal amounts of stressing (manifested by F0, intensity, and duration). For example, the stressed I in ‘‘I bought cat there’’ may not sound equally stressed as the stressed BOUGHT in ‘‘I BOUGHT cat there’’. Therefore, differentiated stress needs to be evaluated and it was expressed in terms of acrosssentence stress. Across-sentence stress was indicated by the average F0, intensity, and vowel duration values obtained from stressed syllable compared with unstressed syllable. For instance, according to Chen et al. (2001), for the across-sentence stress measurement, F0 of the word ‘‘BOUGHT’’ across the four sentences was calculated as the difference between
529
Proficiency in English sentence stress production the mean F0 of the stressed BOUGHT and the average mean F0 of the unstressed bought. DF0 ¼F0BOUGHT
ðF0boughtð1Þ þ F0boughtð2Þ þ F0boughtð3Þ Þ 3
The aforementioned measurements were made from all speech samples recorded, and data were then averaged for numerical analyses.
Int J Speech Lang Pathol Downloaded from informahealthcare.com by University of Hong Kong Libraries on 11/30/11 For personal use only.
Perceptual experiment The English sentences produced by Cantonese ESL speakers were used for the perceptual experiment. The experiment was carried out in a sound-treated room when speech samples were presented to the listeners via high-quality headphones at a comfortable loudness level (*70 dBSPL). Upon listening to each speech sample, the listeners were asked to judge for the placement of stress, degree of stress, and naturalness of the stressed syllable using a 10-point equal-interval scale, with a ‘‘1’’ indicating the worst and a ‘‘10’’ indicating the best production. Answer sheets with the rating scales printed were provided to the listeners. An inter-stimulus pause of 5 seconds was provided to allow sufficient time to complete the perceptual tasks. Prior to the experiment, the listeners were provided with brief practice exercises in order to acquaint themselves with the experimental procedure and setting.
acoustic measurement a second time. For intra-judge reliability, average absolute errors for F0, intensity, and vowel duration were 3.35 Hz, .81 dB, and 10.14 ms. Pearson correlation coefficients for F0, intensity, and vowel duration between the first and second measurements were .96, .82, and .96 (p 5 .01), respectively. Average absolute errors for F0, intensity, and vowel duration for inter-judge measurements were 4.20 Hz, 3.51 dB, and 18.05 ms. Pearson correlation coefficients for measurement of F0, intensity, and vowel duration between the two judges were .72, .80, and .88 (p 5 .01), respectively. To assess the reliability of listeners’ rating of naturalness of the stressed syllables, 10% of the speech samples were randomly selected and presented to the listeners a second time. The scores obtained from the first and second ratings were used to calculate for reliability of rating. Pearson correlation coefficients were .97 and .88 for intra- and interrater reliability, respectively. Statistical analysis To assess the acoustical differences between speech samples produced by Cantonese ESL and English speakers, a mixed-design factorial analysis was used for each measurement. Two-way repeated-measure Analyses of Variance (ANOVAs) with language group as the between-subjects variable and placement of stressed words (I, BOUGHT, CAT, THERE) as within-subjects variable were used for each gender group.
Reliability measures Because of human involvement in F0, duration, and intensity measurements, intra- and inter-judge reliability measures were obtained. Ten per cent of the sentence tokens (96 sentences) were randomly selected from the data corpus for assessment of intrajudge and inter-judge measurement reliability. The selected dataset included speech samples produced by both Cantonese and English male and female speakers. The 96 speech samples were measured by the same investigator a second time for intra-judge reliability measure. The speech samples were measured by a second investigator who was also experienced in
Results Average sentence stress The average F0, intensity, and vowel duration values of sentence stress produced by Cantonese and English speakers are shown in Tables I and II. F0 measurement Male speakers. As the 264 (speaker groups6 stressed words) two-way ANOVA revealed significant interaction effect between language group and
Table I. Mean (M), standard deviation (SD), and range of F0 (Hz), intensity (dB), and duration (ms) for each stressed word produced by Cantonese ESL (CE) and American-English (AE) male speakers. F0 (Hz) Speaker group
Stressed word
Mean (SD)
Intensity (dB) Range
Mean (SD)
Range
Duration (ms) Mean (SD)
Range
CE
I BOUGHT CAT THERE
141 161 153 127
(26) (27) (26) (25)
96–195 108–212 101–210 78–179
67 71 69 68
(4) (4) (5) (5)
56–79 59–97 56–90 53–80
321 141 129 242
(109) (31) (41) (50)
43–540 70–198 57–230 93–373
AE
I BOUGHT CAT THERE
139 143 161 133
(32) (24) (43) (23)
86–205 86–192 79–242 74–186
66 68 68 66
(5) (4) (5) (6)
49–74 50–75 47–75 46–73
139 143 161 133
(32) (24) (43) (23)
38–408 93–291 105–217 72–488
530
M. L. Ng and Y. Chen
Table II. Mean (M), standard deviation (SD), and range of F0 (Hz), intensity (dB), and duration (ms) for each stressed word produced by Cantonese ESL and American-English female speakers. F0 (Hz) Speaker group CE
Int J Speech Lang Pathol Downloaded from informahealthcare.com by University of Hong Kong Libraries on 11/30/11 For personal use only.
AE
Stressed word I BOUGHT CAT THERE I BOUGHT CAT THERE
Mean (SD) 252 292 231 219 225 234 189 207
(31) (35) (29) (23) (35) (44) (34) (36)
Intensity (dB) Range 187–331 165–373 174–360 110–261 148–303 148–332 134–491 73–285
stressed words [F(3, 102)¼5.821, p 5 .01], independent-sample t-test and one-way ANOVA were subsequently carried out. Results indicated no significant difference in F0 of the stressed words produced by male Cantonese ESL and male English speakers (p 4 .01). However, F0 of stressed BOUGHT was significantly higher (p 5 .01) and THERE was significantly lower (p 5 .01) when compared with all remaining three stressed words for male Cantonese ESL speakers, and significant differences were found only between stressed I and stressed CAT, and stressed THERE and CAT in male English speakers (p 5 .01). Female speakers. Similar to male speakers, significant interaction was found between language group and stressed word [F(3,126)¼9.39, p 5 .01]. The subsequent t-test and ANOVA indicated that Cantonese females exhibited a significantly higher F0 (261.78 Hz) than English females (227.81 Hz) in stressed word production (p 5 .01). All stressed words produced by female Cantonese ESL speakers exhibited significantly different F0 (p 5 .01) except between BOUGHT and CAT. The stressed THERE was produced with a significantly lower F0 than other stressed words. Moreover, female English speakers showed significantly lower F0 for the stressed THERE than the other three stressed words (p 5 .01).
Mean (SD) 67 72 68 66 70 71 68 69
(3) (7) (3) (5) (3) (2) (4) (4)
Range 57–82 59–98 55–92 51–121 60–76 59–98 53–76 50–74
Duration (ms) Mean (SD) 325 181 142 301 304 216 167 344
(87) (50) (40) (73) (77) (44) (27) (50)
Range 42–528 60–316 56–240 103–471 37–422 75–305 55–247 124–440
shorter duration than other stressed words (p 5 .01). Male English speakers exhibited significant differences in duration only between I and CAT, and CAT and THERE (p 5 .01). Female speakers. The two-way ANOVA revealed no significant interaction between language group and duration of stressed words [F(3, 126)¼3.812, p 4 .01], and no significant difference in vowel duration [F(1,42)¼2.72, p 4 .01]. However, a significant main effect for stressed word [F(3, 126)¼127, p 5 .01] was found between female Cantonese ESL and female English speakers. Duration measures of all four stressed words was significantly different than the others (p 5 .01), except between stressed I and THERE. I and THERE were significantly longer and CAT was significantly shorter than other stressed words (p 5 .01). Intensity measurement
Duration measurement
Male speakers. The two-way ANOVA showed no significant interaction effect and no significant main effect for language group [F(1, 34)¼1.20, p 4 .01]. However, a significant main effect was found for stressed words [F(3, 102)¼ 12.9, p 5 .01]. For both speaker groups, the stressed word I was associated with a significantly lower intensity and BOUGHT at a significantly greater intensity than the other three stressed words (p 5 .01).
Male speakers. There was a significant interaction effect between language group and stressed words [F(3, 102)¼11.33, p 5 .01]. Subsequent t-test and one-way ANOVA revealed no significant difference in duration of the stressed words produced by male Cantonese ESL and male English speakers (p 4 .01), and significant difference in duration of stressed words between duration of all stressed words except between BOUGHT and CAT produced by male Cantonese ESL speakers (p 5 .01). Cantonese males produced stressed I and THERE with a significantly longer duration than other stressed words (p 5 .01), whereas CAT was produced with a significantly
Female speakers. As a significant interaction effect between language group and stressed words in females was found [F(3, 63)¼11.9, p 5 .01], an independent-sample t-test and one-way ANOVA were carried out. Results indicated no significant difference in intensity of the stressed words produced by female Cantonese ESL and female English speakers (p 4 .01). However, the stressed word BOUGHT was produced with significantly greater intensity than the other three stressed words (p 5 .01) by female Cantonese ESL speakers, but not female English speakers.
Proficiency in English sentence stress production
Int J Speech Lang Pathol Downloaded from informahealthcare.com by University of Hong Kong Libraries on 11/30/11 For personal use only.
Across sentence stress. F0 difference Across-sentence F0 differences associated with male and female speakers are depicted in Figures 1(a) and (b), respectively. Regardless of gender and language group, stressed words were associated with higher relative F0 than unstressed words (as indicated by the positive values). Statistically, two 264 (speaker groups6stressed words) two-way ANOVAs indicated no significant interaction between language group and stressed word [males: F(3, 102)¼3.09, p 4 .01; females: F(3, 126)¼.09, p 4 .01], and no significant main effect for language group for both males and females [males: F(1, 34)¼.62, p 4 .01; females: F(1, 42)¼.258, p 4 .01]. However, a significant main effect was found for stressed word [males: F(3, 102)¼9.42, p 5 .01; females: F(3, 126)¼9.41, p 5 .01]. Consistent across genders, the stressed word CAT showed a significantly higher relative F0 than the other three words (p 5 .01).
531
gender. Two 264 (speaker groups6stressed words) two-way ANOVAs revealed similar findings for male and female speakers. No significant interaction was found between stressed word and language group [males: F(3, 126)¼1.04, p 4 .01; females: F(3, 126)¼1.04, p 4 .01], no significant main effect was found for language group [males: F(1, 34)¼1.88, p 4 .01; females: F(1, 42)¼2.57, p 4 .01]. However, a significant main effect was found for stressed word [males: F(3, 102)¼16.5, p 5 .01; females: F(3, 126)¼8.49, p 5 .01]. For both genders, the stressed word THERE was associated with significantly greater intensity than the other three words (p 5 .01).
Intensity difference The across-sentence intensity differences between the stressed and unstressed words for male and female speakers are shown in Figures 2(a) and (b), respectively. Similar to F0, stressed words were produced with greater intensity than unstressed words, regardless of language and
Duration difference Across-sentence duration differences between stressed and unstressed words produced by male and female speakers are shown in Figures 3(a) and (b), respectively. Similarly, for both males and females, both language groups differentiated stressed words from unstressed words by a longer duration. Consistent with the findings of F0 and intensity, male and female speakers showed similar patterns in across-sentence duration differences. According to the two 264 (speaker groups6 stressed words) two-way ANOVAs, no significant interaction effect was found between stressed word and language group for both males and females
Figure 1. Across-sentence F0 difference associated with stressed and unstressed words produced by (a) male and (b) female Cantonese ESL and English speakers.
Figure 2. Across-sentence intensity difference associated with stressed and unstressed words produced by (a) male and (b) female Cantonese ESL and English speakers.
532
M. L. Ng and Y. Chen
[males: F(3, 126)¼1.04, p 4 .01; females: F(3, 126)¼1.04, p 4 .01]. A significant main effect was found for stressed word [males: F(3, 102)¼62.87, p 5 .01; females: F(3, 126)¼92.59, p 5 .01], but not for language group [males: F(1, 34)¼3.85, p 4 .01; females: F(1, 42)¼0.003, p 4 .01]. The stressed word I had the longest duration (p 5 .01) and CAT the shortest duration (p 5 .01), when compared with other words.
Int J Speech Lang Pathol Downloaded from informahealthcare.com by University of Hong Kong Libraries on 11/30/11 For personal use only.
Perceptual experiment Placement of stress. Average percentage correct identification of stress placement is depicted in Table III. Results showed that the stressed word BOUGHT produced by Cantonese ESL speakers was associated with the highest rating in stress placement (90.6%), followed by THERE (90.3%), and then by CAT (82.0%). The stressed word I was associated with the lowest rating perceived stress placement (77.8%). Listeners tended to confuse it with BOUGHT (15.3%). Degree of stress. Average rating on the perceived degree of stress is represented in Table IV. Results showed that the listeners who were native speakers of
American English perceived the highest degree of stress for THERE (average rating¼6.48), followed by CAT (average rating¼6.18), then by BOUGHT (average rating¼6.05). The stressed I was rated with the lowest degree of stress (average rating¼5.87). Naturalness of stress. Table IV also depicts the average naturalness of stress rating for each stressed word. Results showed that the listeners perceived the stressed I as the most natural (average rating¼5.01), followed by BOUGHT (average rating¼4.61), then by CAT (average rating¼4.44). They perceived the stressed THERE with the least naturalness (average stress¼4.36). Discussion The present study objectively measured Cantonese ESL speakers’ accuracy in producing English sentence stress by using an array of acoustical parameters including F0, duration, and intensity. Fundamental frequency (F0) The present data reveal that, similar to native American English speakers, Cantonese ESL speakers were able to consistently produce stressed words using higher F0 than unstressed words, as evidenced by the positive across-sentence F0 difference. However, the amount of F0 increase used in stressed words demonstrated by female Cantonese ESL speakers exceeded that seen in female English speakers. This finding is similar to that reported by Chen et al. (2001); female Mandarin ESL speakers also showed higher across-sentence F0 than the female English speakers. Such F0 discrepancy may be related to the anthropomorphic differences between Cantonese ESL and English speakers, as supported by Yang (1996) in correlating F0 and formant values with size of laryngeal structures of Caucasians and Asians. However, data from our male Cantonese ESL speakers appear to contradict this suggestion. The lack of significant difference in average and acrosssentence F0 between our male Cantonese and English speakers fails to support this suggestion. The possible anthropomorphic difference between Asians and Caucasians may not be sufficient to explain the significantly higher average and acrosssentence F0 in female Cantonese ESL speakers.
Table III. Average percentage correct identification of stress placement by American-English speaking raters. Stress word presented Stress perceived
Figure 3. Across-sentence duration difference associated with stressed and unstressed words produced by (a) male and (b) female Cantonese ESL and English speakers.
I BOUGHT CAT THERE
I
BOUGHT
CAT
THERE
77.81 15.31 2.19 4.69
4.06 90.63 1.25 4.06
2.19 6.88 82.03 8.91
0.16 4.22 5.31 90.31
Proficiency in English sentence stress production Table IV. Average ratings of degree and naturalness of stress by American-English speaking raters (out of 10). Stressed word
Int J Speech Lang Pathol Downloaded from informahealthcare.com by University of Hong Kong Libraries on 11/30/11 For personal use only.
I BOUGHT CAT THERE
Average rating of degree of stress
Average rating of naturalness of stress
5.87 6.05 6.18 6.48
5.01 4.61 4.44 4.36
Instead, the F0 difference between Cantonese ESL and English speakers may be related to the fact that Cantonese is a tonal language. The higher F0 associated with Cantonese ESL speakers may have originated from the significantly greater F0 fluctuations at the syllable-level than in continuous speech of English (Eady, 1982). The lexical tones in L1 Cantonese may affect Cantonese ESL speakers in producing L2 English stress pattern, particularly in controlling F0 for stressed words. According to Xu (1999), the lexical tone of each Cantonese syllable is the main component determining the height and shape of the F0 contour for that syllable which, at the same time, also affects the adjacent syllables. Since sentence stress alters the F0 contour of the entire sentence, the actual F0 of a sentence is a combination of the tone-related local components and phrase-level long-term F0 variation (Li, Lee, & Qian, 2004; Xu, 1999). In studying Mandarin speakers who were speaking French as a second language, Shen (1990) noticed greater F0 fluctuation in French, possibly due to the interaction between sentence intonation and lexical tones, yielding an increased overall pitch level in L2. Similarly, Ma, Ciocca, and Whitehill (2004) noticed a marked interaction between tone contour and sentence intonation in Cantonese. Perhaps as a result of the interaction of F0 characteristics between a tonal L1 (Cantonese) and a non-tonal L2 (English), Cantonese ESL speakers unconsciously produced stressed words with higher F0 than American English speakers. Duration Our data also indicates that Cantonese ESL speakers differentiated stressed and unstressed English words by using vowel duration. Stressed words were significantly longer than unstressed words, as indicated by both average and across-sentence vowel duration measurements. This lengthening of stressed words by Cantonese ESL speakers is consistent with native English speakers, as well as Mandarin ESL speakers previously reported in the literature (Anderson-Hsieh & Venkatagiri, 1994; Chen et al., 2001). This indicates that Cantonese ESL speakers were making use of duration to contrast stressed and unstressed words in a way comparable to native English speakers. However, when examining the durational characteristics of unstressed syllables, although not
533
statistically significant, Cantonese ESL speakers generally produced unstressed words with a shorter duration than English speakers. This deviation in duration was also seen in Mandarin ESL speakers reported by Chen et al. (2001). Such difference in duration may be attributable to the syllable-timed nature of the L1. As explained by Chen et al., being a syllable-timed language, Mandarin has a relatively regular interval between stressed and unstressed syllables (Roach, 1991), and constant syllable duration across a sentence (Chun, 1982). Accordingly, speakers of a syllable-timed language may find it difficult to manipulate syllable duration to conform to an English sentence stress pattern. This suggestion can be applied to Cantonese ESL speakers, as Cantonese is also a syllable-timed language. Cantonese ESL speakers tended to produce unstressed words with a shorter duration than native English speakers, possibly because of the syllable-times nature of L1 Cantonese. Intensity The lack of significant difference in intensity associated with stressed words between Cantonese ESL and native English speakers indicated that Cantonese ESL speakers were able to differentiate stressed and unstressed words proficiently, in a way similar to English speakers. A similar finding was also found in Mandarin ESL speakers (Chen et al., 2001). It appears that manipulation of intensity to signal stress is easy for non-native speakers of English to learn and grasp. It follows that, among the three acoustical parameters, control of intensity to signal stress might be the first one to be acquired by nonnative speakers. Apparently, this is only a speculation. More longitudinal data is needed regarding which acoustic cue is acquired first by ESL speakers. Based on the above discussion, it is seen that our Cantonese ESL speakers were proficient in achieving English sentence stress, probably more than the Mandarin ESL speakers reported in Chen et al. (2001) as our Cantonese ESL speakers exhibited a comparable durational contrast for English stress than Mandarin ESL speakers. Such difference may be related to the more stringent selection criteria for ESL speakers. In the present study, Cantonese ESL speakers must have achieved at least a grade C in the HKCEE or a D in the HKAL or an A in the speaking section in the University Entrance Test for Speaker of Other Languages. With such requirements, the Cantonese participants can be considered as highly proficient English speakers. As English and Cantonese (Chinese) are the official languages used in Hong Kong, most Cantonese ESL speakers begin attending English classes since kindergarten. The extensive contact with English in Hong Kong and participants’ lengthy ESL learning experience may have contributed to the proficiency in sentence stress pattern produced by Cantonese ESL speakers.
534
M. L. Ng and Y. Chen
Int J Speech Lang Pathol Downloaded from informahealthcare.com by University of Hong Kong Libraries on 11/30/11 For personal use only.
Perceptual ratings of sentence stress produced by Cantonese ESL speakers As indicated by the perceptual rating by native American English speakers, the stressed BOUGHT was judged as the most accurate and the stressed I the least accurate. This is in line with most of the acoustical findings. For example, female Cantonese ESL speakers produced the stressed I with a significantly lower F0 than the stressed BOUGHT and CAT, while they produced the stressed BOUGHT with a significantly higher F0 than the stressed I and THERE. The male Cantonese ESL speakers produced the stressed I with a significantly lower F0 than BOUGHT and CAT, and the stressed BOUGHT with a significantly higher F0 than all other stressed words. In addition, both male and female Cantonese ESL speakers produced the stressed I with a significantly longer duration than the stressed BOUGHT and CAT, while they produced the stressed BOUGHT shorter than the stressed I and THERE. They also produced the stressed BOUGHT louder than other three stressed words. Based on the above discussion, with the combined effects of F0 and intensity, it was not surprising that the native American English raters judged the placement of stressed I with the lowest rating, but BOUGHT with the highest rating. This also agrees with the ratings of degree of stress; listeners judged the stressed I to have the least stress while the stressed BOUGHT the most stress. This might be due to the fact that the word BOUGHT starts with a plosive that is produced with a burst of energy and followed by voicing. The high intra-oral pressure requirement for the word initial plosive may have facilitated stressing the following vowel with higher F0 and greater intensity. However, findings in duration appear to contradict with the average perceptual stress rating. This might indicate that the combined effects of F0 and intensity may be greater than the effect of duration alone in sentence stress. In contrasting stress, the contribution of F0 and intensity combined may be more than duration alone. Yet, this is only a suggested explanation for the contradictory findings for F0, intensity, and duration. Again, further studies are needed to investigate the relative importance of each acoustic cue on the perception of sentence stress in English. It can be seen that there exists great similarity in the way Cantonese and Mandarin ESL speakers produced English sentence stress: both speaker groups were able to contrast stressed and unstressed English words by manipulating F0, duration, and intensity. Even the deviations in stress production were similar: both Cantonese and Mandarin speakers produced shorter unstressed words than English speakers. This is likely to be related to the similarity between Cantonese and Mandarin.
Conclusion With the continuing globalization and internationalization nowadays, being able to speak proficient English is becoming more and more important for daily life communication worldwide including Hong Kong. It is argued that Cantonese ESL speakers would encounter difficulties in speaking everyday English as a second language. Chan and Li (2000) stated that one of the pronunciation problems found in Chinese ESL speakers is to be able to produce English with more natural suprasegmental characteristics including rate of speech, stress, juncture, and intonation. The present results indicate that, due to the influence of Cantonese, Cantonese ESL speakers have some problems in producing English sentence stress. In order to attain more natural English stress, it is suggested that Cantonese ESL speakers learn to control F0, duration, and intensity of stressed and unstressed words. In particular, they should focus more on the increased F0 when producing the stressed words, especially for female Cantonese ESL speakers.
References Anderson-Hsieh, J., & Venkatagiri, H. (1994). Syllable duration and pausing in the speech of Chinese ESL speakers. TESOL Quarterly, 28, 807–812. Chan, A., & Li, D. (2000). English and Cantonese phonology in contrast: Explaining Cantonese ESL learners’ English pronunciation problems. Language, Culture and Curriculum, 13, 1. Chen, Y. (2006). Production of tense-lax contrast by Mandarin speakers of English. Folia Phoniatrica et Logopaedica, 58, 240– 249. Chen, Y., Robb, M. P., Gilbert, H. R., & Lerman, J. W. (2001). A study of sentence stress production in Mandarin speakers of American English. Journal of the Acoustical Society of America, 109, 1681–1690. Cheng, C. (1968). English stress and Chinese tones in Chinese sentences. Phonetica, 18, 77–88. Chun, D. (1982). A contrastive study of the suprasegmental pitch in modern German, American English and Mandarin Chinese. Unpublished doctoral dissertation, University of California, Berkeley. Cooper, W., & Sorensen, J. (1981). Fundamental frequency in sentence production. New York, NY: Springer-Verlag. Cooper, W., Eady, S., & Mueller, P. (1985). Acoustical aspects of contrastive stress in question-answer contexts. Journal of the Acoustical Society of America, 77, 2142–2156. Eady, S. (1982). Differences in the F0 patterns of speech: Tone language versus stress language. Language and Speech, 25, 29– 42. Flege, J. E. (1980). Phonetic approximation in second language acquisition. Language and Learning, 30, 117–134. Flege, J. E., & Davidian, R. (1985). Transfer and developmental processes in adult foreign language speech production. Journal of Applied Psycholinguistics, 5, 323–347. Fry, D. (1955). Duration and intensity as physical correlates of linguistic stress. Journal of Acoustical Society of America, 27, 765–768. HKEAA. (2008a). International Recognition - NARIC (2008). Retrieved from Hong Kong Examinations and Assessment Authority website. Available online at: http://www.hkeaa.edu.hk/en/ir/ Standards_of_HKEAA_qualifications/NARIC/, accessed January 24, 2011.
Int J Speech Lang Pathol Downloaded from informahealthcare.com by University of Hong Kong Libraries on 11/30/11 For personal use only.
Proficiency in English sentence stress production HKEAA. (2008b). International Recognition - IELTS (2004). Retrieved from Hong Kong Examinations and Assessment Authority website. Available online at: http://www.hkeaa.edu. hk/en/ir/Standards_of_HKEAA_qualifications/IELTS/, accessed January 24, 2011. HKEAA. (2008c). International Recognition - IGCSE (2007). Retrieved from Hong Kong Examinations and Assessment Authority website. Available online at: http://www.hkeaa.edu.hk/en/ir/ Standards_of_HKEAA_qualifications/IGCSE/, accessed January 24, 2011. Laver, J. (1994). Principles of phonetics. Cambridge: Cambridge University Press. Li, Y., Lee, T., & Qian, Y. (2004). F0 analysis and modeling for Cantonese text-to-speech. Proceedings of the ISCSLP 2002, 127–130. Ma, J., Ciocca, V., & Whitehill, T. (2004). The effects of intonation patterns on lexical tone production in Cantonese. Proceedings of International Symposium on Tonal Aspects of Languages: With Emphasis on Tone Languages (TAL-2004), 133–136. Major, R. C. (2001). Foreign accent: The ontogeny and phylogeny of second language phonology. Mahwah, NJ: Lawrence Erlbaum Associates.
535
Ng, M., Chen, Y., & Sadaka, J. (2008). Vowel features in Turkish accented English. International Journal of Speech-Language Pathology, 10, 404–413. Raphael, L. J., Borden, G. J., & Harris, K. S. (2007). Speech science primer: Physiology, acoustics and perception of speech (5th ed.). Philadelphia, PA: Lippincott Williams & Wilkins. Roach, P. (1991). English phonetics and phonology: A practical course (2nd ed.). Cambridge: Cambridge University Press. Shen, X. (1990). Ability of learning the prosody of an intonational language by speakers of a tonal language: Chinese speakers learning French prosody. International Review of Applied Linguistics in Language Teaching, 18, 139–152. Tarone, E. (1980). Some influences on the syllable structure of interlanguage phonology. International Review of Applied Linguistics in Language and Teaching, 18, 139–152. Thompson, T., & Gaddes, M. (2005). The importance of teaching pronunciation to adult learners. Asian EFL Journal. Professional Teaching Articles Collection 2005, 179–188. Xu, Y. (1999). Effects of tone and focus on the formation and alignment of F0 contours. Journal of Phonetics, 27, 55–105. Yang, B. (1996). A comparative study of American English and Korean vowels produced by male and female speakers. Journal of Phonetics, 24, 245–261.
