International Journal of Speech-Language Pathology

ISSN: 1754-9507 (Print) 1754-9515 (Online) Journal homepage: http://www.tandfonline.com/loi/iasl20

Cantonese tone production performance of mainstream school children with hearing impairment Karen K. L. Cheung, Ada H. Y. Lau, Joffee H. S. Lam & Kathy Y. S. Lee To cite this article: Karen K. L. Cheung, Ada H. Y. Lau, Joffee H. S. Lam & Kathy Y. S. Lee (2014) Cantonese tone production performance of mainstream school children with hearing impairment, International Journal of Speech-Language Pathology, 16:6, 624-636 To link to this article: http://dx.doi.org/10.3109/17549507.2014.896942

Published online: 27 Mar 2014.

Submit your article to this journal

Article views: 100

View related articles

View Crossmark data

Full Terms & Conditions of access and use can be found at http://www.tandfonline.com/action/journalInformation?journalCode=iasl20 Download by: [University of Idaho]

Date: 05 November 2015, At: 18:12

International Journal of Speech-Language Pathology, 2014; 16(6): 624–636

Cantonese tone production performance of mainstream school children with hearing impairment

KAREN K. L. CHEUNG1,2, ADA H. Y. LAU3, JOFFEE H. S. LAM4 & KATHY Y. S. LEE4 1Department

Downloaded by [University of Idaho] at 18:12 05 November 2015

of Chinese and Bilingual Studies, Faculty of Humanities, The Hong Kong Polytechnic University, Hong Kong, 2Jockey Club Sign Bilingual and Co-enrolment in Deaf Education Programme, Centre for Sign Linguistics and Deaf Studies, The Chinese University of Hong Kong, Hong Kong, 3Division of Speech and Hearing Sciences, Faculty of Education, The University of Hong Kong, Hong Kong, and 4Department of Otorhinolaryngology, Head and Neck Surgery, The Chinese University of Hong Kong, Hong Kong

Abstract This study investigated the Cantonese tone production ability of children with hearing impairment studying in mainstream schools. The participants were 87 Cantonese-speaking children with mild-to-profound degrees of hearing loss aged 5.92–13.58 in Hong Kong. Most of the children were fitted with hearing aids (n ⫽ 65); 17 of them had profound hearing impairment, one who had severe hearing loss had cochlear implantation, and four who had mild hearing loss were without any hearing device. The Hong Kong Cantonese Articulation Test was administered, and the tones produced were rated by two of the authors and a speech-language pathologist. Group effects of tones, hearing loss level, and also an interaction of the two were found to be significant. The children with profound hearing impairment performed significantly worse than most of the other children. Tone 1 was produced most accurately, whereas tone 6 productions were the poorest. No relationship was found between the number of years of mainstreaming and tone production ability. Tone production error pattern revealed that confusion patterns in tone perception coincided with those in production. Tones having a similar fundamental frequency (F0) at the onset also posed difficulty in tone production for children with hearing impairment.

Keywords: Cantonese, tone production, hearing loss, mainstreaming.

Introduction Cantonese tonal system and tone studies Cantonese is a tone language (Matthews & Yip, 1994). It has six distinctive tones with different fundamental frequency (F0) patterns; they differ in F0 height, contour and direction (Gandour, 1981; Matthews & Yip, 1994). Lee, van Hasselt, Chiu, and Cheung (2002b) provided categorization of the six Cantonese tones. Tone 1, tone 3, and tone 6 are level tones, whereas tone 2, tone 4, and tone 5 are contour tones. Tone 1 is a high-level tone, whereas tone 3 is mid-level, and tone 6 is low-level. Tone 2 is a highrising tone, whereas tone 5 is low-rising. Tone 4 is a low falling tone (see Figure 1). Tone signals lexical information and, thus, satisfactory tone perception and production ability are essential for effective communication (Yip, 2002). Despite its importance, tone has not been widely studied and the majority of studies investigated the aspect of tone perception but not production. Moreover, the participants were

usually typically-hearing or children with profound hearing impairment. In view of this, the tone production ability of children with milder degrees of hearing loss is investigated in this study. Cantonese tone perception studies on the typically-hearing population Studies of Cantonese tone perception performance have commonly revealed that tone contrasts involving tone 1 are perceived with ease (Barry, Blamey, Martin, Lee, Tang, Yuen, et al., 2002; Ciocca & Lui, 2003; Lee, Chiu, & van Hasselt, 2002a; Lee et al., 2002b). Barry et al. (2002) proposed that pitch height was crucial in the perception of Cantonese tones, thus the distinctively high average F0 of tone 1 facilitated its perception. On the contrary, small F0 differences would lead to tone discrimination difficulty (Barry et al., 2002; Ciocca & Lui, 2003; Lee et al., 2002a, b). Studies investigating tone contrast perception by the typically-hearing population

Correspondence: Karen K. L. Cheung, FG701, Department of Chinese and Bilingual Studies, Faculty of Humanities, The Hong Kong Polytechnic University, Hung Hom, Hong Kong. E-mail: [email protected] ISSN 1754-9507 print/ISSN 1754-9515 online © 2014 The Speech Pathology Association of Australia Limited Published by Informa UK, Ltd. DOI: 10.3109/17549507.2014.896942

Cantonese tone production

625

Downloaded by [University of Idaho] at 18:12 05 November 2015

Figure 1. Different fundamental frequency of the six Cantonese tones on the vowel [a] (Lee et al., 2002).

have revealed that tone 2/tone 4, tone 2/tone 5, tone 3/tone 6, tone 4/tone 5, and tone 5/tone 6 are perceived least well (Barry et al., 2002; Ciocca & Lui, 2003; Lee et al., 2002a, b). These studies congruently speculated that the close proximity of F0 at the onset between the two tones posed the greatest difficulty for tone discrimination (see Figure 1). A similar contour but small differences in the F0 would also lead to discrimination difficulty, as in the tone 3/tone 6 contrast (Ciocca & Lui, 2003). Cantonese tone perception studies on the population with hearing impairment Similar results were shown in studies on the population with hearing impairment, which found tone 6 to be most difficult to identify (Ching, 1988; Wong & Wong, 2004), and tone 5 contrasts to be most difficult for both pre-lingually deafened children and post-lingually deafened adults with cochlear implants (Barry et al., 2002; Lee, Cheung, Chan, & van Hasselt, 1997). Studies also revealed that the discrimination of tone 2/tone 5, tone 4/tone 6, tone 4/ tone 5, tone 5/tone 6, and tone 3/tone 6 contrasts were performed least well by the children with cochlear implants (Barry et al., 2002; Ciocca, Francis, Aisha, & Wong, 2002; Lee, van Hasselt, & Tong, 2010b; Tse & So, 2012; Wong & Wong, 2004). Lee et al. (2010b) also studied tone discrimination by children with profound hearing impairment fitted with hearing aids and they found that tone 2/tone 5, tone 4/tone 6, and tone 2/tone 4 contrasts were performed least well. Lee et al. (2010b) indicated that F0 similarity at the onset accounted for the perception problem observed in these children with cochlear implants or hearing aids. The difficulty in distinguishing tone 3 and tone 6 was explained by Tse and So (2012), who revealed that difficulty in disguising tones was based only on tone height. Lee et al. (2010b) also suggested that the children with moderate hearing loss were able to perceive tones, as were the typically-hearing children after the use of a hearing device, but not for those with profound hearing loss. These studies indicated that tone

perception by children with hearing loss in general conformed to that of the hearing population at large, and errors were usually made with tones other than tone 1. Ciocca et al. (2002) suggested that the larger separation of F0 between tone 1 and the other tones helped its identification (see Figure 1). They also found that tone 1 was generally produced with a higher overall amplitude level, which might be one of the cues children with hearing impairment exploited for identification and discrimination. However, that does not guarantee that the perception of tone 1 is error-free. Confusion between contour and level tones, including tone 1, was also found in the following studies: Lee et al. (2002b) found the tone 1/tone 2 contrast to be most difficult for children with cochlear implants; Wong and Wong (2004) found that the tone 1/tone 5 contrast was one of the most problematic among cochlear implant receivers; Tse and So (2012) revealed that tone 2/tone 6 and tone 3/tone 5 differentiation was only slightly above chance level. Fok (1974) addressed the confusion between contour and level tones as the inability to correctly perceive contour movement, and this could explain the confusion between level and contour tones contrasts by the cochlear implantees, which was observed in these studies. In sum, tone confusion errors by children with hearing loss were essentially due to their insensitivity to the minute average F0 differences or changes between two tones. Cantonese tone production studies on the typically-hearing population Relatively fewer studies have investigated Cantonese tone production, but their results echo those of tone perception. Tone production case studies revealed that tone 4, tone 5, and tone 6 are differentiated in the later stage of acquisition for typically-hearing children (Tse, 1978; Tse, 1992). Combined with results from other studies with a much larger sample size, tone 1 is consistently reported as the earliest acquired tone (So & Dodd, 1995; To, Cheung, & McLeod, 2013). The possible explanation may arise from the Cantonese phonology. The single and most discussed phonological rule of Cantonese is the

Downloaded by [University of Idaho] at 18:12 05 November 2015

626

K. K. L. Cheung et al.

[⫹ upper, high] tone change at offset due to an underlying [⫹ high] floating tone attached to the syllables (Yip, 1980). In particular, tone 1 is produced in the same manner even after tone change because it is essentially high in upper register all the way from onset to offset (Yip, 1980). Considering its importance in Cantonese phonology, it seems logical that tone 1 should emerge earlier than other tones. In the study by Cheung and Abberton (2000), only one out of 251 typically-hearing children with phonological disorders had a tone production problem by realizing both rising tones (tone 2 and tone 5) as the mid-level tone (tone 3). Moreover, rising tones appeared to be difficult for a very small group of children to acquire naturally (Cheung & Abberton, 2000). Agreeing with both Tse (1978) and Tse (1992), tone 5 was found to be acquired at the latest stage of tone acquisition (Cheung & Abberton, 2000). As with Cheung and Abberton (2000),Yip (2002) found that the production of rising contours was often delayed until late in the first year; this is because a poor control of the laryngeal muscles by infants will lead to a tendency of pitch falling in an utterance. Muscular effort is required to overcome this falling tendency and to sustain a rise (Yip, 2002). Cantonese tone production studies on the population with hearing impairment Two studies by Lee, Tong, and van Hasselt (2007) and Lee, van Hasselt, and Tong (2010a) investigated the tone production of children with cochlear implants and found that the production of tone 4 and tone 5 were most difficult. They suggested a critical age of 4-years-old for cochlear implantation in order to achieve a satisfactory tone production performance. Optimally, children should receive a cochlear implant before 2-years-old, so that they can potentially achieve ˜ 80% accuracy in tone production within one year of starting to use the implant (Lee et al., 2007, 2010a). Compared with typicallyhearing children who are able to master all tones correctly at 2-years-old, children with cochlear implants continue to make errors (Lee et al., 2010a). Khouw and Ciocca (2006) also examined the tone production of adolescents with profound hearing loss fitted with hearing aids, and found that their tone production was in general indistinguishable by average F0 and F0 change in contrast to those by typically-hearing individuals. Because Cantonese relies heavily on F0 contour and height for tonal identification (Fok, 1974; Vance, 1976), hearing listeners of that study generally misclassified the various tones they produced as tone 1, showing that the tones they produced matched the F0 features of tone 1 (i.e., high in average F0 and lack of F0 change). Acoustically, production by adolescents with hearing impairment showed little difference for the six contrastive tones; they had a smaller range of average F0 and tended to have greater variability with regards

to producing F0 change and average F0 of target tones (Khouw & Ciocca, 2006). Relationship between tone perception and production The similarities in the findings between tone perception studies and tone production studies, as well as the tendency for the population with hearing impairment to perceive and produce some of the tones better, led to the speculation that tone perception and production may somehow be related (Xu, Chen, Lu, Zhou, Wang, Liu, et al., 2011). The two ways to improve tone production are by improving tone perception ability and lengthening tone immersion. Improving tone perception ability. Better tone perception ability may be achieved by wearing a hearing device. Significant improvement has been reported for children with hearing impairment fitted with hearing aids and cochlear implants on Cantonese tone perception compared to the unaided conditions (Lee, van Hasselt, & Tong, 2008; Lee et al., 2010b). However, consistent reports have indicated that the current development of hearing devices is not sufficient for tone perception (Barry et al., 2002; Ciocca et al., 2002; Lee et al., 2010b; Wong & Wong, 2004). More specifically, it has been pointed out that the mere amplification of sound signals by hearing aids was less effective in facilitating tone perception for children with profound hearing loss (Lee et al., 2008). In addition, it has been suggested that the persistent difficulties in perceiving pitch in speech by children with cochlear implants could be accounted for by the limitation of the implants in resolving lownumbered harmonics of complex sounds (Ciocca et al., 2002; Wong & Wong, 2004). Instead, these children may have to rely on other (but weaker) cues in periodicity and amplitude information to decrypt average F0 and F0 change (Ciocca et al., 2002; Wong & Wong, 2004). Tse and So (2012) also suggested that the tone awareness abilities of children with hearing impairment were restricted by their auditory perception through cochlear implants, and implantation did not considerably help them in detecting tones. More importantly, as Lee et al. (2010b) and Law and So (2006) indicated, no statistical difference is found between hearing aids amplification and cochlear implantation in the ability of tone differentiation and production. Alternatively, milder degrees of hearing loss may have better tone production ability, yet a lack of investigation in Cantonese prevents us from confirming such an assumption. Lengthening tone immersion though mainstreaming. The length of immersion in the target tones may also play a role in tone production ability, and this can be achieved by the duration of mainstreaming. According to Wamae and Kang’ethe-Kamau (2004), mainstreaming represents “the process of educating the deaf not within the artificial confines of an

Downloaded by [University of Idaho] at 18:12 05 November 2015

Cantonese tone production

institution but within the more natural structure of the public school system” (p. 33). The assimilation of students with hearing impairment in schools should begin early because they have to function in a hearing world (Wamae & Kang’ethe-Kamau, 2004). Among the studies on inclusive education for the hearing-impaired, the results were inconclusive and none of them has ever investigated the speech production ability of Mandarin- or Cantonesespeaking children with hearing impairment, not to mention the effect of mainstreaming on Cantonese tone production ability. Tobey, Geers, Brenner, Altuna, and Gabbert (2003) studied the development of speech production skills in English-speaking children with cochlear implants and the results revealed that children studying in mainstream classrooms, where the auditory–oral mode of communication was primarily used, achieved higher speech production scores, thus suggesting that mainstreaming had a positive effect on the speech production ability of children with hearing impairment. Another study by Most (2007) revealed similar results but in terms of speech intelligibility. Hebrew-speaking children in special classes within regular schools received a significantly lower average speech intelligibility scores than did children who were integrated individually into regular classes. However, Most (2007) also remarked that such a difference in speech intelligibility might not be the direct result of the difference in educational setting. Besides, the communication strategies used, that is, either oral or total communication, were found to produce no significant difference on children’s speech production and vocabulary development post-implantation (Connor, Hieber, Arts, & Zwolan, 2000). The divergence in previous studies may stem from the sample selection, for students enrolling in mainstream schools are usually those who have sufficient hearing to rely heavily on oral communication and still their success varies (Stinson & Antia, 1999). In any case, in the above studies the target languages were not tonal and, thus, the role of mainstreaming on the tone production ability of children with hearing impairment still remains unknown.

627

production performance. Finally, the present study also investigated whether any error pattern could be generated from the produced tones by children with various degrees of hearing loss.

Method Participants A total of 87 children (45 boys and 42 girls) with at least 25 decibels (dB) hearing loss at three frequency averaging of pure-tone (PTA) thresholds (500, 1000, and 2000 Hz) of the better ear were recruited from mainstream primary schools in Hong Kong. They were classified into five hearing loss groups: mild (26–40 dB); moderate (41–55 dB); moderate–severe (56–70 dB); severe (71–90 dB); and profound (⬎ 90 dB) according to the PTA thresholds of their better ear. In addition to hearing loss, no other medical condition affecting their speech and language development was reported. None had an intellectual deficit with an intelligence quotient of less than 70. All children were native Cantonese speakers and were studying in primary schools which used Cantonese as the medium of instruction. Written consent of the principle caregivers was given. These children ranged in age from 5.92–13.58, with a mean of 9.43 and standard deviation of 1.95. Before entering mainstream primary schools in Hong Kong, the children had had 0–3 years of mainstream experience in kindergartens. Therefore, combining that with their mainstream years in primary school, the duration of mainstream experience of these children ranged from 1–9 years (mean ⫽ 5.85, SD ⫽ 2.37). The demographics of the children by different hearing loss level, including their age, gender, school grade, and years of mainstream experience, are summarized in Table I. The current study assumed the result of Lee et al. (2010b) and considered that hearingassistive devices had a null effect. Even so, almost all children with cochlear implants were in the profound hearing loss group, and one was in the severe loss group (see Table I).

Materials Aim and hypothesis In order to aim at uncovering the reasons leading to the difficulty of tone production in children with hearing impairment, the present study investigated three areas: the intrinsic difference between the six Cantonese tones; the degree of hearing loss; and the role of mainstreaming. Based on findings on tone perception and production, it was hypothesized that tone 1 would be performed better than tones 4, 5, and 6. A milder degree of hearing loss would also lead to better tone performance because of relatively better tone perception ability. Longer exposure to mainstreaming may also lead to a better tone

The Hong Kong Cantonese Articulation Test (HKCAT) was adopted in the present study (Cheung, Ng, & To, 2006). This articulation test is primarily used for testing children’s production of 69 syllables in Cantonese. A total of 41 picture stimuli are presented to induce children to produce the target words, and test administrators are required to record the production for each syllable in terms of consonants, vowels (using the International Phonetic Alphabet), and tones. For instance, a picture of a car is presented and the child is asked to name what it is, then the child produces [tshε] (i.e., car in Cantonese), and, if correct, the test administrator will record

628

K. K. L. Cheung et al.

Table I. Demographics of participants according to different hearing loss level. Hearing loss level Demographics n Mean age in years (range) Gender Mean school grade (range) Mean year of mainstreaming (range) Mean age at diagnosis in months (range) Mean age at amplification in months (range) Hearing device used (%) Mean PTA at better ear in dB (range)

Mild

Moderate

Moderate–severe

Severe

Profound

18 10.03 (7.00–12.42) M: 11 4.3 (2–6) 7 (5–9)

18 8.91 (5.92–11.50) M: 9 3.3 (1–6) 5.9 (1–9)

15 9.88 (6.92–13.58) M: 5 3.8 (1–6) 6.8 (4–9)

14 9.27 (6.75–12.42) M: 7 3.4 (1–6) 5.5 (1–9)

22 9.13 (6.58–13.50) M: 13 3 (1–6) 4.4 (1–9)

3.34 (.0–7.00)

2.51 (.0–6.50)

3.18 (.17–7.00)

2.05 (.0–5.50)

.88 (.0–2.25)

4.90 (2.50–8.00)

4.13 (.67–7.17)

4.29 (1.50–9.00)

2.60 (.58–5.67)

1.73 (.0.83–5.00)

n/a: 4 (22.2%) 34.2 (23–42)

HA: 18 (100%) 48.5 (42–57)

HA: 15 (100%) 62.5 (50–85)

HA: 13 (92.9%) 79 (68–90)

HA: 5 (22.7%) 100.8 (90–118)

Downloaded by [University of Idaho] at 18:12 05 November 2015

Note: PTA refers to pure-tone average of hearing sensitivity at 500 Hz, 1000 Hz, and 2000 Hz.

the produced consonant as [tsh], vowel as [ε], and tone as 1. The HKCAT is a standardized test based on results obtained from 1726 children with their ages ranging between 2.33–11.58, as well as from 112 adults. This test was chosen for its proven validity and reliability. It has high content validity because it covers all six Cantonese tones. The test consists of 25 tone 1 characters, 13 tone 2 characters, six tone 3 characters, 14 tone 4 characters, two tone 5 characters, and nine tone 6 characters. The test possesses satisfactory discriminant validity in differentiating the groups of children with and without speech sound disorders. For the tone production measures, the test is able to differentiate children from different age groups as well as from different genders. However, since the present study targets the tone production of children with hearing loss, the results were based entirely on data relating to tones.

Procedures The HKCAT was administered to all 87 participants by either of the two Cantonese-speaking speechlanguage pathologists of the Centre of Sign Linguistics and Deaf Studies in a soundproof or segregated room in a school setting. All participants wore their hearing device during the test. The noise level of the room varied from 45–60 dB. The internal microphone of the Samsung YP-U3 recorder was placed 30–40 cm away from the participant’s mouth. The test administration exactly followed the standard procedure of the HKCAT. Afterwards, the produced tones were rated by three raters by listening twice to the recordings using headphones through a computer in a quiet office, with an interval of 3 months between the two ratings. In the second rating, only five-to-six participants from each hearing loss group were randomly selected for the purpose of examining the intra-rater reliability, and they constituted more than 30% of the whole dataset. The three raters were all native Cantonese speakers; they consisted of one

of the two speech-language pathologists who had administered the test to the participants and two authors with 3 years of research experience with children with hearing impairment. Both authors had previous phonetic training and had received a bachelor degree in linguistics. Each rater rated the incorrectly produced tones as 0 and the correct tones as 1. So, combining the ratings of three raters, each item can be scored from 0–3. This score is the tone production accuracy score for each produced word. The percentage of agreement and the intraclass correlation (ICC) were employed for examining the inter- and intra-rater reliabilities (Portney & Watkins, 2009).

Statistical analyses A series of descriptive statistics and univariate analyses were firstly used to depict the pattern of the data for providing an overview. The major interest was to investigate the relationship between the tone production accuracy of children with hearing impairment and the three predictors which may affect the accuracy. These predictors were the participants’ hearing loss level, their number of years in mainstreaming, and the tone of each word that they produced. The unit of the analysis was the tone production accuracy score of each word produced by the participants. The level of statistical significance was chosen at .05, and SPSS 15.0 was used throughout the statistical analyses. The study design required the 87 participants to produce the same set of 69 words of six tones. In other words, the tone production accuracy was measured for words that were nested within subjects. Instead of a single-level structure, the data structure in this study was organized at two levels (i.e., word and subject), known as a hierarchical data structure. In view of the present hierarchical data structure, the multi-level linear modelling (MLM) technique was used. An advantage of MLM is that it does not require

Downloaded by [University of Idaho] at 18:12 05 November 2015

Cantonese tone production

the assumption of data independence (Tabachnick & Fidell, 2007). Another critical advantage of MLM is that the relationship between response and predictors is allowed to vary across participants. This property is crucial in handling the present dataset; this is because variations in tone production ability among children with hearing impairment are substantial, even when they share common characteristics such as having the same hearing loss level. The present study applied a 2-level random intercepts and slopes model (which is a type of MLM) to study the relationship between tone production accuracy and the predictors. The three predictors in the analysis were organized at two levels according to their nature. The predictor that characterized the word (Tones) was organized at the word level. The subject-level predictors that characterized the participants were their hearing loss level (HL level) and their number of years in a mainstreaming environment (Years of mainstreaming). In the model of this study two levels of equations were used. The wordlevel equation depicted the relationship between tone production accuracy and the tone of the word. The subject-level equations studied how the intercept and slope parameters at word level were influenced by the subject-level predictors, which were the HL level and Years of mainstreaming. All the main effects and the interaction of the predictors were included in the model for detecting statistical significance. All significant effects would be further investigated individually by using the Kruskal-Wallis test and the Mann-Whitney test with Bonferroni correction. Finally, statistical analysis was also carried out to test the tone error pattern based on the entire sample and the children from each hearing loss level. A nonparametric Chi-square test was employed to examine whether or not a target tone error was evenly distributed on the other five tone errors.

Descriptive statistics The descriptive statistics of each predictor are presented in Table II. At the word level, a total of 6003 words consisting of six tones were produced by 87 participants, each participant producing 69 words. Among these, 36.2% of words belonged to tone 1 and 2.9% belonged to tone 5. A non-parametric Chi-square test showed a significant difference between the proportions of words of the six tones (χ2 ⫽ 27.609, df ⫽ 5, p ⬍ .001). A non-parametric Chi-square test showed that there is no statistical significance in the differences between the number of participants in different hearing loss groups (χ2 ⫽ 2.253, df ⫽ 4, p ⫽ .689). Regarding the years of mainstreaming of the participants, the numbers varied from 1–9 years with a median of 6 years and an interquartile range of 4 years. A non-parametric Kruskal-Wallis test showed that the tone production accuracy of six tones is significantly different (χ2 ⫽ 83.575, df ⫽ 5, p ⬍ .001). The accuracy of production of tone 1 was the highest (mean ⫽ 2.93; SD ⫽ .36), followed by tone 2 (mean ⫽ 2.85; SD ⫽ .53), tone 3 (mean ⫽ 2.84; SD ⫽ .54), tone 5 (mean ⫽ 2.84; SD ⫽ .52), tone 4 (mean ⫽ 2.82; SD ⫽ .62), and lastly tone 6 was the lowest (mean ⫽ 2.81; SD ⫽ .57). In Table III, two sets of mean accuracy scores, one at word level and one at subject level, are presented. The Kruskal-Wallis test illustrated that the differences in accuracy scores between children with various degrees of hearing loss are statistically significant both at the word level (χ2 ⫽ 503.561, df ⫽ 4, p ⬍ .001) and the subject level (χ2 ⫽ 31.102, df ⫽ 4, p ⬍.001). As

Table II. Descriptive statistics of predictors at word and subject levels. Variable

Results Rater reliability The accuracy of tone production was rated by a speech-language pathologist and two of the authors. In order to investigate the between-rater and within-rater variation among the ratings, inter-rater and intra-rater reliability was examined by the percentage of agreement and the coefficient of ICC. For the intra-rater reliability of the three raters, the percentages of agreement of the two times of ratings from each of the three raters ranged from 95.5–98.1%. Besides, the ICCs have a range of values from .95–.99. As for the inter-rater reliability, the percentage of agreement between the three raters reached 92.9% and an ICC value of .984. The high values of percentage of agreement (exceeding 70%) (Miles & Huberman, 1994) and the ICCs (exceeding .9) (Portney & Watkins, 2009) evidenced that the raters’ ratings are reliable.

629

Level 1 – Word level (n ⫽ 6003) 1. Tone Tone 1 Tone 2 Tone 3 Tone 4 Tone 5 Tone 6 Totala Level 2 – Subject level (n ⫽ 87) 2. Year of mainstreaming 3. Hearing loss level Mild Moderate Moderate–severe Severe Profound Totalb a Total

Count

%

1131 522 1218 174 783 6003

18.8 8.7 20.3 2.9 13.0 100.0

Min Max Median IQR

1.0

18 18 15 14 22 87

9.0

6.0

20.7 20.7 17.2 16.1 25.3 100.0

number of words produced by all 87 participants. number of participants. IQR, interquartile range. b Total

4.0

630

K. K. L. Cheung et al.

Table III. Tone production accuracy by hearing loss group. Word level (Level 1) (n ⫽ 6003)

Subject level (Level 2) (n ⫽ 87)

Hearing loss group

Mean (SD)

Mean (SD)

Mild Moderate Moderate–severe Severe Profound

2.99 2.94 2.92 2.95 2.62

(.15) (.35) (.36) (.26) (.83)

n 1242 1242 1035 966 1518

2.99 2.94 2.92 2.95 2.62

(.02) (.11) (.08) (.04) (.44)

n 18 18 15 14 22

Downloaded by [University of Idaho] at 18:12 05 November 2015

Note: Tone production accuracy score ranged from .0–3.0.

expected, the participants with mild hearing loss had the highest accuracy in tone production (mean ⫽ 2.99 on both levels) and the profound hearing loss participants had the lowest (mean ⫽ 2.62 on both levels). It is notable that the group with severe hearing loss had higher and less dispersed tone production accuracy relative to the moderate and moderate–severe hearing loss groups. The correlation coefficients between tone production accuracy with years of mainstreaming showed that the number of years of mainstreaming of the participants had a significant correlation with the tone production accuracy at the word level (r ⫽ .034, p ⫽ .009) but not at the subject level (r ⫽ .064, p ⫽ .554). Because the number of years of mainstreaming is a subject-level predictor, its significant correlation coefficient with tone production accuracy at the word level had been seriously exaggerated. The difference in correlations can be treated as evidence that the use of a single-level statistical test may produce biased results.

Tests of main and interaction effects of predictors To test the effects of the predictors of the two levels on tone production accuracy, the random intercepts and slopes model was used, and the results are presented in Table IV. All main effects of the predictors and their cross-level 2-way and 3-way interaction effects were included in the model. The result showed that the main effects of the Tone (F(5, 421.48) ⫽ 5.928, p ⬍ .001) and the Hearing loss group (F(4, 92.64) ⫽ 14.463, p ⬍ .001) were statistically significant predictors on tone production

accuracy. It was found that the cross-level interaction Tone*HL group also contributed a significant amount of explanation power (F(20, 421.47) ⫽ 2.784, p ⬍ .001). The main effect of Year of mainstreaming did not statistically significantly predict the tone production accuracy (F(1, 92.64) ⫽ .448, p ⫽ .505). This result was consistent with the insignificant bivariate correlation (r ⫽ .064, p ⫽ .554) between Year of mainstreaming and the Tone production accuracy score at subject level. In summary, the results of MLM revealed that the tone of the word from the word level, the hearing loss group of the participants from the subject level, and their cross-level interaction were statistically significant predictors of the tone production accuracy of children with hearing impairment. Further investigation on significant effects To further investigate how the significant effects elicited in MLMs affected the hearing-impaired participants’ tone production accuracy, a series of post-hoc comparisons was performed. Figure 2 shows the average tone production accuracy score at the word level by various tones. As aforementioned, an overall statistically significant difference was found among different tones by using the Kruskal-Wallis test (χ2 ⫽ 83.575, df ⫽ 5, p ⬍ .001). The post-hoc multiple comparisons by the Mann-Whitney U-test with Bonferroni adjustment discovered that participants’ accuracy of producing tone 1 words was significantly higher than for all other tones, with p-values of less than .001. Other than those, no other pair of tones showed a difference. The average accuracy scores of various hearing loss groups are presented in Figure 3 and Table III. Significant differences in tone production accuracy at the subject level between hearing loss groups were revealed (χ2 ⫽ 31.102, df ⫽ 4, p ⬍ .001). The post-hoc multiple comparisons elicited that the accuracy of the mild hearing loss group was significantly higher than moderate–severe (p ⬍ .001), severe (p ⫽ .002), and profound (p ⬍ .001) groups. Besides, the accuracy of the profound hearing loss group is significantly lower than the moderate hearing loss group (p ⬍ .001) and the severe hearing loss group (p ⫽ .005).

Table IV. A random intercepts and slopes model for predicting tone production accuracy of participants with hearing impairment. Predictors (Level)

F-value Numerator df Denominator df p-value

Tone (L1) 5.928 HL group (L2) 14.463 Year of MainS (L2) .448 2.784 Tone (L1) * HL group (L2) .274 Tone (L1) * Yr of MainS (L2) 1.825 Yr of MainS (L2) * HL group (L2) Tone (L1) * Yr of MainS (L2) * HL group (L2) 1.377

5 4 1 20 5 4 20

Note: L1 and L2 denote word level and subject level predictors, respectively.

421.467 92.635 92.635 421.467 421.467 92.635 421.467

⬍ .001 ⬍ .001 .505 ⬍ .001 .927 .131 .128

Cantonese tone production

631

Downloaded by [University of Idaho] at 18:12 05 November 2015

Figure 2. Tone production accuracy on word level by tone.

The results of the interaction effect of Tone* Hearing loss group are presented in Figure 4 and Table V. For groups with mild and moderate hearing loss, the average production accuracy between different tones showed no statistical difference. In the moderate–severe and the severe hearing loss groups, the accuracy of tone 1 is significantly higher than that of tone 6. In the profound group, the production accuracy of tone 1 words is significantly higher than the other five tones. Tests of tone error patterns Overall tone error pattern. Production errors, as perceived by listeners, were analysed to reveal significant tone confusion patterns by children with hearing loss. A non-parametric Chi-square test performed on all hearing loss groups using the number of errors as the dependent variable showed that some of the tones had a significantly higher tendency to be produced as another tone. Table VI shows the frequency of tone production errors for all children with hearing loss for each tone and the Chi-square test results. There was a significantly higher tendency

Figure 4. Tone production accuracy by tone on each hearing loss group. MS denotes moderate-severe hearing loss group.

for tone 1 tokens to be produced as tone 3 if errors were made, and vice versa. There was also a significantly higher tendency for tone 2 tokens to be produced as tone 4 if errors were made, and vice versa. For tone 5, errors were made on tone 2, and the confusion is significant. Finally, for tone 6, a significant proportion of errors were made by children who produced it as tone 1. Tone error pattern in each hearing loss group. The production errors perceived were also analysed by each hearing loss group to examine whether significant tone error patterns could be found by using a nonparametric Chi-square test (with Bonferroni’s correction of test significant level p ⬍ .01). From the mild-to-severe hearing loss groups, no distinctive tone production error patterns were found, with an exception in tone 1 of the moderate hearing loss Table V. Post-hoc multiple comparisons on the Tone * Hearing Loss group interaction effect at word level.

Hearing loss group Mild (n ⫽ 1242) Moderate (n ⫽ 1242) Moderate–Severe (n ⫽ 1035) Severe (n ⫽ 966) Profound (n ⫽ 1518)

Figure 3. Tone production accuracy on subject level by hearing loss level. MS denotes moderate-severe hearing loss group.

Tone pairs showing statistically significant difference – – Tone 1 ⬎Tone 6 (p ⬍ .0006) Tone 1 ⬎Tone 6 (p ⬍ .0006) Tone 1 ⬎Tone 2 (p ⬍ .0006); Tone 1 ⬎Tone 3 (p ⬍ .0006); Tone 1 ⬎Tone 4 (p ⬍ .0006); Tone 1 ⬎Tone 5 (p ⬍ .0006); Tone 1 ⬎Tone 6 (p ⬍ .0006)

Note: With Bonferroni correction, p ⬍ .05/(5*15) ⫽ .000,667 is considered as statistical significance. n denotes the total number of words produced by children in the hearing loss group. ⬎ denotes statistically significantly more accurate than.

632

K. K. L. Cheung et al. Table VI. Summary of tone production errors of children with various degrees of hearing loss (n ⫽ 87). Produced tones Target tones Tone Tone Tone Tone Tone Tone

1 2 3 4 5 6

(n ⫽ 2175) (n ⫽ 1131) (n ⫽ 522) (n ⫽ 1218) (n ⫽ 174) (n ⫽ 783)

T1

16 30 30 2 55

T2

T3

T4

T5

T6

χ2

df

p

22

42 13

5 46 11

9 18 3 28

24 36 5 18 7

41.627 32.279 38.000 35.273 9.600 70.283

4 4 4 4 4 4

.000** .000** .000** .000** .048* .000**

11 64 10 7

47 4 40

2 16

9

Downloaded by [University of Idaho] at 18:12 05 November 2015

Note: Chi-square test significant level: * p ⬍ .05. ** p ⬍ .01.

group (p ⫽ .004). Tone 1 of the moderate hearing loss group followed the error pattern of the overall pattern: a significant proportion of errors of tone 1 were confused as tone 3. For the profound hearing loss group, the error patterns also follow the overall pattern in general, with the exception of tone 5, which did not show any significant pattern. Instead, it is equally possible for tone 5 to be incorrectly produced as tone 2 or tone 6 by the profound hearing loss group. This absence of a unique error pattern could be accounted for by the relatively small number of erred tokens (22 in total) of tone 5. In sum, the profound hearing loss group tended to confuse the tone 1/tone 3 contrast and the tone 2/tone 4 contrast. The children in this group also displayed a significant number of errors in producing tone 6 as tone 1.

Discussion Relationship between hearing loss level and tone accuracy The ability to produce tones accurately by children with a mild degree of hearing loss was found to be significantly better than those with moderate–severe, severe, and profound hearing loss. It was suggested that children with a mild-to-moderate degree of hearing loss were able to produce tones with more consistent F0 information and distinguish between the six Cantonese tones better than children with profound hearing loss (Khouw & Ciocca, 2006). One of the possible reasons may lie in the relationship between tone perception and tone production. It is reasonable to speculate that the chances of children with less difficulty perceiving F0 information to produce tones more accurately will be higher than those who have greater difficulty. Studies on Mandarin tone production by American second-language learners (Wang, Jongman, & Sereno, 2003) and children with hearing impairment (Peng, Tomblin, Cheung, Lin, & Wang, 2004; Xu et al., 2011) have established a strong correlation between tone perception and tone production. Xu et al. (2011) found that a good performance in tone identification may be a necessary condition for tone production performance. Such a relationship in Cantonese has not been stud-

ied at present, but assuming an association between tone perception and tone production exists also in Cantonese, analyses on tone perception and tone production performance between different degrees of hearing loss should yield a similar result. Indeed, children with a moderate degree of hearing loss were found to perceive different tones better than those with a greater degree of hearing loss (Lee et al., 2008). However, an investigation of the relationship between tone perception and production is needed before conclusive observations can be made. One may also argue that children in the mild hearing loss group performed better because they were older and had received a longer duration of mainstreaming (Table I), but this argument does not hold in the present case because there is no significant group difference in age (p ⫽ .354) nor in the duration of mainstreaming (p ⫽ .093). The tone production performance by children with profound hearing loss was significantly worse than those in the mild, moderate, and severe groups. It should be noted that 17 out of 18 children with cochlear implants belonged to the profound group. Such performance by children with profound hearing loss may indicate the limitation of the hearingassistive device with regard to tone production. It has been shown that children with profound hearing loss did not greatly benefit from hearing aids (Lee et al., 2008) or from cochlear implants on tone perception (Lee et al., 2010b; Tse & So, 2012); this is because of the inability of the implants to resolve the lowfrequency harmonics by the relatively wide bandpass filters (Ciocca et al., 2002). While it has been suggested that the duration of implant experience has a positive impact on tone production ability for them, implanted children continue to make errors (Lee et al., 2010a). Once again, the relationship between tone perception and production may play some role. The tone perception difficulty arising from a device limitation may lead to imprecise tone production (Lee et al., 2008). Children with profound hearing loss having cochlear implants were also found to display acoustically similar F0 information across different tones (Khouw & Ciocca, 2006). F0 information is the foremost perceptual cue in Cantonese. However, if the F0 cues presented by the hearing device

Downloaded by [University of Idaho] at 18:12 05 November 2015

Cantonese tone production

may turn out to be unreliable for children with profound hearing loss, their low tone accuracy may not be so surprising. Acoustic analysis as further investigation is called for to shed light on this issue. Generally speaking, the remarkable drop in accuracy from the severe group to the profound group suggests that the benefit of hearing devices for children with profound hearing impairment is minimal, or at least less successful than for children with a lesser degree of hearing loss. On the other hand, statistical analysis indicates that the tone production ability of children with moderate–severe hearing loss is not significantly different from those in the profound group. The reason is not entirely clear, but possible reasons could be the relatively older age when this group of children were diagnosed with hearing problems and when they first received amplification, compared with children with moderate, severe, and profound degrees of hearing loss (Table I). Alternatively, it is not that the moderate–severe group performs particularly poorly but that the severe group performs unexpectedly well. Some of the possible reasons could be the earlier age of diagnosis and amplification compared with the children with a milder degree of hearing loss (Table I). There are also factors that cannot be identified in the present study; this group of children might, for instance, have received more special training focusing on tones. Another possibility could be the high individual variation in the profound group where one individual could score up to 99.5% accuracy while another could score only 53.6%, with both of them having cochlear implants and both having experienced amplification before the age of 3 (the child who scored 53.6% was even older than the one scoring 99.5% by 3 years and had received amplification 4 months earlier). This large variation may be due to the variability of the effectiveness of cochlear implantation, which has been suggested in many studies (e.g., Pisoni, Cleary, Geers, & Tobey, 1999; Spencer, 2004; Szagun, 2008; etc.). Nonetheless, the underlying factors, especially those in the pre-implant period, which contribute to this variability are as yet unknown. Anyhow, it is important to note that the actual tone production score of the profound group is on average lower than that of the moderate–severe group, but that it just does not reach a significant difference. Effect of mainstreaming on tone accuracy The number of years of mainstreaming did not have a significant effect on the tone production accuracy of the children with hearing impairment in the present study. The tone production accuracy of these children did not improve with an increasing number of years studying in a mainstream classroom. Tobey et al. (2003) suggested that mainstreaming had a positive effect on the speech production performance of children with hearing impairment. However, no such

633

advantage is observed in the present study, and this agrees with Connor et al. (2000). One reason may arise from the target language studied and the focus here. Whereas previous studies were concerned only with articulations, vocabularies or speech intelligibility in English, the current study focused on the suprasegmental dimension of a tonal language, Cantonese. The results of the current study may reflect that the oral immersion for children with hearing impairment may be advantageous for a non-tonal language or for the segmental dimension of a language only. Another reason may likely be as proposed by Connor et al. (2000), that is, teaching strategy may be a less important factor in the speech production ability of children with hearing impairment than the age of implantation, or than other factors, such as home language, amplification mode, and age at amplification, etc. Cross-linguistic studies of tone perception also suggest that linguistic experience plays an important role (Leather, 1987; Stagray & Downs, 1993, as cited in Francis, Ciocca, Ma, & Fenn, 2008; Wang, Spence, Jongman, & Sereno, 1999). An analysis of these factors on tone production accuracy is beyond the scope of this study, and future research in these areas would shed light on the possible ways to improve the tone production ability of children with hearing impairment. Nonetheless, it would not be sensible to completely disregard the effect of mainstreaming; this is because the current study did not exhaust all the potential factors relating to tone production performance on which mainstreaming might depend. Effect of intrinsic characteristic of Cantonese tones on tone accuracy The results of our study demonstrate that tone has a significant main effect and a significant interaction effect with hearing loss level on tone production accuracy. The intrinsic differences of the six tones are influencing the tone production accuracy of children with hearing impairment. Among the six tones, tone 1 was produced significantly better than all other tones. Taking into account the hearing loss level, tone 1 was produced significantly better than all other tones in the group of children with profound hearing loss, a result that aligns with Khouw and Ciocca (2006). Tone 1 definitely possesses some unique characteristics, which make it distinct from other tones, and these characteristics are able to explain its outstanding production performance. Regarding the ease of production, level tones are relatively easier to produce than contour tones, as the pitch level is constant over a period of time and does not require the need to vary the tension of the laryngeal muscles during the production (Yip, 2002), and, therefore, the absence of such muscular effort may pose less difficulty for children. Nevertheless, the production of the other two level tones, tone 3 and tone 6, was not as good as the production of tone 1. As discussed,

Downloaded by [University of Idaho] at 18:12 05 November 2015

634

K. K. L. Cheung et al.

tone 1 is produced the earliest in children, and this is likely to be the result of the children’s acquisition of Cantonese phonology. The same should hold true for Cantonese-speaking children with hearing impairment. Also, it is already known that both the type and token frequency of level tones in Cantonese significantly outnumber those of contour tones (Lee, 2012), so we would expect children to produce level tones more accurately due to a higher amount of exposure and practice. The relative ease of perception of tone 1 may also be associated with its production. According to the Tonal Sonority Hierarchy proposed by Jiang-King (1999), a high tone is more sonorant than a low tone, and sonority is defined by the loudness of a sound relative to that of other sounds with the same length, stress, and pitch (Ladefoged, 2001, p. 227). As shown in previous studies, tone 1 is the only tone in Cantonese with a generally higher overall amplitude than the other tones (Barry et al., 2002; Ciocca et al., 2002; Fok, 1974). The ease in perceiving tone 1 is supported by the study of Barry et al. (2002), which reported that both hearing-impaired and typically-hearing children were able to successfully discriminate most tone contrasts involving tone 1. Therefore, it may also be argued that the ability in tone perception is likely to affect the tone production performance (Wang et al., 2003; Xu et al., 2011). Another reason may be due to the physical make-up of a child’s vocal tract and larynx height. We know that children have a shorter larynx height and vocal tract compared to adults, and, since a close correlation has been found between vocal tract morphology and speech acoustics (Baer, Gore, Gracco, & Nye, 1991), it seems self-evident that children generally exhibit a higher pitch than adults. The high accuracy of tone 1 and the lower accuracy of tone 3 and especially of tone 6 may also be accounted for by such, but may be less likely since the listeners had years of experience working with children with hearing impairment. The results from the univariate analysis of tone production accuracy by tone revealed that tone 6 tokens were perceived as the least accurately produced, which concurs with previous studies on children with hearing impairment (Ching, 1988; Wong & Wong, 2004). More specifically, for children with moderate–severe to profound degrees of hearing loss, tone 6 was produced significantly worse than tone 1. While both tone 1 and tone 6 are level tones and level tones in theory should be relatively easier to produce (Yip, 2002), there are considerable differences between the two. Unlike tone 1, tone 6 has a relatively small average F0 separation with other low tones (i.e., tone 3, tone 4, and tone 5) (Ciocca et al., 2002; Lee et al., 2010b). On top of that is the large variability of tone production with regards to F0 change and average F0 by children with hearing impairment (Khouw & Ciocca, 2006). Since these children are unable to produce reliably the subtle F0 change and average F0, which are precisely the

important perceptual cues for Cantonese tones, miscategorization of their production is not infrequent. Another possible explanation stems from the difficulty in perception due to the intrinsic feature of tone 6. That may also imply that children are still in the process of developing a full phonological inventory in Cantonese. While it may be true that level tones are easier to produce than contour tones, children have to learn to contrast individual level tones by differentiating average F0 ranges and in theory that may depend on the number of contrasts they have to face. Since there are three level tones in Cantonese in comparison to two rising tones and one falling tone, more effort is required for the successful perception of level tones, especially when the target tone 6 shares the low pitch characteristic with three other tones. Perceived tone confusion pattern Tone errors can be categorized as pitch level confusion and contour confusion. Children with hearing impairment in general produced a level tone incorrectly as another level tone or a contour tone as another contour tone. Pitch level confusion refers to the tone confusion between tone 1, tone 3, and tone 6. Tone 3 and tone 6 were significantly confused as tone 1, whereas tone 1 was significantly confused as tone 3. This confusion pattern among level tones is incongruent with the study by Khouw and Ciocca (2006) on adolescents with hearing impairment. Such confusion shows that children with hearing impairment may have difficulty in producing distinguishable average F0 range for successful perception. This may also imply that these children generally produce a higher-than-average F0 range for level tones that correspond to the intrinsic F0 feature of tone 1. Yet again, this result suggests that a mere perceptual study of the tone production by children with hearing impairment is insufficient and that an acoustic analysis is called for. On the other hand, pitch contour confusions match also with tone perception studies, suggesting that, at least for contour tones, the difficulty in tone perception may also affect tone production accuracy (Lee et al., 2008). The current study showed that tone 2 erred tokens were significantly produced as tone 4 and vice versa. Similar findings were found in studies on typically-hearing children (Lee et al., 2002a, b) and on children with hearing impairment (Lee et al., 2010b). Onset fundamental frequency differs by only 1.36 Hz between tone 2 and tone 4, but they share the largest difference in offset (97.6 Hz) (Lee et al., 2010b). In addition, tone 5 production errors were significantly perceived as tone 2 productions. This is also consistent with studies both on the normal population and on the clinical population (Barry et al., 2002; Ciocca & Lui, 2003; Ciocca et al., 2002). The onset F0 is also small (7.46 Hz) and it differs only slightly in the offset (28.92 Hz) (Lee et al., 2010b). The present result thus pro-

Downloaded by [University of Idaho] at 18:12 05 November 2015

Cantonese tone production

vides more substantial evidence to suggest that similar pitch onsets between tones pose a greater obstacle for children with hearing impairment to discriminate. Their difficulty in producing reliable offset difference may arise from either a lack of fine control of laryngeal muscles to produce contrastive F0 changes in contour tones or a difficulty in perceiving tones with a similar onset (Lee et al., 2002a, b). Further, the fact that the tone confusion patterns are in accordance with those found in the hearing population may suggest that acquisition of tone for the current group of students conforms to the normal development in spite of their slower pace. In contrast, the absence of an error pattern for tone 5 in the profound hearing loss group may be due to a bias in item allocation. The number of tokens of the six tones in the test material is unequal. There are only two tokens of tone 5 words in the whole set of materials, whereas there are 25 tone 1 words. In spite of such limitations, the use of a picture-naming task is believed to be a more appropriate choice for children with hearing loss than using sets of monosyllabic consonant–vowel word cards with all six tones for inducing production since it minimizes the influence of word knowledge and provides a more natural context to children (Khouw & Ciocca, 2006). Similar to the two production studies by Lee et al. (2007, 2010a), the number of tokens for each Cantonese tone in a picture-naming task is more difficult to control since word difficulty and imageability have to be considered also for young children and the frequency of occurrence for Cantonese level tones is in reality much greater than for contour tones (Lee, 2012). Finally, the results of the present study are generated from test materials at the single word level only. The tone production ability of children with hearing impairment in connected speech is still unknown. Designing test stimuli in sentences or connected speech with similar numbers of tokens among the six tones is suggested for future studies. Conclusion The present study found the tone production ability of children with mild-to-severe degrees of hearing loss to be satisfactory. Children with mild and moderate degrees of hearing loss significantly outperformed the children with higher degrees of hearing loss. Tone remains a challenging aspect of speech production by profoundly hearing-impaired children, which may be indicative of the present development of cochlear implants being non-effective in tone perception and distinction. Besides, the results reflect the inadequacy of merely mainstreaming the children with hearing impairment in normal schools in improving their tone production accuracy. The tone production ability of children with hearing impairment was not found to increase with the number of years studying in a mainstream environment. The intrinsic differences of the six tones were

635

also found to be affecting the children’s tone production accuracy. Tone 1 production was the best while tone 6 production was the poorest. An analysis on the tone error patterns suggested that tone confusion patterns in past perception studies coincided with the production error patterns in this study. Moreover, a similar F0 of tone pairs during onset caused confusion for children with hearing impairment. This study points out that tone production is a multi-faceted aspect of speech production and teaching strategy alone is insufficient in explaining such ability in children with hearing impairment. Continuous investigation of other variables that may affect the tone production accuracy of children with hearing impairment is required.

Acknowledgements The authors would like to express their sincere gratitude to the Hong Kong Jockey Club Charity Trust for supporting the research project. We are grateful to all the participating children and schools for their support and co-operation. We would also like to thank Ms Tammy Lau for data collection as well as Ms Emily Lam for data collection and rating of data. Last but not least we would like to thank Dr David Wilmshurst and his team for editing the manuscript.

Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.

References Baer, T., Gore, J. C., Gracco, L. C., & Nye, R. W. (1991). Analysis of vocal tract shape and dimensions using magnetic resonance imaging: Vowels. Journal of the Acoustical Society of America, 90, 799–828. Barry, J. G., Blamey, P. J., Martin, L. F. A., Lee, K. Y. S., Tang, T., Yuen, Y. M., et al. (2002). Tone discrimination in Cantonese-speaking children using a cochlear implant. Clinical Linguistics and Phonetics, 16, 79–99. Cheung, P., & Abberton, E. (2000). Patterns of phonological disability in Cantonese-speaking children in Hong Kong. International Journal of Language and Communication Disorders, 35, 451–473. Cheung, P., Ng, K. H., & To, C. (2006). Hong Kong Cantonese Articulation Test. Hong Kong: Language Information Sciences Research Centre and City University of Hong Kong. Ching, Y. C. T. (1988). Lexical tone perception by Cantonese deaf children. In I. M. Liu, M. J. Chen, & H. C. Chen (Eds.), Cognitive aspects of the Chinese language, vol. 1 (pp. 93–102). Hong Kong: Asian Research Service. Ciocca, V., & Lui, J. Y. K. (2003). The development of the perception of Cantonese lexical tones. Journal of Multilingual Communication Disorders, 1, 141–147. Ciocca, V., Francis, A. L., Aisha, R., & Wong, L. (2002). The perception of Cantonese lexical tones by early-deafened cochlear implantees. Journal of Acoustical Society of America, 111, 2250–2256. Connor, C. M., Hieber, S., Arts, H. A., & Zwolan, T. A. (2000). Speech, vocabulary, and the education of children using

Downloaded by [University of Idaho] at 18:12 05 November 2015

636

K. K. L. Cheung et al.

cochlear implants: Oral or total communication? Journal of Speech, Language, and Hearing Research, 43, 1185–1204. Fok, C. Y. Y. (1974). A perceptual study of tones in Cantonese. Hong Kong: University of Hong Kong, Centre for Asian Studies. Francis, A. L., Ciocca, V., Ma, L., & Fenn, K. (2008). Perceptual learning of Cantonese lexical tones by tone and non-tone language learners. Journal of Phonetics, 36, 268–294. Gandour, J. (1981). Perceptual dimensions of tones: Evidence from Cantonese. Journal of Chinese Linguistics, 9, 20–36. Jiang-King, P. (1999). Tone-vowel interaction in optimality theory. München, Germany: Lincom Europa. Khouw, E., & Ciocca, V. (2006). Acoustic and perceptual study of Cantonese tones produced by profoundly hearing-impaired adolescents. Ear and Hearing, 27, 243–255. Ladefoged, P. (2001). A course in phonetics (4th ed.). Fort Worth, TX: Harcourt College Publishers. Law, Z. W. Y., & So, L. K. H. (2006). Phonological abilities of hearing-impaired Cantonese- speaking children with cochlear implants of hearing aids. Journal of Speech, Language, and Hearing Research, 49, 1342–1353. Leather, J. (1987). F0 pattern inference in the perceptual acquisition of second language tone. In A. James, & J. Leathers (Eds.), Sound patterns in second language acquisition (pp. 59–81). Dordrecht, Netherlands: Foris Publications. Lee, J. L. (2012). Fixed-tone reduplication in Cantonese. McGill Working Papers in Linguistics, 22, 1–16. Lee, K.Y. S., Cheung, D. M. C., Chan, B.Y. T., & van Hasselt, C. A. (1997). Cochlear implantation: Implications for tone language. In I. Honjo, & H. Takahashi (Eds.), Cochlear implant and related sciences update (pp. 254–257). Basel, Germany: Karger. Lee, K. Y. S., Chiu, S. N., & van Hasselt, C. A. (2002a). Tone perception ability of Cantonese-speaking children. Language and Speech, 45, 387–406. Lee, K. Y. S., Tong, M. C. F., & van Hasselt, C. A. (2007). The tone production performance of children receiving cochlear implants at different ages. Ear and Hearing, 28, 34S–37S. Lee, K. Y. S., van Hasselt, C. A., & Tong, M. C. F. (2008). Tone perception in Cantonese- speaking children with hearing aids. Annals of Otology, Rhinology and Laryngology, 117, 313–316. Lee, K. Y. S., van Hasselt, C. A., & Tong, M. C. F. (2010a). Age sensitivity in the acquisition of lexical tone production: An evidence from children with profound congenital hearing impairment after cochlear implantation. Annals of Otology, Rhinology and Laryngology, 119, 258–265. Lee, K. Y. S., van Hasselt, C. A., & Tong, M. C. F. (2010b). Lexical tone perception ability of profoundly hearing-impaired children: Performance of cochlear implant and hearing aid users. Otology & Neurotology, 31, 1079–1087. Lee, K. Y. S., van Hasselt, C. A., Chiu, S. N., & Cheung, D. M. C. (2002b). Cantonese tone perception ability of cochlear implant children in comparison with normal-hearing children. International Journal of Pediatric Otolaryngology, 63, 137–147. Matthews, S., & Yip, V. (Eds.). (1994). Cantonese: A comprehensive grammar. London, UK: Routledge. Miles, M. B., & Huberman, A. M. (Eds.). (1994). Qualitative data analysis: An expanded sourcebook (2nd ed.). Thousand Oaks, CA: Sage. Most, T. (2007). Speech intelligibility, loneliness, and sense of coherence among deaf and hard-of-hearing children in individual inclusion and group inclusion. Journal of Deaf Studies and Deaf Education, 12, 495–503. Peng, S. C., Tomblin, J. B., Cheung, H., Lin, Y. S., & Wang, L. S. (2004). Perception and production of Mandarin tones in prelingually deaf children with cochlear implants. Ear and Hearing, 25, 251–264. Pisoni, D. B., Cleary, M., Geers, A. E., & Tobey, E. A. (1999). Individual differences in effectiveness of cochlear implants in

children who are prelingually deaf: New process measures of performance. Volta Review, 101, 111–164. Portney, L. G., & Watkins, M. P. (Eds.). (2009). Foundation of clinical research: Applications to practice (3rd ed.). Upper Saddle River, NJ: Pearson/Prentice Hall. So, L. K. H., & Dodd, B. J. (1995). The acquisition of phonology by Cantonese-speaking children. Journal of Child Language, 22, 473–495. Spencer, P. E. (2004). Individual differences in language performance after cochlear implantation at one to three years of age: Child, family, and linguistic factors. Journal of Deaf Studies and Deaf Education, 9, 395–412. Stagray, J. R., & Downs, D. (1993). Differential sensitivity for frequency among speakers of a tone and a nontone language. Journal of Chinese Linguistics, 21, 143–163. Stinson, M. S., & Antia, S. D. (1999). Considerations in educating deaf and hard-of-hearing students in inclusive settings. Journal of Deaf Studies and Deaf Education, 4, 164–175. Szagun, G. (2008). The younger the better? Variability in language development of young German-speaking children with cochlear implants. In T. Marinis, A. Papangeli, & V. Stojanovik (Eds.), Proceedings of the Child Language Seminar 2007 – 30th Anniversary (pp. 183–194). Reading, UK: University of Reading Press. Tabachnick, B. G., & Fidell, L. S. (2007). Using multivariate statistics (5th ed.). New York: Allyn and Bacon. To, C. K. S., Cheung, P. S. P., & McLeod, S. (2013). A population study of children’s acquisition of Hong Kong Cantonese consonants, vowels, and tones. Journal of Speech, Language, and Hearing Research, 56, 103–122. Tobey, E. A., Geers, A. E., Brenner, C., Altuna, D., & Gabbert, G. (2003). Factors associated with development of speech production skills in children implanted by age five. Ear and Hearing, 24, 36S–45S. Tse, A. (1992). The Acquisition Process of Cantonese Phonology: A Case Study. Unpublished masters thesis. The University of Hong Kong, Hong Kong SAR. Tse, J. K. P. (1978). Tone acquisition in Cantonese: A longitudinal study. Journal of Child Language, 5, 191–204. Tse, W. T., & So, L. K. H. (2012). Phonological awareness of Cantonese-speaking pre-school children with cochlear implants. International Journal of Speech-Language Pathology, 14, 73–83. Vance, T. J. (1976). An experimental investigation of tone and intonation in Cantonese. Phonetica, 33, 368–392. Wamae, G. M. I., & Kang’ethe-Kamau, R. W. (2004). The concept of inclusive education: Teacher training and acquisition of English language in the hearing impaired. British Journal of Special Education, 31, 33–40. Wang, Y., Jongman, A., & Sereno, J. A. (2003). Acoustic and perceptual evaluation of Mandarin tone productions before and after perceptual training. Journal of Acoustical Society of America, 113, 1033–1043. Wang , Y., Spence, M. M., Jongman, A., & Sereno, J. A. (1999). Training American listeners to perceive Mandarin tones. Journal of Acoustical Society of America, 106, 3649–3658. Wong, A. O. C., & Wong, L. L. N. (2004). Tone perception of Cantonese-speaking prelingually hearing-impaired children with cochlear implants. Otolaryngology-Head and Neck Surgery, 130, 751–758. Xu, L., Chen, X., Lu, H., Zhou, N., Wang , S., Liu, Q., et al. (2011). Tone perception and production in pediatric cochlear implant users. Acta Oto-Laryngologica, 131, 395–398. Yip, M. (1980). The tonal phonology of Chinese. PhD dissertation, MIT. Published 1990, New York: Garland Publishing. Yip, M. (2002). Tone. New York: Cambridge University Press.