Download - Perception and Production of Mandarin Tones in ...

Perception and Production of Mandarin Tones inPrelingually Deaf Children with Cochlear Implants

Shu-Chen Peng, J. Bruce Tomblin, Hintat Cheung, Yung-Song Lin, and Lih-Sheue Wang

Objective: Mandarin is a lexical tone language inwhich four tones are crucial for determining lexicalmeanings. Acquisition of such a tone system may bechallenging to prelingually deaf children with co-chlear implants because, as recent studies haveshown, cochlear implant devices are ineffective inencoding voice pitch information required for tonerecognition. This study aimed to investigate Man-darin tone production and perception skills of chil-dren with cochlear implants.

Design: Thirty prelingually deaf children with co-chlear implants, ages 6;0 (yr;mo) to 12;6, partici-pated. These children received their implants at anaverage age of 5;8, with a range from 2;3 to 10;3. Theaverage length of their cochlear implant experiencewas 3;7, with a range from 1;7 to 6;5. Tasks of toneproduction and tone identification involved a pic-torial protocol of 48 words containing the targetedtones in either monosyllabic or disyllabic forms.

Results: The average scores for tone production was53.09% (SD � 15.42), and for tone identification was72.88% (SD � 19.68; chance level � 50%). Significantdifferences were found in the percentages acrossthe production or identification of tone types ortone pairs. The children with exceptional perfor-mance in tone production tended to also performwell in tone identification. The children’s perfor-mance levels in tone identification and productionwere also discussed in relation to the factors of ageat implantation and length of cochlear implantexperience.

Conclusions: The present results suggest that themajority of prelingually deaf children with cochlearimplants did not master Mandarin tone production.However, a small group of participants demon-strated nearly perfect skills of Mandarin tone pro-duction in addition to tone perception. Thus, it isnecessary to consider factors other than the de-vice’s limitations to explain these high levels ofperformance in the perception and production ofMandarin lexical tones.

(Ear & Hearing 2004;25;251–264)

Cochlear implants are electronic devices thatstimulate the primary auditory nerve fibers to elicitsound perception in individuals with severe to pro-found sensorineural hearing impairments. Manystudies have documented that cochlear implants canfacilitate prelingually deaf children’s perception andproduction of consonants and vowels (e.g., Blamey,Barry, & Jacq, 2001; Kirk, Diefendorf, Riley, &Osberger, 1995; Serry & Blamey, 1999; Tobey, Pan-camo, Staller, Brimacombe, & Beiter, 1991), speechintelligibility (e.g., Chin, Finnegan, & Chung, 2001;Miyamoto, Svirsky, Kirk, et al., 1997; Moog & Geers,1999; O’Donoghue, Nikolopoulos, Archbold, & Tait,1999; Svirsky & Chin, 2000), and language develop-ment (e.g., Robbins, Bollard, & Green, 1999; Rob-bins, 2000; Svirsky, Robbins, Kirk, Pisoni, & Miy-amoto, 2000). Cochlear implant devices, however,have been shown to be ineffective in encoding voicepitch information required for the perception ofsuprasegmental elements such as intonation andtone (e.g., Faulkner, Rosen, & Smith, 2000). In anontonal language such as English, intonation andstress do not often make phonemic distinctions;however, they have various expressive functions andconvey speaker intent (e.g., statement and question)and attitude (e.g., surprise and incredulity). Theperception and production of intonation and stresshave been suggested to be potentially challenging toEnglish-speaking children with cochlear implantsand may consequently affect their speech intelligi-bility (e.g., O’Halpin, 2001).

In tone languages such as Thai, Vietnamese, andMandarin Chinese (Mandarin, hereafter), lexicaldistinctions can be made on the basis of fundamen-tal frequency (F0) patterns (i.e., height and contour)alone. Therefore, changing the perceived F0 patternof a syllable usually results in a change in the lexicalmeaning of that syllable. For example, producingthe syllable ma with a high-level tone means“mother.” but producing the same syllable with ahigh-falling tone means “scold.” There are four ma-jor lexical tones in Mandarin. Based on the F0patterns (i.e., characteristics of the height and thecontour or movement), the four Mandarin lexicaltones can be characterized as: high-level for Tone 1,mid high-rising for Tone 2, low-dipping for Tone 3,and high-falling for Tone 4.

There have been few published studies address-

Department of Speech Pathology and Audiology, University ofIowa, Iowa City, Iowa (S.-C.P., J.B.T.); Graduate Institute ofLinguistics, National Taiwan University, Taipei, Taiwan (H.C.);Department of Otolaryngology, Chi-Mei Medical Center, Tainan,Taiwan, Taipei Medical University, Taipei, Taiwan (Y.-S.L.); andDepartment of Rehabilitation, Min-Sheng Healthcare, Taoyuan,Taiwan (L.-S.W.).

DOI: 10.1097/01.AUD.0000130797.73809.40

0196/0202/04/2503-0251/0 • Ear & Hearing • Copyright © 2004 by Lippincott Williams & Wilkins • Printed in the U.S.A.

251

ing the postimplant tone production skills in prelin-gually deaf children. It was only recently that theinformation regarding tone perception skills in Can-tonese-speaking children with cochlear implants be-came available (e.g., Barry, Blamey, & Martin, 2002;Barry, Blamey, Martin, et al., 2002; Ciocca, Francis,Aisha, & Wong, 2002; Lee, van Hasselt, Chiu, &Cheung, 2002; Wei, Wong, Hui, et al., 2000). Thesestudies indicated that prelingually deaf childrenwith cochlear implants demonstrate difficulty in theperception of Cantonese tones. Given the fact thatthe perception and production of tones tend to bemastered early by children with normal hearing(e.g., Li & Thompson, 1977; Tse, 1978; Yue-Hashi-moto, 1980), it is striking that the acquisition of atone system seems to be particularly challenging toprelingually deaf children who receive cochlear im-plants. For example, Ciocca et al. (2002) investi-gated the identification of Cantonese lexical tones inprelingually deaf children with cochlear implants.Ciocca and colleagues suggested that voice pitch isthe only major suprasegmental cue that determinesCantonese tone perception; neither duration noramplitude plays a significant role in the recognitionof Cantonese lexical tones. Their subjects were 17Cantonese-speaking children, ranging in age from 4to 9 yr (age at implantation ranged from 2.5 to 7.6yr; length of cochlear implant use ranged from 0.9 to3.4 yr). The stimuli comprised natural speech mate-rials in which the segmental sequence ji was pro-duced with the six lexical tones of Cantonese. Eachof these words was randomly produced 10 times. Theresults showed that only two cochlear implant recip-ients performed above the chance level overall. Itwas concluded that early deafened cochlear implantusers had great difficulty in extracting the voicepitch information from natural speech sounds toaccurately identify Cantonese lexical tones.

Similarly, Lee et al. (2002) compared the identi-fication performance of three Cantonese tone pairs(i.e., high-level, high-rising, and low-falling) in 15prelingually deaf children with cochlear implantsand 225 children with normal hearing. Participantswere instructed to point to a corresponding pictureafter a live voice presentation. The average score forchildren with normal hearing was 92% (chance level� 50%), whereas the average score for children withcochlear implants was 64%. Consistent with thefindings of Ciocca et al. (2002), the children withcochlear implants in Lee et al. (2002) appeared tohave difficulty in accurately identifying Cantoneselexical tones.

One plausible source of decreased perception oftones in cochlear implant recipients is the limitationof cochlear implant devices. Recent studies havesuggested that cochlear implants do not effectively

encode voice pitch information necessary for im-planted listeners to accurately perceive or recognizetones (e.g., Mandarin) or intonations (e.g., English)(Faulkner et al., 2000; Xu, Tsai, & Pfingst, 2002).For instance, Faulkner et al. (2000) suggested thatplace-based spectral pitch cues that are within thetypical voice range of fundamental frequencies can-not be delivered to cochlear implant recipients withdevices that adopt the continuous interleaved sam-pling (CIS) speech-coding strategy. This is becausethe individual low-frequency harmonics cannot beresolved by the relatively wide bandpass filters. Inother words, only weak periodicity cues to voicepitch information can be extracted by the CIS speechprocessors. As a result, cochlear implant recipientsare likely to have difficulty in accurately perceivingthe voice pitch of quasiperiodic sounds such asvoiced speech, intonation, and music (Ciocca et al.,2002; Faulkner et al., 2000).

If device limitations were the only contributingfactor for poor tone perception in these children,then we should not expect to see any prelinguallydeaf children with cochlear implants who performwell in tone perception, tone production, or both. Onthe other hand, if such children are found, it is likelythat additional factors play essential roles in thedivergent performance levels of tone production andperception in this population.

In addition to the limitations of the cochlearimplant device, other attribute factors are reportedas crucial or highly related to the speech and lan-guage development following cochlear implantation(e.g., Bollard, Chute, Popp, & Parisier, 1999; Nikolo-poulos, O’Donoghue, & Archbold, 1999; Robbins etal., 1999). These factors are associated with thegreat variability in the postimplant spoken lan-guage performance of prelingually deaf childrenwith cochlear implants and include the duration ofdeafness, age of onset of deafness, age at implanta-tion, duration of cochlear implant use, communica-tion mode, physiological factors such as the numberof surviving spiral ganglion cells, and psychological,educational, or social factors such as patients’ moti-vation or level of intelligence (for a review seeLoizou, 1998; Ouellet & Cohen, 1999).

In the present study, the impact of the two fac-tors, that is, age at implantation and length ofdevice use on the pediatric cochlear implant users’tone production and perception skills, were investi-gated in addition to the Mandarin tone perceptionand production skills in prelingually deaf childrenwith cochlear implants. We sought to answer threespecific research questions: First, how well do chil-dren with cochlear implants perform in tone produc-tion and tone identification tasks? Second, is thereany relation between tone perception and produc-

252 EAR & HEARING / JUNE 2004

tion skills in children with cochlear implants? Third,how are the children’s postimplant tone perceptionand production skills associated with the factors oflength of cochlear implant use and age atimplantation?

METHODS

Participants

Thirty prelingually deaf children (16 boys and 14girls) who received their cochlear implants at theDepartment of Otolaryngology of the Chi-Mei Med-ical Center, Tainan, Taiwan, participated in thepresent study. Their mean age was 9 yr, 3 mo (9;3),with the range from 6;0 to 12;6. The mean length oftheir cochlear implant experience was 3;7, rangingfrom 1;7 to 6;5. The mean age at implantation was5;8, ranging from 2;3 to 10;3. One child (CI-25)became deafened suddenly at 8 mo of age for un-known reasons, and all of the remaining childrenwere congenitally deafened. All children were pre-

operatively identified as being profoundly hearingimpaired. Nineteen children were users of the Nu-cleus 22 device, which used the spectral peak(SPEAK) speech-coding strategy. The remaining 11children received the MED-EL COMBI 40 device,which used the CIS speech-coding strategy. None ofthese children was identified as having concomitantlearning disabilities based on parental reports. Allchildren attended mainstream elementary schoolswith varying degrees of involvement in resourceclassrooms in which oral communication was used.The background information for these 30 children issummarized in Table 1.

Materials

A set of 48 laminated pictures measuring 21 � 15cm2 was used in both the tone production and thetone identification tasks. Each picture showed anobject, animal, adjective, or illustration of a motionrepresenting a word exemplifying a target tone. Halfof the target test items were monosyllabic words

TABLE 1. Summary of 30 participants’ background information

ChildI.D. Sex

Age attest time

(mo)

Age atimplantation

(mo)

Lengthof CI use

(mo)Implanted

earBrand ofCI device

Speech-codingstrategy

*Unaided PTA(dB HL)

PTA of the earwith an implant

at test time(dB HL)

CI-1 Female 93 63 30 Left MED-EL CIS 105.0 16.7CI-2 Male 85 54 31 Left Nucleus SPEAK 91.7 23.3CI-3 Female 72 42 29 Left MED-EL CIS 103.31 26.7CI-4 Female 83 56 28 Right MED-EL CIS 106.71 25.0CI-5 Female 96 65 31 Right MED-EL CIS 96.7 31.7CI-6 Male 84 58 26 Right MED-EL CIS 100.0 28.3CI-7 Female 94 67 28 Right MED-EL CIS 101.7 21.7CI-8 Female 76 58 18 Left MED-EL CIS 103.3 26.7CI-9 Male 104 74 31 Left MED-EL CIS 95.0 26.7CI-10 Male 117 75 42 Left Nucleus SPEAK 90.0 30.0CI-11 Male 138 99 40 Left Nucleus SPEAK 103.3 35.0CI-12 Male 117 77 39 Right Nucleus SPEAK 101.7 48.3CI-13 Male 141 113 29 Left MED-EL CIS 91.7 28.3CI-14 Male 130 106 23 Left MED-EL CIS 103.3 31.7CI-15 Female 150 123 27 Left MED-EL CIS 101.7 28.3CI-16† Female 83 27 56 Left Nucleus SPEAK 1001 30.0CI-17† Male 90 34 56 Left Nucleus SPEAK 1101 21.7CI-18 Male 103 52 51 Right Nucleus SPEAK 81.7 33.3CI-19 Male 106 54 52 Left Nucleus SPEAK 1101 23.3CI-20 Female 105 68 38 Left Nucleus SPEAK 1101 38.3CI-21 Female 111 64 47 Left Nucleus SPEAK 87.5 36.7CI-22 Male 111 52 59 Left Nucleus SPEAK 90.0 31.7CI-23 Female 109 61 48 Left Nucleus SPEAK 1101 33.3CI-24 Female 123 56 66 Left Nucleus SPEAK 105.0 33.3CI-25 Male 142 86 56 Left Nucleus SPEAK 103.3 35.0CI-26 Male 136 75 61 Left Nucleus SPEAK 88.3 28.3CI-27 Male 114 55 59 Left Nucleus SPEAK 1101 46.7CI-28 Female 140 92 48 Left Nucleus SPEAK 105.0 36.7CI-29 Male 146 69 77 Left Nucleus SPEAK 1101 35.0CI-30 Female 141 75 65 Left Nucleus SPEAK 96.7 43.3

* The unaided PTA (in dB HL) was measured before surgery, the most recent audiogram taken for the better ear at the frequencies of 500, 1000, and 2000 Hz.† No reliable ABR waveforms could be recorded at 100 dB nHL for two children, CI-16 and CI-17.CI � cochlear implant; PTA � pure-tone average.

EAR & HEARING, VOL. 25 NO. 3 253

(e.g., qi1, “seven”) and half were disyllabic words(e.g., gong1-ji1, “rooster”). Each picture was used toelicit these children’s production of a certain targettone, and two of the pictures containing a target tonecontrast were paired later to examine the children’sability to identify tones (see Table 2). A 3 � 3 cm2

removable piece of card was attached to the upperright corner of each picture to hide the target nameindicated by Chinese traditional characters alongwith the corresponding phonetic notations, whichincluded an indication of the target tone. This prac-tice encouraged the children to produce spontaneousresponses, as constructed from their own lexicaltone representations. Hence, the card was only re-moved when a child failed to recognize the name ofa certain picture or produced an erroneous tone.

In the tone identification task, a live voice presen-tation was adopted to examine the tone identifica-tion performance of the participants. This attemptwas designed to actively engage the present partic-ipants and maintain the children’s interest levelsbecause each participant also received annual au-diological assessments and the verification of thespeech programming of the cochlear implant deviceon the same day when the tone production andidentification tasks were performed. The use of alive voice allowed more interaction between theexaminer and the child during the task and encour-aged the child to concentrate on the on-going task,thus allowing him or her to perform at an optimallevel.

Procedure

Tone Production Task • Participants were indi-vidually tested by two examiners in a quiet auralrehabilitation room. Production of each tone wasfirst elicited from each child. A spoken model wasprovided by one of the examiners (the first author),who elicited tone production by presenting the childwith the speech materials described above. Theother examiner assisted in setting up the recordingsystem and provided assistance (e.g., interactingwith the child) during the task. The set of 48pictures was presented, one picture at a time, toeach child. Thus, each child produced a total of 48samples of Mandarin tones (i.e., 12 samples for eachtone). The goal of elicitation was to obtain the bestpossible tonal production from each child. With thepresentation of each picture, the child was firstprompted to name the illustration of the picture(e.g., zhe shi she me, “What is this?”). Appropriatereminders such as syntactic or semantic cues wereprovided when necessary. If the child failed to pro-duce the target tone accurately, the 3 � 3 cm2 cardwas removed so that the child was able to read the

Chinese traditional character(s) and the phonemicnotation(s). If this visual cue did not help the childproduce the target tone accurately, the child wouldbe asked to imitate the production of the examiner’sspoken model. The child was then encouraged toproduce the tone again following imitation. Spokenmodels were only occasionally required since allparticipants were able to recognize Mandarin pho-netic notations (based on the parental reports),which are part of the fundamental curriculum forolder preschoolers and first-grade children at ele-mentary schools in Taiwan. The entire task wasrecorded on a Mini-Disk (MD), using a digital MDrecorder (SHARP MD-MT831-S) via a stereo con-denser microphone (AIWA CM-TS22) for furtheranalysis.Tone Identification Task • Each child’s ability ofidentifying the six different tone pairs (i.e., Tone 1versus Tone 2; Tone 1 versus Tone 3; Tone 1 versusTone 4; Tone 2 versus Tone 3; Tone 2 versus Tone 4;and Tone 3 versus Tone 4), using this closed-setidentification task with two pictorial alternativeresponses was examined. Forty-eight pictures werearranged into 24 pairs, and four trials were per-formed for each tone pair (96 test items total) to eachchild. The procedure for examining each tone pairwas arranged into two phases: the preparatoryphase and the testing phase. In the preparatoryphase, the examiner presented the child with thepictures used in the previous tone production taskand the child was first prompted to name eachpicture. If the child did not remember the name of apicture or named the picture inaccurately, the ex-aminer modeled the name for the child and thenprompted the child’s production of the target tone.The child was not prevented from looking for facialcues from the examiner in this preparatory phase.

In the testing phase, the child was asked toindicate the corresponding illustration by pointingto the picture between the pair in front of him/her,for example, na yi ge shi yan3-jing1, “Which picturehas eyes on it?” or na yi ge shi yan3-jing4, “Whichpicture has glasses on it?” The position of the pair ofpictures was fixed, but the order of oral presentationof different target tones was arranged in a counter-balanced order. The examiner’s facial cues werehidden from the child by using a letter-sized (8.5 �11 inch) piece of cardboard in this testing phase. Thechild received one point if she or he identified atarget tone accurately. The other examiner wasresponsible for marking the points that the childreceived. It took approximately 40 to 50 minutes foreach child to complete both tasks. The children werepermitted to take rest breaks whenever they wished.


Scoring and Interjudge Reliability of theTone Production Task

All of the tone production samples were rated bythree native speakers of Mandarin. Two of thejudges were speech pathologists and one was asenior undergraduate student who had taken acourse in phonetics but had no prior experience inlistening to the speech of individuals with hearingimpairments. One of the three judges was involvedin this study other than scoring the children’s toneproductions; the other two judges were not informedof the objectives of the present study. Judges wereinstructed to score each tone production on the basisof the perception of tone properties of the heardwords and minimize the influence of phoneme accu-racy in their judgments of tonality in the complexspeech stimuli. Each scoring sheet listed the 48target tones along with the corresponding Chinesecharacters for the names. An equal appearing inter-

val scale of five values with “1” (completely incor-rect) and “5” (completely correct) marked at theextremes was used to measure the accuracy of eachtarget tone production. Each listener was providedwith the audio recordings of all samples produced byeach child. Judges were instructed to rate the besttrial if a child required more than one trial on aparticular tone production, and they were allowed tolisten to each sample more than one time if theydesired. There was no identifiable information forthe children (e.g., name, chronological age, sex), ordevice-related information (e.g., type of device,speech-coding strategy) on the scoring sheets.

Interjudge reliability coefficients (Shriberg & Lof,1991) were derived with the use of SPSS (v.10.0).The coefficient � for interjudge agreement acrosschildren was 84.37%, which was computed based onall of the 1440 samples produced by all children. Theindividual reliability coefficients, which were com-

TABLE 2. Speech materials used in the tone production and tone identification tasks

Each word is listed with its Chinese character(s). The target tone is marked with a number (1 for tone 1; 2 for tone 2; 3 for tone 3; and 4 for tone 4) to indicate its lexical tone, along with thepin-yin transcription (in italics) and the English translation (second line of each cell). Target tones of the disyllabic test items are underlined.


puted on the basis of all of the 48 samples producedby each child, ranged from 41.16% to 90.85%. Thetone production scores of the three judges wereaveraged for further analysis.

The interjudge reliability was as low as 42% foronly one child (CI-4) and was between 50% and 58%for three children. All of the interjudge reliabilityvalues for the remaining children were between 61%and 91%. Both the characteristics of both listeners(judges) and speakers (children with cochlear im-plants) can contribute to why the interjudge reliabil-ity was low for some children. The production skillsof each tone in some children (including the high-performing ones) did not appear to be stable acrossdifferent words. In addition, listener variability alsoplays a significant role given that the rating wasderived on the based of only three listeners’ judg-ments. Note that the coefficient � for interjudgeagreement across children was 84.37%, and thisvalue was derived on the basis of all samples. Thevalue of 41.16%, on the other hand, was derived onthe basis of the samples of one child (CI-4).

RESULTS

Tone Production Task

Rating values from the judges (ranging from 1 to5) represented the degree of correctness, where 1equaled 0% and 5 equaled to 100%. Each of thevalues was converted to a percent correct value thatrepresented the degree of accuracy. The averagescore for the children’s tone production was 53.09%(SD � 15.42). With regard to the production ofindividual tones, the average score was 62.13% (SD� 19.68) for Tone 1; 42.13% (SD � 17.62) for Tone 2;45.89% (SD � 17.64) for Tone 3; and 62.22% (SD �17.17) for Tone 4. Figure 1 illustrates the box plot ofthe participants’ production scores for the four Man-darin tones. The multivariate analysis of variance(MANOVA), calculated by means of SAS PROCMIXED, showed that the difference in the produc-tion scores of the four tones was statistically signif-icant (F(3,30) � 40.29, p � 0.0001). Post hoc Tukeyadjusted t-tests showed that differences existed be-tween the production scores for Tone 1 and Tone 2(t(30) � 7.48, p � 0.001), Tone 1 and Tone 3 (t(30) �4.66, p � 0.005), Tone 2 and Tone 4 (t(30) � 8.27, p� 0.001), and Tone 3 and Tone 4 (t(30) � 7.87, p �0.001). Accordingly, the tone production scores werehigher for Tone 1 and Tone 4 than for Tone 2 andTone 3. In addition, there were three children whoseoverall scores were 90% or higher (CI-4 � 90.10%;CI-16 � 94.27%; CI-17 � 90.28%).

Tone Identification Task

As with tone production, the participants’ toneidentification scores were also converted into per-centages. The children’s average tone identifica-tion score was 72.88% (SD � 12.81), which wassignificantly higher than the 50% chance level(t(29) � 9.78, p � 0.001). With regard to theidentification of specific tone pairs, the averagescore was 68.96% (SD � 15.78) for Tone 1 versusTone 2; 70% (SD � 18.08) for Tone 1 versus Tone3; 78.96% (SD � 13.88) for Tone 1 versus Tone 4;64.79% (SD � 18.82) for Tone 2 versus Tone 3;78.33% (SD � 20.28) for Tone 2 versus Tone 4; and76.25% (SD � 16.93) for Tone 3 versus Tone 4. Allaverage scores were significantly higher than the50% chance level (all p values � 0.001). Figure 2illustrates the box plot of the children’s identifica-tion scores for the six tone pairs. The MANOVAanalysis, calculated by means of SAS PROCMIXED, showed that the difference in the identi-fication scores of the six tone pairs was statisti-cally significant (F(5,25) � 5.96, p � 0.001). Theaverage score for tone identification was highestfor Tone 1 versus Tone 4. The other identificationscores, in descending order, were Tone 2 versusTone 4, Tone 3 versus Tone 4, Tone 1 versus Tone3, Tone 1 versus Tone 2, and Tone 2 versus Tone 3.Post hoc Tukey adjusted t-tests specified thatsignificant differences existed between the identi-fication scores for Tone 1 versus Tone 2 and Tone

Fig. 1. Box plot illustrates accuracy (in percentage) of eachMandarin tone production in the tone production task. Thex-axis represents the four tones; y-axis displays scores corre-sponding to each tone production. Median is displayed by theline across each grey box; upper and lower bounds of eachgrey box represent quartiles; and the circles stand for theoutliers and the stars represent extreme values.


1 versus Tone 4 (t(30) � 3.21, p � 0.03), Tone 1versus Tone 4 and Tone 2 versus Tone 3 (t(30) �4.02, p � 0.004), Tone 2 versus Tone 3 and Tone 2versus Tone 4 (t(30) � 3.57, p � 0.01), and Tone 2versus Tone 3 and Tone 3 versus Tone 4 (t(30) �3.85, p � 0.007). The general trend of this set ofanalyses indicated that the identification scoreswere significantly higher for the pairs that con-tained Tone 4 than for the pairs without this tone.

Correlations Between Tone Production andTone Identification

A statistically significant Pearson correlationcoefficient (r) was found between the overall aver-age scores for tone production and for tone identi-fication (r � 0.44, p � 0.015). Figure 3 illustratesa scatterplot of the children’s scores for toneproduction and for tone identification. As Figure 3shows, three children (CI-4, CI-16, and CI-17)performed substantially better than the remain-ing children. These three high-performing chil-dren’s average tone production scores were 90.10%(CI-4), 94.27% (CI-16), and 90.28% (CI-17); theircorresponding average tone identification scoreswere 90.63%, 92.71%, and 93.75%. When thesethree children’s data points were excluded fromthis set of analyses, the Pearson correlation coef-ficient did not achieve a statistically significantlevel (r � 0.01, p � 0.98).

Correlations Between the Children’sPerformance in Tone Identification and ToneProduction and the Factors of Age atImplantation and Length of CochlearImplant Experience

The Pearson correlation coefficient between theoverall tone identification score and age at implan-tation was not statistically significant (r � �0.20,p � 0.29), but there was a statistically significantrelation between the overall tone production scoreand age at implantation (r � �0.58, p � 0.001).Figure 4 shows the children’s tone production scoresas a function of age at implantation. Moreover, thecorrelation coefficients achieved a statistically sig-nificant level (� � 0.05) between age at implantationand the score for each tone production. The Pearsoncorrelation coefficient between age at implantationand the average score for Tone 1 was �0.37 (p �0.047); for Tone 2 was �0.55 (p � 0.002); for Tone 3was �0.64 (p � 0.001); and for Tone 4 was �0.44(p � 0.014).

The correlation analyses between the overall av-erage tone production and tone identification scoresand the factor of length of cochlear implant userevealed that none of the Pearson correlation coeffi-cients achieved a statistically significant level (bothp values � 0.05).

Fig. 2. Box plot displays accuracy (in percentage) of theidentification of the six tone pairs in the tone identificationtask. The x-axis represents the six tone pairs; y-axis displaysscores corresponding to the identification of each tone pair.Median is displayed by the line across each grey box; upperand lower bounds of each grey box represent quartiles.

Fig. 3. Scatterplot shows distribution of accuracy (in percent-age) of children’s tone production and tone identification.The x-axis represents overall tone production accuracy; y-axisdisplays overall tone identification accuracy. Children withthe MED-EL device are marked with a cross; those with theNucleus device are marked with a circle. Six children aremarked with their identification number (CI-4, CI-16, CI-17,CI-6, CI-25, CI-30); see text for a detailed discussion.


The two types of cochlear implant device(MED-EL and Nucleus) were not implanted in chil-dren across the same period of time (11 childrenwere users of the MED-EL device and had beenusing their implants for 18 to 30 mo, whereas theother 19 children were users of the Nucleus deviceand had been using their implants for 31 to 77 mo attest time). The average age at implantation for thechildren with the Nucleus device was 64.5 mo (SD �18.2) and for those with the MED-EL device was 75mo (SD � 26.5). The difference in age at implanta-tion between the two subgroups of children did notachieve a significant level (t(28) � 1.29, p � 0.21).The overall average tone identification was 72.15%(SD � 13.70) in children with the Nucleus deviceand was 74.15% (SD � 11.64) in those with theMED-EL device. The overall average tone produc-tion was 63.53% (SD � 11.59) in the Nucleus groupand was 60.66% (SD � 13.94) in the MED-EL group.Note that in the present study, the length of cochlearimplant use for most of the children with the Nu-cleus device was longer than 30 mo, whereas theduration for most of those with the children with theMED-EL device was less than 30 mo. With theassumption that longer cochlear implant use doesnot reduce tone perception and tone productionskills, one-tailed, two-sample t-tests were conductedto test whether the MED-EL group performed lesswell than the Nucleus group on the tone identifica-tion and tone production accuracy. The results indi-

cated that none of the average tone identification ortone production scores in the cochlear implant recip-ients with the MED-EL device were lower thanthose with the Nucleus device (tone identification:t(28) � 0.61, p � 0.55; tone production: t(28) ��0.41, p � 0.69). Similarly, no statistically signifi-cant difference was found for the production scoresfor particular tones or for the identification scoresfor particular tone pairs between two groups ofchildren with different types of devices (all p values� 0.1).

Pearson correlation coefficients were then exam-ined between the scores for tone production or toneidentification and the children’s length of cochlearimplant use on the basis of the device type. Nostatistically significant difference was found be-tween tone production scores in both the MED-ELgroup and the Nucleus group. Moreover, no statisti-cally significant correlation was found between theoverall tone production scores or tone identificationaccuracy and length of cochlear implant use inchildren who used the MED-EL device (both p val-ues � 0.5). In children who received the Nucleusdevice, however, the Pearson correlation coefficientbetween the overall tone identification score andlength of cochlear implant use was statisticallysignificant (r � 0.50; p � 0.03). Figure 5 shows thedistribution of the accuracy of the children’s toneidentification as a function of length of device expe-rience. Further analyses indicated that in children

Fig. 5. Scatterplot represents distribution of accuracy (inpercentage) of tone identification (y-axis) as a function oflength of cochlear implant use (x-axis). Children with theMED-EL device are marked with a cross; those with theNucleus device are marked with a circle.

Fig. 4. Scatterplot illustrates distribution of accuracy (inpercentage) of tone production (y-axis) as a function of age atimplantation (x-axis). Children with the MED-EL device aremarked with a cross; those with the Nucleus device aremarked with a circle.


with the Nucleus device, the Pearson correlationcoefficient achieved a statistically significant levelbetween length of cochlear implant use and theaverage score of the identification of the pair, Tone 2versus Tone 3 (r � 0.60, p � 0.007). In addition, thecorrelation coefficient between length of cochlearimplant use and the identification score of the pair,Tone 2 versus Tone 4, approached the significantlevel (r � 0.45, p � 0.052).

DISCUSSION

Tone Production Task

Results of the tone production task revealed thatas a group, the participants’ overall average toneproduction accuracy was approximately 53%. Theaverage scores for individual tone production werenot equally distributed among the four tones. Theproduction scores were higher for Tone 1 and Tone 4than for Tone 2 and Tone 3. In other words, prelin-gually deaf children with cochlear implants weremore proficient in the production of Tone 1 (thehigh-level tone), and Tone 4 (the high-falling tone)than in the production of Tone 2 (the mid high-risingtone) and Tone 3 (the low-dipping tone). This pat-tern is consistent with that of the acquisition of toneproduction exhibited in children with normalhearing.

Children with normal hearing become proficientin a tone system rather early in life, and the acqui-sition of such a tone system is no slower than that ofconsonant and vowel systems. In a longitudinalstudy, Li and Thompson (1977) investigated Manda-rin tone development in 17 children with normalhearing who were 18 to 36 mo of age. The authorsnoted that the acquisition of the Mandarin tonesystem occurred very early during the 7-mo obser-vation period in these children and preceded themastery of consonants and vowels. Similarly, Yue-Hashimoto’s (1980) case study reported that thesubject was able to distinguish Mandarin tonesconsistently in all monosyllabic and a majority ofdisyllabic words by the age of 20 mo. The authorconcluded that the rate of tone acquisition is muchfaster than that of consonants or vowels given thatby 20 mo of age, the child had not yet been proficientin the Mandarin phoneme system.

Certain tone types have been demonstrated to bemastered earlier than others (e.g., Li & Thompson,1977; Su, 1985; Tse, 1978). Tse (1978) investigatedCantonese-speaking children’s acquisition of a tonesystem. The author suggested a universal principlefor the order of tone acquisition: level tones (e.g.,Mandarin Tone 1) are acquired earlier than contourtones (e.g., Mandarin Tone 2 and Tone 3), and fallingtones (e.g., Mandarin Tone 4) are acquired before

rising tones (e.g., Mandarin Tone 2). Similarly, Su(1985) investigated two Mandarin-speaking chil-dren’s (ages 14 and 17 mo) tone development andfound that the pattern of the acquisition order ofspecific tones is consistent with that in Tse (1978);that is, Tone 1 and Tone 4 are mastered earlier thanTone 2 and Tone 3. Su (1985) suggested that Tone 2and Tone 3 are often confused with each other, andthus more difficult to discriminate than Tone 1 andTone 4 for young children with normal hearing. Insummary, despite the fact that the criteria that havebeen adopted to determine the acquisition of a tonesystem are not the same across studies, the findingsin the literature have unambiguously suggestedthat the Mandarin tone system is acquired early (byage 2;6) with respect to its production by childrenwith normal hearing. After this age, even thoughTone 2 and Tone 3 may still be confusing to thesechildren (Lee, 1993), typically young children do notencounter much difficulty in acquiring a tonesystem.

The results revealed from the tone productiontask are consistent with the findings in childrenwith normal hearing. That is, prelingually deafchildren with cochlear implants are more proficientin the production of Tone 4 (the high-falling) thanTone 2 (the mid high-rising). In children with nor-mal hearing, the production of rising tone or intona-tion is generally acquired later than their fallingcounterparts (e.g., Cruttenden, 1981; Li & Thomp-son, 1977). The difficulty of producing rising tones orintonation may be associated with their physiologi-cal correlates such as vocal effort (e.g., Li & Thomp-son, 1977; Lieberman, 1967; Snow, 1998). Further-more, the phonological distinction of naturalness(marked versus unmarked; Yavas, 1998) betweenfalling and rising tones may also explain the reasonwhy Tone 3, which is comprised of a more complexpitch contour than the other two tones, tended to bemore challenging for these children to produce accu-rately than Tone 1 and Tone 4. Interestingly, an-other more challenging tone for the children toproduce was Tone 2, which is conventionally char-acterized as a rising tone (e.g., Chao, 1968). Thistone, however, has been suggested to have a slightdipping acoustical property (Fon, 1997; Ho, 1976;Shih, 1988). Although the dipping contour of Tone 2does not appear to be as consistent as it is in Tone 3,the dipping property of the contours in both tonesmay at least partially contribute to the source ofambiguity between Tone 2 and Tone 3 in terms ofboth perception and production. Taken together, theintrinsic properties of Mandarin tones are associ-ated with the unequal performance levels of theproduction of specific tones in children with cochlearimplants.


Tone Identification Task

In the present study, a live voice presentation wasadopted to examine the tone identification perfor-mance of the participants to increase the children’sinterest levels. The live voice presentation would beespecially appropriate for the purpose of the toneidentification task, which was designed to investigatethe children’s lexical tone perception of Mandarin. Thetone identification task was essentially a linguisticallydriven, word identification test. Hence, any allophonicvariance should be acceptable and tolerable to ourparticipants in that they are native speakers of theMandarin language.

Note, however, the use of the live voice presenta-tion may have some limitations. For instance, whena comparison is made across children, it is possiblethat the test item variability may influence both thereliability and validity of the results. Additionally,with the use of a live voice presentation, the speechproduction of individual talkers can vary consider-ably from trial to trial, especially when trials areconducted on different days or with different partic-ipants. These limitations can be diminished with theuse of recorded speech stimuli.

The results of the tone identification task showedthat the overall average tone identification accuracywas about 73%, which was statistically higher thanthe chance level (50%). As a group, the children’saverage performance level in the tone identificationtask may not be very satisfactory. However, therewere six children (see Fig. 3) who achieved anaverage identification score of 89% or above. Thesechildren (6 of 30) of the prelingually deaf cochlearimplant recipients were able to identify Mandarintones from a pair accurately. The remaining chil-dren, on the other hand, were less capable of accu-rately identifying a perceived tone from a pair thatthey heard.

The identification scores of specific tone pairswere not equally distributed among the six pairs. Ingeneral, the average scores tended to be higher forthe tone pairs which contained Tone 4 than the pairsthat did not contain this tone. The intrinsic charac-teristics of this tone may again help to account forthese results. That is, Tone 4 (the high-falling tone)has the shortest duration among the four tones inMandarin perceptually (Whalen & Xu, 1992; Xu etal., 2002; Yang, 1989). Duration is perceived mainlyon the basis of the slow time-intensity, temporalenvelope cues of speech signals ranging from 2 to 50Hz (Rosen, 1992). This low-frequency temporal en-velope cue of speech signals can be preserved viacochlear implant devices (Green, Faulkner, &Rosen, 2002). Hence, it is possible that the childrenwith cochlear implants were able to effectively use

the available temporal envelope cues to distinguishTone 4 from the other tones because of its shortestduration among the four Mandarin tones. This canbe supported by the present findings in which thechildren with cochlear implants were able to pro-duce both Tone 1 and Tone 4 more accurately thanTone 2 and Tone 3, yet only pairs with Tone 4 (butnot Tone 1) were better identified. It is likely thatthe children with cochlear implants were able to usethe duration cues to distinguish Tone 4 from othertones in a pair. In contrast, Tone 1, lacking such apatent duration cue, was not identified easily whenit was paired with Tone 2 or Tone 3.

The developmental pattern of tone acquisition inhearing-impaired children who do not use cochlearimplants is quite similar to that of children withnormal hearing, but the acquisition rate tends to beslower. Chen (1986) investigated the acquisition ofMandarin tones in three children with hearing im-pairments. The author found that the order of toneacquisition in these children with hearing impair-ments did not differ from that in children withnormal hearing, that is, Tone 1 and Tone 4 wereacquired earlier than Tone 2 and Tone 3. However,these hearing-impaired children tended to acquirethe Mandarin tone system slowly. By the time thesechildren were 4 or 5 yr of age, they had not yet fullyacquired the Mandarin tone system.

Similar to hearing-impaired children who did notuse cochlear implants, prelingually deaf childrenwith cochlear implants did not appear to master theMandarin tone system in terms of perception andproduction. According to the present participants’performance levels in the production of specifictones, Tone 2 and Tone 3 were more challenging forthese children to produce accurately than were Tone1 or Tone 4. With respect to the identification ofspecific tone pairs, these children showed greatdifficulty in accurately identifying tones that werenot contrasted with Tone 4. In summary, the presentresults suggest that in comparison with the toneacquisition pattern in children with normal hearing,prelingually deaf children with cochlear implantsdemonstrate a delayed pattern of tone acquisition.

Relations Between Tone Production andTone Identification

The present results indicated that as a group, thechildren’s performance levels in the tone productionand tone identification tasks exhibited a slightlypositive correlation. However, when the scores of thethree high-performing children were excluded fromanalyses, the Pearson correlation coefficient nolonger achieved a statistically significant level (al-pha � 0.05) for the remaining 27 children. The three


children (CI-4, CI-16, and CI-17) who received rela-tively high scores (90% and above) for tone produc-tion were among the participants who also identifiedtones the best (89% and above). On the other hand,the children (e.g., CI-30, CI-6, and CI-25) who ob-tained high scores (89% and above) in the toneidentification task did not appear to be able toproduce tones very accurately. Specifically, two chil-dren’s (CI-6 and CI-25) average tone identificationscores were approximately 89%, but their averagetone production scores were about 55%. Similarly,one child (CI-30) received an average score of 91.67%for tone identification, but her score for tone produc-tion was as low as 40.45%. Taken together, althoughthese results indicated the lack of a significantPearson correlation coefficient between the scoresfor tone production and tone identification in themajority of these children (N � 27), some evidence ofan association between the participants’ identifica-tion and production performance levels is shown bythe performance of the children with exceptionalperformance in tone production. These children withexceptional performance in tone production alsoperformed well in tone identification. Thus, goodperformance in tone identification may be a neces-sary condition for that in tone production. However,the ability to correctly identify a tone from a pair didnot imply the ability to produce these tones accu-rately, therefore good performance in tone identifi-cation is not a sufficient condition for that in toneproduction.

In addition, as Figure 3 reveals, since the threehigh-performing children who received a relativelyhigh score (90% and above) in tone production alsoreceived a high score (90% and above) in tone iden-tification, presumably high proficiency in tone pro-duction (again, 90% and above) implies high profi-ciency in tone identification. Additionally, therewere six children who performed at a reasonablyhigh level, namely, 89% or above, the limitations ofcochlear implant devices in encoding voice pitchinformation may not exclusively contribute to thelimited skills of the perception or production ofMandarin tones in many of the prelingually deafchildren with cochlear implants.

The subsequent question is, what are the factorsthat may account for this finding that some of theprelingually deaf children with cochlear implantswere able to produce or identify tones more accu-rately than others? Most of our participants werecongenitally deafened (except for CI-25), and theydid not exhibit much difference in their preoperativehearing thresholds. According to the present results,the two children (CI-16; CI-17) among the threehigh-performing participants were also among theones who received their implants at relatively

younger ages (2;3 for CI-16; 2;10 for CI-17), whereasthe two children (CI-25, CI-30) who received highscores in the tone identification task but not in thetone production task received their implants at arelatively older age (7;2 for CI-25; 6;3 for CI-30). It islikely that there was some kind of association be-tween the performance levels in tone production andage at implantation. This provisional speculationwas further examined by conducting the next set ofanalyses specifically on the possible associationsbetween the children’s performances levels in thetwo tasks and the factor of age at implantation.

Relations Between the Children’s ToneIdentification and Production Skills and theFactors of Age at Implantation and Length ofCochlear Implant Experience

The present results revealed a significantly neg-ative correlation between the overall tone produc-tion score and age at implantation. In other words,the younger a prelingually deaf child received acochlear implant, the better his/her performancelevels in tone production. Moreover, a significantlypositive correlation was found between the overalltone identification score and length of cochlear im-plant use in children who had used their implantsfor more than 30 mo (those with the Nucleus device)at test time. In other words, the ability to identifytones accurately was associated with these chil-dren’s longer use of cochlear implant devices in thisgroup of children (N � 19).

The positive correlation only existed in childrenwith the Nucleus device but not in the ones with theMED-EL device. Because there was a relativelylimited range (13 mo) for cochlear implant use in thechildren with MED-EL devices and a relativelyextensive range (47 mo) for the Nucleus group, thesmaller and more restricted sample in the MED-ELgroup may diminish the chances of finding a signif-icant correlation between the children’s perfor-mance level and length of cochlear implant use.

It is possible that the present result was a conse-quence of the different types of cochlear implantdevice the children used rather than a result of thedifference in length of cochlear implant use. In otherwords, the effect of device types might have beenconfounded with the factor of length of cochlearimplant use. This hypothesis, however, was notsupported by our present results. The present re-sults indicated that children with the MED-EL de-vice (the ones with less than 31 mo of device expe-rience) did not demonstrate poorer performancelevels than the ones with the Nucleus device (theones with more than 30 mo of device experience).Moreover, even though the MED-EL device was amore recently implanted system for some of the


present participants, the average age at implanta-tion for the MED-EL group did not differ from thatof the Nucleus group. Thus, it is seemingly reason-able to rule out the potential confounding betweenthe effect of device types and length of cochlearimplant experience.

As just noted, when the two subgroups of chil-dren’s performance levels in tone identification werecompared, children with the MED-EL device, de-spite shorter length of use, performed as well asthose with the Nucleus device. These results mayimply that the rate of the acquisition of tone percep-tion was more rapid for the MED-EL group than theNucleus group. It is plausible that these differencesin device types may be due to different speech-coding strategies used. That is, the MED-EL deviceused the CIS speech-coding strategy, which has amuch higher stimulation rate than the SPEAKspeech-coding strategy that was used by the Nucleusdevice (for a review see Loizou, 1998). Note, how-ever, verification of this assumption requires addi-tional evidence such as longitudinal data and isbeyond the scope of the present study.

Nevertheless, the present results suggested thatthe there is no direct correspondence between thetone perception and production skills in the majorityof prelingually deaf children with cochlear implants.Moreover, length of cochlear implant use and age atimplantation may be associated with the postim-plant development of tone perception and tone pro-duction skills of the children with cochlear implants.The identification scores were positively correlatedwith length of cochlear implant experience in theNucleus group. There was a significant negativecorrelation between age at implantation and chil-dren’s performance levels in tone production. Thetone production performance levels, however, werenot significantly associated with length of cochlearimplant use. Taken together, these results suggestthat tone perception and tone production skills didnot develop in a parallel fashion in these prelin-gually deaf children with cochlear implants.

The findings of the association between our par-ticipants’ tone production performance levels andage at implantation are consistent with the findingsof Sharma et al. (2002). Sharma and colleaguesexamined the consequences of different ages of im-plantation on the development of the human centralauditory system and found that children who wereimplanted before age 3;6 demonstrated age-appro-priate P1 latency responses within the first severalmonths after implantation. Those who received theirimplants after 7 yr of age did not show normalcortical response latencies to speech. The authorsconcluded that human central auditory system re-

mains maximally plastic before the age of 3;6, andplasticity may diminish after this age.

Tone acquisition occurs rather early in speechdevelopment. Since the fluctuation of F0 is notdisplayed in observable speech movements, produc-tion of tones provides fewer visual cues than conso-nants and vowels. Accordingly, acquisition of tonesmay rely heavily on the integrity of the auditorysystem. Most of our participants who received theirimplants later than 4 yr of age did not demonstrategood tone production performance. An exception isthat one child (CI-4), who received her implant at 56mo old, performed extraordinarily in the tone pro-duction task. This result is still compatible with thefindings of Sharma et al. (2002) in which some of thechildren who received their implants between ages 4and 7 yr still demonstrated age-appropriate P1 la-tency responses.

There are some potential limitations in thepresent study that may need to be considered infuture studies. First, the raters were explicitly in-structed to rate the accuracy only on the basis of thetonality in the complex stimuli. Note, however, wedid not exclude the potential influence of phonemeaccuracy on the judgments of tone production accu-racy, and therefore it is likely that the presentresults might have been biased by the participants’accuracy in phoneme production. This problem canbe resolved by removing the higher frequencies fromthe children’s speech samples so that only the fun-damental remains. Second, the present study onlyexamined the potential relation between the chil-dren’s performance levels and the factors of age atimplantation and length of cochlear implant usewhile other variables were controlled. We can notexclude the impact of these variables (such as edu-cational, social, or learning variables) on pediatriccochlear implant recipients’ postimplant tone per-ception or production skills. Third, this study aimedto provide the descriptive data of the pediatriccochlear implant users’ postimplant tone perceptionand production skills, and for this purpose we didnot perform acoustic analyses on the children’s pro-duction data. It is certainly crucial to perform suchanalyses to help us understand the intrinsic proper-ties of the production of Mandarin tones in thesechildren. For example, if the children accomplishtone identification on the basis of durational cuesrather than the F0 patterns (height and contour),then they may also be equally good at producing thecues that they are able to hear (e.g., duration).

In summary, the majority of prelingually deafchildren with cochlear implants did not appear to beable to accurately produce Mandarin tones. Toneproduction or identification skills were not acquiredevenly among tones or tone pairs by these children.


The intrinsic properties of specific tones may con-tribute to why certain tones or tone pairs are moredifficult than others to produce or identify accu-rately. The children with exceptional performance intone production tended to also perform relativelywell in tone identification. However, the ability toaccurately identify a tone from a pair did not implyhigh accuracy in tone production. Moreover, thefactors of length of cochlear implant use and age atimplantation are not equally associated with thechildren’s postimplant tone production and percep-tion skills. There was a small group of prelinguallydeaf children with cochlear implants who demon-strated nearly perfect skills of Mandarin tone pro-duction in addition to tone perception. Thus, it isnecessary to consider factors other than the device’slimitations to elucidate the high levels of perfor-mance in the pediatric cochlear implant users’ per-ception and production of Mandarin lexical tones.

ACKNOWLEDGMENTS:An earlier version of this manuscript was presented at the 7th

International Cochlear Implant Conference held in Manchester,U.K. The authors appreciate all children and their parents whoparticipated in this study. We gratefully acknowledge the assis-tance from the audiologists at the Department of Otolaryngologyof the Chi-Mei Medical Center. Sincere thanks are sent toMei-Ling Shen for her help in gathering and verifying the patientinformation in Taiwan, as well as to the kind help from GeorgeMorgan, Robin Barrow and Vanessa Shaw in proofreading thisreport. We wish to thank Dr. Nelson Lu for his advice on thestatistical analyses applied to the present study. Finally, weearnestly appreciate the most useful suggestions and commentsprovided by Dr. Mario Svirsky, Dr. Melanie Matthies, Dr. SuWooi Teoh, and our two anonymous reviewers.

Address for correspondence: Yung-Song Lin, M.D., Taipei Medi-cal University, Department of Otolaryngology, Chi Mei MedicalCenter, 901 Chung Hwa Road, Yung Kan City, Tainan County,Taiwan, R.O.C. E-mail: [email protected]

February 12, 2003 January 21, 2004

REFERENCES

Barry, J. G., Blamey, P. J., & Martin, L. F. A. (2002). Amultidimensional scaling analysis of tone discrimination abil-ity in Cantonese-speaking children using a cochlear implant,Clinical Linguistics and Phonetics, 16, 101–113.

Barry, J. G., Blamey, P. J., Martin, L. F. A., Lee, K. Y. S., Tang,T., Ming, Y. Y., & van Hasselt, C. A. (2002). Tone discrimina-tion in Cantonese-speaking children using a cochlear implant.Clinical Linguistics and Phonetics, 16, 79–99.

Blamey, P. J., Barry, J. G., & Jacq, P. (2001). Phonetic inventorydevelopment in young cochlear implant users 6 years postop-eration. Journal of Speech, Language, and Hearing Research,44, 73–79.

Bollard, P. M., Chute, P. M., Popp, A., & Parisier, S. C. (1999).Specific language growth in young children using the CLAR-ION cochlear implant. The Annals of Otology, Rhinology &Laryngology, 177 (Suppl.), 119–123.

Chao, Y. -R. (1968). A Grammar of Spoken Chinese. Berkeley, CA:University of California: Berkeley & Angeles.

Chen, J. L. (1986). Phonological Acquisition in Mandarin byHearing-impaired Children: A Longitudinal Case Study. Fu-Jen Catholic University: Unpublished M. A. Thesis.

Chin, S. B., Finnegan, K. R., & Chung, B. A. (2001). Relationshipsamong types of speech intelligibility in pediatric users ofcochlear implants. Journal of Communication Disorders, 34,187–205.

Ciocca, V., Francis, A. L., Aisha, R., & Wong, L. (2002). Theperception of Cantonese lexical tones by early-deafened co-chlear implantees. The Journal of the Acoustical Society ofAmerica, 111, 2250–2256.

Cruttenden, A. (1981). Falls and rises: meanings and universals.Journal of Linguistics, 17, 77–91.

Faulkner, A., Rosen, S., & Smith, C. (2000). Effects of the salienceof pitch and periodicity information on the intelligibility offour-channel vocoded speech: Implications for cochlear im-plants. The Journal of the Acoustical Society of America, 108,1877–1887.

Fon, J. (1997). What are tones really like? An acoustic-based studyof Taiwan Mandarin Tones. National Taiwan University: Un-published M. A. Thesis.

Green, T., Faulkner, A., & Rosen, S. (2002). Spectral and tempo-ral cues to pitch in noise-excited vocoder simulation of contin-uous-interleaved-sampling cochlear implants. The Journal ofthe Acoustical Society of America, 112, 2155–2164.

Ho, A. T. (1976). The acoustic variation of Mandarin tones.Phonetica, 33, 353–367.

Kirk, K. I., Diefendorf, E., Riley, A., & Osberger, M. J. (1995).Consonant production by children with multichannel cochlearimplants or hearing aids. Advances in Oto-Rhino-Laryngology,50, 154–159.

Lee, K. Y. S., van Hasselt, C. A., Chiu, S. N., & Cheung, D. M. C.(2002). Cantonese tone perception ability of cochlear implantchildren in comparison with normal-hearing children. Interna-tional Journal of Pediatric Otorhinolaryngology, 63, 137–147.

Lee, T. H. -T. (1993). Theoretical issues in language developmentand Chinese child language. In C. -T. J. Huang & Y.-H. A. Li(Eds.), New Horizons in Chinese Linguistics (pp. 293–356).Boston: Kluwer Academic Publishers.

Li, C. N., & Thompson, S. A. (1977). The acquisition of tone inMandarin-speaking children. Journal of Child Language, 4,185–199.

Lieberman, P. (1967). Intonation, perception, and language. Cam-bridge, MA: MIT Press.

Loizou, P. C. (1998). Introduction to cochlear implants. IEEEEngineering in Medicine and Biology Magazine: the QuarterlyMagazine of the Engineering in Medicine & Biology Society, 18,32–42.

Miyamoto, R. T., Svirsky, M., Kirk, K. I., Robbins, A. M., Todd, S.,& Riley, A. (1997). Speech intelligibility in children withmultichannel cochlear implants. Advances in Oto-Rhino-Lar-yngology, 52, 198–203.

Moog, J. S., & Geers, A. E. (1999). Speech and language acquisi-tion in young children after cochlear implantation. Otolaryn-gologic Clinics of North America, 32, 1127–1141.

Nikolopoulos, T. P., O’Donoghue, G. M., & Archbold, S. (1999).Age at implantation: its importance in pediatric cochlear im-plantation. Laryngoscope, 109, 595–599.

O’Halpin, R. (2001). Intonation issues in the speech of hearingimpaired children: Analysis, transcription and remediation.Clinical Linguistics and Phonetics, 15, 529–550.

O’Donoghue, G. M., Nikolopoulos, T. P., Archbold, S. M., & Tait,M. (1999). Cochlear implants in young children: The relation-ship between speech perception and speech intelligibility. Earand Hearing, 20, 419–425.


Ouellet, C., & Cohen, H. (1999). Speech and language develop-ment following cochlear implantation. Journal of Neurolin-guistics, 12, 271–288.

Robbins, A. M. (2000). Language development. In S. B. Waltzman& N. L. Cohen (Eds.) Cochlear Implants (pp. 269–283). NewYork: Thieme Medical Publishers.

Robbins, A. M., Bollard, P. M., & Green, J. (1999). Languagedevelopment in children implanted with the CLARION co-chlear implant. The Annals of Otology, Rhinology & Laryngol-ogy, 177 (Suppl.), 113–118.

Rosen, S. (1992). Temporal information in speech: Acoustic,auditory and linguistic aspects. Philosophical Transactions ofthe Royal Society of London Series B—Biological Sciences, 336,367–373.

Serry, T. A., & Blamey, P. J. (1999). A 4-year investigation into phoneticinventory development in young cochlear implant users. Journal ofSpeech, Language, and Hearing Research, 42, 141–154.

Sharma, A., Dorman, M. F., & Spahr, A. J. (2002). A sensitiveperiod for the development of the central auditory system inchildren with cochlear implants: Implications for age of im-plantation. Ear and Hearing, 23, 532–539.

Shih, C. -L. (1988). Tone and intonation in Mandarin. Workingpapers of the Cornell Phonetics Laboratory, 3, 83–109.

Shriberg, L. D., & Lof, G. L. (1991). Reliability studies in broadand narrow phonetic transcription. Clinical Linguistics &Phonetics, 5, 225–279.

Snow, D. (1998). Children’s imitations of intonation contours: arerising tones more difficult than falling tones? Journal ofSpeech, Language, and Hearing Research, 41, 576–587.

Su, A. (1985). The Acquisition of Mandarin Phonology by Taiwan-ese Children. Fu-Jen Catholic University: Unpublished M. A.Thesis.

Svirsky, M. A., & Chin, S. B. (2000). Speech production. In S. B.Waltzman & N. L. Cohen (Eds.), Cochlear Implants (pp. 293–309). New York: Thieme Medical Publishers.

Svirsky, M. A., Robbins, A. M., Kirk, K. I., Pisoni, D. B., &Miyamoto, R. T. (2000). Language development in profoundlydeaf children with cochlear implants. Psychological Science,11, 153–158.

Tobey, E. A., Pancamo, S., Staller, S. J., Brimacombe, J. A., &Beiter, A. L. (1991). Consonant production in children receiv-ing a multichannel cochlear implant. Ear and Hearing, 12,23–31.

Tse, J. K. (1978). Tone acquisition in Cantonese: a longitudinalcase study. Journal of Child Language, 5, 191–204.

Wei, W. I., Wong, R., Hui, Y., Au, D. K., Wong, B. Y., Ho, W. K.,Tsang, A., Kung, P., & Chung, E. (2000). Chinese tonal lan-guage rehabilitation following cochlear implantation in chil-dren. Acta Oto-Laryngologica, 120, 218–221.

Whalen, D. H., & Xu, Y. (1992). Information for Mandarin tonesin the amplitude contour and in brief segments. Phonetica, 49,25–47.

Xu, L., Tsai, Y., & Pfingst, B. E. (2002). Features of stimulationaffecting tonal-speech perception: implications for cochlearprostheses. The Journal of the Acoustical Society of America,112, 247–258.

Yang, Y. -F. (1989). The vowels and the perception of Chinesetones. Acta Psychologica Sinica, 34, 29–34.

Yavas, M. S. (1998). Naturalness and markedness. PhonologyDevelopment and Disorders (pp. 169–191). San Diego: SingularPublishing Group.

Yue-Hashimoto, A. O. (1980). Word play in language acquisition:a Mandarin case. Journal of Chinese Linguistics, 8, 181–204.
