a comparison of vowel productions in children with cochlear implants

Fax +41 61 306 12 34E-Mail [email protected]

Original Paper

Folia Phoniatr Logop 2011;63:154–160

DOI: 10.1159/000318879

A Comparison of Vowel Productions in Prelingually Deaf Children Using Cochlear Implants, Severe Hearing-Impaired Children Using Conventional Hearing Aids and Normal-Hearing Children

N. Baudonck a K. Van Lierde a I. Dhooge a P. Corthals a, b

a Department of Otorhinolaryngology, Audiologic and Logopaedic Sciences, Ghent University, and

b Health Care Department, University College Ghent, Ghent , Belgium

Introduction

Speech in prelingually deaf children is characterized by deficits in articulation of consonants and vowels [1] . Vowel production by hearing-impaired children con-tains certain types of articulatory errors such as diph-thongation, substitutions and neutralization, and supra-segmental errors such as nasalization and hoarseness [1–5] .

Cochlear implantation has become a standard proce-dure in the treatment of prelingually deaf children. Even though these implants primarily facilitate speech percep-tion, they are also an important aid to the development of several aspects of speech production such as overall intelligibility [6] , suprasegmental features [7] and the production of consonants [8] and vowels [9] . Serry and Blamey [9] found that the production of vowels, and es-pecially of monophthongs, improved after implantation. Monophthongs were acquired earlier and more accurate-ly than diphthongs and consonants, suggesting a relative ease of production of monophthongs compared to the other sound classes [9] . This difference between vowel and consonant production was confirmed by Van Lierde et al. [8] , who found normal vowel production in prelin-gually deaf children with cochlear implants (CI), whereas consonant production showed several deviations such as distortions, substitutions and omissions. However, in the study by Tobey et al. [10] , the accuracy of phoneme pro-

Key Words

Vowel production � Deafness, prelingual � Children �

Cochlear implants

Abstract

Objective: The purpose of this study was to compare vowel

productions by deaf cochlear implant (CI) children, hearing-

impaired hearing aid (HA) children and normal-hearing (NH)

children. Patients and Methods: 73 children [mean age: 9; 14

years (years;months)] participated: 40 deaf CI children, 34

moderately to profoundly hearing-impaired HA children

and 42 NH children. For the 3 corner vowels [a], [i] and [u], F 1 ,

F 2 and the intrasubject SD were measured using the Praat

software. Spectral separation between these vowel for-

mants and vowel space were calculated. Results: The sig-

nificant effects in the CI group all pertain to a higher intra-

subject variability in formant values, whereas the significant

effects in the HA group all pertain to lower formant values.

Both hearing-impaired subgroups showed a tendency to-

ward greater intervowel distances and vowel space. Conclu-sion: Several subtle deviations in the vowel production of

deaf CI children and hearing-impaired HA children could be

established, using a well-defined acoustic analysis. CI chil-

dren as well as HA children in this study tended to overar-

ticulate, which hypothetically can be explained by a lack of

auditory feedback and an attempt to compensate it by pro-

prioceptive feedback during articulatory maneuvers.

Copyright © 2010 S. Karger AG, Basel

Published online: October 8, 2010

Nele Baudonck UZ Gent 2P1, Dienst Logopedie De Pintelaan 185 BE–9000 Ghent (Belgium) Tel. +32 9332 4462, Fax +32 9332 4993, E-Mail nele.baudonck @ ugent.be

© 2010 S. Karger AG, Basel1021–7762/11/0633–0154$38.00/0

Accessible online at:www.karger.com/fpl

Vowel Production in Children with CI Folia Phoniatr Logop 2011;63:154–160 155

duction in CI children was even lower for vowels (61.6%) than for consonants (68%).

To describe the quality of vowel production, one can apply either a subjective descriptive analysis such as pho-netic transcription, or an objective acoustic analysis of formant frequencies. The first two formants (F 1 and F 2 ) are considered to be the most important because, based on those two formants, a listener will be able to identify a specific vowel [11] . Determination of F 1 and F 2 frequen-cies of vowels offers the possibility to describe vowels in terms of high/low and front/back placement of the tongue in the oral cavity. The vowels [a], [i] and [u] represent the extreme articulatory positions of the tongue in English as well as in Dutch [12] . Representing the formant frequen-cies of these vowels in an F 1 :F 2 diagram yields a so-called ‘vowel triangle’. This triangle is a graphic representation of the articulation space for vowel production with [a], [i] and [u] acting as ‘corner vowels’. In hearing-impaired pa-tients, several abnormalities concerning formant fre-quencies and vowel space are described, such as central-ization of the first two formants [13] and small articula-tion movements, which implies a reduction in articulation space and vowel triangle size [2] . Yet, children with CI experience show vowel features closer to normal values. Uchanski and Geers [14] compared the second formant frequencies of the American English anterior vowel [i] and posterior vowel [$] of CI children with those of nor-mal-hearing (NH) children and concluded that nearly all CI children produced the vowels with F 2 values within normal limits (88 and 87% of the children, respectively, for [i] and [$]). Also the CI children in the study by Cam-pisi et al. [15] showed F 1 [a] and F 2 [a] values comparable to those of NH children. Yet, in the study by Liker et al. [16] , the mean F 1 [a] of the CI children was significantly lower in comparison with NH controls at 2 of the 3 test moments, indicating reduced movements of the jaw, re-sulting in a smaller vowel space. The CI children also showed fronted vowel space due to consistently higher F 2 frequencies in comparison with the NH controls. The F 2 frequencies were higher for the back vowels, indicating that the vowel space in the CI children is, on average, smaller in comparison with the controls. Liker et al. [16] explained the fronting of the vowel space by the tendency of therapists, family and the children themselves to move the articulation to where it is more visible. Hocevar-Boltezar et al. [17] found a decrease in F 1 of the Slovenian vowels [u] and [i] in prelingually deaf children 6 months after implantation. Those changes caused better phono-logical differentiation between both vowels and a postim-plant vowel space expansion.

It seems that acoustic analysis can reveal more and dif-ferent information on vowel production in hearing-im-paired children than transcriptions can do. Acoustic analyses of vowel formants may reveal more subtle dis-tinctions which are not crossing the phoneme bound-aries. In the study by Horga and Liker [18] , CI children approached the vowel space of NH children more than profoundly deaf hearing aid (HA) children. The main purpose of the present study was to compare the vowel production of prelingually deaf CI children with moder-ately to profoundly hearing-impaired HA children who had better hearing thresholds, and NH children. An acoustical analysis of the three corner vowels [a], [i] and [u] was made using the Praat software [19] . In a previous study by Baudonck et al. [20] concerning consonant pro-duction, deaf CI children were also compared with mod-erately to profoundly deaf HA children. In this study, the CI children performed better although the HA children had better thresholds. Based on those results concerning consonant production and the improved vowel produc-tion after implantation as described in other studies, the authors of the present study hypothesize that vowel pro-duction by deaf CI children approaches the vowel pro-duction by NH children at least as much as the produc-tion by moderately to profoundly hearing-impaired HA children does.

Subjects and Methods

The protocol was approved by the ethics committee of the Uni-versity of Ghent (reference No.: 06017) and was in accordance with the ethical standard stipulated in the Helsinki Declaration for research involving human subjects.

Subjects Seventy-four prelingually hearing-impaired children, all en-

rolled in Flemish oral/aural rehabilitation programs before the age of 3 years, were selected to participate in this study. They all suffered from nonsyndromal congenital hearing loss, and each child received its first HA before the age of 3 years. A minimal nonverbal intelligence score of 80 and use of the Dutch oral com-munication mode was required. Forty prelingually deaf children [mean age: 8; 8 years (years;months)] received a multichannel CI at an average age of 2; 8 years. Thirty-one of them were implanted with the Nucleus implant (Cochlear Corporation), 5 with the Di-gisonic implant (Neurelec) and 4 with the Clarion implant (Ad-vanced Bionics). Thirty-four children (mean age: 9; 6 years) were bilateral conventional HA users with a moderate-to-profound hearing loss. All children had at least 2 years of experience with their current device (HA or CI), which was fitted by experienced audiologists. During testing, all devices were adjusted to each child’s usual amplification and processing settings. The control group consisted of 42 NH children with a mean age of 9; 3 years.

Baudonck /Van Lierde /Dhooge /Corthals Folia Phoniatr Logop 2011;63:154–160 156

For the 3 subgroups, information concerning chronological age and, where appropriate, age at first HA fitting, the most recent better-ear unaided hearing threshold (pure tone audiometry, PTA), the most recent free-field aided hearing threshold, and age at implantation is provided in table 1 .

A Mann-Whitney U test showed that mean chronological age, duration of initial hearing loss and mean nonverbal intelligence quotient of the CI and HA groups did not differ significantly (p 1 0.05). Regarding the audiometric data, recent better-ear unaided thresholds were significantly better in both HA groups (p ! 0.001) compared to the CI group, while the aided thresholds in the free field condition showed no marked difference between the HA and CI groups.

Recording Procedure All recordings were made in a quiet room at the children’s

school or in the University Hospital Ghent. Speech samples were elicited by means of a picture naming test [21] in which children are asked to name black-and-white drawings of common objects and actions in order to elicit instances of all Dutch single speech sounds in all permissible syllable positions. Those speech samples were recorded directly on a portable computer (Toshiba M70 with Harman Kardon sound card), under standard conditions using a condenser stereo microphone (Sony ECM-MS907). For the re-cording, the examiners used the Praat software [19] with a sam-pling frequency of 44,100 Hz.

Acoustic Analyses For the 3 corner vowels [a], [i] and [u] of each child, 10 mid-

vowel fragments with stable formant patterns were selected from the same 10 monosyllabic words or stressed syllables of bisyllabic words. The midvowel segments were selected using visual inspec-tion of the oscillogram and the spectrogram. Fragments showing the most stable formant patterns were identified, extracted with-out formant transitions, and regrouped into one comprehensive sound fragment (mean total duration: 0.81 s), which was then sub-mitted to the formant detection algorithm. The Burg algorithm for formant detection was used with a 0.025-second gaussian analysis window (effective duration) and +6 dB preemphasis from 50 Hz onwards. The formant search range was limited to 8,000 Hz to adapt the algorithm to the dimensions of children’s vocal tracts. For each vowel fragment of each child, a specific amount of linear predictive coding coefficients was determined, based on visual inspection of the spectrogram and spectrum, and on intrasubject variability in formant values (the number of coefficients was cho-

sen in such a manner that the formant SD of all analysis frames were as low as possible). For further verification, the obtained for-mant values were compared with the formant values from Eng-lish-speaking children of comparable ages, as described by Vor-perian and Kent [22] as well as with formant values from Dutch men as described by Pols [23] . In case of doubt, because of a salient contrast with the values from the latter references (taking into ac-count language and age differences) or because of high intra-subject variability, another set of 10 midvowel segments was se-lected and the whole procedure repeated in order to find a plau-sible formant value.

The 50th percentile value of the first (F 1 ) and second formant frequency (F 2 ) as well as the intrasubject SD were measured for each vowel of each child using the Praat software [19] . The euclid-ian distance between the corner vowels along the axes of an F 1 :F 2 scatter plot as well as the surface area of the vowel triangle were calculated. To guarantee a standardized protocol for all calcula-tions, a Praat script (by P.C.) was systematically applied. Those calculations were completed accurately by the first author (N.B.) individually, consistently following the above-described proce-dure. As a control procedure, a remeasurement of formant values was made by the last author (P.C.) for 8 randomly selected chil-dren (10%). For F 1 , the mean difference between the values by both examiners was 20 Hz (range: 3–61 Hz). For F 2 , the mean differ-ence between the values by both examiners was 58 Hz (range: 2–101 Hz).

Statistical Analyses SPSS for Windows (version 15.0) was used for the statistical

analysis. The results of the CI children were compared with the results of the HA and NH children. The Kolmogorov-Smirnov test ( � = 0.01) and the Shapiro-Wilk test were performed to study the distribution of the variables. The variables ‘F 2 [a], [i] and [u]’ and ‘intrasubject SD F 1 and intrasubject SD F 2 [a], [i] and [u]’ de-viated from a normal distribution (p ! 0.05 on either the Kol-mogorov-Smirnov or the Shapiro-Wilk test), whereas the remain-ing variables were normally distributed. For the comparisons be-tween the 3 subgroups, a one-way ANOVA in combination with a post hoc Scheffé test was applied in case of normal distribution. In case of a deviation from a normal distribution, Kruskal-Wallis tests were executed to document possible differences between the 3 subgroups. When the obtained p was smaller than 0.05, pairwise comparisons were executed using a Mann-Whitney procedure. The significance level was set at � = 0.05 in all tests.

Table 1. Demographic information on participants in this study

Number Chronological ageyears

Age at deafnessmonths

Age at first HAmonths

Unaided thresholdPTA dB HL

Aided threshold PTA dB HL

Age at implan-tation, months

NH children 42 9;3 (4;3–15;3) – – – – –HA children 34 9;5 (4;1–14;2) 0 (0–0) 13 (2–36) 80 (58–105) 34 (17–50) –CI children 40 8;8 (4;2–15;5) 0 (0–10) 11 (1–36) 109 (73–120) 33 (11–50) 32 (6–131)

V alues are means with ranges in parentheses.


Results

Table 2 represents the mean formant frequencies, the mean intrasubject SD of the formant frequencies, the mean intervowel distances, the mean vowel space and the p values of the comparisons between the 3 subgroups. Concerning vowel [a], the CI children did not show sig-nificantly different formant values compared to the NH children, while there was a significantly lower F 2 in the HA children compared to the NH children. The CI chil-dren, on the other hand, showed a slightly higher but sig-nificant intrasubject variability for both formants com-pared to the NH children, as shown by the higher SD. No significant differences were detected between the CI and the HA groups for vowel [a] parameters. For the vowel [i], a scarcely significant lower F 2 value was found for the HA group compared to the CI group. No differences could be found between the formants of the vowel [i] of CI and NH children. Concerning the vowel [u], there were no statisti-cally significant differences between both hearing-im-paired groups. The HA children showed a significantly lower F 1 and F 2 [u] than the NH children. In the CI group, a higher intrasubject variability in the F 1 [u] was found. As seen in table 2 , the hearing-impaired children system-atically showed the highest values for all intervowel dis-

tances /i-u/, /a-i/ and /u-a/ as well as for the vowel space, whereas the NH children systematically had the lowest values. In both hearing-impaired groups, a tendency to-ward an increased vowel space compared to NH children was found. The /a-u/ difference between the NH children and both hearing-impaired subgroups, and the /a-i/ dif-ference between the CI and NH children were statisti-cally significant, while no significant differences between the HA and CI children were found. The vowel triangles of the 3 subgroups are represented in figure 1 .

Discussion

The main purpose of the present study was to compare prelingually deaf CI children with moderately to pro-foundly hearing-impaired HA children and NH children with regard to vowel production. An acoustical analysis of the three corner vowels [a], [i] and [u] was made using the Praat software. The hypothesis was that vowel pro-duction in CI children would be closer to normal than in HA children. Generally speaking, both hearing-impaired groups showed several significant differences when com-pared to the NH group. The significant effects in the CI group all pertain to a higher intrasubject variability in

Table 2. Formant frequencies, intrasubject SD of the formant frequencies, intervowel distances, vowel space of the acoustic analysis and p of comparisons between 3 subgroups

CI(n = 40)

HA(n = 34)

NH(n = 42)

Kruskal-Wallis/ANOVA

CI-NHpost hoc

HA-NHpost hoc

CI-HApost hoc

p p p p

F1 [a], Hz 9928156 9838150 9328161 0.182Intrasubject SD (F1) 127848 116862 97840 0.024* 0.026* 0.253 0.640

F2 [a], Hz 1,8338240 1,7718240 1,9338257 0.017* 0.188 0.019* 0.557Intrasubject SD (F2) 191867 162878 154851 0.034* 0.045* 0.882 0.170

F1 [i], Hz 400865 401850 422856 0.166Intrasubject SD (F1) 67857 60825 54815 0.258

F2 [i], Hz 3,0408249 2,9028227 3,0268242 0.030* 0.965 0.082 0.052Intrasubject SD (F2) 212882 199870 175859 0.062

F1 [u], Hz 431875 411866 464858 0.003* 0.087 0.003* 0.426Intrasubject SD (F1) 65823 59818 58820.1 0.008* 0.008* 0.170 0.537

F2 [u], Hz 1,1478223 1,0388189 1,2148254 0.004* 0.403 0.004* 0.122Intrasubject SD (F2) 267891 230874 232883 0.086

Vowel space, kHz2 5398153 5428165 4448200 0.020* 0.054 0.055 0.996i-u 1,8838289 1,8638333 1,8088351 0.488a-i 5918157 5828146 5058158 0.023* 0.043* 0.095 0.971u-a 5608147 5738123 4628155 0.001* 0.010* 0.005* 0.967

F or each formant of each vowel, the mean intrasubject variability is given. For all mean values, the SD is given. * p < 0.05.


formant values, whereas the significant effects in the HA group all pertain to lower formant values.

The first remarkable formant differences were related to the vowel [a]. Compared to NH children, the F 2 [a] was lower in the HA children. The CI children showed a high-er intrasubject variability for both formants of the vowel [a], as measured by SD, which hypothetically indicates some articulatory vacillation in the production of [a] even after implantation. However, based on the fact that the F 2 values of the CI children in the present study resemble normal values more than those of the HA children do, one can conclude that cochlear implantation probably has a subtly positive impact on the place of articulation of [a]. In the study by Campisi et al. [15] , the values of CI children were compared with age-matched control data [24] . In agreement with the present study, no differences between CI and NH children were found for formant fre-quencies of [a].

Concerning the formants of the front vowel [i] and dorsal vowel [u], the CI children in this study did not show significant differences from their NH peers. This is in agreement with Uchanski and Geers [14] , who found that nearly all of 181 CI patients produced the front vow-el [i] and back vowel [$] with F 2 values within normal limits. On the other hand, for the HA children in the

present study, both formants of the vowel [u] were de-creased in comparison with normal values. Hypotheti-cally, the decreased F 1 [u] can be explained by the fact that HA children usually show more residual hearing at low frequencies. On that account, it seems that they seek to produce the first formant, which is also the most intense among formants, as low as possible in an attempt to en-hance auditory feedback. In the CI children, the F 1 [u] was also lower compared to the NH children although this difference was not significant. As was the case for [a], CI children showed an increased intrasubject variability in F 1 [u], as measured by SD.

The decreased F 2 [u] in the HA children can indicate a tendency toward a more dorsal articulation of the vow-el [u]. The production of the [u] turns out to be a poten-tially difficult speech parameter for hearing-impaired children, especially for children using HA. Concerning the front vowel [i], the F 2 was slightly decreased in the HA children in comparison with NH and CI children. This trend was also seen for the vowels [a] and [u]. It seems to support Boone’s theory that hearing-impaired children tend to keep their tongues too far back in their mouths [25] . Nevertheless, one should pay attention when consid-ering only the means of values for F 2 in the three sub-groups in the present study. Although significantly lower mean F 2 results were found in the HA group, a huge over-lap between the three subgroups cannot be denied when looking at the range and SD. This means that although, on average, the HA group shows decreased mean F 2 val-ues in comparison with the normal hearing controls, many hearing-impaired children still perform within normal ranges.

In the CI children, no significant differences were found for F 2 frequencies, thus backing or fronting of the vowel space was not observed. Liker et al. [16] , however, found fronted vowel space in CI children due to consis-tently higher F 2 frequencies. In their study, F 2 frequencies of back vowels were clearly increased, yielding vowel space values for CI children that were, on average, small-er than in NH controls. This concomitant reduction in vowel space is also in contrast with the results of the pres-ent study. The increase in F 1 [a] values, although not sig-nificant, caused more salient /i-a/ and /u-a/ contrasts in both hearing-impaired groups of the present study, in comparison with the NH subjects. As a consequence, the area of the vowel triangle tends to be larger in the CI as well as in the HA groups. In some respect, these results are rather surprising and contradictory to previous re-search in which a reduction in vowel triangle size was found in hearing-impaired patients [13] .

1,500

3,500

1,000

3,000 2,500 2,000 1,500 1,000 500

(O

PE

N)

1st

form

an

t(C

LOS

ED

)

( FRONT) 2nd formant (BACK ) scri

pt

P.C

.

CI childrenHA childrenNH children

5,000

Co

lor

ve

rsio

n a

va

ila

ble

on

lin

e

Fig. 1. Vowel triangle in CI, HA and NH children.


This contradiction with previous research partly may have been brought about by language differences, but hy-pothetically, the phenomenon of an increased vowel space could also be explained by a tendency of therapists and family to exaggerate their articulation movements in order to facilitate speech reading. Probably, hearing-im-paired children imitate this trend and even use the exag-gerated movements themselves in order to enhance pro-prioceptive feedback. Thus, these unexpected results can be explained by the same simple rationale, i.e. excessive articulation movements which hearing-impaired chil-dren probably use, imitating their models in order to im-prove visual and proprioceptive feedback. Indeed, since neonatal hearing screening is well organized in Flanders, rehabilitation typically starts at a very early age. Hearing-impaired children receive speech therapy from early childhood on and are trained intensively to compensate speech production deficits. Liker et al. [16] found a small-er vowel space for CI children in comparison with NH controls. It is worth mentioning that the children in the study by Huber et al. [24] started rehabilitation at a later age (mean: 4; 5 years; range: 1; 10–7; 8 years) than the par-ticipants in the present study (before the age of 3 years). Hypothetically, early and intensive articulation training in order to compensate speech production deficits could have led to overarticulation in the hearing-impaired chil-dren of this study. Yet, overarticulation of vowels has no harmful effect on speech intelligibility.

Besides processing spoken utterances in a purely lin-guistic way by extracting canonical categories such as phoneme and word units, listeners simultaneously en-code nonlinguistic perceptual characteristics [26] . Those are audible deviations from the pronunciation standard that do not necessarily interfere with a correct identifica-tion of speech sound categories or overall intelligibility. For instance, the degree of fine-grained variation in syn-thesized speech is linked to both intelligibility and natu-ralness ratings [27] . The formant values and dispersions in the speech of the CI and HA children of the present study are possible candidates too. Thus, besides being correlated with intelligibility, the observed subtle acous-tic formant distinctions, some of which can remain un-detected by the naked ear, may perhaps signal deficits in ‘naturalness’ of a child’s speech. When it comes to vowel production in early-rehabilitated hearing-impaired chil-dren, further research should pay more attention to speech ‘naturalness’ rather than to intelligibility.

The authors are aware that the variability in the sev-eral subgroups is great, and that the number of possibly influencing factors is considerable. Multiple variables can

affect the speech performance of children. The age at on-set of deafness, age at first hearing device, age at implan-tation, type of device and degree of hearing loss are known variables in this study. In contrast, variables such as the quality of an educational program, children’s and parents’ self-efficacy beliefs [28] , children’s motivation [29] as well as other family, school and community factors are unknown and are all factors which can influence speech production. Since there was a considerable range of age at implantation relative to age at time of testing in this study, there was also a wide range of CI experience. It would be very interesting to investigate the influence of CI experience on the evolution of vowel formants. Yet, in the present study, the subgroups were too small to com-pare reliably when children were regrouped on the basis of a clear difference in months of CI experience. A sub-stantial number of patients are needed to avoid variabil-ity and find significant differences. However, such a number of patients are not available in Flanders. It should also be noted that the CI group in Flanders – and, as a consequence, the participants in this study – consists of unilaterally as well as bilaterally implanted children. To what extend bilateral implantation has an impact on the speech production skills of prelingually deaf children is a subject of further research.

In conclusion, we can state that, in this study, several subtle deviations in the vowel production of prelingually deaf children could be established using an objective acoustic analysis. Children using HA showed a tendency toward a more dorsal articulation of the vowels, which was not seen in the CI children. Both hearing-impaired groups showed a tendency toward overcompensation while articulating, possibly following the example of par-ents and therapists and, perhaps, as a side effect of early and intensive rehabilitation. This conclusion can be ex-plained by a lack of auditory feedback and the search for more proprioceptive feedback during articulatory ma-neuvers.

Acknowledgments

The authors wish to express their deep appreciation for the hearing-impaired subjects who spent their time on this research. We gratefully acknowledge the contribution of the speech and language pathologists of the rehabilitation centers Sint-Lieven-spoort Ghent, De Poolster Brussels and Overleie Kortrijk and the schools Spermalie Bruges and Kids Hasselt. Thanks to Birgit Phil-ips and Evelien D’Haeseleer for their advice and help. N. Baudonck is funded by the Research-Foundation Flanders.


References

1 Smith CR: Residual hearing and speech pro-duction in deaf children. J Speech Hear Res 1975; 18: 795–811.

2 Angelocci AA, Kopp GA, Holbrook A: The vowel formants of deaf and normal-hearing eleven-to-fourteen-year-old boys. J Speech Hear Disord 1964; 29: 156–160.

3 Hudgins CV, Numbers GD: An investigation of the intelligibility of the speech of the deaf. Genet Psychol Monogr 1942; 25: 289–392.

4 Markides A: The speech of deaf and partial-ly-hearing children with special reference to factors affecting intelligibility. Br J Disord Commun 1970; 5: 126–140.

5 Nober EH: Articulation of the deaf. Except Child 1967; 33: 611–621.

6 Allen MC, Nikolopoulos TP, O’Donoghue GM: Speech intelligibility in children after cochlear implantation. Am J Otol 1998; 19: 742–746.

7 Tye-Murray N, Spencer L, Woodworth GG: Acquisition of speech by children who have prolonged cochlear implant experience. J Speech Hear Res 1995; 38: 327–337.

8 van Lierde KM, Vinck BM, Baudonck N, de Vel E, Dhooge I: Comparison of the overall intelligibility, articulation, resonance, and voice characteristics between children using cochlear implants and those using bilateral hearing aids: a pilot study. Int J Audiol 2005; 8: 452–465.

9 Serry TA, Blamey PJ: A 4-year investigation into phonetic inventory development in young cochlear implant users. J Speech Lang Hear Res 1999; 42: 141–154.

10 Tobey EA, Geers AE, Brenner C, Altuna D, Gabbert G: Factors associated with develop-ment of speech production skills in children implanted by age five. Ear Hear 2003; 24: 36S–45S.

11 Steven KN: Articulatory-acoustic-auditory relationships; in Hardcastle WJ, Laver J (eds): The Handbook of Phonetic Sciences. Oxford, Blackwell, 1999, pp 462–506.

12 Pols LCW, van der Kamp LJT, Plomp R: Per-ceptual and physical space of vowel sounds. J Acoust Soc Am 1969; 46: 458–467.

13 Monsen R: Normal and reduced phonologi-cal space: the production of English vowels by deaf adolescents. J Phon 1976; 4: 189–198.

14 Uchanski RM, Geers AE: Acoustic charac-teristics of the speech of young cochlear im-plant users: a comparison with normal-hear-ing age mates. Ear Hear 2003; 24: 90S–105S.

15 Campisi P, Low AJ, Papsin BC, Mount RJ, Harrison RV: Multidimensional voice pro-gram analysis in profoundly deaf children: quantifying frequency and amplitude con-trol. Percept Mot Skills 2006; 103: 40–50.

16 Liker M, Mildner V, Sindija B: Acoustic anal-ysis of the speech of children with cochlear implants: a longitudinal study. Clin Linguist Phon 2007; 21: 1–11.

17 Hocevar-Boltezar I, Boltezar M, Zargi M: The influence of cochlear implantation on vowel articulation. Wien Klin Wochenschr 2008; 120: 228–233.

18 Horga D, Liker M: Voice and pronunciation of cochlear implant speakers. Clin Linguist Phon 2006; 20: 211–217.

19 Boersma P, Weenink D: Praat: doing phonet-ics by computer (version 5.1.05). http://www.praat.org/ (accessed May 1, 2009).

20 Baudonck N, Dhooge I, D’Haeseleer E, van Lierde K: A comparison of consonant pro-ductions between children using cochlear implants or hearing aids. Int J Pediatr Oto-rhinolaryngol 2010; 74: 416–421.

21 van Borsel J: Articulation in Down’s syn-drome adolescents and adults. Eur J Disord Commun 1996; 31: 415–444.

22 Vorperian HK, Kent RD: Vowel acoustic space development in children: a synthesis of acoustic and anatomic data. J Speech Lang Hear Res 2007; 50: 1510–1545.

23 Pols LCW: Spectral Analysis and Identifi-cation of Dutch Vowels in Monosyllabic Words. Amsterdam, VU University, 1977.

24 Huber JE, Stathopoulos ET, Curione GM, Ash TA, Johnson K: Formants of children, women, and men: the effects of vocal inten-sity variation. J Acoust Soc Am 1999; 106: 1532–1542.

25 Boone D: Modification of the voices of deaf children. Volta Rev 1966; 68: 686–692.

26 Nygaard LC: Perceptual integration of lin-guistic and non-linguistic properties of speech; in Pisoni DB, Remez RE (eds): The Handbook of Speech Perception. Oxford, Blackwell, 2005, pp 391–413.

27 Wouters J, Macon MW: Effects of prosodic factors on spectral dynamics. 2. Synthesis. J Acoust Soc Am 2002; 111: 428–438.

28 Bandura A, Barbaranelli C, Caprara GV, Pastorelli C: Multifaceted impact of self-effi-cacy beliefs on academic functioning. Child Dev 1996; 67: 206–222.

29 Eccles JS, Wigfield A, Schiefele U: Motiva-tion to succeed; in Eisenberg N (ed): Hand-book of Child Psychology. New York, Wiley, 1998, pp 1051–1075.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

a comparison of vowel productions in children with cochlear implants

Documents