a Speech Production Changes with the Use of a Multichannel Cochlear Implant in a Postlingually Hearing Impaired Adult Sandy Cummings Emily Groenewald Rene Hugo Department of Communication Pathology University of Pretoria Lida Miiller Department of Speech Pathology and Audiology University of Stellenbosch Mike van der Linde Department of Information Technology Division of Academic Computing University of Pretoria ABSTRACT Profoundly deaf cochlear implant users provide an interesting population in which to assess the role of distorted auditory feedback in speech, since their electrically stimulated hearing is significantly different from normal hearing. The aim of the study was to evaluate, by means of spectrograph^ and listener analyses, the speech production changes in a postlingually deafened adult with the use of a multichannel cochlear implant over time, compared to that of hearing aids as well as no-amplification. The results indicated significant improvements in the use of suprasegmental speech features as well in the production of specific segmental features of speech. OPSOMMING Uitermatige dowe pasiente met kogleere inplantings vorm 'n interessante populasie waarby die invloed van versteurde ouditiewe terugvoer op spraak ondersoek kan word, aangesien elektries-gestimuleerde gehoor betekenisvol verskil van normale gehoor. Die doel van die studie was 'n ondersoek na die invloed van 'n multikanaal kogleere inplanting oor 'n verloop van tyd, in vergelyking met geen ouditiewe versterking en versterking dmv gehoorapparate op spraakproduksie van 'n postlinguale dowe volwassene. 'n Vergelykende spektrografiese en luisteraaranalise is uitgevoer. Die resultate het betekenisvolle verbeteringe getoon in die suprasegmentele, sowel as spesifieke segmentele eienskappe van spraak. INTRODUCTION Research on the benefits of cochlear implants have in the past focused primarily on the speech perception of the postlingually deafened implant user (Dowell, Brown, Seligman & Clark, 1985; Dowell, Seligman, Blarney & Clark, 1987; Eddington, 1983; Cohen, Waltzman & Shapiro, 1985 and Schindler, Kessler, Rebscher, Yanda & Jackler, 1986). Improvements in en- vironmental sound recognition as well as improvements in auditory sensitivity (Thielemeir, Eisenberg & Brimacombe, 1982; Tyler, Gantz, McCabe, Lowder, Otto & Preece, 1985) have also been widely documented. An increasing number of researchers have begun to place emphasis on the speech production characteristics of postlingually deafened implant wearers, (Iler-Kirk & Edgerton, 1983; Leder, Spitzer, Milner, Flevaris- Phillips, Richardson & Kirchner, 1986; Tartter, Chute & Hellman, 1989; Waldstein, 1990; Lane & Webster, 1991; Economou, Tartter, Chute & Hellman, 1992; Perkell, Lane, Svirsky & Webster, 1992; Svirsky, Lane, Perkell & Wosniak, 1992). The lack of research directed at the speech production changes could be due to the fact that a postlingually acquired deafness does not necessarily lead to problems of speech degeneration (Ling, 1976). Studies designed to investigate the speech Die Suid-Afrikaanse Tydskrif vir Kommunikasieafwykings, Vol. 41, 1994 R ep ro du ce d by S ab in et G at ew ay u nd er li ce nc e gr an te d by th e P ub lis he r (d at ed 2 01 2) 4 Sandy Cummings, Emily Groenewald, Rene Hugo, Lida Miiller, Mike van der Linde characteristics of postlingually deaf subjects have, how- ever, found that longterm total auditory deprivation in hearing adults resulted in eventual deterioration into flat, unmodulated and dysprosodic voice, (Cowie, Doug- las-Cowie & Kerr, 1982). Waldstein (1990) investigated some effects of postlingual deafness on speech by ex- ploring selected properties of consonants, vowels and s u p r a s e g m e n t a l s in the speech of seven totally, postlingually deafened individuals. Their results dem- onstrated that postlingual deafness affects the produc- tion of all the classes of speech sounds. Re-introduc- tion of "auditory input" with the electrical stimulation of the auditory nerve should result in changes in speech production. These changes in speech production are pri- marily reliant on the type of information derived from stimulation and finally the implant listener's articula- tory adaptation to prolonged profound deafness, (Perkell et al., 1992). The most common changes in speech production with the use of multichannel cochlear implants can be de- scribed in terms of suprasegmental and segmental changes, although previous studies have indicated that the suprasegmental features are affected the most due to the presence of a profound hearing loss, (Leder et al., 1986). Waldstein, (1990) indicated that the results of his study do not support the hypothesis that the suprasegmental properties of speech show the greatest effects of a loss of auditory feedback, while the segmen- tal properties remain relatively unaffected. "Rather, auditory feedback appears to be implicated in monitor- ing and maintaining phonetic precision in all classes of speech sounds." Waldstein, (1990:211). Various s t u d i e s i n v e s t i g a t i n g the effects of multichannel cochlear implants on speech production have indicated that significant changes in speech pro- duction can occur with the use of a multichannel im- plant. Lane & Webster, (1991) have reviewed some of the studies investigating the speech of postlingually deafened speakers including several studies on the ef- fects of cochlear implants. They conclude that the ma- jority of studies implicate a role for audition in regulat- ing, directly or indirectly, several speech properties in- cluding voice quality, voice aspiration, vocal duration, pitch, intonation, stress, tempo, nasality and frication and plosive articulation, (Iler-Kirk & Edgerton 1983; Leder et al., 1986; Tartter et al., 1989; Economou et al., 1992; Perkell et al., 1992). METHOD AIMS The aim of the present study was to evaluate the speech production changes brought about by the use of a multichannel cochlear implant over time, in an adult with a profound postlingual sensorineural hearing loss. The subject's speech production skills were evaluated pre- and post-operatively by means of spectrographic and listener analyses. These evaluations included: - the spectrographic evaluation of the suprasegmental features of speech, in terms of sentence duration, fun- damental frequency variations and word stress (am- plification). - the spectrographic evaluation of the segmental fea- tures of speech, namely vowels and consonants. The vowels were evaluated in terms of vowel length and the formant relationships between F1 and F2. The con- sonants were evaluated in terms of the spectral noise band frequencies for fricatives, the spectral frequency ranges for the plosive sections of the stop sounds and the relative amplitude peaks of both the fricatives and plosives. - the listener's analysis of speech production was evalu- ated in terms of pitch variation, and vocal and pausal duration. - the results of the speech production changes with the multichannel cochlear implant over time were com- pared to the subject's speech production without any amplification and with binaural hearing aids. RESEARCH METHODOLOGY A single-case research study using a time series ex- perimental design was used in order to evaluate the speech p r o d u c t i o n changes with the use of a multichannel cochlear implant in an adult with a postlingual sensorineural hearing loss. The data was obtained by performing evaluations in the following phases: Pre-implant: No-amplification (NA); hearing aid (HA) Post-implant: 0-months (CI-0), 3-months (CI.3) and 6-months (CI-6) cochlear implant SUBJECT The subject (MC) lost his hearing due to meningitis at 19 years of age. He was implanted five years later. His pre-implant audiograms indicated a bilaterally to- tal sensorineural hearing loss with minimal benefit from binaural hearing aids as well as a vibrotactile device. His right ear was implanted on 31 October 1991 with the Cochlear 22-channel implant. All electrodes were easily inserted. The electrodes were programmed in the Bipolar + 1 mode using the MPEAK coding strategy. The subject (MC) was an Afrikaans speaker using a non- standard dialect. IMPLANT AND SPEECH PROCESSOR t The cochlear implant used in the present ^tudy is known as the Mini-System 22 comprising 22 pure plati- num bands supported on a flexible silicone rubber car- rier. The speech processor utilized is referred to as the MSP (Mini Speech Processor). The MPEAK coding strat- egy extracts and codes F1 and F2, where the F1 is rep- resented by the dominant spectral peak in the range from approximately 300-1000Hz and F2 between 800- 4000Hz. The estimate of F1 from the acoustic signal controls the selection of an apical electrode pair, while the estimate of F2 controls the selection of a basal elec- trode pair. The spectral energy in the regions of 2000- 2800Hz, 2800-4000Hz and above 4000Hz are presented to three basally located electrodes. The fundamental or voicing frequency determines the pulse rate of the stimu- lation and the amplitude of the acoustic signal deter- mines the stimulus intensity, (Cochlear, 1989). The South African Journal of Communication Disorders, Vol. 41, 1994 R ep ro du ce d by S ab in et G at ew ay u nd er li ce nc e gr an te d by th e P ub lis he r (d at ed 2 01 2) Speech Production Changes with the Use of a Multichannel Cochlear Implant in a Postlingually Hearing Impaired Adult HEARING AIDS The two Phonak PPC-L-4 hearing aids were utilized in the pre-implant evaluation condition. The maximum output of the hearing aids is 142 dB SSPL with a fre- quency range of 140 - 4900 Hz. The subject used the hearing aid on volume 2,5 with the following settings: LC - 1 , HC - 5, G - 6 and SSPL - 6. The subject also wore new standard acrylic ear moulds. STIMULI AND EQUIPMENT The data for the evaluation of the subject's speech production skills were determined objectively and sub- jectively by means of spectrographic and listener analy- ses respectively. For the spectrographic analysis of the suprasegmental features of speech, seven sentences vary- ing in length and including the various typical prosodic features of statements, questions and commands, were selected to be spectrographically analyzed. For the spectrographic analysis of the segmental features, eight isolated single vowels; / i /, / a /, = / y /, / oe /, / u /, / a /, / ο / & / ε / in CVC combinations were selected to be analyzed. The vowels selected were the most common vowels utilized in Afrikaans where the transition to diphtongs does not take place. The consonants selected were those consonants which are either ommitted or distorted in the presence of a profound hearing loss (Nickerson, 1975). The plosives, / t / f / k / , / p / and the fricatives / s / and / χ / were used in the final position and / s / and / f / were used in the initial position of the word. These consonants were analyzed within the con- text of two different words and preceded or followed by different vowels. For the listener analysis, the subject was required to read a paragraph in which the length, language, content and complexity was appropriate for the purpose of the evaluation as well as the subject's reading skills. The specified material wfas read into a Dynamic M 0 2 NCC microphone (15cm from the microphone) in an I AC soundproof and sound reverberation free unit. The re- cordings were made on a Direct Head Casette Deck. Nachamitchi, model 682 2X The speaker's distance from the microphone as well as the input attenuation were kept c o n s t a n t t h r o u g h o u t all the r e c o r d i n g s . Spectrographic analyses wrere conducted using the KAY DSP Sonagraph, Model 5500. The analysed data was graphically displayed on aiNEC/Multisync colour moni- tor. PROCEDURE Recording of data Recordings of all the stimulus words and sentences were made without any amplification of the signal in the pre-implant condition, with binaural hearing aids in the pre-implant condition, the following day after switch-on i.e., 0- months; 3-months and 6-months post- implant. Spectrographic analysis For the analysis of the suprasegmental features, a combination analysis setup of the sonagraph, whereby the waveform, amplitude and pitch trace of the sen- tences were displayed, was used for the measurements. The sentence duration was measured by applying time cursors on the speech wave display. Pitch variation was determined by using the frequency cursor readings on the computed pitch curves at the highest and lowest peaks in the pitch trace. Word stress or amplification was measured by the time cursors at the highest peak of computed amplitude curves. The analysis of segmental features of specific speech sounds focused on the duration and formant frequen- cies of vowels and the spectral characteristics of the frica- tives and the plosive energy of stop consonants. The vowels were analyzed up to 4000Hz and the consonants up to 8000Hz. A combination display of a wideband spectrogram in conjunction with a narrowband ampli- tude spectrum of the computed spectral energy between two time cursors was used. Time cursors were used in order to determine the length of the vowel. For the meas- urements of the formant frequencies, a stable middle portion of each vowel was selected in order to exclude the transitions to the consonants. Frequency cursors, in conjunction with the time cursors were utilized in order to determine the formant frequencies. The consonants were analyzed in terms of the spec- tral noise band frequencies for the fricatives (i.e., mini- mum and maximum frequencies of the fricatives), the spectral frequency range for the plosive sections of the stop sound (i.e., minimum and maximum frequencies of the plosive sections of the stop sounds) and the rela- tive amplitude peaks of the fricatives and plosives. Time and frequency cursors were used to determine the fre- quency boundaries as well as the relative amplitude peaks. Listener analysis For the listener's evaluation of the subject's speech production, a rating scale for listener's evaluation of pitch variation and vocal and pausal duration was used, (Coles, 1990). The recordings were presented to four trained listeners via Tandberg Educational headphones in a language laboratory by means of a Tandberg Edu- cational IS 10 cassette player. The listeners were asked to evaluate pitch variation, vocal duration and pausal duration by means of a five point rating scale (Appen- dix A). RESULTS RESULTS OF THE SPEECH PRODUCTION EVALUATION USING SPECTROGRAPHIC ANALY- SIS. It is important to note that there were no marked abnormal characteristics pertaining to the subject's suprasegmental and segmental features as found in some postlingually deafened adults. Spectrographic analyses were executed in order to detect any degen- eration of speech features which had possibly occurred during the period of profound deafness as well as any changes in speech production brought about by the use of the cochlear implant. Die Suid-Afrikaanse Tydskrif vir Kommunikasieafwykings, Vol. 41, 1994 R ep ro du ce d by S ab in et G at ew ay u nd er li ce nc e gr an te d by th e P ub lis he r (d at ed 2 01 2) 6 Sandy Cummings, Emily Groenewald, Rene Hugo, Lida M l l e r , Mike van der Linde Suprasegmental features The results of sentence duration across various au- ditory input situations are presented in Table 1. When analyzed, the majority of the sentences showed an increase in sentence duration at the six month cochlear implant level, as opposed to the no-amplifica- tion and hearing aid conditions. The increases in sen- tence duration occurred at either the cochlear implant 0-month or 3-month situation, with a highest mean duration for the seven sentences at 0-months. A quali- tative analysis of the individual sentences does, how- ever, give logical explanations for these results. The interpretation and explanation of the above findings will be given in the discussion of results. As can be seen in Table 2, the highest relative ampli- tude peaks for stressed words showed a small, however, consistent increase in the amplitude range with the use of the cochlear implant, with the values ranging from 33,ldB, 37,4dB and 37dB during the 0-, 3- and 6-month cochlear implant intervals, compared to 31,7dB and 34dB obtained during the no-amplification and hearing aid conditions respectively. The values for all.sentences obtained during the 3-month interval showed an in- crease in amplitude peaks when compared to the 0- month interval. During the 6-month interval, however, there was a slight decrease in amplitude peaks for sen- tences two, four and six when compared to the 3-month cochlear implant interval. These values compared to the 0-month interval did, however, show an increase in amplitude range. The results of the fundamental frequency (F0) varia- tions, including the minimum and maximum F0 values as well as the differences in F0 across the various audi- tory input situations, are presented in Table 3. As can be seen in Table 3, the mean difference in fundamental frequency variations for the no-amplifica- tion (49,7Hz) and hearing aid (48,7Hz) situations did not differ significantly from one another. Surprisingly, the results for the 0-month cochlear implant interval showed a decrease in F0 variation (44,4Hz). A signifi- cant increase in the mean F0 variation occurred at the 3-month interval (110,7Hz), indicating a possible over compensation in the use of varying F0. The results ob- tained at the 6-month interval (86,5Hz) also showed an increase in F0 variation when compared to the no-am- plification and hearing aid situations, however, the vari- ation was less and possibly more normalized, than ob- served during the 3-month interval. Segmental features - Vowels Due to variations in vowel duration for each of the vowels within the three cochlear implant evaluation intervals, a mean vowel duration for each of the eight vowels during the cochlear implant intervals was cal- culated. The results of the evaluation of vowel dura- tion presented in Table 4, did not indicate significant Table 1. Sentence duration (seconds) for sentences across various auditory input situations. Sentence number N.A. H.A. CI-0 CI-3 CI-6 Mean values CI 0, 3, 6 1 0,8719 0,9750 0,9937 0,9031 0,8844 0,927 2 1,153 1,272 1,438 1,472 1,387 1,432 3 1,700 1,727 2,425 1,993 1,878 2,098 4 1,537 1,691 1,628 1,600 1,547 1,591 5 1,237 1,575 1,456 1,359 1,322 1,379 6 0,9469 0,9688 0,9562 1,031 0,9312 0,972 7 0,8031 0,9600 1,006 0,9750 0,9125 0,964 Table 2. Highest amplitude peaks (dB) for stressed words in sentences across various auditory input Sentence number N.A H.A CI-0 CI-3 CI-6 1 25 35 33 37 37 2 24 32 30 37 35 3 37 34 31 36 37 4 34 30 33 36 35 5 34 34 37 40 41 6 33 35 32 39 37 7 35 38 36 37 37 Mean amplitude peaks (dB) 31,7 34 33,1 37,4 37,0 The South African Journal of Communication Disorders, Vol. 41, 1994 R ep ro du ce d by S ab in et G at ew ay u nd er li ce nc e gr an te d by th e P ub lis he r (d at ed 2 01 2) e e c h Production Changes with the Use of a Multichannel Cochlear Implant in a Postlingually Hearing Impaired Adult differences between the various auditory input situa- tions. For all the vowels examined except for the vowel /y/, however, when a mean value for the 0-, 3- and 6- m o n t h evaluations was calculated, the latter did show a decrease in vowel duration when compared to either the no-amplification or hearing aid condition. The formant frequencies of the vowels analyzed, i.e., F1 and F2, presented in Table 5, did not differ from one another significantly across the various auditory input situations. Compared to the normal values of F1 and F2 in the vowels of male Afrikaans speakers, (Van der Merwe, Groenewald, Van Aardt, Tesner & Grimbeek, 1993), the formant frequencies of four of the vowels i.e. / i /, / a /, = / u / and / ε / fell within the normal limits. In the four remaining vowels i.e. / oe /, / ο / and /oe/ exam- ined except for / y /, the F2 values fell within the nor- mal limits for these vowels. The F1 values, however, 7 were measured at higher frequencies than the normal limits for F l . The results in the present study did, how- ever, indicate that the F l and F2 values did remain rela- tively consistent across the various auditory input situ- ations as was observed with the other four vowels ex- amined. - Consonants As can be seen in Table 6 for the majority of the con- sonants examined in the present study, the use of the cochlear implant did show an increase in the relative amplitude peaks measured for consonant production i.e., plosives and fricatives. The relative amplitude peaks for the consonants in the various auditory input situa- tions ranged from 0-46dB in the no-amplification situa- tion, 0-50dB in the hearing aid situation, 19-5ldB in Table 3. Minimum and maximum fundamental frequency values (Fo-Hz) and difference in fundamental frequency (Δ Fo) for sentences across various auditory input situations. Sentence number Max Fo Min Fo N.A N.A Δ Fo H.A H.A Δ Fo CI-0 CI-0 Δ Fo CI-3 CI-3 Δ Fo CI-6 CI-6 Δ Fo 1 Max 162 45 157 70 134 34 243 137 222 103 1 Min 117 45 87 70 100 34 106 137 119 103 2 Max 132 43 129 45 124 38 204 113 182 80 2 Min 89 43 84 45 86 38 91 113 102 80 3 Max 170 ' 89 150 62 129 48 213 115 213 89 3 Min 81 ' 89 88 62 81 48 98 115 124 89 4 Max 144 28 128 30 138 35 182 56 186 69 4 Min 116 28 98 30 103 35 126 56 117 69 5 Max 179 73 165 72 186 89 262 176 256 149 5 Min 106 73 93 72 97 89 86 176 107 149 6 Max t 144 27 142 42 124 23 196 79 165 54 6 Min 1 117 27 100 42 101 23 117 79 111 54 7 Max 162 43 167 60 144 44 222 99 204 62 7 Min 119 43 107 60 100 44 123 99 142 62 Mean difference in Fo ! 49,7 54,4 44,4 110,7 86,5 Table 4. Vowel duration (seconds) for vowels across various auditory input situations Word Vowel N.A. H.A. CI-0 CI-3 CI-6 Mean mier /i / 0,4812 0,4016 0,4187 0,2469 0,3125 0,3260 waak /a/ 0,2187 0,3719 0,2813 0,3875 0,3562 0,3416 duur /y/ 0,3375 0,3156 0,4812 0,3312 0,4590 0,4238 dop /d/ 0,1469 0,1344 0,1500 0,1312 0,1094 0,1302 toer /u/ 0,5875 0,5594 0,4906 0,4094 0,4031 0,4343 pen /ε/ 0,1656 0,1563 0,1781 0,1187 0,1156 0,1374 rug /oe/ 0,1219 0,1656 0,1594 0,1281 0,1563 0,1479 mis /a/ 0,0781 0,1312 0,1344 0,0962 0,0734 0,1013 Die Suid-Afrikaanse Tydskrif vir Kommunikasieafwykings, Vol. 41, 1994 R ep ro du ce d by S ab in et G at ew ay u nd er li ce nc e gr an te d by th e P ub lis he r (d at ed 2 01 2) 8 Sandy Cummings, Emily Groenewald, Rene Hugo, Lida Muller, Mike van der Linde Table 5. Formant frequencies - Hz (F/Fj) for vowels across various auditory input situations Word Vowel f / f 2 N.A H.A CI-0 CI-3 CI-6 Mean Ci 0, 3, 6 Normal F'/F '•• ·.. mier /i / F. 300 260 300 300 280 293 245 2180 2190 2160 2320 2380 2287 2186 waak (a) FI 760 620 770 640 600 670 679 f 2 1180 1160 1180 1220 1060 1153 1113 duur /y/ 260 240 220 390 290 300 * f 2 2180 2160 2080 2310 2470 2287 * dop Ν FI 580 280 540 540 590 557 373 f 2 1020 1080 1180 1040 1040 1087 805 toer /u/ F , 260 250 260 390 360 337 266 F 2 780 920 780 800 1010 863 961 pen /ε/ F , 500 480 500 560 530 530 353 F 2 1960 2040 2060 2100 2000 2053 2055 rug /ce/ F, 640 540 560 580 580 573 429 F 2 1280 1200 1240 1160 1040 1147 1314 mis /a/ F, 580 540 540 540 600 560 507 f 2 1400 1520 1480 1470 1330 1427 1514 Table 6: Relative amplitude peaks (dB) for consonants in CVC word combinations tory input situations Word Consonant Position No amplification Hearing Aid CI 0 months CI 3 months CI 6 months lied /t/ Final 0 38 23 35 48 pit /t/ Final 29 18 31 47 j 48 loop /p/ Final 0 19 22 23 ! 21 druip /p/ Final 0 0 19 31 j 29 bek /k/ Final 26 0 20 37 1 39 rok PeJ Final 0 20 21 33 ' 36 kous /s/ Final 31 43 48 44 48 mos /s/ Final 44 43 48 44 49 rug /x/ Final 33 43 27 38 36 saag /x/ Final 29 35 34 33 31 sag /s/ Initial 43 50 51 29 54 saag /s/ Initial 46 48 48 54 52 vaak m Initial 19 34 22 34 28 vuur m Initial 33 34 26 29 32 The South African Journal of Communication Disorders, Vol. 41, 1994 R ep ro du ce d by S ab in et G at ew ay u nd er li ce nc e gr an te d by th e P ub lis he r (d at ed 2 01 2) e e c h Production Changes with the Use of a Multichannel Cochlear Implant in a Postlingually Hearing Impaired Adult the 0-month cochlear implant interval, 23-54dB in the 3 - m o n t h interval and 21-54dB in the 6-month interval. Three of the consonant productions in the final position j e / 1 /, / k / and / ρ / were not produced with any audi- ble'plosive release in the no-amplification and hearing aid situations therefore having no measurable ampli- tude peaks i.e., OdB. The results of the mean spectral frequency ranges for the plosive energy of the plosives i.e., / 1 /, / ρ /, and / k / and the spectral noise band frequencies for the frica- tives i.e., / s /, I x l and / f /, presented in Table 7, did show a nar- rowing of these frequency ranges for the majority of con- sonants evaluated in the three cochlear implant inter- vals when compared to the hearing aid conditions. Once again there were no measurable spectral frequency ranges (0Hz) for three of the consonants i.e., / ρ /, / t / and / k /, as there was no audible plosive release in the final position during the no-amplification and hearing aid conditions. 9 Results of the speech production evaluation using listener analysis. The results of the listener's evaluation of the sub- ject's speech production are presented in Table 8. The trained listeners rated the subject's spontaneous speech at the various auditory input situations, according to the rating scale used. These ratings occurred in a lan- guage laboratory. As can be seen in Table 8 the listeners rated the sub- ject's use of pitch variation as being monotone in the no-amplification, hearing aid and 0-month cochlear implant conditions. At the 3-month cochlear implant interval, the listeners rated the subject's speech as hav- ing little variation and the 6-month cochlear implant interval the listeners rated the subject's use of pitch variation as being normal. With regard to vocal dura- tion, the listeners evaluated the subject's vocal dura- tion in the no-amplification, hearing aid and cochlear implant 0-month interval as being longer than normal. Table 7. Minimum and maximum frequencies for consonants in CVC word combinations across various auditory input situations. Word Consonant Position Minimum No Amp HA CI-0 CI-3 CI-6 maximum lied /t/ Final Min 0 1760 2460 1680 2880 Max 0 8000* 4900 8000* 6040 pit 111 Final Min 3440 1760 2480 2720 3400 Max 6800 3160 8000* 8000* 8000* loop /p/ Final Min 0 3560 1080 1160 3680 Max 0 4720 8000* 7960 6440 druip /p/ Final Min 0 0 1040 1240 980 Max 0 0 8000* 8000* 8000* bek PeJ Final Min 880 0 1060 960 880 Max 8000* 0 5000 5600 6240 rok PeJ Final Min 0 960 880 680 640 Max 0 4280 5020 8000* 3680 kous Isl | Final Min 2420 2240 2500 2720 3200 Max 8000* 8000* 8000* 8000* 8000* mos Isl ! Final Min 2220 2360 2480 2560 3160 Max 8000* 8000* 8000* 8000* 8000* rug Ixl Final Min 800 760 2480 2640 2320 Max 8000* 8000* 8000* 8000* 6960 saag Ixl Final Min 2460 880 2060 2760 2280 Max 8000* 8000* 8000* 8000* 7440 sag Isl Initial Min 2180 3200 1080 2800 3320 1 Max 8000* 8000* 8000* 8000* 8000* saag Isl Initial Min 1000 3320 1060 2960 3000 Max 8000* 8000* 8000* 8000* 8000* vaak Ifl Initial Min 2100 3440 3480 1360 3680 Max 8000* 8000* 8000* 8000* 8000* vuur Ifl Initial Min 1360 3520 1620 2880 3600 Max 8000* 8000* 8000* 8000* 7360 Die Suid-Afrikaanse Tydskrif vir Kommunikasieafwykings, Vol. 41, 1994 R ep ro du ce d by S ab in et G at ew ay u nd er li ce nc e gr an te d by th e P ub lis he r (d at ed 2 01 2) 10 Sandy Cummings, Emily Groenewald, Ren Hugo, Lida M l l e r , Mike van der Linde At the 3-month and 6-month cochlear implant intervals, the subject's vocal duration was evaluated as being nor- mal. As far as pausal duration is concerned, the listen- ers evaluated the use of pausal duration as being pri- marily longer than normal in the no-amplification and hearing aid conditions. At the cochlear implant 0-month interval, two of the listeners rated the pausal duration as being normal, with the remaining listeners describ- ing it as being either abnormally long or longer than normal. The pausal duration at the 3-month interval was evaluated by the majority of the listeners as being longer than normal. One listener judged it to be nor- mal. At the 6-month cochlear implant interval, however, the subject's use of pausal duration was evaluated as being normal. As can be seen in Table 8, listener two was the only listener who differed from the other listeners. The Pearson Correlation Co-efficient and t-test was used in order to determine whether or not listener two differed significantly from the other listeners, i.e., one, three and four. These results indicated that there were no sig- nificant differences between the listeners ratings and that all the listeners ratings correlated significantly with one another. DISCUSSION SPECTROGRAPHIC ANALYSIS Suprasegmental features Prior to having the cochlear implant, it was noticed subjectively that the subject was inclined to increase his utterance length. The latter being a characteristic of profoundly deaf individuals, as noted by Nickerson, (1975). Taking this into account, it was expected that when utterance length was spectrographically analyzed, the results obtained with the cochlear implant would show a decrease in duration, when compared to the no- amplification and hearing aid condition. Results ob- tained from the evaluation of sentence duration (Table 1), however, did not at first yield any significant data. In fact, the mean values calculated for the cochlear im- plant condition were in all instances longer than the values obtained in the no-amplification situation, as well as the majority of values obtained during the hear- ing aid condition. These results are in direct contrast to the sentence duration characteristics of profoundly deaf individuals. These individuals tend to speak at a slower rate than what is considered normal speed, (Waldstein, 1990). The values measured were consequently com- pared with the spectrograms obtained for each of the sentences in the various auditory input situations. These comparisons indicated that the reason for the lower sentence duration values in the no-amplification situation and the higher sentence duration values in the cochlear implant situations was that in many in- stances the plosive release of final consonants of the last words spoken in the sentences were often omitted due to the absence of auditory feedback. The utterance length measured did therefore not include these omit- ted plosive portions of the consonants. This resulted in shorter sentence duration measurements. During the three cochlear implant intervals the subject was aware of these final consonants due to the improved speech information being provided by the cochlear implant as well as improved auditory feedback, and the subsequent production of these final consonants productions re- sulted in a measured increase in the utterance length. As far as word stress (amplification) within the sen- tences is concerned, the use of the cochlear implant re- sulted in an increase in the highest relative amplitude peaks for stressed words when compared to the no-am- plification and hearing aid situations (Table 2). The specific words within the sentences which were stressed remained consistent throughout all the auditory input situations. The relative amplitude peaks of these words increased as the subject was receiving increasing Evaluation parameters Listeners No-Amplification: Listeners ratings Hearing Aid: Listener ratings Cochlear Implant 0-months: Listener ratings Cochlear implant 3-months: Listener ratings Cochlear implant 6 months: Listener ratings Pitch L.l 1 1 1 2 3 Variation L.2 1 2 2 3 3 L.3 1 1 1 2 3 L.4 1 1 1 2 3 Vocal L.l 2 2 2 3 "3 Duration L.2 2 2 2 3 4 L.3 2 2 2 3 3 L.4 2 2 2 3 / 3 Pausal L.l 2 3 3 2 3 Duration L.2 1 2 2 2 3 L.3 2 2 1 2 3 L.4 2 2 3 '3 3 Tabel 8. Rating scale for listeners' evaluation of pitch variation, vocal duration and pausal duration- results . The South African Journal of Communication Disorders, Vol 41, 1994 R ep ro du ce d by S ab in et G at ew ay u nd er li ce nc e gr an te d by th e P ub lis he r (d at ed 2 01 2) e e c h Production Changes with the Use of a Multichannel Cochlear Implant in a Postlingually Hearing Impaired Adult 11 a m o u n t s of intensity information via the cochlear im- nlant. These alterations in loudness are coded by the multichannel cochlear implant as changes in current level, (Tobey & Hasenstab, 1991). An individual's abil- ity to make use of intensity variations within utterances p r e v e n t s the tendency to become monotone and rein- forces the normal patterns of intonation in speech, (Iler- Kirk & Edgerton, 1983). In terms of the pitch variations in FO, there appeared to be an over compensation in the use of varying funda- mental frequency at the 3-month interval, with a slight decrease in variation at the 6-month interval, resulting in a more normalized version of fundamental frequency v a r i a t i o n s . The subject was, therefore, at the 3-month interval beginning to perceive the increased spectral information being provided by the implant as well as the variations in FO as coded by the stimulation rate delivered to a given channel, (Tobey & Hasenstab, 1991). These variations in FO were in turn providing increased suprasegmental information required in the use of ques- tions, as was emphasizing certain words within a sen- tence in order to either alter the meaning of the sen- tence or emphasize a specific word within the sentence, (Iler-Kirk & Edgerton, 1983). It is also interesting to note that the words in the sentences which were pro- duced with the highest FO pitch remained consistent throughout all the auditory input situations. These words produced with the highest FO pitch were also the words produced with the highest amplitude peaks. These results indicate that the subject is utilizing the suprasegmental features of speech accurately with the aim of emphasizing a specific word in an utterance. Segmental features - Vowels The overall decrease in vowel duration with the use of the cochlear implant is 'confirmed by the results ob- tained by Tartter et al., (1989) who also found a decrease in vowel length in a postlingually deafened individual during the first year of usejwith a multichannel cochlear implant. The decrease in vowel length can be attrib- uted to the improved temporal coding mechanisms pro- vided by the cochlear implant, which in turn, improve and enhance auditory feedback monitoring, implicated in regulating the phonetic precision of segmental and suprasegmental characteristics of speech, (Waldstein, 1990; Svirsky et al., 1992). As far as the relationship between the vowel formant frequencies is concerned, the results of the F1 and F2 values across the various auditory input situations did not show any significant differences (Table 5). The re- lationship between the F1 and F2 values for the eight different vowels indicated that the vowel,productions across the various auditory input situations occupy well defined loci as expected. These results verify the mini- mal changes in the relationship between F1 and F2 across the various auditory input situations, indicating relatively stable vowel productions. To further enhance the interpretation of these results, the mean F1 and F2 cochlear implant values were compared with the mean values obtained from normal hearing Afrikaans speak- ing male subjects, (Van der Merwe et al., 1993). The mean F1 values obtained in the cochlear implant condi- tion for each vowel were consistently higher than the normal F1 values. Tartter et al., (1989) in their study indicated that their subjects exhibited lower values af- ter a period of use with the cochlear implant. A possible explanation for the higher F1 values obtained in the present study, could be attributed to the fact that the subject was not a standard Afrikaans speaker. Varia- tions from the norm are often found when a dialect of a language is spoken, (Sommerstein, 1977). By taking the overall results into account, it can therefore be as- sumed that the subject's profound hearing loss did not result in deviant vowel productions. - Consonants With regard to the relative amplitude peaks for con- sonants in CVC word combinations across the various auditory input situations, the most significant increases in amplitude peaks were for the plosive sections of the stop consonants (Table 6). These significant increases in amplitude peaks can be interpreted in terms of the absence, in the first instance, of the plosive releases in the no-amplification situations for the consonants / ρ /, / 1 / and / k / as well as the consonants / ρ / and / k / in the hearing aid situation, to the presence in the sec- ond instance of plosive releases for these consonants during the cochlear implant intervals. For the frica- tives in the initial and final positions, the majority of amplitude peaks in either the 3- or 6-month cochlear implant conditions are higher than the amplitude peaks observed in the no-amplification or hearing aid condi- tions, indicating improved accuracy in consonant pro- duction. This can be attributed to two factors. In the first instance, the cochlear implant is providing in- creased spectral information which in turn allows the subject to perceive these consonants auditorily, thereby improving the auditory feedback mechanism of the sub- ject's own productions, (Dorman, Soli, Dankowski, Smith, McCandless & Parkin, 1990). In the second place, the cochlear implant is resulting in increased accuracy in the articulation of consonants, which con- sequently results in an increase in the intensity of the production. The spectral frequency ranges of the plosive sections of the stop consonants during the 6-month cochlear im- plant interval did show a decrease in the frequency re- gions occupied for the majority of the productions when compared with the no-amplification condition. A de- crease in frequency range is a direct result of a more concentrated release of the plosive section of the stop consonant, thereby improving the accuracy of produc- tions. Minifie (1973), mentions that the noise bursts of the stop consonants are dependent upon the volumes within the vocal tract, which participate in resonance of the source energies. These volumes are primarily determined by the vocal tract occlusion. It can there- fore be assumed that the more concentrated production of the plosive sections of the stop sounds, can be attrib- uted to the correct placing of vocal tract occlusion. The latter is a result of the improved auditory feedback mechanisms provided by the cochlear implant itself. As far as the fricatives are concerned, the spectral noise bands with the cochlear implant at the 6-month inter- val also showed a decrease in the frequency ranges measured for each fricative when compared to the no- Die Suid-Afrikaanse Tydskrif vir Kommunikasieafwykings, Vol. 41, 1994 R ep ro du ce d by S ab in et G at ew ay u nd er li ce nc e gr an te d by th e P ub lis he r (d at ed 2 01 2) 12 Sandy Cummings, Emily Groenewald, Rene Hugo, Lida Miiller, Mike van der Linde amplification situation. These spectral noise band lim- its measured for the fricatives fell within the noise band limits of normal speakers, (Baken, 1987). The majority of the fricatives showed an increase in the minimum frequency with the use of the cochlear implant. These findings resulted in a narrowing of the spectral noise band frequencies which once again indicated the con- centration of energy as a result of marked vocal tract constriction, resulting in increased accuracy of fricative production (Minifie, 1973). - LISTENER ANALYSIS The results of the listener's rating of pitch variation indicated that the subject was beginning to use pitch and intensity information being provided by the cochlear implant. These results are supported by Iler-Kirk & Edgerton, (1983); Dowell et al., (1985); Leder et al., (1986); and Tyler & Kelsay, (1990), who also reported improvements in voice control with the use of a multichannel cochlear implant. These improvements can be attributed to the pitch information being pro- vided by the cochlear implant resulting in improved auditory feedback of the subject's utterances. As far as vocal duration and pausal duration are concerned, the improvements which were made could once again be attributed to the improvements in auditory feedback resulting in improved voice control and temporal reso- lution, (Leder et al., 1986; Tyler & Kelsay, 1990; Tartter et al., 1989). The latter improvements correlate well with the results obtained during the suprasegmental analysis. CONCLUSION In conclusion, the results of the spectrographic analy- ses indicated that the information provided by the cochlear implant resulted in the improved use of suprasegmental features by the subject. There was an overall decrease in sentence length resulting in less drawn out speech. For the sentences where there was an increase in sentence length, the presence of conso- nants in the final position during the cochlear implant intervals as opposed to the absence thereof during the no-amplification and hearing aid situations, provided a logical explanation for this occurrence. The use of in- creased word stress measured in terms of relative am- plitude as well as an increase in the variation of funda- mental frequency, typically resulted in less monotone and more variable speech production. As far as the segmental features are concerned, the results of the vowel analysis indicated that the cochlear implant was providing improved auditory feedback which subsequently resulted in a decrease in vowel length over time. The relationship between the first and second formants for the various vowels did not show any significant differences across the various auditory input situations. These results were expected as the subject's vowel productions were not perceived as being deviant when subjectively compared to the same vowel productions by normal hearing individuals. The over- all increase in relative amplitude peaks for the conso- nants investigated as well as the narrowing of the spec- tral frequency ranges and the spectral noise band fre- quencies indicated that the use of the cochlear implant was resulting in far more accurate consonant produc- tions. The results of the listener's analysis of speech pro- duction indicated that the multichannel cochlear im- plant confirmed a significant improvement in the use of suprasegmental features in the subject's speech pro- duction as perceived by the listeners. The overall im- provement in the use of suprasegmental speech features can be attributed to the normalization in the use of pitch variation as well as vocal and pausal duration with the cochlear implant over time. Finally, the "hearing" provided by the cochlear im- plant is considered to have two major roles in main- taining the communicative effectiveness of the produc- tion mechanisms in adults. In the first instance, self hearing helps to calibrate production mechanisms by monitoring relations between the implant user's own articulations and his/her acoustic output. In the second instance, the speaker can validate his/her acoustic out- put by observing the behaviour of the listeners and by detecting discrepancies between his own speech and that of the listeners, (Perkell et al., 1992). REFERENCES Baken, R.J. (1987). Clinical Measurement of Speech and Voice. Massachusetts: College- Hill Press, Inc. Cochlear Pty Limited. (1989). Mini System 22: Audiologists Handbook. Sydney, Australia. Cohen, N.L., Waltzman, S.B. & Shapiro, W.H. (1985) Clinical trials with a 2 2 - c h a n n e l c o c h l e a r p r o s t h e s i s : In Laryngoscope, 95: 1448-1454. Coles, L.D. (1990). The effect of certain suprasegmental features of the hearing impaired child's speech intelligibility. Unpublished Master's Thesis, University of Pretoria. Cowie, R., Douglas-Cowie, E. & Kerr, A.G. (1982). A study of speech deterioration in post-lingually deafened adults. In Journal of Laryngology and Otology, 96: 101-112. Dorman, M.F., Soli, S., Dankowski, K., Smith, L., McCandless, G. & Parkin, J. (1990). Acoustic cues for consonant identification by patients who use the Ineraid cochlear implant. In Journal of the Acoustical Society of America, 88: 2074-2079. Dowell, R.C., Brown, A.M., Seligman, P.M. & Clark, G.M. (1985). Patient results for a multiple-channel cochlear prosthesis. In Schindler, R.A. & Merzenich, M.M. (eds.): Cochlear Implants. New York: Raven Press, ι Dowell, R.C., Seligman, P.M., Blarney, P.J. & Clark, G.M. (1987). Speech perception using a two formant 22-electrode cochlear prosthesis in quiet and in noise|. In Acta Otolaryngology (Stockholm), 104: 439-446. > Economou, Α., Tartter, V.C., Chute, P.M. & Hellman, S.A. (1992). Speech changes following reimplantation from a single-channel to a multichannel cochlear implant. In Journal of the Acoustical Society of America, 92: 1310-1323. Eddington, D.K. (1983). Speech recognition in deaf subjects with multi-channel intracochlear electrodes. In Parkins, C.W. & Anderson, S.W. (eds.): Cochlear Prosthesis: An international symposium, Annals of the New York Academy of Sciences, 405: 48-63. Iler-Kirk, K. & Edgerton, B.J. (1983). The effects if cochlear implant use on voice parameters. In Otolaryngologic Clinics of North America, 16: 281-292. Lane, H. & Webster, J. (1991). Speech deterioration in postlingually deafened adults. In Journal of the Acoustical Society of America, 89: 859-866. Leder, S.B., Spitzer, J.B., Milner, P., Flevaris-Phillips, C., Richardson, F. & Kirchner, J.C. (1986). Reacquisition of contrastive stress in an adventitiously deaf speaker using a single-channel cochlear implant. In Journal of the Acoustical Society of America, 79: 1967-1974. Ling D. (1976). Speech and the Hearing-Impaired Child: Theory and Practice, Washington, D.C.: The Alexander The South African Journal of Communication Disorders, Vol. 41, 1994 R ep ro du ce d by S ab in et G at ew ay u nd er li ce nc e gr an te d by th e P ub lis he r (d at ed 2 01 2) c o e e c h P r o d u c t i o n C h a n g e s with the U s e of a M u l t i c h a n n e l Cochlear I m p l a n t in a P o s t l i n g u a l l y H e a r i n g I m p a i r e d A d u l t 13 Graham Bell Association for the Deaf. Minifie. F.D. (1973). Speech acoustics. In Minifie, F.D., Hixon, j J. & Williams, F. (eds.): Normal Aspects of Speech, Hearing and Language. Englewood Cliffs, New Jersey: Prentice-Hall, Inc. N i c k e r s o n , R.B. (1975). Characteristics of the speech of deaf p e r s o n s . In The Volta Review, 77: 342-362. Perkell, J., Lane, H., Svirsky, M. & Webster, J. (1992). Speech of c o c h l e a r implant patients: A Longitudinal study of vowel production. In Journal of the Acoustical Society of America, 91: 2961-2978. Schindler, R.A., Kessler, D.K., Rebscher, S.J., Yanda, J.L. & Jackler, R.K. (1986). The UCSF/STORZ multichannel cochlear implant: Patient results. In Laryngoscope, 96: 597- 603. S o m m e r s t e i n , A.H. (1977). Modern Phonology. Edward Arnold Publishers. Svirsky, M.A., Lane, H., Perkell, J.S. & Wozniak, J. (1992) Effects of short-term auditory deprivation on speech production in adult cochlear implant users. In Journal of the Acoustical Society of America, 92: 1284-1300. Tartter, V.C., Chute, P.M. & Hellman, S.A. (1989). The speech of a postlingually deafened teenager during the first year of use of a multichannel cochlear implant. In Journal of the Acoustical Society of America, 86: 2113-2121. Thielemeir, M.A., Eisenberg, C.S. & Brimacombe, J.A. (1982) Audiological results with the cochlear implant. In Annals of Otology, Rhinology and Laryngology, 91, (Suppl. 91): 62- 66. Tobey, E.A. & Hasenstab, M.S. (1991). Effects of a Nucleus multichannel cochlear implant upon speech production in children. In Ear and Hearing, 12: 48-54. Tyler, R.S., Gantz., B.J., McCabe, B.F., Lowder, M.W., Otto. S.R. & Preece, J.P. (1985). Audiological results with two single-channel cochlear implants. In Annals of Otology, Rhinology and Laryngology, 94: 133-139. Tyler, R.S. & Kelsay, D. (1990). Advantages and disadvantages reported by some of the better cochlear implant patients. In The American Journal of Otology, 11: 282-288. Van der Merwe, Α., Groenewald, E., Van Aardt, D., Tesner, H.E.C. & Grimbeek, R,J. (1993). Die formant-patrone van Afrikaanse vokale soos geproduseer deur manlike sprekers. In South African Journal of Linguistics, 11(2): 71-79. Waldstein, R.S. (1990). Effects of postlingual deafness on speech production: Implications for the role of auditory feedback. In Journal of the Acoustical Society of America, 88: 2099-2114. A p p e n d i x A. R a t i n g s c a l e for the l i s t e n e r ' s e v a l u a t i o n of p i t c h v a r i a t i o n a n d v o c a l a n d p a u s a l d u r a t i o n Description o f feature as perceived by listener 1 2 3 4 5 ω l-c pitch v a r i a t i o n m o n o t o n e little v a r i a t i o n normal too m u c h v a r i a t i o n a b n o r m a l v a r i a t i o n cd