Research Note

The Effects of Behavioral Speech Therapyon Speech Sound Production With Aduits

Who Have Cochiear ImpiantsFrances M. Pomaville^ and Chris N. Kladopoulos''

Purpose: In this study, the authors examined the treatmentefficacy of a behavioral speech therapy protocol for adultcochiear implant recipients.Method: The authors used a multiple-baseline, across-behaviors and -participants design to examine theeffectiveness of a therapy program based on behavioralprinciples and methods to improve the production oftarget speech sounds in 3 adults with cochiear implants.The authors included probe items in a baseline protocolto assess generalization of target speech sounds tountrained exemplars. Pretest and posttest scores fromthe Arizona Articulation Proficiency Scale, Third Revision(Arizona-3; Fudala, 2000) and measurement of speecherrors during spontaneous speech were compared,providing additional measures of target behaviorgeneralization.

Results: The results of this study provided preliminaryevidence supporting the overall effectiveness and efficiencyof a behavioral speech therapy program in increasing percentcon-ect speech sound production in adult cochiear implantrecipients. The generalization of newly trained speech skillsto untrained words and to spontaneous speech wasdemonstrated.Conclusion: These preliminary findings support theapplication of behavioral speech therapy techniques fortraining speech sound production in adults with cochiearimplants. Implications for future research and thedevelopment of aural rehabilitation programs for adultcochiear implant recipients are discussed.

Key Words: adults, aural rehabilitation, speech production,speech-language pathology, cochiear implants

A s a result of technological advances over the past 4decades, cochiear implants (CIs) are now a viableoption for many adults who experience bilateral,moderate-to-profound sensorineural hearing loss thatimpairs speech recognition and functional communicationabilities and who demonstrate limited improvement withtraditional amplification (Gifford, 2011). Adult CI candi-dates may experience a range of communication deficits,depending on the age of onset and on the severity of theirhearing loss. Adults with prelingual deafness often demon-strate significantly impaired speech production (Busby,Roberts, Tong, & Clark, 1991; McGarr & Osberger, 1978),while those with postlingual deafness often experience agradual deterioration of speech sound production (Abberton

"California State University, Fresno''Capella University, Minneapolis, MinnesotaCorrespondence to Frances M. Pomaville: fpomavil@csufresno.eduEditor: Jody KreimanAssociate Editor: Ewa JacewiczReceived January 11, 2012Revision received July 1, 2012Accepted August 28, 2012DOI: 10.1044/1092-4388(2012/12-0017)

et al., 1985; Leder & Spitzer, 1990; Ramsden, 1981;Zimmerman & Rettaliata, 1981). This speech deteriorationmay result in marginally normal speech, which can carrysignificant social and vocational penalties and can negativelyaffect an individual's self-image (Cowie, Douglas-Cowie,Phil, & Kerr, 1982; Edgerton, 1985).

The focus of most research that examines the speechproduction outcomes of adult CI users has been withpostlingually deafened adults who have experience producingintelligible speech (e.g., Edgerton, 1985; Kishon-Rabin,Taitelbaum, Tobin, & Hildesheimer, 1999; Langereis,Bosman, van Olphen, & Smoorenburg, 1997; Svirsky, Lane,Perkell, & Webster, 1992; Vick et al., 2001; Waters, 1986).Little research has been conducted to examine the speechproduction characteristics of prelingually deafened adult CIusers (e.g.. Busby et al., 1991; Dawson et al., 1995; Vicket al, 2001; Wong, 2007). There is substantial individualvariability in the rate of improvement and the level of speechperformance attained by adult CI users (National Institutesof Health, 1995).

Although technological advances have resulted inimproved patient outcomes for many CI users, technology isonly one contributor to a user's success. Aural rehabilitation(AR) refers to the treatment of hearing loss and often includescomponents such as auditory training, speechreading,

Journal of Speech, Language, and Hearing Research Vol. 56 531-541 fiprW 2013 American Speech-Language-Hearing Association 531

speech therapy, education, and counseling (Boothroyd, 2007;Sweetow & Palmer, 2005). AR programs vary greatly in termsof their content and implementation, and empirical evidenceregarding the efficacy of individual components is lacking.Recent advances in CI technology have led to a renewedinterest in the application of AR programs to CI recipients(Bloom, 2004; Rossi-Katz & Arehart, 2011). The presentstudy focused on one aspect of AR: speech therapy. To date,the effects of various therapy approaches that seek to improvespeech production in adult Cl users have not been welldocumented (Busby et al., 1991; Pantelemidou, Herman, &Thomas, 2003; Waters, 1986; Wong, 2007). Severalresearchers have emphasized the need to consider speechtherapy as part of the AR process (Dawson et al., 1995;Edgerton, 1985; Waters, 1986; Wayner, 2000; Wong, 2007).Yet, studies conducted by Rossi-Katz and Arehart (2011) andby Tucci, Lambert, and Ruth (1990) found that many CIrecipients received a limited amount of AR because they livedfar from their CI center or lacked motivation, family support,or funding.

The purpose of this study was to evaluate the efficacyof behavioral speech therapy for improving speech soundproduction in adult CI recipients with prelingual or earlydeafness. Behavioral speech therapy procedures that havebeen effective in treating a variety of communicative disorderswere extended to adult CI recipients. This study addressed tworesearch questions. First, does the percent correct productionof target speech sounds improve with participation in abehavioral speech therapy program when compared withbaseline measures? Second, does the percent correct produc-tion of target speech sounds during spontaneous speechimprove after participation in a behavioral speech therapyprogram when compared with pretest measures?

MethodThis study used a multiple-baseline, across-behaviors

and -participants design to determine the efl'ect of behavioralspeech therapy on percent correct speech productions. Stablebaselines were established for all target speech sounds.Treatment was then initiated for one of the target sounds,while baseline conditions were continued for the others. Theinitiation of treatment for the flrst target speech sound wasthen staggered across sessions from one participant toanother. For each participant, treatment for the second andthird target speech sounds was initiated when a performancecriterion was met for the previous target speech sound. Thisprocedure resulted in a staggered initiation of treatmentacross target responses for each participant and a staggeredinitiation of treatment across participants. A comparison ofpercent correct speech sounds between baseline and treat-ment phases was used to determine treatment effectiveness.In addition, experimental probes were administered tomeasure generalization to untrained stimuli.

ParticipantsThree adults with CIs participated in the study.

Participants were not excluded based on gender, etiology of

deafness, or make and model of CI. Unpaid volunteers wererecruited by distributing an informational flier to CI surgerycenters in California and by posting the flier at the Deaf andHard of Hearing Service Center in Fresno and at CaliforniaState University, Fresno (CSUF), where the research wasconducted. Two of the participants (Participants A and B)were professionals (i.e., a teacher and a nurse, respectively),and one was a high school student (Participant C). All threereported using both American Sign Language and spokenEnglish to communicate throughout the day. Participant A,a man, was diagnosed with a bilateral, severe-to-profoundsensorineural hearing loss at age 2 years and used hearing aidsuntil a single CI was placed in June 2002. Participant B, awoman, was diagnosed with a moderate hearing loss at age3 years and began using hearing aids. Hearing and speechdeteriorated over several years, and in 2006, she wasdiagnosed with a bilateral, profound sensorineural hearingloss. A CI was placed in the right ear in January 2007 and inthe left ear in March 2009. Participant C, a woman, wasdiagnosed with a bilateral, profound sensorineural hearingloss at age 18 months and used hearing aids until a singleCI was placed in June 2005. At the time of this study.Participants A and C had had their CIs for 7.5 and 4.5 years,respectively. Participant B had the first CI for 3 years and thesecond CI for 1 year prior to the study. All three participantsreported having received speech therapy as children andyoung adults prior to receiving their implants.

Data Collection and Treatment ProceduresThis research was conducted with the approval of the

Institutional Review Board of the governing institutionsand conformed to all standards of human subject research.All sessions were conducted in an individual treatment roomat the Speech, Language, and Hearing Clinic at CSUF.All sessions were audiotaped.

Prescreening and posttest procedures. Prescreening(pretest) and posttraining (posttest) procedures consisted ofadministering the Arizona Articulation Proficiency Scale,Third Revision (Arizona-3; Fudala, 2000) and measuringspeech errors during spontaneous speech. Administration ofthe Arizona-3 was consistent with the instructions providedin the test manual, resulting in a list of speech sound errorsthat occurred during the test. The researcher then engagedthe examinee in conversation to collect at least 50 utterances ofspontaneous verbal communication. Sample lengths rangedfrom 50 to 66 utterances. The researcher used the audiotape tocreate a written transcript of the examinee's speech sample.Misarticulations were transcribed using the InternationalPhonetic Alphabet. This resulted in a list of speech sounderrors that occurred during spontaneous speech. Next, thenumber of correct productions for each of the error speechsounds was divided by the total number of opportunities forproduction of that sound to calculate the percent accuracyfor the production of each error speech sound.

The results of the prescreening procedures showed thatall three participants demonstrated multiple speech sounderrors and reduced intelligibility. Table 1 presents a list ofspeech sound errors identified during the prescreening. For

532 Journal of Speech, Language, and Hearing Research Vol. 56 531-541 April 2013

Tabie 1. Speech sound errors for each participant, identifiedduring prescreening.

Participant Speech sound errors

/r/ and /r/-blends (tr, dr, kr, gr, br, fr, thr), /s/ and /s/-blends (sp, st, sk, ts), N and /l/-blends (pi), /tf/,/dj/, /z/, /e/, /r]/, /a-/ and /a-/-diphthongs

/r/ and /r/-blends (tr, kr, gr, br, pr, thr), /s/ and /s/-blends (sw, st, sk, ts), //, /d3/, /z/, iy, /vj, /y-/, /^/,and /a'Adiphthongs

/r/ and /r/-blends (tr, dr, str, gr, br, pr, fr, thr), /s/ and/sAblends (st, sk, ts, ks), / I / , / /, /^/, /zJ, //, /G/,/r)/, />/, /a-/, and /a-Adiphthongs

each participant, three speech sounds from their list of speechsound errors were selected and targeted for therapy. Theselection of target speech sounds was based on choosingsounds that were least Ukely to infiuence one another andmost lcely to affect intelUgibility. Clients' preferencesregarding the sounds that bothered them the most were alsoconsidered. In addition, pretest and posttest results from theArizona-3 and the spontaneous speech sample provided ameans to compare speech sound production, at the word andconversational levels, before and after treatment.

Baseline procedures. A baseUne protocol was createdfor each participant that consisted of 90 stimulus items: 20words that were used for training and 10 words that were notused for training (i.e., generalization probe items) for each ofthe three target speech sounds. Generalization probe wordswere included to measure stimulus-response generaUzationto untrained exemplars. On the basis of the findings ofSkelton (2004), training and probe exemplars were selectedthat contained the target speech sound in all positions of thewords, as well as in blends where appropriate. Words alsovaried from one to five syllables in length. A list of thetreatment and probe exemplars for each participant isprovided in the online supplemental materials. Untrainedexemplars were intermixed with the trained exemplars, andall exemplars were presented in a randomized order duringthe baseUne phase and after each treatment session. AUexemplars required evoked, word-level productions of thetarget speech sound. No corrective feedback or reinforce-ment was provided during the baseUne or treatment probeprocedures. Using this baseUne measure, a stable baselinelevel of percent correct production of each of the three targetspeech sounds was estabUshed for aU participants prior toinitiating treatment. Each participant was asked to read thewords on the list, and each production was scored as correct(+) or incorrect (-). The obtained scores were used tocalculate the percent of correctly produced target speechsounds. A stable baseline performance across sessions wasdefined as no more than a 10% difference in percent correctproductions for three consecutive sessions. Generalizationprobe items were not considered when determining thestability of baseline measures.

Treatment pliase procedures. Participants were sche-duled for two 1-hour sessions per week for the duration of

the study (4-5 weeks). Throughout treatment, the audibiUtyof aU verbal models and cues was confirmed through clientreport. Once stable baselines were established for allparticipants, treatment was initiated for the first targetspeech sound with the first participant. Correct productionof the first target sound was modeled, and visual, verbal, andtacte prompts were provided to faciUtate correct articu-latory placement to produce that sound. Once a participantdemonstrated correct articulatory placement, therapy pro-ceeded using the foUowing discrete trial procedures.

The researcher modeled the target speech sound inisolation or in a consonant-vowel (CV) syllable, and theparticipant provided an imitative response. Each imitatedresponse was followed by immediate positive feedback for acorrect production (i.e., the target response) or correctivefeedback for an incorrect producdon. Positive feedbackincluded verbal praise or nonverbal confirmation such as ahead nod. Corrective feedback included saying, "No, that isnot correct," describing what was incorrect about theproduction, and possibly providing verbal instructions or avisual demonstration to facilitate a correct production. Eachmodeled response was scored as correct (+) or incorrect (-).Once a participant produced five consecutive correctmodeled responses, the response mode shifted to evokedresponses. The participant was then provided with a writtensymbol representing the target speech sound or CV syllable.The participant produced the target speech sound or syUablein response to the written symbol. Each evoked response wasfoUowed by immediate feedback, as previously described.If, at any time, the participant produced two consecutiveincorrect evoked responses, modeUng was reinstated untilfive consecutive correct modeled productions were obtained.Once the participant achieved 80% accuracy for evokedproductions of the target speech sound in isolation or CVsyllables, treatment shifted to the word level.

Word-level treatment progressed from modeled toevoked trials, as previously described. InitiaUy, the partici-pant was provided with a card containing the written targetword whe the researcher modeled the first target word.Once the participant produced five consecutive correctmodeled responses, the response mode shifted to evokedresponses in which the participant was provided with a cardcontaining the written target word but no model. Once theparticipant demonstrated 10 consecutive correct evokedproductions of the first target word, training began on thenext word for that target speech sound. Treatment continuedin this manner with additional exemplars unt the treatmentcriterion was obtained for the first target speech sound. Thetreatment criterion was reached when the percent accuracyfor production of the target speech sound in trained wordsincreased by a minimum of 50% as compared with thesession with the highest percent correct speech productionduring the baseUne phase. For each participant, treatmentended when the treatment criteria for all three target speechsounds were met.

Once the treatment criterion was obtained for the firsttarget speech sound, treatment began with the second targetspeech sound. The treatment protocol for the second and

Pomaviile & Kiadopoulos: Speech Therapy With Adults Who Have Cochiear Implants 533

third target speech sounds was identical to the procedureused for the first target speech sound. The initiation oftreatment for the first target speech sound was then staggeredacross sessions from one participant to another. Thisprocedure resulted in a staggered initiation of treatmentacross target responses for each participant and a staggeredinitiation of treatment across participants. After the studywas concluded, data summary sheets for all three targetspeech sounds that indicated the percent correct speechproduction for all baseline, treatment, and probe measure-ments taken during the study were produced and areavailable in the online supplemental materials.

Interrater ReliabilityInterrater reliability of correct and incorrect speech

productions for each target response was established byhaving a second researcher independently score responses tobaseline and probe items using an audiotape of the session.This occurred for one randomly selected baseline session andtwo randomly selected treatment sessions for each partici-pant and for all administrations of the Arizona-3. For eachof the sessions, word-by-word recordings of correct andincorrect speech sounds were compared among raters, and apercentage agreement score was calculated. Interrateragreement for the pretest and posttest Arizona-3 measures,and for baseline and treatment phase measures, rangedfrom 87% to 95.6%, from 93.3%) to 96.7%, and from 90%to 93.3%, respectively, thus demonstrating a high level ofinterrater reliability.

ResultsMultiple Baselines Across Behaviors

A comparison of percent correct speech productionbetween baseline and treatment phases provided evidence tosupport the overall effectiveness of the therapy program.Figures 1, 2, and 3 present percent correct speech produc-tion of treatment (solid lines) and probe (dashed Unes) wordsas a function of sessions for Participants A, B, and C,respectively. Baseline phase data for each target behaviorappear to the left of the vertical line, while treatment phasedata appear to the right of this line. These figures show thatfor each participant, baseline performance for all three targetspeech sounds was stable prior to the initiation of treatmentand that each target behavior met the treatment criterionprior to the initiation of treatment for the next targetbehavior. The data presented for all three participants show adesired treatment effect in that percent correct production ofthe target speech sounds remained stable during the baselinephase and increased when the intervention was applied.

Target speech sounds selected for Participant A were/s/, /f/, and Irl, introduced in that order. Figure 1 shows thattreatment was initiated for the first target speech sound (/s/)during Session 4, for the second target speech sound (/^ Z)during Session 5, and for the third target speech sound(/r/) during Session 7. The percent correct speech production

for hl increased to criterion during Session 10, at which timetreatment for Participant A was discontinued.

Target speech sounds selected for Participant B were/y/, /s/, and hi, introduced in that order. Figure 2 showsthat treatment was initiated for the first target speech sound(/f/) during Session 5, for the second target speech sound {Isi)during Session 6, and for the third target speech sound {hi)during Session 9. The percent correct speech productionfor hi increased to criterion during Session 11, at which timetreatment for Participant B was discontinued.

Target speech sounds selected for Participant C were/Q/, lij, and /z/, introduced in that order. Figure 3 showsthat treatment was initiated for the first target speech sound(/r)/) during Session 6, for the second target speech sound{Idpj) during Session 7, and for the third target speechsound {Izl) during Session 9. The percent correct speechproduction for 111 increased to criterion during Session 11, atwhich time treatment for Participant C was discontinued.

In addition to the visual inspection of graphical data,the calculation of effect size can quantify the outcomes forindividual participants when a single-subject research designis used (Beeson & Robey, 2006). Effect size was calculatedto quantify the magnitude of change in level of performancefor each investigated target behavior with the exception oftarget behavior 2 for Participant 3 because of a lack ofbaseline variance in this case (see Table 2). Effect size wascalculated using a variation of Cohen's d statistic:

, XA2-XAi

where Aj and A\ designate posttreatment and pretreatmentperiods, respectively, X is the mean of the data collected in aperiod, and SA is the corresponding SD. This version of Buskand Serlin's (1992) d statistic {d\) is a reliable estimator ofeffect size when pretreatment variance is a nonzero value. Inaddition, the effect sizes for each series of data points can beaveraged to represent the treatment effect for each singleparticipant by using a weighted mean, as described byBeeson and Robey. Using this calculation, weighted dstatistics of 20.97, 24.32, and 27.95 were determined forParticipants A, B, and C, respectively. It is not possible tocode these treatment effects as small, medium, or largebecause standards for interpretation (i.e., benchmarks) forspeech production research have not yet been established.Nevertheless, these results combined with visual inspection ofthe data presented in Figures 1, 2, and 3 reveal a positivetrend in percent correct speech production as a result oftreatment. Furthermore, the magnitude of change betweenprescreening and posttreatment measures (effect size)appears to be meaningful and may contribute to theestablishment of benchmarks in this area.

Multiple Baselines Across ParticipantsFigure 4 shows percent correct speech production as a

function of session for each target speech sound and eachparticipant. The top, middle, and bottom panels illustrateParticipants A, B, and C, respectively. The vertical lines

534 Journal of Speech, Language, and Hearing Research Vol. 56 531-541 April 2013

Figure 1. Percent correct speech production as a function of the session (baseline, treatment, andgeneralization probes) for Participant A across three target behaviors.

Target Behavior 1 : Isl

100 1o 90 -S 80 -W ''O "^ g 60 -o o 50 -b-o 40 2 30 -c- 20 -g 10 -S 0

Target Behavior 2: /tJ/

> Target Behavior 1

- - Target Behavior 1 Probe

10 11 12Session

.co

apee

1

IB oO "OO 2^ Q-

Figure 2. Percent correct speech production as a function of the session (baseline, treatment, andgeneralization probes) for Participant B across three target behaviors.

Target Behavior 1 : /tj/

o0)0a.W c

2o TJ

O 20.

I

100908070605040302010O

Target Behavior 2: /s/

- -

Target Behavior 1 - Target Behavior 1 Probe

5 6 7Session

9 10 11 12

oO)

t 3o -0o 0)

100908070605040302010O

Target Behavior 2- Target Behavior 2 Probe

6 7 8 9 10 11 12Session

Target Behavior 3: /r/

oa>

t :3o "o

O E

1009080706050403020100

5 6 7 8Session

Target Behavior 3- -*- - Target Behavior 3 Probe

9 10 11 12

correct speech production for the untrained words wasgreater than for the trained words. This may be the result ofthe number of probe items included on the baseline probeprotocol (10 probe items vs. 20 treatment exemplars).Despite these exceptions, the percent correct speech pro-duction and overall pattem of performance for the trainedand untrained words was similar during most treatmentsessions.

Generalization to Spontaneous SpeechPretest and posttest measures of the percent correct

productions of target speech sounds taken during sponta-neous speech were compared to determine whether produc-tion of target speech sounds during spontaneous speechimproved after participation in the behavioral speech therapyprogram. Table 3 shows the number of opportunities each

536 Journal of Speech, Language, and Hearing Research Vol. 56 531-541 April 2013

Figure 3. Percent correct speech production as a function of the session (baseline, treatment, andgeneralization probes) for Participant C across three target behaviors.

Target Behavior 1 : IT\I

Q.

2

O 2(UIQ.

100908070605040302010O

Target Behavior 1

- Target Behavior 1 Probe

I 1

7 8

Session

10 11 12

Target Behavior 2:

o(U0)

10090 -80 -70 -60 -SO -40 -30 -20 -10 -0

Target Behavior 3: /z/

Target Behavior 2

- Target Behavior 2 Probe

9 10 11 12

Session

o

a.

rt oo .^

O P0)

(U3.

100 -90 -80 -70 -60 -50 -40 -30 -20 -10 -0 -

Target Behavior 3

- -A- - Target Behavior 3 Probe

Session

participant had to produce each target speech sound and thepercent correct production of target speech sounds in eachspeech sample. The results presented in the table show thatall participants improved their productions of the targetspeech sounds at the conversational speech level even thoughtreatment focused on single-word productions. The overallamount of improvement ranged from 12.7 percentage points{lf)l for Participant C) to 83.3 percentage points {1^1 forParticipant B).

DiscussionIn this preliminary study, the efficacy of using

behavioral speech therapy to improve speech sound pro-duction in three adults with CIs was investigated. Accordingto the American Speech-Language-Hearing Association'slevels of evidence for studies of treatment efficacy, thismultiple-baseline design study would be rated at Level lib:a well-designed quasi-experimental study (ASHA, 2004).

Pomaville & Kladopoulos: Speech Therapy With Adults Who Have Cochlear Implants 537

Table 2. Analysis of data for effect

Participant

A

B

C

Target Behavior

123123123

size calculations.

Mean A,

28.337.54.168.758.05.62

18.0

1.25

MeanA2

90.839070*89.1690.065*9195

*70

iMean A2 - Mean A.,

62.582.565.8480.4182.059.3873

68.75

SDA^

2.882.885.842.62.735.622.73

2.31

di

21.728.6411.2732.1630.0310.5626.73

29.76

Wofe. Dashes indicate "data not available": There was no variance in the baseiine measures for TargetBehavior 2 in Participant 3: therefore, the effect size caiouiation could not be completed. A-^ = baseline phase;A2 = posttreatment phase.*Posttreatment figures were based on a single measure for the third target behaviors.

Visual inspection of the baseline and treatment datapresented for the three participants shows an increase inpercent correct speech productions after treatment onset foreach target speech sound using a multiple baseline acrosstarget speech sounds for each participant (see Figures 1, 2,and 3) and across participants (see Figure 4), thus demon-strating a desired treatment effect. Effect size calculationsfurther support the magnitude of change. These findings areconsistent with several other studies that suggested speechtherapy, in conjunction with the CI, may provide many adultCI recipients an opportunity to improve their speechproduction and communicative abilities (Busby et al., 1991;Dawson et al., 1995; Pantelemidou et al, 2003; Waters, 1986;Wong, 2007; Zwolan, Kileny, & Telian, 1996). This isrelevant because deaf speakers, including those with mar-ginaUy normal speech, may experience lower self-image orsignificant social and vocational penalties (Cowie et al., 1982;Edgerton, 1985). Thus, improved speech production maypotentially improve quality of Ufe in these areas.

The replication of this finding using multiple baselinesacross behaviors and participants provided evidence that theintervention rather than some extraneous event was respon-sible for the change in percent correct speech production.Prior to the study, all three participants demonstratedmultiple speech sound errors even after being implanted forat least 3 years. None of them reported recent changes intheir speech production, and all three demonstrated stablebaselines. Therefore, the findings were consistent with thoseof Dawson et al. (1995), Waters (1986), Wayner (2000), andWong (2007), who concluded that the improved auditionprovided by the CI alone is often not adequate to ensureoptimal communication.

In regard to treatment efficiency, each participant metthe treatment criterion for his or her respective individualtarget speech sounds within four sessions and met thetreatment criterion for aU three target sounds within sevensessions. Treatment efficiency was also supported bygeneraUzation of target speech sounds to untrained stimuUand to conversational speech, even though treatment onlyoccurred at the word level. GeneraUzation of improved

speech production to untrained words was confirmedthrough use of an intermixed probe (Hegde, 2003), as seen inFigures 1, 2, and 3. The generalization of correct speechsound production to spontaneous speech was estabUshedthrough comparing the speech errors identified during theanalysis of spontaneous speech samples coUected duringprescreening and posttest procedures (see Table 3). Theamount of improvement varied across target speech soundsand was likely influenced by the participants' motivation andabiUty to monitor their spontaneous speech, the particularspeech sound, and variabiUty in the amount of training timedevoted to each speech sound during treatment. OveraU,however, these findings support the generalization ofimproved speech sound production in trained words tountrained words and to spontaneous conversational speech.

The concern over treatment efficiency has led to thedevelopment of concurrent treatment in which trainingoccurs on the full range of presumed easy-to-hard exemplarswithin each session, intermixed and in random order(Skelton, 2004). Skelton found that concurrent treatment, asopposed to hierarchical task sequencing, promoted general-ization of the target behavior and treatment efficiency.Training and probe exemplars for this study varied from oneto five syUables in length and contained the target speechsound in all positions of words and in blends whereappropriate. Although treatment in the present study focusedon word-level productions, the findings further support thenotion that phonemes do not need to be trained separately forvarious positions within the word, for blends, or for wordsof various lengths. The speed at which the participants met thetreatment criterion for each target behavior was rapid despitethe variabiUty of exemplar types, as was generaUzation tountrained stimuU and spontaneous speech. Therefore, thepresent findings provide preliminary evidence to supportthe premise that behavioral speech therapy using mixedexemplars is both effective and efficient in improving theproduction of target speech sounds in adults with CIs.

Rossi-Katz and Arehart (2011) found that the majorityof audiologists who responded to their survey did notconsider speech-language evaluation results when directing

5 3 8 Journal of Speech, Language, and Hearing Research Vol. 56 531-541 April 2013

Figure 4. Percent correct target speech sound production as a function of session (baseline and interventionphases) for each target behavior and participant in a multiple-baseline, across-behaviors and -participants design.The staggered lines represent the initiation of treatment for each target behavior. TB = target behavior.

Participant A

2.w

rre

o

ent

o0}

1009080 1

c 70 -g 60^ 50 i

T32 40 -1'^ 30

20100

Participant B

z > ^

10090

(UQ.W

,3

I9O

Per

cots

odu

80 -70605040 -30 -20 -10 -

Participant C

100g 90S 80c 70^0 .2 60 -

1 - ^^O 2 40

CDQ .

9 10 11 12

rehabilitation for their older CI patients. Tucci et al. (1990)found that aural rehabilitation programs for adult CIrecipients were sparse and that those offered by the CIcenters often did not include speech production trainingas a primary component. In addition, many CI recipientsdid not receive additional AR services because of the longtravel distances to their CI centers. The experiences ofthe participants in this study were consistent with resultsreported by Rossi-Katz and Arehart and by Tucci et al. in

that they were treated at three different CI centers, yetnone of them participated in comprehensive AR programsthat included speech therapy. In addition, all of them hadto travel at least 2.5 hr to their respective centers. Thesefindings support the need for more community-based ARprograms separate from CI centers.

Although these findings are preliminary, the identifi-cation of an efficacious approach to improving speechsound production in adult CI recipients contributes to the

Pomaville & Kladopoulos: Speech Therapy With Adults Who Have Cochlear Implants 539

Table 3. Number of opportunities to produce target speech sounds and the percent correct production oftarget speech sounds in the spontaneous conversational speech samples during the prescreening (pretest)and posttest.

Participant

Participant AIsl///r/

Participant B///s/M

Participant C/ry/cfe//z/

Number of opportunities toproduce target speech sounds

Prescreening

941318

46730

138

43

Posttest

1241128

66916

145

34

Percent correct productionof target speech sounds

Prescreening

24.460.08

11.1

0.0017.940.0

23.00.00

27.9

Posttest

76.672.750.0

83.352.253.3

35.760.047.0

establishment of an evidence-based speech therapy protocolfor this group of individuals. This should encourage morespeech-language pathologists to develop similar therapyprograms for this population and should encourage CIcenters to include behavioral speech therapy as a primarycomponent in existing AR programs.

Limitations and Future DirectionsThe primary limitation of the present study is the

inability to infer from the sample to a larger populationowing to the small sample size. The multiple baselines acrossspeech sounds and across participants do serve as a form ofreplication; however, inferring the present findings to thepopulation of adult CI recipients is limited by using threeparticipants. Therefore, additional replication of the pre-sented methods will be required to demonstrate externalvalidity of the findings. In addition, the study would havebeen strengthened by extending the duration of treatmentso that the participants would be required to maintain thetreatment criterion for the third target behavior acrossadditional sessions. This would have permitted an analysis ofthe stability of percent correct speech productions over timefor all target responses. Moreover, the study would havebeen strengthened by collecting follow-up measures ofpercent correct speech production for training and probewords and for spontaneous speech some time after theposttreatment assessment was completed to examine main-tenance and generalization of trained speech productionsover time.

The results of this study suggest several possibilities forfuture research. Replication of the present and similarresearch findings can further substantiate the efficacy ofbehavioral speech therapy for CI recipients. Specifically,discrete trial speech production training with differentparticipants (e.g., ages, abilities), settings, and speech soundscan contribute to the scope of application of the presenttreatment procedure.

AcknowledgmentsWe thank Katie Nemeth, the graduate student who served as

our second researcher during the collection of interrater reliabilitydata, and Don Freed, department chair at California StateUniversity, Fresno.

ReferencesAbberton, E., Fourcin, A., Rosen, S., Walliker, J., Howard, D.,

Moore, B., Frampton, S. (1985). Speech perceptual andproductive rehabilitation in electro-cochlear stimulation.In R. Schlinder & M. Merzenick (Eds.), Cochlear implants(pp. 527-537). New York, NY: Raven Press.

American Speeeh-Language-Hearing Association. (2004). Evidence-ba.Hed practice in communication disorders: An introduction.Retrieved from www.asha.org/docs/html/TR2004-00001.html

Beeson, P. M., & Robey, R. R. (2006). Evaluating single-subjecttreatment research: Lessons learned from the aphasia literature.Neuropsychology Review, 16, 161-169.

Bloom, S. (2004). Technological advances raise prospects for aresurgence in use of auditory training. The Hearing Journal, 57.19-24.

Boothroyd, A. (2007). Adult aural rehabilitation: What is it and doesit work? Trends in Amplification, 11, 6371.

Busby, P. A., Roberts, S. A., Tong, Y. C, & Clark, G. M. (1991).Results of speech perception and speech production training forthree prelingually deaf patients using a multiple electrodecochlear implant. British Journal of Audiology, 25. 291-302.

Busk, P., & Serlin, R. (1992). Meta-analysis for single-case research.In T. Kratochwl & J. Levin (Eds.), Single-case research designand analysis: New directions for psychology and education(pp. 197-198). Hillsdale, NJ:' Erlbaum.

Cowie, R., Douglas-Cowie, E., Phil, D., & Kerr, A. (1982). A study ofspeech deterioration in post-lingually deafened adults. Journal ofLaryngology and Otology, 96, 101-112.

Dawson, P. W., Blamey, P. J., Dettman, S. J., Rowland, L. C,Barker, E. J., Tobey, E. A., Cowen, R. (1995). A clinical reporton speech production of cochlear implant users. Ear andHearing, 16, 551-561.

Edgerton, E. (1985). Rehabilitation and training of postlingually deafadult cochlear implant patients. Seminars in Hearing, 6. 65-88.

540 Journal of Speech, Language, and Hearing Research Vol. 56 531-541 April 2013

Fudaia, J. B. (2000). Arizona-3: Arizona Articulation ProficiencyScale, Third Revision. Los Angeles, CA: Western PsychologicalServices.

Gifford, R. H. (2011). Who is a cochiear implant candidate? TheHearing Journal 64. 16-22. doi:10.1097/01.HJ.0000399149.53245.bl.

Hegde, M. N. (2003). Clinical research in communicative disorders:Principles and strategies (3rd ed.). Austin, TX: Pro-Ed.

Kishon-Rabin, L., Taitelbaum, R., Tobin, Y., & Hildesheimer, M.(1999). The effect of partially restored hearing on speechproduction of postlingually deafened adults with multichannelcochiear implants. The Journal of the Acoustical Society ofAmerica. 106. 2843-2857.

Langereis, M. C, Bosman, A. J., van Olphen, A. F., & Smoorenburg,G. F. (1997). Changes in vowel quality in postlingually deafenedcochiear implant users. Audiology. 36. 279-297.

Leder, S. B., & Spitzer, J. B. (1990). A perceptual evaluation of thespeech of adventitiously deaf adult males. Ear and Hearing. 11,169-175.

McGarr, N. S., & Osberger, M. J. (1978). Pitch deviancy andintelligibility of deaf speech. Journal of CommunicationDisorders. 13. 419^41.

National Institutes of Health. (1995). Cochiear implants in adultsand children. Consensus Development Conference Statement,13(2), 1-30. Retrieved from www.nim.nih.gov/medlineplus

Pantelemidou, V., Herman, R., & Thomas, J. (2003). Efficacy ofspeech intervention using electropalatography with a cochiearimplant user. Clinical Linguistics & Phonetics. 17. 383-392.

Ramsden, R. (1981). Rehabilitation of the suddetily deaf adult. Ear.Nose and Throat. 60, 49-54.

Rossi-Katz, J., & Arehart, K. (2011). Survey of audiologic serviceprovision to older adults with cochiear implants. AmericanJoumal of Audiology, 20. 84-89.

Skelton, S. L. (2004). Concurrent task sequencing in single-phonemephonologic treatment and generalization. Journal ofCommunication Disorders. 37, 131-155.

Svirsky, M., Lane, H., PerkeU, J., & Webster, J. (1992). Speech ofcochiear implant patients: Results of a short-term auditorydeprivation study. The Journal of the Acoustical Society ofAmerica, 92, 1284-1300.

Sweetow, R., & Palmer, C. (2005). Efficacy of individual auditorytraining in adults: A systematic review of the evidence. Journal ofthe American Academy of Audiology, 16, 494-504.

Tucci, D., Lambert, P., & Ruth, R. (1990). Trends in rehabilitationafter cochiear implantation. Archives of Otolaryngology-Head &Neck Surgery. 116. 571-574.

Vick, J., Lane, H., Perkell, J., Matthies, M., Gould, J., & Zandipour,M. (2001). Covariation of cochiear implant users' perceptionand production of vowel contrasts and their identification bylisteners with normal hearing. Journal of Speech, Language,and Hearing Research. 44. 1257-1267.

Waters, T. (1986). Speech therapy with cochiear implant wearers.British Journal of Audiology, 20, 35^3.

Wayner, D. S. (2000). Audiologic rehabilitation for adults withcochiear implants. Seminars in Hearing, 21, 245-256.

Wong, P. (2007). Changes in speech production in an eady deafenedadult with a cochiear implant. International Joumal of Language& Communication Disorders. 42. 387-405.

Zimmermann, G., & Rettaliata, P. (1981). Articulatory pattems ofan adventitiously deaf speaker: Implications for the role ofauditory information in speech production. Journal of Speechand Hearing Research, 24, 169-178.

Zwolan, T. A., Kileny, P. R., & Telian, S. A. (1996). Self-report ofcochlear implant use and satisfaction by prelingually deafenedadults. Ear and Hearing. 17, 198-208.

Pomaville & Kladopoulos: Speech Therapy With Adults Who Have Cochiear Implants 541

Copyright of Journal of Speech, Language & Hearing Research is the property of AmericanSpeech-Language-Hearing Association and its content may not be copied or emailed tomultiple sites or posted to a listserv without the copyright holder's express written permission.However, users may print, download, or email articles for individual use.