Top t-ang DisordersVol. 25, No. 3, pp. 207-219© 20()'i Lippmcotr Williams & Wilkins, Inc.

Stimulability and TreatmentSuccess

Susan Rvacbew, PhD

This article addresses 2 questions of importance to the treatment of speech sound disorders:(1) When selecting treatment targets, is it best to begin with the most or the least stimulable po-tential phoneme targets? (2) When treating imstimulable phonemes, which treatment procedureswill result in the best outcome? A summary of the findings from 3 randomized controlled trials isprovided. In these studies, outcomes were generally better when stimulable targets were treated;however, outcomes for unstimulahle targets were improved by including phonemic perceptiontraining alongside phonetic placement procedures in the treatment program. The clinician musttake final responsibility for judging the applicability of these re,search findings to each individualcase. Clinical decisions should be made after discussing the known benefits and ri,sks of any giventreatment practice with the client and/or the client s family. Key v/ords: ei'ieience based practice,phonological disorder, speech perception, speech sound disorder, speech therapy, stimulability

STIMUIJVBILITY reflects a child's ability- tocorrectly imitate a given phoneme when

provided with the instruction to "watch atidlisten" followed by models of the phoneme,usually in the context of nonsense syllables orsimple real words (for an historical overviewof the clinical application of stimtilability-testing procedures, see Powell & Miccio,1996). Stimuiability testing is recommendedas an essential part of an assessment, evenwhen the "clock is running" and the clinicianis forced to make departures from ideal as-sessment practices (e.g., Bleile, 2002; Miccio,2002).

Prom the School of Communication Sciences andDisorders, Mcdill University. Montreal, Quebec,Canada.

The author helped develop and has a financial interestin Speech Assessment and Interactive Learning Systems,a treatment procedure that was evaluated in several oftbe studies summarized in this article.

This article was prepared with support fmm the Cana-dian Language and Literacy Research Netu^ork (wivw.cllnet.ca). Portions of this article were /nvsented as partofaseminaron "Evidence-Based Practice: The Why andthe How" chaired fry Amy Weiss and Rehecca McCavley,at the 2004 annual convention ofthe A merican Speech-Langtiage-Hearing Association.

Corresponding author: .Susan Rvachew, PhD, School ofCommunication Sciences and Disorders, McGill Univer-sity, 1266 Pine Avenue West, Montreal, Quebec, CanadaH3G IA8 (e-mail: susan.rvacheu>@mcgillca).

One reason for including stimulability test-ing in an assessment of a child's pbonologicalskills is that the results may have clear prog-nostic indications. Improvements in articula-tory accuracy for an unmastered phoneme aremore likely to occur over a given time pe-riod when the child is stimulable, rather thanunstimulable, for accurate production ofthespeech sound in question (e.g., Miccio, Elbert,& Forrest, 1999; Tyler, 1996), Furthermore,children who have generally higher levelsof stimulability show better progress overtime than children who have generally lowerlevels of stimulability (eg., Sommers, Leiss,Delp, Gerber, Fundrella, Smith, Revucky,Ellis, & Haley, 1967), Tliese findings applyregardless of whether children show typ-ical or delayed speech development andregardless of whether they are receivingintervention.

Tbere is a consensus that stimulability testrestilts also have implications for treatmentplanning in general and for target selection inparticular. There is considerable controversyabout the clinical implications of stimulabil-ity or a lack thereof for the design of an et-fective phonology intervention, however (foran historical overview of this controversy, seePowell & Miccio, 1996), Basic research onstimulability presents a problem for clini-cal decision making because the apparent

207

208 TOPICS IN LANGUAGE DISORDERS/JULY-SEPTEMBER 2005

implications of the basic research findingsare contradictory: if stimulable phonemes im-prove even without treatment, the logicalcotirse of action might be to avoid thesephonemes as treatment targets and focus in-tervention on the unstimulahle phonemes; onthe other hand, if unstimulable phonemes donot improve even when they are treated, anequally logical course of action might be toavoid them as treatment targets and focus in-tervention on the stimulable phonemes. Re-solving the issue requires the clinician to re-view the relevant literature while applying theprinciples of evidence-based practice to iden-tify the course of action that is likely to leadto the best clinical outcome for a given client.The purpose of this article is to examine tbeempirical evidence relating to treatment ofunstimulable and stimulable phonemes.

THEORIES ABOUT THE NATUREOF STIMULABIUTY

Traditionally, stimulabilit>' was viewed asevidence of the structural and functional in-tegrity of tbe child's speech mechanism. Amore recent theory posits that the abilityto imitate a speech sound reflects underlyingphonological knowledge about the pboneme(Powell & Miccio, 1996), Phonological knowl-edge includes an understanding of the lin-guistic features tbat form a given pboneme.the contrastive relationships between thephoneme and otber phonemes, and thepbonotactic environments in wbicb a givenpboneme may occur. Underlying phonolog-ical knowledge is, in turn, seen as the ba-sis tor generalization of articulatory accu-racy both within and across phonemes andphoneme classes. Tliis theoretical perspectiveis linked to tbe suggestion that unstimulablephonemes should be the focus of interventionbecause teaching tbese sounds provides anopportunity to introduce new phonologicalknowledge that will reorganize the child'sphonological system and result in enhancedgeneralization of learning from treated to un-treated phonemes.

Lof (1996) examined the correlates of stim-ulability in an effort to illuminate the rela-tionship between stimulability and phonolog-ical knowledge. Thirty children's underlyingphonological knowledge for each of severalcommonly misarticulated phonemes was de-termined by assessing phonemic perceptionusing a procedure recommended by Locke(1980). Specifically the child was asked tojudge the correctness of target words pro-duced by an adult, eitber correctly or with asimulated error (i.e., if tbe child has correctphonemic perception for the /I/ phoneme,[lif] should be perceived as a correct produc-tion of tbe word "leaf but [wil] should bejudged to be incorrect), Lof found that all pos-sible combinations of phonemic perceptionperformance and stimulability occurred. Forexample, tbe same child might perceive /k/and be stimulable for tbis pboneme, perceive/r/ and not be stimulable for this phoneme,misperceive /I/ and be stimulable for tbisphoneme, and misperceive /s/ but not bestimulable for this pboneme. Rvachew, Rafaat,and Martin (1999) replicated this finding withtwo additional samples of cbildren, using asimilar test procedure except that recordingsof authentic rather than simulated error pro-ductions were used for tbe phonemic percep-tion task. These data, summarized in Table 1,demonstrate that underlying knowledge of aphonological contrast and stimulability are in-dependent constructs.

Lof (1996) found that stimulability was pos-itively associated with greater visibility of thetarget sound, older age of the child, higher so-cioeconomic status of the family, and greateroverall speech-imitation abilities of the childas reported by the parent, Lof speculatedthat stimulability may also reflect child focus(Kwiatkowski & Shriberg, 1993), meaning tbepropensity of the child to attend to the rele-vant aspects of the model combined with thechild's motivation to change his or ber pro-duction accuracy.

These researcb findings suggest that stim-ulability and phonemic perception are inde-pendent abilities that are each required forphoneme acquisition as illustrated in Figure 1.

Stimulability and Treatment Success 209

Table 1. Relationship between phonemic perception and stimulability across three studies and53 children*

Study

L«f(1996)Rvachew ctRvachew etTotal

alal

(1999),(1999),

StudyStudy

12

Perceive andstimulable

6101026

Not perceiveand not

stimulable

1032

15

Perceivebut not

stimulable

916

16

Not perceivebut stimulable

578

20

*The values represeni numbers of phoneme,s Lorresponding to each combination of perceptual abiliw and stimulability.The totals sum to greater than tbe number of subjects because multiple phonemes were testt-d per subject. The totalnumber of phonemes assessed, summed across the 53 children, was 77.

Stimulability reflects the structural and func-tional integrity of the speech mechanism, thechild s access to visual, tactile, and kines-thetic information about the required articula-tory gestures, imitation skills, and child focus.Phonemic perception arises from the struc-tural and functional integrity of the child sauditory and speech perception mechanisms,the child's access to appropriate speech in-put, and the cognitive-linguistic skills re-

quired for processing and learning from thatinput.

The framework illustrated in Figure 1 alsoreflects the finding that stimulability doesnot by itself lead directly and inevitably tophoneme acquisition, even when the child re-ceives therapy for remediation of the speechsound errors (Miccio etal., 1999; Tyler, 1996).Rvachew et al. (1999) showed that improve-ments in speech sound accuracy for a given

structural and functional integrity ofthe auditory and speech perceptionmechanisms

Appropriate input (many and variedexemplars of contrasting phonemecategories).

Cognitive-linguistic processingabilities

Phonemicperception

Phonemeacquisition

Structural and functional integrity of 'the speech mechanism.

Appropriate input (visual, tactile, andkinaesthetic information about therequired articulatory gestures).

Imitation skills.

Focus and motivation.

Stimulability

Figure 1. An illustration of the proposed relationships among phonemic perception, stimulability, andphoneme acquisition.

210 TOPICS IN LANGUAGE DisoRDERsyJiiLY-SEPTEMBER 2005

phoneme were most likely to occur when thechild demonstrated stimulability and phone-mic perception prior to the onset of a 12-weekphonology intervention (improvements wereobserved for 64% of phonemes that were per-ceived and stimulable and 0% of phonemesthat were not perceived and not stimulable).

A theory of the process by which phone-mic perception and stimulability might inter-act to produce phoneme acquisition is in-stantiated in a neural network model calledDirections in auditory space to Velocities inArticulator space (DIVA; Guenther. 1995). Acentral feature of this model is that the goalof articulation is assumed to be the achieve-ment of a particular acoustic-phonetic prod-uct (as opposed to a specific constellation ofarticulatory gestures). The learner uses audi-tory feedback to identify deviations from thedesired acoustic-phonetie target space andthen adapts the articulatory gestures to bothachieve and maintaiti articulatory accuracy,even as the vocal tract is undergoing devel-opmental changes in size and shape (Callan,Kent, Guenther, & Vorperian, 2000). For ex-ample, the /JV phoneme is characterized notby a specific peak frequency of noise en-ergy, but rather by a range of frequencies thatare low relative to those peak frequenciesthat characterize the /s/ phoneme. Feedbackof the acoustic-phonetic product of speechmovements allows the child to learn to pro-duce the varying constellations of articulatorygestures that will result in this phoneme in dif-ferent phonetic contexts (e.g., greater tongueretraction is required in the context of un-rounded than rounded vowels).

Figure 1 also suggests some interventionprocedures that may be effective. The emer-gence of phonemic perception and stimula-biiity each require certain inputs to the childthat are fully under the control of the clini-cian. Many empirical investigations have re-vealed the type of input that is requiredfor the induction of categorical perceptionof phonemes (Guenther, Husain, Cohen, &Shinn-Cunningham, 1999; Maye, Werker, &Gerken, 2002; Rvachew, 1994). Specifically,the learner must hear a broadly vaiying distri-

bution of exemplars of contrasting phonemecategories. For example, when teaching thechild to perceive the contrast between /s / and/J/, it is best to present a large number of ex-emplars of words containing /s / and wordscontaining /J/, ideally produced by differenttalkers. Most of the words that the child hearsshould represent the prototypical exemplarsof the two sound categories (i.e., lower fre-quency fricative noise for /J/ and much higherfrequency fricative noise of/s/). However, it isalso important to present some exemplars ofthese phonemes that are less than ideal, eventhose approaching the category boundary be-tween the two sounds.

The inputs required for achieving stimula-bility should provide information about theappropriate positioning of the articulatorsand help the child focus attention on tac-tile and kinesthetic feedback associated withthe production of the appropriate articula-tory gestures. Modelling, shaping, and pho-netic placement techniques are suggested inSecord (1981) and have been shown to beuseful for remediating speech sound errors(Powell, Elbert, Miccio, Strike-Roussos, &Brasseur, 1998).

EVALUATING THE EVTOENCE

Although many treatment efficacy studiesthat are relevant to the topic of stimula-bility have been published, most of thesehave employed experimental (e.g., multiplebaseline) or nonexperimental (e.g., multipleprobe) single-subject designs and thus are notparticularly well suited to the two researchquestions considered here (Gierut, Elbert, &Dinnsen, 1987; Gierut, Morrisette, Hughes, &Rowlands, 2001; Powell, Elbert, & Dinnsen,1991; Tyler & Figurski, 1994). A full reviewand critique of all the relevant treatment ef-ficacy studies is beyond the scope of tbis ar-ticle. However, a few studies involving a ran-domized controlled design will be discussed.A randomized controlled trial is essential forstudying this topic because it is the only de-sign that can help differentiate maturationfrom generalization and treatment effects (for

Stimulabiiity and Treatment Success 211

further discussion of research design issues,see Rvachew & Nowak, 2001).

STIMULABILITY AND TARGETSELECTION STRATEGY

Rvachew and Nowak (2001) investigatedthe relative effectiveness of two distinct targetselection strategies on treatment outcomes,using a randomized controlled design. Forty-eight 4-year-(3ld children witb moderate or se-vere speech sound disorders received a pre-treatment assessment (Al), a (>week block oftreatment targeting two consonant phonemes(bUx:k 1), a second assessment (A2), another6-week block of treatment targeting two ad-ditional consonant phonemes (block 2), anda posttreatment asse.ssment (A3), All assess-ments were conducted by a speech-langtiagepathologist who was blind to the child sgroupassignment and treatment targets. The treat-ment approach during each block was tra-ditional, with phonetic placement, shaping,and modelling techniques used to establishstimulability at the syllable level. Once stim-ulability in syllables was achieved, the childwas provided with opportunities for artic-ulation practice in imitated words, sponta-neous words, imitated patterned sentences,spontaneous patterned sentences, imitatedsentences, and spontaneous sentences. Allchildren began the treatment program at thestimulation phase and progressed from onetreatment step to the next upon achieving80% correct responding.

The phonemes that were targeted dur-ing the two treatment blocks depended onwhether the child was assigned at randomto receive treatment for the most stimula-ble, early developing sounds (ME group) orthe least stimulable, late developing sounds(LL group). At each of the three assessments,the child was asked to imitate all items fromthe Productive Phonological Knowledge Pro-file (Gierut et al., 1987), which provides 19opportunities to produce most of the Englishconsonants. The child's responses were usedto rank the child's unmastered phonemes ac-ct)txling to type of Productive Phonological

Knowledge. Gierut et al. described six typesof productive phonological knowledge thatreflect the consistency with which the childarticulates a given phoneme across word po-sitions. Potential treatment targets were thosephonemes that were at type 4 (mastered inat least one but not all three word posi-tions), type 5 (inconsistently correct but notmastered in any word position), or type 6(never correct). Rvachew and Nowak consid-ered type 6 pbonemes in their study to be un-stimulable because the child failed to producethe phoneme correctly even after 15 opportu-nities to imitate it in simple words. These po-tential treatment targets were further rankedwithin each type of Productive PhonologicalKnowledge according to the 90th percentileage of acquisition, according to Smit, Hand,Freilinger, Bernthal, and Bird (1990). In eachtreatment block, children assigned to the MEgroup received treatment for the two soundsthat were most stimulable and earliest devel-oping, with the proviso that the two soundsnot share the same manner class (e.g., /k/and /f/ would be selected as treatment targetsrather than /k/ and /g/). Children assignedto the LL group received treatment for thetwo sounds that were least stimulable and lat-est developing, again with the constraint thatthe two sounds not share the same mannerclass (e.g., /r/ and /s/ would be selected astreatment targets rather than /s/ and /J/), Thecriteria for choosing targets for children inthe LL condition effectively ensured that thisgroup would receive treatment for unstimula-ble sounds, at least during the first treatmentblock. The most frequently targeted soundsduring the first treatment block for the MEgroup were /k/ and /f/. The most frequentlytargeted consonants during the first treatmentblock for the LL group were /s/ and /r/,

Figtire 2 shows mean group response accu-racy for three sets of consonants from the Pro-ductive Phonological Knowledge Profile forthe three assessments, with the ME group'sperfonnance shown in the left panel (2a) andthe LL group's performance shown in theright panel (2h). The three sets of consonantsillustrated in Figure 2a are the most stimulable

212

a.

14

cQ. 10(A

TOPICS IN LANGUAGE DISORDERS>(JULY-SEPTEMBER 2005

b.

Treatment provided

Treatment not provided

Block 1 target

Block 2 target

Never treated

Al A2

Assessment

A3

14

0) 1^VIco

Treatment provided

Treatment not provided

Never treated

Block 2 target

Block 1 target

Al A2

Assessment

A3

Figure 2. Mean ntimber of correct responses for certain consonants on the Productive PhonologicalKnowledge Profile, administered at three assessment points: prior to treatment (Al), after the first treat-ment block (A2), and after the second treatment block {A3). Tlie panel on the left (a) shows the resultsfor children in the ME group, and the panel on the ri^ht (b) shows the results for the children in the LLgroup. (Based on data from Rvachew & Nowak, 2001).

phonemes (treated in the first block), the sec-ond most stimulable phonemes (treated inthe second block), and the least stimulablephonemes (never treated in the ME group).The three sets of consonants illustrated inFigure 2b are the least stimulable phonemes(treated in the first block), the second leaststimulable phonemes (treated in the secondblock), and the most stimulabie phonemes(never treated in the LL group).

Expressing the change in scores from oneassessment to the next in the form of meanresidualized gain scores allows for a numberof different comparisons of relative rate ofchange, after taking into account the child'spretreatment level of productive accuracyfor these phonemes (specifically, a residual-ized gain score is the difference between thechild s actual posttreatment probe score andthe score that would be predicted from thechild s pretreatment probe score, as deter-mined by a regression equation derived from

the pretreatment and posttreatment probescores for all 48 children).

First, the rate of change during blockswhen the phonemes were treated was com-pared with the rate of change observed dur-ing blocks when the phonemes were nottreated. These analyses were conducted sep-arately within each group and cotifirmed thedifferences in rate of change that are read-ily observable in Figure 2. For the ME group,the rate of change for tbe most stimulablephonemes was higher during the block wbenthese pbonemes were treated, than duringthe block when these phonemes were nottreated; 0.44 versus 0.05: K95) = 2.60, p =.011; see solid lines versus dashed lines forblocks 1 and 2 targets in Figure 2a. For the LLgroup, the rate of change for the least stimu-lable phonemes was higher during the blockwhen these phonemes were treated, than dur-ing the block when these phonemes werenot treated; 0.08 versus -0.56; /(95) = 4.58,

Stimulability and Treatment Success 213

p = ,000; see solid lines versus dashed linesfor blocks 1 and 2 targets in Figure 2b, In otherwords, gains were greater for block 1 targetsduring block 1 than during block 2. whereasgains were greater for block 2 targets duringblock 2 than during block 1, This was true forboth groups.

Ihe second analysis considered rate ofchange for treated phonemes across the MEand LL groups. The rate of change in produc-tion accuracy from A1 to A3 for treated stim-ulable phonemes was greater than the rate ofchange for treated tinstimtilable phonemes;0,25 versus -0.51; K9i) = 3-78, p = .000;highest line in Figure 2a versus lowest line inFigure 2b,

Tbe tbird analysis considered the relativerate of change for stimulable pbonemes acrossthe ME and LL groups. From Al to A3, tbe rateof change shown by the ME group for treatedstimulable phonemes was not significantlygreater than the rate of change sbown by theLL group for untreated stimulable phonemes;0,25 verstis 0,21; rt:94) = 0.22,/? = ,829; high-est line in Figure 2a versus bigbest line inFigure 2b.

The final analysis considered the rate ofchange for unstimulable phonemes across theME and the LL groups. From Al to A3, therate of change shown by the ME group for un-treated unstimulable pbonemes was actuallysignificantly greater than the rate of changeshown by tbe LL group for treated unstimu-lable phonemes; 0,05 versus -0,51; ?(94) =2,97, p = ,000; lowest line in Figure 2a ver-sus lowest line on Figure 2b, Overall, treat-ment progress was very poor for unstimu-lable pbonemes but treating the stimulablephonemes first appeared to facilitate improve-ments for unstimulable phonemes in the MEgroup.

Other more global measures of cbange inthis study, including Percentage of Conso-nants Correct, did not reveal any significantdifferences in outcomes between tbe twogroups of participants. Therefore, it is nol sur-prising tbat tbe most appropriate inteqire-tation of tbese findings is a matter of con-tinuing debate (Morrisette & Gierut, 2003:

Rvaehew & Nowak, 2003), However, there isno doubt that, in this study, which involved atraditional treatment approach, children weremost likely to master tbe targeted consonantwben it was stimulable (fiirther studies are re-quired to determine whether this same out-come would be acbieved in the context of aphonological intervention involving minimalpairs procedures).

Parents and clinicians were most satisfiedwith the outcome of the ME target selectionstrategy. Targeting stimuiable sounds first didnot prevent tbe children from making spon-taneous improvements toward correct pro-duction of other more difficult phonemes; infact, tbe ME group showed greater progressfor untreated unstimulable pbonemes thandid the LL group for treated unstimulablephonemes. Absolutely no improvement wasobserved for 45% of treated unstimulable tar-gets. For these reasons, Rvacbew and Nowak(2001) concluded that the most prudent strat-egy for most children is to target stimulablephonemes first. On the other hand, 10% ofdiildren in the LL group actually mastered atreated unstimulable target sound, Stimulabil-ity should never be the only variable underconsideration when selecting tbe most appro-priate pbonemes to remediate.

EFFECTIVE TREATMENT OFUNSTIMUIABLE PHONEMES

There are many cases when it is necessaryto target a phoneme for wbicb tbe child is un-stimulable, Tbe phoneme that has the mostimpact on intelligibilit>- may be unstimulabiefor example. If an entire class of phonemes ismissing from the child's repertoire, you willbe forced to introduce one or more membersof this class even if the child is unstimulablefor all of the relevant phonemes. Therefore,it is important to identify treatment proce-dures that will result in better outcomes forunstimulable pbonemes than those observedby Rvacbew et al. (1999) and by Rvacbew andNowak (2001). To follow is a summary of two.studies sbowing that phonemic perception

214 TOPICS IN LANGUAGE DISORDERS/JULY-SEPTEMBER 2005

intervention can facilitate the acquisition ofstimulability and eventual mastery of targetsounds.

The framework shown in Figure 1 suggeststhat acquisition of a given phoneme requiresaccurate phonemic perception of the tar-geted phoneme contrast and stimulability forthe target phone. Rvachew (1994) conducteda randomized control trial to test this hypoth-esis. In this study, six once-weekly treatmentsessions were provided to 27 preschoolerswith speech sound disorders, all of whomwere completely unstimulable for /J/. Eachtreatment session for all children consistedof 10 min of phonemic perception trainingimplemented in a computer-game format,followed by 20 min of articulation therapydirected at the /J/ phoneme. The treatmentapproach during the articulation therapy partof the sessions was traditional, with phoneticplacement, shaping and modelling techniquesused to establish .stimulability at the isola-tion and syllable levels. If stimulahility wasachieved, the treatment program progressedin steps from imitated words thnnigh progressively longer units of speech. All childrenbegan the treatment program at the stimula-tion phase and progressed from one treatmentstep to the next upon achieving 90% correctresponding. The phonemic perception train-ing procedure was also the same for all chil-dren: the child was asked to listen to recordedwords and identify those words that weregood exemplars of a particular target. Thestimuli that the children listened to dependedon the child s group assignment, howeverThe experimental group listened to a varietyof exemplars of the word "shoe" recordedfrom adults and children, with half the wordsbeing correct exemplars and the other half ofthe words representing common misarticula-tionsofthis word (e.g., [tu], [SU], [̂ U]). Whenthe 'i:hild accurately identified a word as being"shoe"or "not shoe," interesting cartoon char-acters appeared on the computer screen toreinforce the children for correct responding.

This study involved two control groups.The first control group listened to repeatedpresentations of a single prototypical record-

ing of the word "shoe" and a single prototypical recording of the word "moo" andresponded as described above for the exper-imental group. This control group was in-cluded to confirm the importance of present-ing children with a variety of exemplars ofthe target phoneme, as opposed to present-ing only perfect or prototypical exemplars ofthe target sound.

The second control group listened to thewords "cat" and "Pete" produced correctlyby a single adult talker. These children wereexpected to identify correct versions of theword "cat."The clinician who conducted thearticulation therapy was blind to the child'sgroup assignment. This control condition wasnot expected to impact the children's speechperception skills. Rather, it provided a controlfor other aspects of the phonemic perceptiontraining procedure (e.g., extra therapy timeand individual attention, the opportunity toplay a fun computer game).

Outcome assessments were also conductedby a blind observer On average, the childrenin the experimental group achieved stimula-bilitj' and then learned to produce the /J/phoneme at the level of spontaneous words.In contrast, only one child in the second con-trol group achieved stimulability at the iso-lated sound level. The remaining children inthis group experienced no gain even after6 weeks of therapy. The performance of thegroup that heard "shoe" and moo" was in-termediate between the experimental groupand the second control group. Percent correctresponding on a spontaneous object namingprobe also confirmed that the phonemic per-ception training procedure facilitated the chil-dren s acquisition of this difficult phoneme.Furthermore, presenting the children with avariety of good quality and lesser quality ex-emplars of the target word produced by mul-tiple child and adult talkers was particularlyeffeetive.

Subsequently, a commercial version of thephonemic perception training program thatwas administered to the experimental groupin Rvachew (1994) was developed and ex-panded to cover additional phonemes in the

Stimulahility and Treatment Success 215

word initial and word final positions (SpeechAssessment and Interactive Learning System[SAII^]; AVAAZ Innovations, Inc., 1994). Arecent mndomized controlled trial demon-strated that this program enhanced sound pro-duction learning by children receiving phono-logical interventions (Rvachew, Nowak, &Cloutier, 2004). In this study, 34 pre-school-age children with moderate or severe speechsound disorders were randomly assigned toan experimental treatment program or a con-trol program that was offered as a supple-ment to their regular speech therapy sessions.Each child received 16 once-weekly speechtherapy sessions in which the child's clini-cian selected treatment targets and providedthe interventions that she felt were appropri-ate. After each of these sessions, the childreceived an additional 10-min interventionthat was administered hy a student researchassistant. The children assigned to the ex-perimental condition received the SAILS pro-gram targeting eight consonants. The childrenassigned to the control condition listenedto computerized books. Again, treating clini-cians were blind to the child s group assign-ment as was the speech-language pathologistwho conducted the outcome assessments.The posttreatment assessment revealed thatthe experimental group demonstrated signif-icantly greater articulation accuracy in single-word naming and in conversation than didthe control group. Furthermore, a follow-upassessment 1 year later, when the childrenwere approaching the end of kindergarten, re-vealed that 50% of experimental group chil-dren had achieved age-appropriate articula-tion skills whereas only 19% of control groupchildren had achieved articulation skills thatwould be considered tj'pical.

One of the outcome measures was Percent-age of Consonants Correct in conversation forspecific phonemes that were not mastered bythe children prior to treatment, calculated be-fore and after treatment. Ihe children whoreceived the SAILS program made greater im-provements in articulation accuracy for thesephonemes than children who were assignedto the control condition. This benefit to the

SAILS group was observed for 11 of the 13difficult sounds that were monitored in thisstudy (see Figure 1 of Rvachew et al., 2004).

Although the study by Rvachew et al.(2004) was nt)t directly concerned with stim-ulability, details of the pretreatment and post-treatment conversation samples suggestedthat this combination of articulation therapyand phonemic perception training enhancedoutcomes for unstimulable phonemes.Figure 3 shows the change in percent correctarticulation of certain consonants from thepretreatment to the posttreatment assess-ment, as described above. However, in thiscase, the changes in articulation accuracy areshown as a function of pretreatment inven-tory status. In other words, the figure showschange in articulation accuracy separately forphonemes that were present in the child'sinventor}' prior to treatment (and thus clearlystimulable) and for phonemes that werenever produced in the child s inventory priorto treatment (and thus not likely to have beenstimulable). Figure 3a shows that children inthe SAILS group made a 27% improvementin /J/ accuracy regardless of whether thisphoneme was present in the child's pretreat-ment conversational sample. Children in thecontrol group who were stimulable for /J/made more than twice the gain in productionaccuracy for this phoneme than did controlgroup children who never produced thissound prior to treatment. The results for/e/ and A/ indicated that change from nocorrect productions to at least some correctproductions occurred only for children whoreceived the SAILS intervention.

CONCLUSIONS

l\vo questions were asked about tbetreatment of stimulable and unstimulabiephoneme targets: (1) When selecting treat-ment targets, is it best to begin with the mostor the least stimulable potential phonemetargets? (2) When treating unstimulablephonemes, which treatment procedures areassociated with the highest probability of

216 TOPICS IN LANGUAGE DISORDERS/JULY-SEPTEMBER 2005

a. 50

45

40

35

30

15

10

5

0

27 27

1

• /// in inventory |

D Ijl not in inventory

12

^^M 5

SAILS Control

50

45

40

35«?• 30n0 25

1 20I 15

10

5

0

38

^ ^ ^ 1 14

• IQl in inventory

D IQl not in inventory

12

^ ^ ^ 1 0SAILS Control

50

45

40

35o? 30

20

15

10

25

• Ai/in inventory

n/j/not in inventory

SAILS Control

Figure 3 . Frf treatment to posttreatment improvements in percent correct production of certain conso-

nants tor children who received the SAILS phonemic perception training program versus children who

were assigned to the contn>l group, as a function of pretreatment inventory stati:s. (Based on data from

Rvachew et aL, 2(M)4).

Stimulability and Treatment Success 217

achieving accurate articulation of the newphoncmt? With regard to the first question,a nindoniized controlled trial led to the con-clusion that treatment of the most stimulahlepotential targets is likely to result in a greaterrate of change for the treated phonemethan is treatment of the least stimulablepotential targets. The results of this study alsosuggested that a target selection strategy thatbegins with the most stimulable and earliestdeveloping phonemes will facilitate thespontaneous emergence of unstimulablephonemes. However, these findings reflectonly a single study. Greater confidence inthese conclusions could come from replica-tions and extensions of this study in additionalrandomized controlled trials. Studies that ex-amine other combinations of target selectioncriteria would be valuable. It is possiblethat the most effective targets, in terms ofpromoting system wide change, might bethose phonemes that are stimulable butrarely produced correctly (in other words,phonemes representing Type 5 ProductivePhonological Knowledge). Other possibleideal targets might be unstimulable but earlydeveloping phonemes or stimulable late de-veloping pbonemes. These hypotheses havenot yet been investigated experimentally.

With regard to the second question, a num-ber of studies have sbown that a combina-tion ot phonemic perception and stimtila-bility training enhances children's responseto treatment Clamieson & Rvachew, 1992;Rvacbew, 1989, 1994; Rvachew et al,, 1999,2004), This enhancement of sound produc-tion learning is at least as dramatic for un-stimulable pbonemes as it is for stimulablephonemes. The DIVA model (Callan et al,.2000) suggests that tbis enhancement oc-curs because tbe SAILS phonemic percep-tion training procedure provides the childwith a stable acoustic-phonetic target forthe phoneme being learned. This underlyingacoustic-phonetic target allows the child todiscover the mapping between tbe phoneme,the appropriate articulatory gestures, and theacoustic-phonetic outcome of those articula-tory gestures.

This specific hypothesis about tbe mecha-nism by which the SAILS program enhancessound production learning has not been ex-perimentally confirmed, however. Other ex-planations are possible. The treatment effectmay be due to a more generalized processwhereby the child learns to attend more care-fully to speech input. Alternatively, tbe pri-mary lesson that tbe child learns may be tomake judgments about tbe accuracy of hisor ber own speech. Treatment approacliesthat require the child to self-monitor maybe more effective than traditional clinician-monitoring procedures, regardless of whetherthe SAILS program is a component of tbe ther-apy. Furthermore, these studies of tbe efficacyof the SAILS program involved either a tradi-tional behaviorist approach or a cycles phono-logical process approach to treatment (e,g,,Hodson & Paden, 1983). Tbe efficacy of theSAILS phonemic perception training programhas not been evaluated in combination witba more litiguistic approach involving minimalpairs procedures (eg,, Dean fit Howell, 1986),Neither has the SAILS treatment program beenevaluated in relation to tbe multiple phonemeprogram designed by Miccio to facilitateacquisition of stimulability (Miccio, 2005;Miccio & Elbert, 1996), Furtber studies re-garding the efficacy of phonemic perceptiontraining to facilitate the acquisition of unstim-ulable pbonemes are required.

It may be frustrating to clinicians that re-searchers always concltide tbat more re-search is required," This inevitable conclu-sion does not mean that tbe available researchcannot support clinical decision making. Ran-domized controlled trials do provide goodinformation about tbe likely outcome of agiven treatment practice tbr a given popula-tion of clients. Tbe speech-language pathol-ogist must make a judgment about the ex-tent to which the research findings apply toa specific child in a specific clinical context.The clinician should also engage tbe child andthe family in a discussion about the poten-tial benefits and risks associated with the ap-plication of a given clinical practice. The re-search reviewed here suggests that there is

218 TOPICS IN LANGUAGE DISORDERS/̂ ULY-SEPTEMBER 2005

a very strong tisk of achieving no gain aftersix or more weeks of therapy when unstim-ulable targets are treated using a behavioristapproach. However, there may be many situa-

tions in which the potential benefits outweighthe known risk. The choice of treatment ap-proach for a given child is ultimately up to theclinician.

REFERENCES

AVAAZ Innovations, Inc. (1994). Speecb Assessment amiInteractii'e t.eaming System (Version 1.2). London,Ontario, Canada: Author.

Bleile, K. {2(M>2). Evaluating aniculation and phonolog-ical disorders when the clock is running, Ameri-can Journal of Speech-Language Pathology. 11, 243-249.

Callan, D. E,, Kent, R. D., (iuenther, E H., & Vorperian,H. K. (2(XX)). An auditory-feedback-based neural net-work model of speech production that is robust to de-velopmental changes in the size and shape of the ar-ticulatory system,/oHfTJtf/ of Speech, Language, andHearing Research. 4X 721-738,

Dean, E., & Howell, J. (1986), Developing linguisticawareness: A theoretically based approach to phono-logical disorders. British foumal of Disorders ofCommunication, 31, 225-238.

Gierut, J, A., Elbert, M,, & Dinnsen, D, A. (1987). A func-tional analysis of phonological knowledge and gener-alization learning in misarticulating children./ouma/of Speech and Hearing Research, JO, 462-479.

Gierut, J. A., Morrisette, M. L, Hughes, M. T, & Rowlands.S. (2()()l). Phonological treatment efficacy and devel-opmental norms, language. Speech, and Hearing Ser-t'ices in Schools, 27. 215-230.

Guenther, E H. (1995). Speech sound acquisition, coar-ticulation, and rate effects in a neural network modelof speech production. Psychological Review, I02Q>),594-621.

Guenther, E H., Husain, E T., Cohen, M. A., & ShinnCunningham, B. G. (1999)- Effects of categorizationand discrimination training un auditory perceptualspace, foumal of the Acoustical Society of America,106, 2900-2912.

Hodson, B. W., & Faden, E, P (1983). Targeting inteUigi-ble speech: A pbonological apfnoacb to remediation.Boston: College Hill,

Jamieson, D, G,, & Rvachew, S, (1992). Remediation ofspeech production cmirs with sound identificationtraining, you rruj/ of Speech-Language Pathology andAudiology, 16, 201-210.

Kwiatkowski, J,, & Shriberg, L D. (1993). Speech nor-malization in devetopmenUil phonological disoRlers:A retnjspective study of capability-focus tlieory. Lan-guage, Speech, and Hearing Services in Sdmols, 24,10-18.

Locke, J. L. (1980). The iniierence of speech perceptionin the phonologicalty disordered child. Some clinically

novel pniccdurcs, their ust, .some findings,yo«ni<// ofSpeech and Hearing Disorders. 45, 445-468.

Lof, G. L. (1996). Factors associated with speech-soundstimulability. foumal of Communication Disorders,29. 255-278.

Maye, J,, Werker, J. E, & Geriten, L (2(K)2). Infant sensi-tivity to distributional information can affect phoneticdiscrimination. Cognition, 82, BIOI-BIII.

Miccio, A. W (2<W2). Clinical problem solving: Assess-ment of phonological disorders. American foumal ofSpeech-Language Pathology, II, 221-220,

Miccio, A. W. (2(M)5). Components of phonological assess-ment. In A. Kamhi & K. Pollock (Eds), Phonologicaldisorders in children: Assessment and intervention(pp. 163-173) Baltimore;, MD: Brookes Publishing Co,

Miccio, A. W., & Elbert, M. (1996), Enhancing stimulabil-ity: A treatment program, youmtf/ of CommunicationDisorders, 29, 335-351.

Miccio, A. W., Elbert, M., & Forrest, K, (1999), Therelationship between stimulability and phonologicalacquisition in children with n<»rmally developing anddisordered phonologies, American foumal ofSpeech-Ixinguage Pathology. 8, 347-363.

Morrisette, M. L., & Gierut, J. A. (2003). Unified treat-ment recommendations: A response to Rvachew andNowak, Jttumal of Speech, Language, and HearingResearch, 46(2}. 383-385,

Powell, T W., Elben, M,, & Dinnsen, D, A. (1991), Stimula-bility as a foctor in the phonological generalization ofmisarticulating preschool children, foumal of Speechand Hearing Researcb, 34, 1318-1328,

Powell, T W, Elbert, M., Miccio, A. W., Strikc-Roussos,C, & Brasseur, J, (1998), Facilitating [s] pnxluctionin young children: An experimental evaluation of mo-toric and conceptual treatment approaches. ClinicalLinguistics & Phonetics, /2(2), 127-146.

Powell, T. W., & Miccio, A. W. (1996), Stimulability: Auseful clinical tool, foumal of Communication Dis-orders, 29, 237-253,

Rvachew, S. (1989). Remediation of speech productionemirs with sound identification training. Journal ofSpeech-Language Pathology and Audiology, 16, 201-2()H,

Rvachew, S. (1994). Speech perception tr^ning can fa-cilitate sound production learning. yoHrn«/ of Speechand Hearing Research, 37, ^41-^51.

Rvachew, S., & Nowak, M. (2001), The effect of tar-get selection strategy on sound production

Stimulability and Treatment Success 219

Journal of speech. Language, and Hearing Research.•i-*, 610-623.

Rvachew, S,, & Nowak, M. (2(K)3). Cllinical fiuttomcsas a function of tai^et selection strategy: A reply toMorrisette and Gicrut.foumat of Speech, Language,and Hearing Research. 46(2), 386-389.

Rvachew, S., Nowak, M., & Cloutier, G. (2(X)4). Effectof phonemic perception training on the speech pro-duction and phonological awareness skills of childrenwith expressive phonological delay. American four-nal of Speech-lxtnguage Pathology, l.i, 2^0-263.

Rvachew, S., Rafaat, S., & Manin, M. (1999). Stimulabil-it), speech perception and the treatment of phonolog-ical disorders. American foumal of Speech-LanguagePathology^ S. ,^3-34,

Secord, W. (198!). Eliciting sounds: Techniques for clin-icians. San Antonio, TX: Psychological (^iorporation.

Smit, A. B., Hand, L., Frcilinger,J. J., Bemthal, J. E., & Bird,A. (1990). The Iowa articulation norms project and itsNebraska replication. _/o« ma/ of Speech and HearingDisorders, 55, 779-798.

Sommers, R. K., Lciss, R. H., Delp, M., Gerber, A.,Fundrella, D., Smith, R., Revucky, M., Ellis, D., & Haley,V. (1967). Factors related to the effectiveness of articu-lation therapy for kindei^arten, first, and second gradechildren, yourwa/ of Speech and Hearing Research,13. 428-437.

lyicr, A. A. (1996). Assessing stimulability' in toddlers.Journal of Communication Disorders, 29. 279-297.

Tyler, A- A., & Figurski, Ci. R. (1994). Phonetic inventorychanges after treating distinctions along an implica-tional hierarchy. Clinic^ Linguistics & Phonetics, 8,91-107.