child ’ s other language, given that only ~ 5.2% (6300 of 120,744) of ASHA-certifi ed SLPs report being bilingual. Of these 6300 bilingual SLPs, ~ 2700 speak Spanish (ASHA, 2010).

Although the regular delivery of treatment in English to bilingual children with SSD is convenient, it might not be warranted by the evidence. Research shows that Spanish – English bilingual children suc-ceed academically at rates higher than children with limited English profi ciency or children from Latino backgrounds who do not speak Spanish (Rumberger & Larson, 1998). Research indicates that academic instruction in the non-English (i.e., home) language helps rather than hinders the acquisition of English (e.g., Campos, 1995; Cobo-Lewis, Eilers, Pearson, & Umbel, 2002; Kohnert, 2013), and, in fact, leads to higher academic achievement in English (e.g., Thomas & Collier, 2003). For treatment of child language disorders, current best practices argue for providing intervention in the home language as well as in English (Kohnert, 2013; Perozzi & Sanchez, 1992; Restrepo, Morgan, & Thompson, 2013).

It is not known whether intervention in only English is an effective method of delivering speech

Introduction

Speech sound disorders (SSDs) affect ~ 680,000 children in the US, including ~ 10% of pre-school and school-age children (Gierut, 1998). For children with SSDs (whether phonological, articulatory, and/or motor planning in origin), receiving treatment is crucial for later academic success. Between 50 – 70% of school-age children with a SSD exhibit academic diffi culty and often require other remedial services (Gierut, 1998). Children ’ s quality-of-life is affected by having a speech sound disorder (Markham & Dean, 2006; Markham, van Laar, Gibbard, & Dean, 2009), with the frustration from not being understood often resulting in negative social effects (McCormack, McLeod, McAllister, & Harrison, 2010).

A growing number of children in the US are in bilingual language environments and ~ 12% of the US population speaks Spanish (Shin & Kominski, 2010). For bilingual children with SSD, the challenges for receiving adequate treatment are considerable. Although 92% of US school-based speech-language pathologists (SLPs) indicate having children with SSDs on their caseloads, they are more likely to provide treatment only in English, not in the

Correspondence: Christina Gildersleeve-Neumann, Speech & Hearing Sciences, PO Box 751, Portland State University, Portland, OR 97207-0751, USA. Email: cegn@pdx.edu

International Journal of Speech-Language Pathology, 2014; Early Online: 1–15

ISSN 1754-9507 print/ISSN 1754-9515 online © 2014 The Speech Pathology Association of Australia LimitedPublished by Informa UK, Ltd.DOI: 10.3109/17549507.2014.898093

Cross-linguistic generalization in the treatment of two sequential Spanish – English bilingual children with speech sound disorders

CHRISTINA GILDERSLEEVE-NEUMANN 1 & BRIAN A. GOLDSTEIN 2

1 Portland State University, Portland, OR, USA, and 2 La Salle University, Philadelphia, PA, USA

Abstract The effect of bilingual service delivery on treatment of speech sound disorders (SSDs) in bilingual children is largely unknown. Bilingual children with SSDs are typically provided intervention in only one language, although research sug-gests dual-language instruction for language disorders is best practice for bilinguals. This study examined cross-linguistic generalization of bilingual intervention in treatment of two 5-year-old sequential bilingual boys with SSDs (one with Childhood Apraxia of Speech), hypothesizing that selecting and treating targets in both languages would result in signifi -cant overall change in their English and Spanish speech systems. A multiple baseline across behaviours design was used to measure treatment effectiveness for two targets per child. Children received treatment 2 – 3 times per week for 8 weeks and in Spanish for at least 2 of every 3 days. Ongoing treatment performance was measured in probes in both languages; overall speech skills were compared pre- and post-treatment. Both children ’ s speech improved in both languages with similar magnitude; there was improvement in some non-treated errors. Thus, treating both languages had an overall positive effect on these bilingual children ’ s speech. Future bilingual intervention research should explore alternating treatments designs, effi ciency of monolingual vs bilingual treatment, different language and bilingual backgrounds, and between-group comparisons.

Keywords: Intervention , speech sound disorder , bilingualism.

Int J

Spe

ech

Lan

g Pa

thol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

24.2

1.19

8.14

8 on

05/

05/1

4Fo

r pe

rson

al u

se o

nly.

2 C. Gildersleeve-Neumann & B. A. Goldstein

therapy to bilingual children with SSDs. Although it is accepted that bilingual children have differentiated linguistic systems by age 2;0 (Keshavarz & Ingram, 2002), the two speech sound systems in bilingual children are likely inter-linked, which would allow intervention effects in one language to generalize to the other language (e.g., Fabiano-Smith & Goldstein, 2010; Paradis, 2001). However, the degree of such generalization between speech systems for bilingual children with SSD after receiving treatment for their disorder is unexplored. This study examines cross-linguistic generalization of speech skills as well as the effect of bilingual intervention on two children ’ s bilingual speech systems.

Theoretical framework and its application to bilinguals

Dynamic systems theory (DST) provides the theo-retical foundation for this investigation of bilingual intervention for SSD, although the theory is not being tested directly. DST explores how seemingly autonomous components are inter-connected in developmental stages of complex biological systems (Thelen & Smith, 1995; Thelen, Kelso, & Fogel, 1987), including speech (Thelen, 1991). The com-plexity of speech development from a DST perspec-tive guides our intervention approach, which utilizes articulatory, phonological, and metalinguistic fac-tors. DST also guides our hypothesis that the two languages of a bilingual child are linked in develop-ment and that intervention that addresses both lan-guages will have cross-linguistic effects on speech development. We propose that bilingual intervention can also be guided by DST by selecting speech tar-gets that affect both languages and by providing dual-language intervention.

DST has been explored in treating SSDs (Rvachew & Bernhardt, 2010). In a randomized control trial, Rvachew and Bernhardt treated 48 children with SSDs, randomly assigning them to two groups based on the complexity of their treatment targets. Results indicated that the children assigned to the group focusing on less complex targets made more progress than those in the group focusing on more complex targets. Specifi cally, a more complex form, /t ∫ /, emerged from less complex characteristics of earlier speech development.

In development, dynamic interaction between inter-linked components is expected to result in skill progression. However, DST also accounts for U-shaped change: surface-level regressions concur-rent with underlying progress in the complexity of higher-level abilities (Gershkoff-Stowe & Thelen, 2004). This U-shaped change has been noted in the language skills of developing bilinguals (Kohnert, 2013). In bilinguals, assumptions of DST suggest underlying, dynamic interactions between languages would result in cross-linguistic generalization during

development. Such cross-linguistic generalization would allow bilingual children to transfer phonologi-cal and articulatory knowledge from one language to increase accuracy of productions in the other. This cross-linguistic generalization helps explain why a bilingual child, who does not speak either of their two languages as much as monolingual children do, deve-lops speech at about the same rate as monolinguals (e.g., Fabiano-Smith & Goldstein, 2010; Gildersleeve-Neumann & Wright, 2010; Gildersleeve-Neumann, Pe ñ a, Davis, & Kester, 2009). Our study examines cross-linguistic generalization in bilingual treatment for two Spanish – English bilingual children to explore bilingual treatment effectiveness and dynamic inter-actions across languages.

Evidence for intra-linguistic generalization in SSD treatment in monolingual children

The expectation for intra-linguistic generalization is a basic assumption of speech-language therapy and many intervention studies on monolingual children with SSD have systematically investigated the rate and types of intra-language generalization. Tyler, Edwards, and Saxman (1987) examined two phono-logical process-based approaches with four children ranging in age from 3;1 – 5;1. Treatment lasted for 6 – 8 weeks and included 12 – 16 total sessions. A gen-eralization probe was created to track progress on treated and untreated sounds. The intervention resulted in an increase in performance on treated and untreated sounds. Performance on treated sounds was greater than that for untreated sounds. As expected, cross-class generalization did occur for these participants, as sounds across classes were chosen as intervention targets.

Gierut and colleagues have evaluated a number of factors related to intra-linguistic generalization in treatment studies with monolingual, English-speaking children with SSD (Gierut & Champion, 2001; Gierut, Elbert, & Dinnsen, 1987; Gierut, Morrisette, Hughes, & Rowland, 1996). Overall, these studies suggest that targeting more complex aspects of the speech system (e.g., sounds excluded from the inventory, later-developing sounds, non-stimulable sounds, more marked sounds) results in greater system-wide changes than does targeting less complex aspects of the speech system (sounds included in the inventory, early-developing sounds, stimulable sounds, less marked sounds). However, in a randomized design treatment study of 48 children with moderate-to-severe SSD, Rvachew and Nowak (2001) found that children who received interven-tion focusing on early-developing sounds showed greater improvement than those participants whose targets were later-developing. Moreover, greater intra-linguistic generalization did not occur for those receiving intervention on later-developing sounds. Rvachew and Bernhardt ’ s analysis of treatment

Int J

Spe

ech

Lan

g Pa

thol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

24.2

1.19

8.14

8 on

05/

05/1

4Fo

r pe

rson

al u

se o

nly.

Cross-linguistic generalization in speech treatment 3

response suggests that “ new complex behaviours can emerge when intervention serves to stabilize sub-component skills ” as predicted by DST (Rvachew & Bernhardt, 2010, p. 37). From a bilingual perspective, these stable sub-components could be interpreted as having greater inter-language associations and being more salient than those having lesser inter-language associations.

Childhood apraxia of speech (CAS) is an SSD resulting primarily from motor planning diffi culties (ASHA, 2007). Studies of monolingual children with CAS suggest intense, frequent, and long-term treat-ment that emphasizes the motor planning of speech, and that generalization of treatment targets to new phonological properties is limited (ASHA, 2007; Ballard, Robin, McCabe, & McDonald, 2010; Strand, Stoeckel, & Baas, 2006). Strand and Debertine (2000) used a single-subject design with a 5-year-old with CAS, training a limited set of functional phrases in short intense sessions 4 days per week. Target phrases improved, but no generalization was shown. In a single-subject design study with four children with CAS, Strand et al. (2006) trained a small number of phrases 2-times per day, 5 days per week. After 6 weeks, three of the four children demonstrated improvement and limited generalization effects. Also using a single-subject design, Edeal and Gildersleeve-Neumann (2011) treated two children with CAS between 2 – 3-times per week for 40-minute sessions. High intensity treatment (100 � productions in 20 minutes) resulted in the most generalization of speech skills. Research has not demonstrated the best treatment strategies for generalization across languages in bilingual children with CAS, although intense articulatory treatment, likely focusing on early-developing sounds to build a stable foundation for further speech development, builds on the fi ndings from monolingual research on children with CAS.

Cross-linguistic generalization in treatment for SSD in bilingual children

Only a few studies have examined cross-linguistic generalization in the treatment of bilingual children with SSD; no studies have examined cross-linguistic generalization in children with CAS. Holm, Ozanne, and Dodd (1997) treated a 5-year-old Cantonese – English speaking boy who heard and spoke Canton-ese at home until the age of 3;3, when he entered school. The child was treated in English-only, in two phases. The fi rst phase targeted distorted /s/, and the second phase targeted cluster reduction. Cross-linguistic generalization was mixed. Findings indicated that treatment on /s/ resulted in changes on /s/ in both languages; however, the treatment of cluster reduction only improved cluster production in English and not in Cantonese. These results are confounded by the commonly-accepted belief that

Cantonese does not contain clusters in its phonology (So, 2007).

Holm and Dodd (2001), using a pre-/post-case study design, examined cross-linguistic generaliza-tion in one Cantonese – English and one Punjabi – English speaking child, both fi rst exposed to English around age 3. In the Cantonese – English child, treatment initially consisted of articulation therapy in English, focusing on distortion of interdental /s/. The authors hypothesized that generalization would occur to /s/ in Cantonese, given that the error was not language-specifi c. Results indicated that, after 7 weeks of intervention, the child improved in English on the target sound and sounds not directly targeted: / ∫ / and clusters. Generalization occurred in Canton-ese as well; the child produced /s/, / ts /, / ts h / in Cantonese with 70% accuracy (percentage accuracy of those segments prior to intervention was not indi-cated). Subsequently, the child received eight weekly sessions of phonological therapy. Cluster reduction and gliding in English decreased after intervention. The authors noted that generalization across lan-guages did not occur, although they did not provide post-intervention data in Cantonese supporting this position. With the second child (Punjabi – English), Holm and Dodd used a core vocabulary intervention approach to carry out English-only treatment. This approach emphasized sound production and phono-logical awareness skills. Results indicated increased consonant accuracy in both English and Punjabi immediately after and 2 weeks post-treatment.

Ray (2002) used a cognitive-linguistic treatment approach with a 5-year-old trilingual (Hindi, Gujarati, and English) boy with a mild SSD. The child, born in the US, was exposed to Hindi and Gujarati from birth. At age 4, he was “ formally introduced ” to English at school, but he had been exposed to it prior to that time. Minimal contrast therapy in English was used to target multiple error patterns (fi nal consonant deletion, gliding, cluster reduction). Results indi-cated that treatment was effective in decreasing the percentage-of-occurrence of error patterns and increasing consonant accuracy in all three languages, and improving the child ’ s percentile rank on a stan-dardized English articulation assessment. Although phonological skills were examined in all languages, the effect of clinician and home practice was not accounted for, nor was it clear if and how generaliza-tion took place across the child ’ s languages.

Limitations of previous research

Although the few existing studies focusing on reme-diating SSD in bilingual children indicate that treatment in English promotes generalization, they were not designed to examine the therapeutic ben-efi ts of cross-linguistic generalization. In all cases, treatment took place only in English, even though the children were bilingual or trilingual. As noted for other areas of language, improvement in skills

Int J

Spe

ech

Lan

g Pa

thol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

24.2

1.19

8.14

8 on

05/

05/1

4Fo

r pe

rson

al u

se o

nly.

4 C. Gildersleeve-Neumann & B. A. Goldstein

in the home (non-English) language resulted in improvement of skills in English. It remains unclear if these generalization effects were only uni-directional, leaving an open question as to whether generalization effects occur when treatment is in the home language and English. We also do not fully under-stand how cross-linguistic generalization operates.

Findings from multilingual treatment studies to date have limited generalizability for a variety of rea-sons. The case studies have not utilized experimental designs or control conditions, making it diffi cult to determine if the treatments themselves resulted in the changes and whether the fi ndings generalize to other bilingual children. None of the studies reported measures of treatment fi delity. Few details on the specifi c treatment approaches have been specifi ed, target choices are often not described or justifi ed, and the defi nition of generalization employed has not been provided. In addition, it is unknown whether there would be greater overall, or more rapid, improvement if the treatment had fi rst occurred in the home or community language, or bilingually, or if it had been based on the choice of treatment targets. Also, effects of other variables known to infl uence intra- and cross-linguistic generalization in SSD treatment (e.g., language profi ciency, SSD severity) are often not controlled. Changes in the children ’ s phonological system not targeted as part of intervention are not typically measured. Finally, no studies of cross-linguistic generalization in bilin-gual children with CAS have been published.

In summary, results from treatment studies of SSDs on monolingual and bilingual children suggest that intervention highlighting language-specifi c interactions during speech intervention can yield intra- and cross-linguistic generalization of that speech information. Results also indicate that inter-actions focusing on intra- and/or cross-linguistic generalization can bring about changes in the whole speech system because of that input during interven-tion. These general fi ndings have proven to be robust in monolingual English-speaking children. However, the exact nature of the importance of these variables in SSD remediation in bilingual children has not been measured. We hypothesize that selecting targets with strong interaction capacity (phonological prop-erties whose errors affect both languages at relatively high error rates) and treating targets in both lan-guages will result in cross-linguistic generalization, as measured by signifi cant increases in phoneme accuracy and phonetic complexity and signifi cant decreases in the percentage-of-occurrence of error patterns in both English and Spanish.

Method

Participants

The children in this study were two boys: Carlos and Rub é n (pseudonyms). Both boys were diagnosed

with SSDs in their educational settings prior to study onset; Rub é n also has CAS. They were born in the US in predominantly Spanish home language envi-ronments. Both boys were referred to the university clinic by their school-based SLPs. Both boys had received their initial assessment at their educational setting prior to this treatment study and had received intervention for ~ 6 months in English. The boys were generally healthy.

Carlos was 5;8 at study onset. Carlos had just completed a full-day kindergarten in a Spanish – English bilingual immersion school. He lived with his parents, 8-year-old brother, and 1-year-old sister. Carlos ’ parents were originally from Mexico and spoke and understood little English; all spoken and written communication with Carlos ’ parents was conducted in Spanish, based on parental preference. Carlos was fi rst exposed to English regularly 2 years prior to this study. Carlos occasionally spoke limited English with his older brother when they were with English-speaking friends; in all other settings Carlos spoke Spanish. He had been diagnosed with a mod-erate SSD ~ 9 months prior to study onset.

Rub é n was 5;6 at the onset of this study. He lived with his parents who were originally from Mexico, and his 1-year-old sister. Rub é n ’ s family spoke no English at home, nor did his younger sister. All spoken and written communication with Rub é n ’ s parents was in Spanish, by parental preference. Rub é n ’ s father understood and spoke English mod-erately well; his mother reported not speaking or understanding English. Rub é n had fi rst been exposed to English regularly 1 year prior to the study when he entered an English-only Head Start classroom for 4 hours per day, 3 days per week. Rub é n had been diagnosed with a severe SSD at 3;6; his school clinician also suspected CAS, which was confi rmed during this study. Rub é n had received speech ther-apy in English once per week in his classroom.

Design and procedures

As part of this treatment effi cacy study, three outcome variables were measured pre- and post-intervention: phoneme accuracy; phonetic complexity; and pho-nological error patterns. Two treatment targets, Target 1 (TRG1) and Target 2 (TRG2), were selected for each child based on individual assessment results. Carlos and Rub é n received treatment 2 – 3-times per week for 8 weeks. Treatment was provided in Spanish for at least 2 out of every 3 days; English was pro-vided on the remaining days. To determine treatment effects during the intervention, we used a multiple-probe across behaviours design, a type of single-subject experimental design. Baseline measurements (the control conditions) for multiple behaviours allow for description of pre-treatment levels and predictions for future levels of performance with the intervention (Kazdin, 2011). A strength of the single-subject design is that the effects of treatment

Int J

Spe

ech

Lan

g Pa

thol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

24.2

1.19

8.14

8 on

05/

05/1

4Fo

r pe

rson

al u

se o

nly.

Cross-linguistic generalization in speech treatment 5

can be measured in an individual child ’ s intervention (Horner, Carre, Halle, McGee, Odom, & Wolery, 2005), lessening the threats to validity for individuals that group treatment research in controlled settings have.

We established baseline for each child ’ s treatment targets in both Spanish and English. Treatment for TRG1 began fi rst while baseline data continued to be collected for TRG2. During this fi rst treatment phase, non-targeted language activities were com-pleted during treatment time set aside for TRG2. Eventually, both TRG1 and TRG2 were treated for both children (and no additional language activities were conducted). The targets selected were different for each child. The criterion for target selection was existence of the error in both languages and error pattern rate, with developmental appropriateness considered in the selection of errors with high rates of occurrence (Rvachew & Nowak, 2001). Probe data were collected for the two treated targets. Stimuli at the appropriate level (sound, word, or phrase) were presented to each child with a request to repeat the stimuli. The probes were randomly selected for collection at the start of each session.

Treatment sessions were 50 minutes long, divided as follows: 5-minute probes of treated target behaviours, 20 minutes for each treatment target (randomly ordered), 5 minutes for metalinguistic treatment generalization across languages: discus-sion and application of treatment information to tar-gets in the non-treated language. In Phase 1, only TRG1 was treated; time allotted for TRG2 was spent on non-goal receptive language activities (e.g., sequencing, colour, and attribute identifi cations). Phase 1 was complete when the child achieved a pre-determined probe accuracy criterion for the level they were at with TRG1. During Phase 2, both TRG1 and TRG2 were treated. Treatment involved articulatory and phonological strategies for both children; for the child with CAS, the cueing strategies effective in treatment for motor planning diffi culties were added to facilitate productions (e.g., Edeal & Gildersleeve-Neumann, 2011; Strand et al., 2006). All assessment and treatment procedures were admin-istered to each child by a graduate student clinician under the supervision of the fi rst author. The student clinician was trained in the intervention protocol. Procedures included four consecutive phases of direct individual interactions with participants: eligibility assessment (analysis of which determined treatment targets), baseline testing, treatment, and post-treatment speech testing. Each phase is described below.

Phase 1: Eligibility assessment. To determine appropri-ateness for a speech production study, we conducted a full assessment of each boy ’ s communication system prior to treatment in the clinic setting. We gathered detailed case history information for each child from parent surveys, evaluating developmental

history, language use, and parent concerns related to speech. The speech development questions on the unpublished parent survey allowed Likert scale com-parisons between responses. Previous research has found strong correlations between this parent survey and percentage of consonants correct (PCC) in Spanish (Stertzbach & Gildersleeve-Neumann, 2005) and English (Powers, 2010).

Both Rub é n and Carlos passed an oral peripheral examination and pure-tone hearing screenings at 20 dB HL in each ear at 1000, 2000, and 4000 Hz. Informal evaluation confi rmed age-typical cognition, voice, and fl uency. Rub é n was shy, which initially affected appropriate social communication; however, both boys were determined to have typical social communication skills. Language skills were evalu-ated using formal and informal assessment measures. Rub é n and Carlos were administered the Spanish Clinical Evaluation of Language Fundamentals-Preschool 2 (Wiig, Secord, & Semel, 2009). In keep-ing with best practices for use of standardized assessments for bilingual children, scores are not reported because of the test ’ s low validity, specifi city, and sensitivity for bilingual children. Nevertheless, the performance of both boys fell within the typical confi dence interval for monolingual Spanish speak-ers. These fi ndings were corroborated with observa-tions during conversation and narrative tasks in the test administration, suggesting that Rub é n ’ s and Carlos ’ receptive language skills were within normal limits.

Speech production was assessed through stan-dardized but not norm-referenced assessment mea-sures. These measures included single-word and sentence-level articulation and phonology tasks in Spanish and English, using the Phonological and Articulatory Bilingual Assessment (PABA) (Gildersleeve-Neumann, 2014). The PABA is a research protocol with English and Spanish compo-nents that are administered on different days. It contains single word and sentence components with stimuli that represent the children ’ s vocabu-lary, includes 1-, 2-, 3-, and 4-syllable words, all language-appropriate phonemes (except English / ʒ / because of its low frequency in children ’ s vocabu-lary), and within- and across-syllable consonant sequences frequent in each language. The single word portion of the PABA is a picture identifi cation task. There are 136 English words and 115 Spanish words. If children do not identify the pictures spon-taneously, the clinician produces the picture name in a delayed imitation format; if the child does not produce the word in delayed imitation, direct imitation is utilized. The imitated sentence sub-components have three trial sentences and 18 sen-tences with pictures refl ecting all key words. If the child does not produce every word in the imitated sentence, the non-produced words are removed from the target sentence for analysis. All speech samples were digitally recorded in audio and video

Int J

Spe

ech

Lan

g Pa

thol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

24.2

1.19

8.14

8 on

05/

05/1

4Fo

r pe

rson

al u

se o

nly.

6 C. Gildersleeve-Neumann & B. A. Goldstein

in Spanish and/or English were collected. Probe sets for each target were not collected in every session. Probe data collection took ~ 5 minutes. Treatment for Target 1. Twenty minutes were 2. spent on TRG1. This period began with a brief review of the target and then drill play activities to elicit 40 – 50 attempts of accurate production targets in functional words and phrases as appropriate for treatment level. Treatment for Target 2. Twenty minutes were 3. spent on TRG2. The same activity types and frequency of production goals used for TRG1 were used for TRG2. Carry-over discussion of treatment goals to 4. language not treated. The fi nal 5 minutes of each session focused on the meta-linguistic application of treatment goals to the non-targeted language.

TREATMENT PHASE 1

In Phase 1, the fi rst target to reach a stable baseline based on target probe measurement was treated. A baseline was considered stable if there was an unchanging or increasing error rate for the target over three baseline probes. If both treatment targets had similar baseline stability, TRG1 was selected by the fi rst author. In Phase 1 only TRG1 was treated; a non-speech language activity was completed during the session time allotted for TRG2. Criterion for Phase 1 completion was individually determined and was not necessarily probe-related.

TREATMENT PHASE 2

Once the initial criterion for TRG1 was reached, a new criterion for TRG1 was set for Phase 2 and treatment for TRG2 began, also with an established probe criterion. TRG1 and TRG2 were each treated for 20 minutes; order of target treatment within a session was randomized. Probes for TRG1 and TRG2 were collected regularly at the beginning of each session to note response to treatment in both languages and to monitor error rates for the untreated target, when applicable.

APPROACH

There are several individual treatment procedures that have been shown to be effective in facilitating speech improvements in young children with SSD (Gierut, 2005). Our treatment approach combined features common to successful articulatory and pho-nological treatment for monolingual children. Core features to this approach were: (a) promoting child ’ s meta- and perceptual awareness of specifi c session speech goals; (b) using developmentally-appropriate and engaging practice activities that facilitate drill play conditions, most effective for speech sound

formats for later transcription. Speech skills were evaluated in Spanish and English separately.

Phase 2: Pre-treatment target selection and baseline criterion. Results from the speech eligibility assess-ment were used for four fundamental purposes: (1) to select individual treatment targets for each child, (2) to provide the fi rst of minimally three target measurements to serve as a baseline for treatment effi cacy data collection, (3) to provide pre- and post-measurements for three dependent variables in Spanish and English (PCC, percentage of vowels correct (PVC), and phonetic complexity), and (4) to provide a pre- and post-treatment measure for comparing to treatment performance.

TARGET SELECTION

For each child, two targets were addressed. Criteria for target selection were the existence of the error in both languages and the error rate (emphasizing higher rates), with developmental appropriateness considered in selection (Rvachew & Bernhardt, 2010; Rvachew & Nowak, 2001). Targets were addressed in both languages. Probe stimuli were individualized for each child and target. Probe stim-uli for a particular target changed if short-term objectives were achieved.

Phase 3: Treatment INTENSITY

Treatment intensity is a critical variable in successful speech therapy that affects outcomes. Intense treat-ment more likely results in the desired change (Duhon, Mesmer, Atkins, Greguson, & Olinger, 2009). Research suggests that greater frequency of production targets results in increased accuracy in probe data as well as generalization to non-targeted utterances (Edeal & Gildersleeve-Neumann, 2011).

There are many ways that intensity is interpreted in the treatment literature (Maas, Robin, Hula, Freedman, Wulf, Ballard, et al., 2008; Warren, Fey, & Yoder, 2007). Treatment intensity differences include individual vs group intervention (Duhon et al., 2009), total number of treatment hours (e.g., Lovaas, 1987), number of elicited target productions during therapy (Edeal & Gildersleeve-Neumann, 2011), and number of hours per week of treatment (Robey, 1998). As described below, we controlled treatment duration and frequency, number of speech productions per treatment target, and treatment form.

TREATMENT DESIGN

The organization of procedures in a single 50-minute session was as follows:

Collect probe data for treatment goals. 1. During each session, sets of 10 probe stimuli

Int J

Spe

ech

Lan

g Pa

thol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

24.2

1.19

8.14

8 on

05/

05/1

4Fo

r pe

rson

al u

se o

nly.

Cross-linguistic generalization in speech treatment 7

treatment in pre-school-aged children; (c) including articulatory and phonological components to each treatment session with cueing strategies that have been demonstrated as effective for children with motor planning diffi culties (Edeal & Gildersleeve-Neumann, 2011; Strand & Debertine, 2000); and (d) utilizing intensive production practice on targets embedded in functional words and phrases (Elbert, Powell, & Swartzlander, 1991; Skelton, 2004). Our approach assumed that children need extensive practice in each treatment session to be successful with new motor patterns they had not yet produced successfully (Edeal & Gildersleeve-Neumann, 2011; Skelton, 2004).

Although the exact treatment protocol depended on each child ’ s target stimulability, treatment involved articulatory placement training, as well as cueing and feedback to ensure accurate production in isolation or single syllables. After the child produced the sound or syllable shape accurately, the target was practiced in functional words and phrases in which phonetic context complexity was monitored. A variety of words and phrases of varying lengths were utilized. If the child made errors, utterance com-plexity was simplifi ed until success was re-achieved; at that point, the level of phonetic complexity was again increased.

Concurrently, discussion of phonological con-trasts was utilized to increase phonological know-ledge of target behaviours. We utilized minimal pair strategies (e.g., Weiner, 1981) because they have been used frequently in treating SSDs in chil-dren when phonological contrasts are neutralized (e.g., Saben & Ingham, 1991; Williams, 2000). We used minimal pairs when possible to contrast a tar-get sound to the children ’ s productions that were in error. Combining articulatory and phonological strategies has been utilized in previous studies in the treatment of children with SSDs (e.g., Saben & Ingham, 1991; Tyler et al., 1987; Weiner, 1981). As targets were produced with more accuracy, practice in a variety of utterance lengths was used to support generalization (Skelton, 2004). This varied practice was used unless there was a regres-sion in target accuracy; in these cases, blocked practice at a shorter less complex level was con-ducted until success was achieved again. Frequency of 40 – 50 productions in functional words and phrases per 20-minute period was targeted and consistently achieved. To assure that length of ses-sion components was adhered to, a timer was used. This production frequency goal was monitored through a hand-clicking mechanism employed by the therapist.

TREATMENT FIDELITY

All sessions were videotaped and reviewed by the fi rst author to verify that the session included each of the core treatment features specifi ed above.

Phase 4: Post-treatment speech testing POST-TREATMENT PROBE MEASURES

The same probe types and administration were used post-treatment to determine if treatment progress was maintained. Single-word and sentence speech testing administered prior to treatment were repeated to determine children ’ s post-treatment performance on the three dependent variables: phoneme accu-racy; phonetic complexity; and error patterns in Spanish and English. Pre-treatment and post-treatment scores were compared for each child to measure change.

Coding, scoring and analysis PRE- AND POST-TREATMENT SPEECH TESTING

The single words and imitated sentences were recorded digitally and uploaded for computer playback. All single-word and sentence speech samples were pho-netically transcribed by trained individuals who were profi cient in Spanish and English to ensure accurate transcription of English and non-English phonemes. The transcribed data were entered into Logical International Phonetics Program software (Oller & Delgado, 2000) to analyse and compare Spanish and English sound properties. Compari-sons were made to the adult target, taking dialect into account.

Pre- and post-treatment performance for each child were determined as follows in the single word and sentence tasks in both languages, examining (1) phoneme accuracy using PCC (Shriberg, Austin, Lewis, McSweeney, & Wilson, 1997) and PVC (Shriberg, 1993) for English and Spanish, (2) aver-age utterance complexity in both languages, 1 using Index of Phonetic Complexity (IPC) (Jakielski, 1998), and (3) phonological error pattern percentage-of-occurrence. We determined treatment effect size measuring the standard mean difference for all data points (SMD ALL ), determining the mean difference in probe data points for baseline and treatment peri-ods and dividing by the standard deviation, for each child ’ s TRG1 and TRG2 single word probe data and for the combined total. The use of SMD ALL effect size allows for inclusion of all data points. Effect sizes were interpreted using Cohen ’ s d as follows: small (.02 – .049), medium (.50 – .79), and large (.80 or greater) (Cohen, 1988).

TRANSCRIPTION RELIABILITY

Inter-judge and intra-judge reliability for broad and narrow transcription (Shriberg & Lof, 1991) was completed on the pre- and post-assessment mea-sures and the probe words. All productions were re-transcribed by the fi rst author and were subjected to established methods for determining inter-judge and intra-judge reliability (Tyler, Lewis, & Welch, 2003). Overall inter-transcriber reliability was 96%.

Int J

Spe

ech

Lan

g Pa

thol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

24.2

1.19

8.14

8 on

05/

05/1

4Fo

r pe

rson

al u

se o

nly.

8 C. Gildersleeve-Neumann & B. A. Goldstein

Results

Carlos

Carlos was ~ 25% intelligible in conversation in Spanish. On the day of the English assessment, Carlos spoke in English only when asked to repeat the imitated sentences task, choosing to respond in Spanish at all other times. Carlos completed both the word and sentence imitation tasks for his speech assessment. Carlos spoke more clearly in single word tasks; in multisyllabic words and running speech, intelligibility decreased dramatically, primarily because Carlos produced few true consonants in longer utterances. Carlos ’ intelligibility increased only slightly when he attempted to self-correct dur-ing communication breakdowns.

In the analysis of Carlos ’ speech, we considered English or Spanish sounds in his phonetic inventory if they were produced in three different types of utterances across the single word and imitated sen-tences tasks. Carlos produced all Spanish vowel and consonant phonemes except for the alveolar trill /r/. Carlos also produced the Spanish consonant allo-phones [ β , ð , ]. In Spanish, Carlos produced the following English phonemes: / / four times, / ε / twice and / ∫ / and / ɹ / once. Carlos had a complete English vowel inventory. The English consonants that Carlos did not yet produce were /v, θ , ∫ /; none of these sounds are phonemic in Spanish. Carlos produced the following Spanish sounds in his English: the voiced velar fricative [ ] once and the voiceless velar fricative /x/ nine times. He produced a variety of word shapes in both languages, including initial and medial clusters, multisyllabic words, and fi nal con-sonants. Accuracy rates for consonants and vowels and frequency of segmental and syllabic error patterns were analysed. Table I compares Carlos ’ pre- and post-treatment accuracy and the highest frequency patterned and individual phoneme errors in single words and sentences affecting both lan-guages. Language-specifi c high frequency errors

were trill (80% errors in words, 100% in sentences) and tap (57% words, 92% sentences) production. In English, derhoticization (36% words, 44% sentences) and gliding (40% words, 38% sentences) occurred with high frequency.

After analysis of Carlos ’ pre-treatment speech characteristics, two syllable-level treatment targets were selected for treatment in Spanish and English. TRG1 was /s/ � consonant cluster (abutting conso-nants) production in initial and medial word posi-tion (/s/ � clusters only occur in medial position in Spanish). TRG2 was medial and fi nal consonant and consonant cluster production, not including /s/ clusters. Word and sentence probe stimuli were developed for Carlos from age-appropriate words demonstrating the targeted errors. If a selected word had been used in treatment in the previous two ses-sions, a different randomly selected word from the list was probed. Prior to probe presentations, Carlos was told one time that the words contained his new sounds and that he should repeat the words and sentences the teacher said. Baseline probes were col-lected prior to treatment onset in the clinical setting; treatment probes were collected at the beginning of each session. Type of probe data was randomly selected each day; each type was not collected dur-ing every session.

Carlos ’ baseline data were collected in both lan-guage environments as appropriate. Treatment was conducted in Spanish during sessions 5, 6, 7, 9, 10, 11, 13, 14, 15, 17, 18, and 19. Treatment was con-ducted in English during sessions 8, 12, and 16. Cross-linguistic generalization activities occurred at the end of every treatment session. Treatment for TRG1, /s/ clusters, began in session fi ve after a stable baseline was achieved. After baseline, TRG1 was treated for 15 sessions for 300 minutes total. Probe data for TRG1 are shown in Figure 1 for single words and in Figure 2 for imitated sentences. As Figure 1 shows, Carlos did not produce /s/ � clusters in single words or sentences in either Spanish or English for

Table I. Carlos ’ pre- and post-treatment accuracy and error pattern frequency in words and sentences.

Spanish English

Words Sentences Words Sentences

Pre Post Pre Post Pre Post Pre Post

Accuracy measures Percentage Consonants Correct 73.2 86.4 63.0 76.6 60.9 78.8 51.9 56.1Percentage Vowels Correct 95.9 97.0 88.6 95.1 80.6 90.3 85.5 82.1Index of Phonetic Complexity 1.91 2.32 1.07 1.40 1.92 2.25 .95 1.09Percentage Initial Consonants Correct 86.6 90.0 80.0 90.7 76.8 85.3 69.9 75.3Percentage Medial Consonants Correct 83.3 85.4 60.7 83.3 73.6 82.6 69.6 66.7Percentage Final Consonants Correct 44.4 84.4 18.4 50.0 46.2 70.2 27.3 43.3 Error rates /s/ 47.3 7.4 75.0 43.7 73.3 26.7 54.5 70.0Stopping 7.6 1.9 1.6 1.6 8.7 7.5 6.8 4.9Cluster Reduction 53.3 21.7 55.3 48.8 38.4 26.8 37.5 30.2Final Consonant Deletion 52.8 9.4 79.0 50.0 36.3 20.2 65.2 48.3

Int J

Spe

ech

Lan

g Pa

thol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

24.2

1.19

8.14

8 on

05/

05/1

4Fo

r pe

rson

al u

se o

nly.

Cross-linguistic generalization in speech treatment 9

Figure 1. Carlos ’ word probe data measurements for Target 1 (TRG1): Accuracy rates for initial and medial /s/-clusters. Probe data were collected prior to therapy during each session.

Figure 2. Carlos ’ sentence probe data measurements for TRG1: Accuracy rates for initial and medial /s/-clusters.

the fi rst six sessions. After treatment was initiated, Spanish probe word accuracy increased to 50% in the 7 th session, briefl y decreasing to 30% accuracy in session 8, increasing to 70% in session 9, and stabilizing at 90% in words from sessions 16 – 19. As shown in Figure 2, Carlos ’ production of /s/ clusters also improved in Spanish sentences, increasing in accuracy to 10% in the 8 th session, 20% in session 11, 60% in session 14, 70% in session 17, and end-ing at 50% in session 19.

In English, treatment effects were not as rapid; however, /s/ cluster productions increased in accu-racy in both words and sentences. In both environ-ments, /s/ clusters were produced with 0% accuracy throughout the baseline period (see Figures 1 and 2). During treatment, word accuracy for TRG1 increased fi rst in session 8 to 20%, fell again to 0% in session 9, rose to 30% in session 13, 50% in session 16, and returned to 30% in session 19. In English sentences, TRG1 accuracy remained at 0% until session 17

when it was 30%; TRG1 accuracy was 20% in session 19.

Treatment for TRG2, medial and fi nal consonant and consonant cluster production (not including /s/ clusters) began in session 8 after a stable baseline was achieved and TRG1 was produced with 25% or greater accuracy for two sessions. TRG2 was treated for 11 sessions for 220 minutes total. Probe data for Carlos ’ TRG2 are shown in Figure 3 for single words and Figure 4 for imitated sentences. As Figure 3 shows, Carlos produced medial and fi nal consonants in Spanish with ~ 30% accuracy prior to treatment; this accuracy was lower in sessions 3 and 4, and decreased to the lowest level of 10% in session 6. After treatment for TRG2 was initiated in session 8, TRG2 increased to 58% in Spanish single words in session 9, hovering between 55 – 60% for the remain-ing sessions. In Spanish sentences, TRG2 was pro-duced with 20% accuracy during the baseline phase;

Figure 3. Carlos ’ word probe data measurements for TRG2: Accuracy rates for medial and fi nal consonants (excluding /s/-clusters).

Figure 4. Carlos ’ sentence probe data measurements for TRG2: Accuracy rates for medial and fi nal consonants (excluding /s/-clusters).

Int J

Spe

ech

Lan

g Pa

thol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

24.2

1.19

8.14

8 on

05/

05/1

4Fo

r pe

rson

al u

se o

nly.

10 C. Gildersleeve-Neumann & B. A. Goldstein

after treatment was initiated, TRG2 probe data accu-racy increased to 35% in session 9, decreased to 0% in session 12 and 10% in session 15, then increased to 40% and 50% in sessions 18 and 19, respectively. In English words, probe data accuracy for TRG2 was similar to Spanish, ~ 30%. After intervention was ini-tiated, TRG2 single word accuracy increased to 50% in words in session 12, then increased to ~ 60% accu-racy in sessions 14, 17, and 19. In English sentences, TRG2 accuracy for sentences was 0% during the baseline phase. TRG2 accuracy increased to 20% in session 9, returned to 0% in session 12, increased to 10% in session 15, and then increased to 40% in session 19. The effect size for baseline and treatment phases for single words was calculated using Cohen ’ s d. The difference in accuracy for Carlos during treat-ment in both Spanish and English was large for TRG1 ( d � 1.594), TRG2 ( d � 4.168), and overall ( d � 1.475).

Table I compares pre- and post-treatment speech samples. PCC increased in both languages and in single words and sentences. PCC increased 18% in English and 13% in Spanish words. PVC increased 10% in English single words but did not change in Spanish words. PVC increased 6% in Spanish sen-tences, but decreased 2% in English sentences. In words and sentences in both languages, phonetic complexity increased. The percentage of initial, medial, and fi nal consonants correct increased dra-matically in most situations, particularly for fi nal consonants. Medial word position did not show a dramatic increase in consonant accuracy in Spanish words (only 2% increase) and in English sentences, where a 3% decrease in accuracy was observed.

We treated /s/-clusters in initial and medial word position in both languages and medial and fi nal con-sonants excluding /s/ clusters. We examined the effects of these on /s- production and related error patterns. As shown in Table I, error frequency for the phoneme /s/ decreased in Spanish words by 40%, in Spanish sentences by 31%, and in English words by 46%. Errors on /s/ increased by 15% in English sentences.

It is possible that the increase in errors in English sentences refl ected a U-shaped change as suggested by Dynamic Systems Theory (DST, Gershkoff-Stowe & Thelen, 2004). In the pre-assessment, Carlos short-ened many target English sentences and, by the end of therapy, Carlos was attempting the more complex English words. The frequency of cluster reduction and fi nal consonant deletion decreased in both lan-guages in Carlos ’ sentences and words. The smallest decrease was 7% for cluster reduction in English sentences; the largest was a 42% decrease in fi nal consonant deletion in Spanish words.

Rub é n

During the assessment, Rub é n spoke few sentences unless prompted; in sentences and single words Rub é n was ~ 10% intelligible in Spanish. During English assessment, Rub é n only spoke in English during the single word task; most words were attempted only after the examiner provided a model. Rub é n attempted but did not successfully complete the imitated sentences task in Spanish or English; both were discontinued after multiple attempts with fewer than three or four words.

Because we only have single word and, thus, less data for Rub é n, we considered a sound to be included in his phonetic inventory if he produced it two or more times in different types of utterances in the single word task. Rub é n produced all Spanish vowel and all consonant phonemes except the following: the voiceless alveolar fricative /s/, the voiceless velar fric-ative /x/, the alveolar tap / ɾ /, and the alveolar trill /r/. Rub é n produced the Spanish consonant allophones [ β ], but not the [ ð ] or the [ ]. In his Spanish, Rub é n produced the following English phonemes: /v/ and / ɹ / once, / ʌ / twice, and / ʊ / once. Rub é n had a complete English vowel inventory except for the rhotic vowel / ɚ /. The English consonants that Rub é n did not pro-duce were / θ , z, s, ʃ , h, ɹ /; of these sounds, only /s/ is phonemic in Spanish. Rub é n produced the following Spanish sounds in English: the bilabial fricative [ β ]

Table II. Comparison of pre- and post-treatment assessment results: Rub é n.

Spanish English

Pre Post Pre Post

Accuracy measures Percentage Consonants Correct 40.2 47.8 38.6 54.6Percentage Vowels Correct 87.2 91.5 70.6 83.9Index of Phonetic Complexity 1.08 1.3 1.3 1.8Percentage Initial Consonants Correct 49.3 66.2 61.9 76.5Percentage Medial Consonants Correct 39.2 42.2 46.8 45.8Percentage Final Consonants Correct 11.1 22.6 22.1 52.3 Error rates /s/ 100.0 88.5 100.0 84.1Stopping 33.3 21.1 13.9 12.1Final Consonant Deletion 77.8 71.0 48.8 32.6Initial Cluster Reduction 85.7 84.2 76.7 69.0Medial Cluster Reduction 66.7 56.9 45.5 38.9

Int J

Spe

ech

Lan

g Pa

thol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

24.2

1.19

8.14

8 on

05/

05/1

4Fo

r pe

rson

al u

se o

nly.

Cross-linguistic generalization in speech treatment 11

position. Practice rates were delayed in therapy as needed. These treatment strategies are in line with dynamic temporal and tactile cueing, shown successful in monolingual children with CAS (e.g., Edeal & Gild-ersleeve-Neumann, 2011; Maas & Farinella, 2012; Strand et al., 2006). Phonological strategies were also included in therapy. Metaphonological activities to encourage cross-linguistic gene ralization occurred at the end of every treatment session.

Treatment for TRG1, initial and fi nal /s/ singletons, began in session four after achieving a stable baseline. TRG1 was treated for 20 sessions for 400 minutes total. Figure 5 shows probe data for single words. There were two phases of probe data for TRG1 treat-ment: direct imitation (phase 1) and delayed imitation (phase 2). As Figure 5 shows, Rub é n pro-duced /s/ with 0% accuracy during baseline and through session 6 in both languages. Starting in ses-sion 7, /s/ production increased in accuracy in direct imitation in Spanish and English, staying above 50% accuracy starting in session 14. Production accuracy increased more rapidly in Spanish than in English. Starting in session 19, probe data were collected for target utterances produced in delayed imitation. After an initial slight decrease in accuracy in Spanish, Rub é n produced target words with at least 75% accu-racy in both languages.

Treatment for TRG2, stop � liquid cluster pro-duction, began in session 19, at the start of Phase 2 for TRG1. TRG2 was treated for seven sessions for a total of 124 minutes of treatment. Probe data for Rub é n ’ s TRG2 are shown in Figure 6 for single words. Rub é n produced stop � liquid clusters with 0% accuracy in Spanish and English during baseline; target words were produced with between 10 – 20% accuracy in treatment sessions 2 – 4 in both languages. In sessions 5 – 7, TRG2 words were produced with 60% accuracy or greater. Effect size for baseline and treatment phases for single words

once and the low central vowel /a/ 14 times. He pro-duced primarily CV syllable shapes in both Spanish and English, although a few initial and medial clus-ters and fi nal consonants were produced in both lan-guages. Accuracy rates for consonants and vowels as well as frequency of segmental and syllabic error were analysed. Table II compares Rub é n ’ s pre- and post-treatment accuracy and the highest frequency errors in single words affecting both languages. In addition, some language-specifi c errors were observed with high frequency. In Spanish, these were trill (100%) and tap (98%) substitution; gliding (33%) and der-hoticization (100%) were frequent in English.

Rub é n exhibited characteristics of moderate CAS in assessment and treatment: inconsistent produc-tion of consonants and vowels in repeated produc-tion of words, diffi culty with motor planning, including metathesis of syllables, groping, diffi culty with changing motor patterns, increased error rates in longer utterances, and unusual and frequent vowel errors. Rub é n ’ s speech was also characterized by unusually slow rate and lengthened or abrupt transi-tions within syllables.

After pre-treatment speech characteristics were analysed, one segmental and one syllable-level error were selected for treatment in both languages. TRG1 was /s/. Rub é n did not produce /s/ at the onset of treatment and his initial treatment focused on pro-ducing an approximation of /s/ in functional one- and two-syllable words in both languages. While the initially-planned probe sentences were collected, because of Rub é n ’ s diffi culty in /s/ production these probe words and sentences were at 0%. TRG2 was word-initial and medial stop � liquid consonant clus-ter production. Probe data at the single word level were collected at the beginning of each session.

Rub é n ’ s baseline probe data were collected in both language environments. Originally probe data (baseline and treatment) were intended to be in randomly selected single words. However, since Rub é n produced both TRG1 and TRG2 with 0% accuracy in probe words and in all speech prior to therapy, probe data used for comparison were altered so treatment effects could be monitored. Thus, baseline and probe data were production of TRG1 and TRG2 in one- and two-syllable single words. Phase 1 for TRG1 was percentage accuracy for single words with /s/ produced in direct imitation; Phase 2 for TRG1 was percentage accuracy for single words with /s/ in delayed imitation.

Rub é n ’ s treatment was conducted in Spanish during sessions 4, 5, 6, 7, 9, 10, 12, 14, 15, 17, 18, 19, 22, 23, and 24. Treatment was conducted in English during sessions 8, 11, 13, 16, and 20. Rub é n required intensive cueing and feedback for target productions. Rub é n infrequently produced /s/ without distortions; /s/ was counted as correctly produced when it was an alveolar fricative, even if slightly distorted. Rub é n exhibited motor planning diffi culties, requiring visual, tactile, and occasionally auditory cues for articulatory placement. He often produced the target sounds in the wrong word

Figure 5. Rub é n ’ s probe data measurements for TRG1: Accuracy rates for /s/ production in one- and two-syllable words.

Int J

Spe

ech

Lan

g Pa

thol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

24.2

1.19

8.14

8 on

05/

05/1

4Fo

r pe

rson

al u

se o

nly.

12 C. Gildersleeve-Neumann & B. A. Goldstein

Figure 6. Rub é n ’ s probe data measurements for TRG2: Accuracy rates for initial and medial Stop � Liquid clusters in one- and two-syllable words.

for Rub é n ’ s was calculated using Cohen ’ s d . The difference in accuracy for Rub é n ’ s during treatment in both Spanish and English was large for TRG1 ( d � 2.355), TRG2 ( d � 3.173) and overall ( d � 2.644).

A comparison of Rub é n ’ s pre- and post-treatment speech samples is shown in Table II. PCC increased by 8% in Spanish and by 16% in English. PVC increased by 4% in Spanish and 13% in English. In words and sentences in both languages, phonetic complexity (frequency of later-developing sounds and complex word shapes) increased. The percent-age of initial consonants produced correctly increased by 17% in Spanish and 14% in English. The percent-age of fi nal consonants produced correctly increased by 12% in Spanish and 30% in English. Medial con-sonants did not change in English and increased in accuracy in Spanish by 3%.

The frequency of error in the phoneme /s/ and error patterns affected by the treatment targets was compared. Error frequency for /s/ decreased by 11.5% in Spanish, from 100% to 88.5%, and 16% in English, from 100% to 84%. Stopping decreased by 11% in Spanish and almost 2% in English. Final consonant deletion decreased from 78% to 71% in Spanish and from 49% to 33% in English. Also, initial and medial cluster reduction decreased in post-assessment (although not treated, many of these clusters also include /s/). Initial cluster reduc-tion did not change much in Spanish, remaining at ~ 85%, but decreased by 7% in English, to 39%. Medial cluster reduction decreased from 67% to 57% in Spanish and from 46% to 39% in English.

Discussion

This study explored the effectiveness of bilingual therapy in two bilingual Spanish – English boys with

SSDs. We hypothesized that selecting targets that affected both languages at relatively high rates and providing intervention in both languages would result in cross-linguistic generalization, measured by signifi cant increases in phoneme accuracy and pho-netic complexity and signifi cant decreases in the percentage-of-occurrence of error patterns in both languages. The hypothesis supports a dynamic sys-tems perspective to speech development. The multi-ple-baseline across behaviours design provided a controlled experimental condition for each child ’ s response to intervention.

Our study shows that bilingual therapy for SSDs targeting phonological representation, articulation, and metaphonological skills can be effective in impro-ving the child ’ s overall sound system. Results indi-cated that the children ’ s speech improved in both languages, with a medium-to-large effect size depend-ing on treatment target observed in both children. Although the accuracy of treated targets improved in all cases, increases in overall accuracy were also noted, particularly overall consonant and vowel accuracy and utterance-level complexity. Gains were similar in mag-nitude in both languages, even with the Spanish lan-guage emphasis in treatment. We recognize that our research design does not allow for effi ciency measures, leaving us unable to determine the strength of the gains with an English-only or predominantly-English treatment model. Nevertheless, the reports from pre-vious therapy (discussed below), in which the children made limited gains in their primarily English treat-ment, lead us to believe that focusing on the children ’ s fi rst language and their stronger language resulted in greater overall treatment gains than an emphasis on their weaker language would have provided.

We also found that our treated targets resulted in improvement in non-treated phonological error pat-terns, such as fi nal consonant deletion, which decreased in both children, particularly in English. While one might interpret these changes as evidence of maturity, external factors suggest this is not the case. Both Car-los and Rub é n had participated in treatment for 6 months prior to the study. Their previous therapy was once per week in small groups. No control of number of productions per session was calculated. Parent and school therapist progress reports suggest little to no improvement in speech production goals during this time period. Also, TRG2 accuracy for both children did not change in error frequency during the control periods. We propose that the many positive effects on non-treated behaviours (e.g., decreases in non-target patterns, increases in PVC and PCC in English and Spanish) in our treatment in Spanish and English is in keeping with a dynamic systems perspec-tive and results from cross-linguistic generalization of treatment, treatment frequency and session length, and the number of target productions per session.

Five minutes of each treatment session were spent demonstrating to Carlos and Rub é n how the speech characteristic practiced that day in one language

Int J

Spe

ech

Lan

g Pa

thol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

24.2

1.19

8.14

8 on

05/

05/1

4Fo

r pe

rson

al u

se o

nly.

Cross-linguistic generalization in speech treatment 13

also applied to properties of the other language. The importance of meta-phonological tasks in treatment for bilingual SSDs has not been explored previously, and it is diffi cult to know if the improve-ment of targets in both languages resulted from the meta-phonological generalization, the dual-language treatment, or both. Further studies should explore these possibilities.

One unexpected fi nding was the increase in vowel accuracy in both children. This change was more dramatic in English. While we did not specifi cally treat vowels nor did we assume that vowel error was an aspect of either child ’ s speech sound disorder, it appears that the clear treatment models provided in English may have helped both boys build more accu-rate perceptual and production representations for English vowels (Rvachew & Nowak, 2004).

Our treatment emphasized Spanish, with approxi-mately three of four sessions conducted in Spanish. The decision to primarily focus on Spanish was motivated by the sequential bilingual situation of both children — they were in Spanish home settings, spent time primarily with family and friends who speak Spanish, and demonstrated stronger receptive and expressive Spanish skills. As noted in the results section, Rub é n had both a segmental goal, the pho-neme /s/, and a syllable-level goal, the production of stop � liquid clusters. Both goals for Carlos addressed phonotactic complexity, particularly in longer utter-ances. Goals were chosen for multiple reasons: func-tional load in both languages, frequency in both Spanish and English, and developmental appropri-ateness. It is important to consider the importance of the targets you select in a bilingual child as they apply across languages.

Results showed that targeting associations between languages results in cross-linguistic generalization and suggest the interconnectedness of seemingly autonomous languages, an inter-connectedness spec-ifi ed by DST (Thelen & Smith, 1995). In these two children, such generalization resulted in their ability to use skills in one language for more accurate pro-ductions in the other (Paradis, 2001). Both children showed improvement in both languages, not only on target structures but also on their phonological sys-tems as a whole. For example, both overall consonant and vowel accuracy increased from pre- to post-treat-ment, indicating system-wide changes that went beyond improvement on the treatment targets them-selves (although accuracy on those targets improved as well). Such target-specifi c and system-wide change shows at least some evidence for DST in that pho-nological properties which are closely associated result in greater cross-linguistic generalization across languages. This cross-linguistic generalization has been noted in studies examining the phonological skills of Spanish – English (e.g., Fabiano-Smith & Goldstein, 2010; Gildersleeve-Neumann et al., 2009) and Russian – English (Gildersleeve-Neumann & Wright, 2010) bilingual children. Albeit limited at

this point, there is converging evidence from treat-ment and non-treatment studies supporting cross-linguistic generalization, as would be indicated by Dynamic Systems Theory.

Our study of bilingual intervention showed that treating both languages had an overall positive effect on the speech system of sequential bilingual children with SSDs. We observed fairly equal change in Span-ish and English, even though the ratio of therapy was ~ 3:1, Spanish-to-English. Thus, treating the stronger language appeared to have a greater impact on the children ’ s overall system. We also found that an increase in English vowel accuracy in at least some sequential bilingual children can occur without targeted vowel therapy.

As a postscript, Rub é n ’ s and Carlos ’ changing lin-guistic environments highlight the bilingual child ’ s shifts in opportunities, necessities for both languages throughout development, and emphasize the impor-tance of considering the whole child in treatment language choices. Both children continued in the clinic for a year after this study. Rub é n ’ s family then moved 1 hour away and he left our clinic. However, after 6 months Rub é n ’ s family returned, driving twice a week after school for services. Rub é n ’ s par-ents believed his English-only school setting and English-only therapy were resulting in loss of speech skill in Spanish and English. Rub é n ’ s teacher noted after his return to bilingual therapy that his English speech skills improved. Carlos also moved: his family returned to Mexico a year after his treatment ended. Although we have lost contact, we are happy that Carlos received services in Spanish, since he currently relies on his Spanish more than English skills in his educational and social communication. For many bilingual children, the home language is often more important than English in many communication settings and can have greater importance in future situations.

Our results show that bilingual treatment was effective in treating speech sound disorder in two Spanish – English children. Our treatment language ratio emphasized Spanish over English and resulted in improvement in both languages. This Spanish emphasis was based on the ages, length of exposure to English, and home and school profi les of Carlos and Rub é n. The ratio will not be the same for all bilingual children or all sequential bilingual children; instead, language histories for each client will need to be considered in determining the ideal ratio.

Our study is limited by its scope; there are no other treatment data for bilingual Spanish – English children with SSD. Clearly, extensive treatment effi -cacy research needs to be completed to understand the best language environments for treatment of bilingual children. Bilingual children differ in mul-tiple ways and the language needs in treatment will vary. The understanding of intervention for bilingual children is in its infancy. A variety of studies are needed to determine how treatment needs differ for

Int J

Spe

ech

Lan

g Pa

thol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

24.2

1.19

8.14

8 on

05/

05/1

4Fo

r pe

rson

al u

se o

nly.

14 C. Gildersleeve-Neumann & B. A. Goldstein

simultaneous and sequential bilinguals, for different language typology combinations, and for children who have concomitant disorders, among other fac-tors. Research should include more single-subject design studies to explore change in individual chil-dren and randomized control studies to explore group treatment effects. To better serve bilingual children, we need to understand the effectiveness as well as effi ciency of English-only, Spanish-only, and bilingual therapy for bilingual children.

Note

IPC of each word is derived by calculating the 1. point value for each of eight complexity factors (consonant place, consonant manner, vowel type, word length, word shape, cluster produc-tion, type of cluster, and intra-word consonant sequencing). IPC has been validated in English (Howell, Au-Yeung, Yaruss, & Eldridge, 2006) and Spanish (Howell & Au-Yeung, 2007).

Acknowledgments

We are grateful to Claudia Meyer and Lisa Arakelian for their contributions to treatment and to the families who took part in this study.

Declaration of interest: The authors report no confl icts of interest. The authors alone are respon-sible for the content and writing of the paper.

References

American Speech-Language-Hearing Association . (2007) . Childhood apraxia of speech [Technical Report] . ASHA . www.asha.org/policy. Rockville, MD.

American Speech-Language-Hearing Association . (2010) . Demographic profi le of ASHA members providing bilingual and Spanish-language services . ASHA . http://www.asha.org/uploaded Files/Demographic-Profile-Bilingual-Spanish-Service-Members.pdf.

Ballard , K. J. , Robin , D. A. , McCabe , P. , & McDonald , J . (2010) . A treatment for dysprosody in childhood apraxia of speech . Journal of Speech, Language, and Hearing Research , 53 , 1227 – 1245 .

Campos , S. J . (1995) . The Carpinteria Preschool Program: A long-term effects study . In E. Garcia , & B. McLaughlin (Eds.) , Meeting the Challenge of Linguistic and Cultural Diversity in Early Childhood Education . New York: Teacher’s College .

Cobo-Lewis , A. B. , Eilers , R. E. , Pearson , B. Z. , & Umbel , V. C . (2002) . Interdependence of Spanish and English knowledge in language and literacy among bilingual children . In D. K. Oller , & R.E. Eilers (Eds.) , Language and literacy in bilingual children (pp . 118 – 134) . Clevedon, UK: Multilingual Mat-ters .

Cohen , J . (1988) . Statistical power analysis for the behavioral sciences . Hillsdale, NJ: Erlbaum .

Duhon , G. J., Mesmer, E. M., Atkins, M. E., Greguson, L. A., & Olinger, E. S. (2009). Quantifying intervention intensity: A systematic approach to evaluating student response to increasing intervention frequency . Journal of Behavioral Educa-tion , 18 , 101 – 118 .

Edeal , D. , & Gildersleeve-Neumann , C. E . (2011) . The impor-tance of production frequency in therapy for childhood apraxia of speech . American Journal of Speech-Language Pathology , 20 , 95 – 110 .

Elbert , M. , Powell , T. W. , and Swartzlander , P . (1991) . Toward a technology of generalization. How many exemplars are suffi -cient? Journal of Speech and Hearing Research , 34 , 81 – 87 .

Fabiano-Smith , L. , & Goldstein , B . (2010) . Phonological acquisi-tion in bilingual Spanish-English speaking children . Journal of Speech, Language, and Hearing Research , 53 , 160 – 178 .

Gershkoff-Stowe , L. , & Thelen , E. , (2004) . U-shaped changes in behavior: A dynamic systems perspective . Journal of Cognition & Development , 5 , 11 – 36 .

Gierut , J . (1998) . Treatment effi cacy: Functional phonological disorders in children . Journal of Speech, Language, and Hearing Research , 41 , S85 – S100 .

Gierut , J . (2005) . Phonological intervention: The how or the what? In A. Kamhi, & K. Pollock (Eds.), Phonological disorders in children: Clinical decision making in assessment and intervention (pp. 201 – 210) . Baltimore, MD: Brookes Publishing .

Gierut , J. , & Champion , A . (2001) . Syllable onsets II: Three-element clusters in phonological treatment . Journal of Speech, Language, and Hearing Research , 44 , 886 – 904 .

Gierut , J. , Elbert , M. , & Dinnsen , D . (1987) . A functional analy-sis of phonological knowledge and generalization learning in misarticulating children . Journal of Speech and Hearing Research , 30 , 462 – 479 .

Gierut , J. , Morrisette , M. , Hughes , M. , & Rowland , S . (1996) . Phonological treatment effi cacy and developmental norms . Language, Speech, and Hearing Services in Schools , 27 , 215 – 220 .

Gildersleeve-Neumann , C. E . (2014) . Phonology and Articulation Bilingual Assessment (PABA). [Unpublished assessment tool] . Portland, OR: Portland State University .

Gildersleeve-Neumann , C. E. , & Wright , K. E . (2010) . English phonological acquisition in 3- to 5-year-old children learning Russian and English . Language, Speech, and Hearing Services in Schools , 41 , 429 – 444 .

Gildersleeve-Neumann , C. E. , Pe ñ a , E. P. , Davis , B. L. , & Kester , E. S . (2009) . Effects on L1 during early acquisition of L2: Speech changes in Spanish at fi rst English contact . Bilingualism: Language and Cognition , 12 , 259 – 272 .

Holm , A. , & Dodd , B . (2001) . Comparison of cross-language generalisation following speech therapy . Folia Phoniatrica et Logopaedica , 53 , 166 – 172 .

Holm , A. , Ozanne , A. , & Dodd , B . (1997) . Effi cacy of intervention for a bilingual child making articulation and phonological errors . International Journal of Bilingualism , 1 , 55 – 69 .

Horner , R. H. , Carr , E. G. , Halle , J. , McGee , G. , Odom , S. , & Wolery , M . (2005) . The use of single-subject research to iden-tify evidence-based practice in special education . Exceptional Children , 71 , 165 – 179 .

Howell , P. , & Au-Yeung , J . (2007) . Phonetic complexity and stut-tering in Spanish . Clinical Linguistic and Phonetics , 21 , 111 – 127 .

Howell , P. , Au-Yeung , J. , Yaruss , S. J. , & Eldridge , K . (2006) . Phonetic diffi culty and stuttering in English . Clinical Linguis-tics and Phonetics , 20 , 703 – 716 .

Jakielski , K. J . (1998) . Motor organization in the acquisition of consonant clusters . Unpublished doctoral dissertation, Univer-sity of Texas at Austin. Ann Arbor, MI: UMI Dissertation services.

Kazdin , A. E . (2011) . Single-case research designs: methods for clinical and applied settings . New York: Oxford University Press .

Keshavarz , M. , & Ingram , D . (2002) . The early phonological development of a Farsi-English bilingual child . International Journal of Bilingualism , 6 , 255 – 269 .

Kohnert , K . (2013) . Language disorders in bilingual children and adults (2 nd ed.) . San Diego, CA: Plural Publishing .

Lovaas , O. I . (1987) . Behavioral treatment and normal educa-tional and intellectual functioning in young autistic children. Journal of Consulting and Clinical Psychology , 55 , 3 – 9 .

Int J

Spe

ech

Lan

g Pa

thol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

24.2

1.19

8.14

8 on

05/

05/1

4Fo

r pe

rson

al u

se o

nly.

Cross-linguistic generalization in speech treatment 15

Shriberg , L. D . (1993) . Four new speech and prosody-voice meas-ures for genetics research and other studies in developmental phonological disorders . Journal of Speech and Hearing Research , 36 , 105 – 140 .

Shriberg , L. D. , & Lof , G. L . (1991) . Reliability studies in broad and narrow phonetic transcription . Clinical Linguistics and Pho-netics , 5 , 225 – 279 .

Shriberg , L. D. , Austin , D. , Lewis , B. , McSweeney , J. , & Wilson , D . (1997) . The percentage of consonants correct (PCC) metric: Extensions and reliability data . Journal of Speech, Language, and Hearing Research , 40 , 708 – 722 .

Skelton , S. L . (2004) . Concurrent task sequencing in single-phoneme phonologic treatment and generalization . Journal of Communication Disorders , 37 , 131 – 155 .

So , L. K. H. (2007) . Cantonese speech acquisition . In S. McLeod (Ed.) , The international guide to speech acquisition (pp . 313 – 326) . Clifton Park, NY: Thomson Delmar Learning .

Stertzbach , J. B. , & Gildersleeve-Neumann , C. E . (2005) . Parent report as a speech screening tool in Spanish-speaking children . Technical session presented at the annual meeting of ASHA, San Diego, CA .

Strand , E. A. , & Debertine , P . (2000) . The effi cacy of an integral stimulation intervention with developmental apraxia of speech . Journal of Medical Speech-Language Pathology , 8 , 295 – 300 .

Strand , E. A. , Stoeckel , R. , & Baas , B . (2006) . Treatment of severe childhood apraxia of speech: A treatment effi cacy study . Journal of Medical Speech-Language Pathology , 14 , 297 – 306 .

Thelen , E . (1991) . Motor aspects of emergent speech: A dynamic approach . In N. A. Krasnegor , D. M. Rumbaugh , R. L. Schiefelbusch , & M. Studdert-Kennedy (Eds.) , Biological and behavioral determinants of language development (pp. 339 – 362) . Hillsdale, NJ: Lawrence Erlbaum Associates .

Thelen , E. , & Smith , L. B . (1995) . A dynamic systems approach to the development of cognition and action . Cambridge, MA: A Bradford Book, MIT Press .

Thelen , E. , Kelso , J. A. S. , & Fogel , A . (1987) . Self-organizing systems and infant motor development . Developmental Review , 7 , 39 – 65 .

Thomas , W. P. , & Collier , V. P . (2003) . The multiple benefi ts of dual language . Educational Leadership , 61 , 61 – 64 .

Tyler , A. A. , Edwards , M. L. , & Saxman , J. H . (1987) . Clinical application of two phonologically-based treatment procedures . Journal of Speech and Hearing Disorders , 52 , 393 – 409 .

Tyler , A. A. , Lewis , K. E. , & Welch , C. M . (2003) . Predictors of phonological change following intervention. American Journal of Speech-Language Pathology , 12 , 289 – 296 .

Warren , S. F. , Fey , M. E. , & Yoder , P. J . (2007) . Differential treatment intensity research: A missing link to creating opti-mally effective communication interventions. Mental Retar-dation and Developmental Disabilities Research Reviews , 13 , 70 – 77 .

Weiner , F . (1981) . Treatment of phonological disability using the method of meaningful minimal contrasts: Two case stud-ies. Journal of Speech and Hearing Disorders , 46 , 97 – 103 .

Wiig , E. , Secord , W. A. , & Semel , E . (2009) . The Clinical Evaluation of Language Fundamentals-Preschool 2nd Edition, Spanish . San Antonio, TX: Harcourt Assessment .

Williams , A. L . (2000) . Multiple oppositions: Case studies of variables in phonological intervention . American Journal of Speech-Language Pathology , 9 , 289 – 299 .

Maas , E. , & Farinella , K. A . (2012) . Random versus blocked prac-tice in treatment for childhood apraxia of speech . Journal of Speech, Language, Hearing and Research , 55 , 561 – 578 .

Maas , E. , Robin , D. A. , Hula , S. N. A. , Freedman , S. N. , Wulf , G. , Ballard , K. J. , et al . (2008) . Principles of motor learning in treatment of motor speech disorders . American Journal of Speech-Language Pathology , 17 , 277 – 298 .

Markham , C. , & Dean , T . (2006) . Parents ’ and professionals ’ perceptions of quality of life in children with speech and language diffi culty . International Journal of Disorders of Language and Communication , 41 , 189 – 212 .

Markham , C. , van Laar , D. , Gibbard , D. , & Dean , T . (2009) . Children with speech, language and communication needs: Their perceptions of their quality of life . International Journal of Language and Communication Disorders , 44 , 748 – 768 .

McCormack , J. , McLeod , S. , McAllister , L. , & Harrison , L. J . (2010) . My speech problem, your listening problem, and my frustration: The experience of living with childhood speech impairment . Language, Speech, and Hearing Services in Schools , 41 , 379 – 392 .

Oller , K. , & Delgado , R . (2000) . Logical International Phonetics Programs (v. 2.02) . Miami, FL: Intelligent Hearing Systems .

Paradis , J . (2001) . Do bilingual two-year-olds have separate phono-logical systems? International Journal of Bilingualism , 5 , 19 – 38 .

Perozzi , J. A. , & Sanchez , M. I. C . (1992) . The effect of instruction in L1 on receptive acquition of L2 for bilingual children with language delay . Language, Speech, and Hearing Services in Schools , 23 , 348 – 352 .

Powers , B. L . (2010) . Bilingual Spanish-English speaking 4-year-old children: English normative data and correlations with parent reports. (Unpublished master ’ s thesis). Portland State University, Portland, OR.

Ray , J . (2002) . Treating phonological disorders in a multilingual child: A case study . American Journal of Speech-Language Pathology , 11 , 305 – 315 .

Restrepo , M. A. , Morgan , G. , & Thompson , M . (2013) . The effi cacy of vocabulary intervention for dual language learners with language impairment . Journal of Speech, Language, and Hearing Research , 56 , 748 – 765 .

Robey , R. R . (1998) . A meta-analysis of clinical outcomes in the treatment of aphasia . Journal of Speech, Language, and Hearing Research , 41 , 172 – 187 .

Rumberger , R. W. , & Larson , K. A . (1998) . Toward explaining dif-ferences in educational achievement among Mexican American language-minority students . Sociology of Education , 71 , 68 – 92 .

Rvachew , S. , & Bernhardt , B. M . (2010) . Clinical implications of dynamic systems theory for phonological development. American Journal of Speech Language Pathology , 19 , 34 – 50 .

Rvachew , S. , & Nowak , M . (2001) . The effect of target-selection strategy on phonological learning . Journal of Speech, Language, and Hearing Research , 44 , 610 – 623 .

Rvachew , S. , & Nowak , M . (2004) . Effect of phonemic perception training on the speech production and phonological awareness skills of children with expressive phonological delay . American Journal of Speech-Language Pathology , 13 , 250 – 263 .

Saben , C. , & Ingham , J . (1991) . The effects of minimal pairs treatment on the speech-sound production of two children with phonologic disorders . Journal of Speech and Hearing Research , 34 , 1023 – 1040 .

Shin , H. B. , & Kominski , R. A . (2010) . Language use in the United States: 2007. American community survey reports ACS-12 . Washington, DC: U. S. Census Bureau .

Int J

Spe

ech

Lan

g Pa

thol

Dow

nloa

ded

from

info

rmah

ealth

care

.com

by

24.2

1.19

8.14

8 on

05/

05/1

4Fo

r pe

rson

al u

se o

nly.