Complimentary Author PDF: Not for Broad Dissemination

A Treatment for Dysprosody

in Childhood Apraxia of Speech

Kirrie J. Ballard
University of Sydney,
New South Wales, Australia Purpose: Dysprosody is considered a core feature of childhood apraxia of speech
(CAS), especially impaired production of lexical stress. Few studies have tested the
effects of intervention for dysprosody. This Phase II study with 3 children investigated
Donald A. Robin
the efficacy of a treatment targeting improved control of relative syllable durations in
The University of Texas Health Sciences 3-syllable nonwords representing strong-weak (SW) and weak-strong ( WS) stress
Center at San Antonio and patterns (e.g., BAtigu or baTIgu). Treatment sessions were structured along the
The University of Texas principles of motor learning ( PML) approach.
at San Antonio Method: Three children, age 7 to 10 years, with mild to moderate CAS and
normal language development participated in an intensive 3-week treatment.
Patricia McCabe Within-participant designs with multiple baselines across participants and behaviors
Jeannie McDonald were used to examine acquisition, generalization, and maintenance of skill.
University of Sydney Results: All children improved in their ability to control relative duration of syllables in
SW and WS nonwords. Improvement was also noted in control of loudness and pitch
contrasts. Treatment effects generalized to untreated nonword stimuli, but minimal
change was seen in production of real words.
Conclusion: Findings support the efficacy of this approach for improving production
of lexical stress contrasts. Structuring the intervention according to the PML approach
likely stimulated strong maintenance and generalization effects.
KEY WORDS: childhood apraxia of speech, prosody, lexical stress, treatment,
motor learning

T
here are few studies on the efficacy of treatment for childhood apraxia
of speech (CAS). Historically, clinicians have labeled these children
as difficult to treat because of limited or slow progress (Forrest, 2003;
Hall, Jordan, & Robin, 2007) and a lack of research evidence to guide
treatment approaches (Morgan & Vogel, 2008). Nevertheless, the symp-
toms of CAS typically persist throughout life. Thus, the development
of innovative approaches to speech treatment in CAS is critical. To that
end, we designed a novel treatment approach that targets the perceived
dysprosody that is a predominant feature of CAS (American Speech-
Language-Hearing Association [ASHA], 2007).
Though there has been debate over the diagnostic markers of CAS,
ASHA (2007) published a position statement to serve as a unifying guide
for research on this population. Three diagnostic characteristics were
identified in the statement as having wide support, reflecting impaired
programming of the spatial and temporal parameters of speech move-
ments. These are “(a) inconsistent errors on consonants and vowels in
repeated productions of syllables or words, ( b) lengthened and disrupted
coarticulatory transitions between sounds and syllables, and (c) inappro-
priate prosody, especially in the realization of lexical or phrasal stress
[emphasis added]” (ASHA, 2007, p. 2). The latter two features underlie
the common perception that individuals with CAS segment their speech,

Journal of Speech, Language, and Hearing Research • Vol. 53 • 1227–1245 • October 2010 • D American Speech-Language-Hearing Association 1227
 Complimentary Author PDF: Not for Broad Dissemination

as if talking syllable by syllable (Robin, Maas, Sanberg, are consistent with those of Shriberg and colleagues, but
& Schmidt, 2007). Interestingly, these features are also they did not comment about whether their acoustic mea-
central to the adult form of apraxia of speech (Duffy, 2005; sures correlated well with perceptual judgments of lex-
McNeil, Robin, & Schmidt, 1997). ical stress. Munson et al. (2003) performed both acoustic
Previous studies that have examined the supra- and perceptual measures of CVCV nonword productions
segmental disruptions in CAS have found abnormal pro- from five CAS and five phonologically delayed ( PD) chil-
duction of lexical, or metrical, stress as measured by dren between the ages of 3;9 (years;months) and 8;11.
perceptual judgments and/or acoustic variables (Munson, They reported that all children were able to mark WS
Bjorum, & Windsor, 2003; Nijland et al., 2003; Shriberg, nonwords with durational contrast and SW nonwords
Aram, & Kwiatkowski, 1997; Skinder, Strand, & Mignerey, with pitch and intensity contrasts. However, the pro-
1999; Velleman & Shriberg, 1999). Lexical stress is pho- ductions of the CAS children were perceived by listeners
netically realized through manipulation of three vari- to be less accurate in stress placement than the PD chil-
ables: duration of the vowel (ms), vocal intensity (dB), dren. Munson et al. did not comment on how acoustic
and fundamental frequency (F0 in Hz; Kager, 2007). In measures may have correlated with degree of perceptual
English, stressed syllables have longer vowel durations accuracy.
and higher peak intensity and peak F0. Multisyllabic In summary, perceptual and acoustic measurement
words in English tend to have an alternating stressed- studies have supported the hypothesis that CAS is a
destressed pattern across syllables. Nouns typically fol- speech motor disorder that, in part, affects control of
low a trochaic pattern of strong-weak (SW) stress, while temporal parameters of speech movements that under-
verbs more typically have an iambic pattern of weak- lie production of prosodic features at the syllable level.
strong ( WS). Children as young as 12 months of age have However, the relationship between acoustic and per-
the capacity to manipulate these phonetic features in their ceptual measures has not been fully explored. This is an
prelinguistic productions (Davis, MacNeilage, Matyear, & interesting question, given that perceptual judgments
Powell, 2000). Pollock, Brammer, and Hageman (1993) are often considered the gold standard in assessment
analyzed SW and WS CVCV nonword productions in chil- and treatment planning for speech disorders (Duffy,
dren age 2 to 4 years. They reported that the 2-year-olds 2005).
were able to signal lexical stress accurately over CVCV CAS intervention studies have focused primarily on
nonwords through relative vowel duration. The duration improving segmental errors, with few targeting impair-
contrast became more pronounced between 2 and 3 years, ments of coarticulation or prosody. For example, interven-
primarily through shortening of the unstressed syllable. tions have focused on accuracy of phonemes (Williams &
Although 2-year-olds were not signaling stress through Stephens, 2004), often taking a linguistic approach based
pitch and intensity, by 3 years of age children were able on training phonological processes (Powell, 1996) or pho-
to do so. There was little change observed in relative du- nological awareness (Moriarty & Gillon, 2006). Other
ration, pitch, and intensity measures from 3 to 4 years of studies have included children with severe speech im-
age. pairment and very young children and so have focused
The speech motor impairment of CAS appears to on developing intelligible production of core vocabularies
interfere with development of the fine rapid control of (Strand, Stoeckel, & Bass, 2006) or augmentative and
articulatory muscles that is required for expression of alternative forms of communication (Cumley & Swanson,
subtle lexical stress contrasts across syllables (Shriberg 1999). However, it is important to note that Strand and
et al., 1997; Skinder et al., 1999). In one study, 52% of colleagues (2006) advocate for varying prosody during
53 children with suspected CAS were perceived to use treatment for articulatory accuracy of words. In their
excessive, equal, or misplaced stress in connected speech Dynamic Temporal and Tactile Cueing treatment ap-
samples, which was a much higher incidence than the proach, they routinely have children produce words with
10% level for a broader population of children with speech randomly varying intonational patterns, as modeled by
delay (Shriberg et al., 1997). While Velleman and Shriberg the clinician. This team has not specifically measured or
(1999) found that children with CAS were perceived to reported changes in prosody in the children treated.
produce the same types of lexical stress errors as chil- Given the agreement among clinicians and re-
dren with non-CAS speech delay, they noted that the searchers that CAS is an impairment of speech motor
stress errors persisted in children with CAS as late as control and that dysprosody is a core symptom of CAS
14 years. Nijland et al. (2003) performed acoustic mea- (ASHA, 2007), it is logical to develop treatments to im-
sures of word and segment durations in children with prove speech motor skills underlying production of pro-
CAS. They found significantly longer durations in CAS sodic contrasts. Though treatments for dysprosody exist
than normally developing peers, and, unlike peers, the (Hall et al., 2007; for a review, see Hargrove, Anderson, &
CAS children did not shorten vowel duration in weaker Jones, 2010), no studies have experimentally tested in-
stressed initial syllables. The findings of Nijland et al. tervention for prosody in CAS (Morgan & Vogel, 2008).

1228 Journal of Speech, Language, and Hearing Research • Vol. 53 • 1227–1245 • October 2010
 Complimentary Author PDF: Not for Broad Dissemination

Furthermore, few interventions for CAS or production of treatment targets rapid and fluent production of lexical
prosodic contrasts have been cast in a motor learning stress contrasts in multisyllabic strings. It is assumed
framework. The principles of motor learning ( PML) ap- that impaired lexical stress production in CAS stems
proach provides a strong starting point (Schmidt & Lee, from a deficit in rapid and fluent control of temporal and
2005). The PML approach was developed within the spatial parameters of articulator movement required to
Schema Theory of Motor Control and Learning (Schmidt, produce the fine variations in duration, vocal intensity,
1975). Over the past 30 years, several hundred published and F0 across syllables. The approach uses multisyllabic
studies have led to the development of a set of simple strings to provide intensive practice in transitioning
principles that facilitate maintenance and generaliza- between segments and syllables with varying stress
tion of trained motor skills in children and adults (Maas patterns. Nonwords are selected because they (a) allow
et al., 2008; Schmidt & Lee, 2005). These studies have random sequencing of syllables to increase the variety
largely considered limb movements in healthy individuals, of transitions practiced and ( b) simulate encountering
but an increasing number have shown similar effects in novel words and rapidly planning movements and move-
adults with limb and speech motor impairments (see ment sequences without the influence of learned motor
Maas et al., 2008, for a review). plans or linguistic representations. In this instance, the
PML guide the structure and frequency of practice level selected for treatment was three-syllable varied
and the provision of augmented feedback. Existing ther- nonsense strings; at the commencement of the treat-
apy programs for CAS can be easily adapted to a PML ment, the three participants were able to produce these
structure without losing their core elements such as strings with high segmental accuracy. Therefore, the
the tactile and kinesthetic cues of the Prompts for Re- effects of the treatment on production of lexical stress
structuring Oral-Muscular Phonetic Targets (PROMPT) could be examined independently of the influence of
method. segmental errors. However, the children produced the
strings with relatively equal duration, intensity, and
Principles of practice structure that facilitate long-
pitch levels across the first and second syllables in both
term learning of motor skills include (a) high intensity of
SW and WS forms, giving rise to the perception of syl-
practice, ( b) training multiple and varied skills in par-
lable segregation.
allel with (c) random ordering of stimuli within session,
and (d) initiating training at high levels of task or stim- Our treatment approach was guided by the PML
ulus complexity. In relation to the complexity principle, approach in the choice of complex and varied stimuli,
it has been shown that training a motor skill of given high-intensity practice, random order of stimulus pre-
complexity (e.g., st consonant blends) results in gener- sentation during practice, and low frequency of KR
alization down the hierarchy (i.e., to s and t) but not up feedback. Stimulus complexity was defined by number
the hierarchy (i.e., str blends; Maas et al., 2008). Prin- of syllables and different phonemes in a string (Schneider
ciples of feedback frequency and structure that facilitate & Frens, 2005). High-complexity stimuli were novel four-
long-term motor learning include (a) restricting feed- syllable strings with three different consonants and vowels
back to information on simple accuracy (i.e., knowledge (e.g., butagitu), midcomplexity stimuli were three-syllable
of results, or KR feedback) rather than detailed perfor- varied strings (e.g., butagi), and low-complexity stimuli
mance characteristics (i.e., knowledge of performance, or were three-syllable strings with varied consonants (e.g.,
KP feedback), ( b) providing feedback on only 50% of re- bataga). As a central problem of CAS is planning varied
sponses, and (c) inserting a 3–5-s silent pause between a sequences of syllables for fluent production, it was pre-
response and the provision of feedback. Generally, the dicted that manipulating prosody in novel and varied
principles serve to increase the difficulty of the practice three-syllable sequences would be challenging. The spe-
task and encourage self-evaluation of responses. These cific hypotheses were as follows:
principles show high overlap with the recently devel- 1. Treatment, structured in accordance with PML, will
oped principles of experience-dependent neural plastic- result in improved ability to produce two contrasting
ity (Kleim & Jones, 2008; Ludlow et al., 2008). Several stress patterns (i.e., SW and WS) in three-syllable
treatment efficacy studies have supported use of PML varied strings (e.g., BAtigu and baTIgu). That is, the
with acquired apraxia of speech (Austermann Hula, children would move from a largely unimodal dis-
Robin, Maas, Ballard, & Schmidt, 2008; Ballard, Maas, tribution (e.g., equal stress on both SW and WS
& Robin, 2007; Knock, Ballard, Robin, & Schmidt, 2000; items) to a bimodal distribution (i.e., SW and WS
Maas, Barlow, Robin, & Shapiro, 2002) and CAS (Strand stress) for relative duration, peak intensity, and peak
et al., 2006). F0 across adjacent syllables in a string.
The current study is a Phase II trial testing the 2. Treatment effects will generalize to less complex
efficacy of a new treatment approach in three cases of three-syllable strings (e.g., BAtaga) but not to more
CAS, using within-subject experimental designs. The complex four-syllable strings (e.g., BAtiguta).

Ballard et al.: Prosody Treatment for CAS 1229

 Complimentary Author PDF: Not for Broad Dissemination

3. Treatment effects may generalize to untreated SW their mother, who was a qualified elementary school read-
and WS three-syllable real words, as they contain ing teacher. Assessment of language skills using the Aus-
the same number of syllables, and existing lexical tralian version of the Clinical Evaluation of Language
representations and semantic linkages may facili- Fundamentals, Fourth Edition (CELF–4; Semel, Wiig, &
tate production. However, generalization may not Secord, 2006) confirmed language skills within normal
occur because the selected stimuli (see Appendix) limits or higher for all children. No formal cognitive as-
contain on average 7.2 phonemes, including conso- sessments were administered. However, CELF–4 scores
nant clusters, compared with six singleton phonemes correlate moderately to highly with measures of intelli-
in the treated stimuli. gence in children (Pearson Education, 2008). Results of
4. Treatment and generalization effects will be re- language and speech testing are presented in Tables 1
tained at 4 weeks posttreatment, with performance and 2.
better than baseline levels. All three children had previously received tradi-
tional articulation therapy, working on one sound at a
5. Acoustic measures of relative duration, intensity,
time while moving from sounds in isolation and simple
and /or pitch will correlate highly with perceptual
words up through connected speech. Individual therapy
judgments of lexical stress, demonstrating that
was provided one to two times per week, and the family
these variables accurately capture the perceived
strictly adhered to a daily home practice regimen through-
lexical stress errors and treatment-related changes
out. M1 attended therapy from age 3;2 to 6;9, working
of CAS speakers.
sequentially on velar plosives, fricatives, affricates, and
liquids. F1 attended therapy from age 2;10 to 7;0 and
worked on plosives, fricatives, affricates, and liquids. M2
Method attended therapy from 3;9 to 5;6, working sequentially
on fricatives, affricates, and liquids. In no case was there
Participants direct treatment of prosody. Treatment was terminated
Three children ( M1: male, age 10;10; F1: female, age when speech sound accuracy was at an age-appropriate
9;2; and M2: male, age 7;8) from a single family were level. Of note, all children also received occupational
referred by their parents, in response to a recruitment therapy for fine motor skills over the same period as
advertisement for a broader CAS treatment study. M1 speech therapy, and M2 was still receiving occupational
and F1 were left-handed, although there was no family therapy at the time of this study. Physical therapy had
history of left-handedness. All had a history of severe not been advised. The children had not received speech
speech sound disorder ( F1 > M1 > M2) in the context of therapy in the 2 years prior to the study. Upon enroll-
normal receptive and expressive language skills, with no ment, they produced only a few segmental errors that
documented hearing or visual impairment. Currently, were all sound distortions, and their speech intelligibility
clinical diagnosis of CAS is based solely on perceptual was 100%. However, their parents reported that family
judgment; there is no standardized test available. As and friends described their speech as robotic, with M2
such, diagnosis of CAS was based on presence of the most affected. Following the position statement on CAS
core perceptual features of CAS (ASHA, 2007) during the (ASHA, 2007), lexical stress was the focus here, as one
speech tasks of the Motor Speech Examination (i.e., aspect of speech naturalness.
diadochokinesis, production of mono- and multisyllabic
words, and connected speech; Duffy, 2005); the Goldman
Fristoe Test of Articulation—Second Edition (Goldman Stimuli
& Fristoe, 2000); the Children’s Test of Nonword Rep- Four sets of stimuli were created (see Appendix) and
etition (Gathercole & Baddeley, 1996), for M1 and F1 used with all the children. Set 1 included the to-be-
only; and the Inconsistency Assessment from the Diag- treated three-syllable nonsense strings containing three
nostic Evaluation of Articulation and Phonology (Dodd, different plosive consonants and three different long
Hua, Crosbie, Holm, & Ozanne, 2002), for M2 only. vowels. There were 36 possible CVCVCV combinations
There was unanimous agreement on CAS diagnosis (e.g., batigu, butiga), and these were duplicated to make
between the first three authors, all experienced speech- a set of 72 with two different stress conditions: 36 with
language pathologists. No orofacial structural abnormali- an SW stress pattern over the first two syllables and 36
ties, muscle weakness, or altered muscle tone and reflexes with a WS pattern. Twenty-five of the syllable strings
were identified; this assessment ruled out a frank dys- with SW and 25 with WS were randomly selected for
arthria ( Motor Speech Examination; Duffy, 2005). None treatment. Set 2 contained the remaining 22 strings that
of the children had been identified with, or received ser- were left untreated to test for generalization of treat-
vices for, language reading or intellectual impairment. ment effects to untreated exemplars of the treated be-
The children were reading above age level, as reported by havior. Set 3 included an additional 20 syllable strings

1230 Journal of Speech, Language, and Hearing Research • Vol. 53 • 1227–1245 • October 2010
 Complimentary Author PDF: Not for Broad Dissemination

Table 1. Results of pretreatment speech and language testing for the three participants (M1, F1, and M2).

M1 F1 M2

Standard Standard Standard

Test score Percentile Results score Percentile Results score Percentile Results

CELF–4 Receptive Language subtests

Word Classes 11 63 WNL 13 84 WNL 13 84 WNL
Concepts and Following Directions 11 63 WNL 12 75 WNL 7 16 WNL
Sentence Structure — — — — — — 10 50 WNL
Receptive Language score 106 66 WNL 115 84 WNL 98 45 WNL
CELF–4 Expressive Language subtests
Formulated Sentences 14 91 > NL 15 95 > NL 11 63 WNL
Word Classes 15 95 > NL 9 37 WNL — — —
Word Structure — — — — — — 14 91 > NL
Recalling Sentences 14 91 > NL 15 95 > NL 17 99 > NL
Expressive Language score 126 96 > NL 118 88 > NL 124 95 > NL
Raw score Raw score Raw score
Children’s Test of Nonword 31/40 < NL 26/40 < NL — —
Repetition (Gathercole &
Baddeley, 1996)
Diagnostic Evaluation of — — — — 10/25 Mild
Articulation and Phonology:
Inconsistency Assessment
(Dodd et al., 2002)
Errors Errors Errors
Goldman Fristoe Test of Mild distortion of vowels and Mild distortion of vowels, /s/, Distortion of vowels, /r/, and
Articulation—Second Edition /r/; voicing of /f/; schwa and /r/; devoicing of medial /w/; devoicing of [th];
(Goldman & Fristoe, 2000) insertion in some /l/ blends /v/; schwa insertion in some audible nasal air emission
/l/ blends on final /m/

Note. CELF–4 = Clinical Evaluation of Language Fundamentals, Fourth Edition (Semel et al., 2006); WNL = within normal limits; NL = normal limits.
Dash indicates not administered or not appropriate for child’s chronological age.

generated to test for generalization to untreated but three syllables and SW/WS stress pattern) but were more
related behaviors. These included 10 less complex three- complex in number of phonemes (M = 7.2) and presence of
syllable strings with plosives (e.g., bataga, in both SW consonant blends and later developing phonemes.
and WS forms, to total 20 stimuli) and 10 more complex All stimuli were presented orthographically in
four-syllable strings (e.g., butigabi, in both SW and WS 24-point Times New Roman font on individual cards.
forms, to total 20 stimuli). Set 4 included 10 familiar real Strong syllables in the nonwords were indicated with
words (e.g., motorbike) that shared features of Set 1 (i.e., bold font, and vowels were written as they sounded (e.g.,
SW: baateegoo; WS: baateegoo).

Table 2. Presence of characteristics of childhood apraxia of speech,

based on perceptual judgment by three experienced speech-language Design and Procedures
pathologists.
Single-subject designs with multiple baselines
Feature of apraxia of speech M1 F1 M2 across behaviors and participants were used. Partici-
pants had three ( M2) or four ( M1 and F1) baseline tests
Slow speech rate Mild Mild Mild-moderate on the speech behaviors of interest. Treatment was then
Equal stress across syllables Mild Mild Moderate
provided in 60-min sessions 4 days per week for 3 weeks
Syllable segregation Mild Mild Moderate
during a school vacation. Baselines and experimental
Inconsistent errors Mild Mild Mild-moderate
and retention probes were administered to measure treat-
Repetitions and revisions Mild Mild Mild
Sound distortions Mild Mild Mild ment effects, generalization to untreated items, and ex-
perimental control.

Ballard et al.: Prosody Treatment for CAS 1231

 Complimentary Author PDF: Not for Broad Dissemination

Baseline and Experimental Probes The clinicians defined these terms to the participants (i.e.,
emphasis referred to getting the durational contrast; flu-
Each baseline test involved the participants read-
ency to speaking at a habitual rate without pauses, hesi-
ing aloud a randomly selected set of 10 treated strings,
tations, or repetitions; and loudness to overall intensity
as well as 10 untreated strings of the same complexity,
level for the sentence). They referred to each term as they
10 less complex strings, 10 more complex strings, and
provided the KR and KP feedback for each response (e.g.,
10 real words, for a total probe of 50 productions. Ex-
“Emphasis and fluency were spot on, but your voice was
perimental probes identical to the baseline were ad-
too loud”). Prepractice continued until the participant
ministered after every fourth treatment session, prior
had produced five consecutive target sentences correctly,
to any treatment for the day to avoid the influence of
as perceived by the clinician, without a model.
recent clinician feedback and rehearsal on performance.
Three experimental probes were completed during the Practice. The practice part of each session followed
treatment phase, with the third probe being the first immediately after the prepractice. Between 100 and 120
posttreatment test. A fourth probe was given 4 weeks practice trials were completed per session, comprising
posttreatment to test retention of any treatment and 10–12 trials for each of the 10 treated strings. Syllable
generalization effects. strings, embedded in carrier phrases, were presented
orthographically and read aloud by the participants.
Stress pattern was indicated with bold font on the strong
Treatment syllable. All stimuli were presented in random order,
Treatment sessions began once all baseline tests and KR feedback was provided on 50% of responses
had been completed. One student clinician was assigned fading from 100% on the first 10 trials to 10% on the final
to each participant and administered all treatment ses- 10 trials. Participants were told that they would not re-
sions and experimental probes. All sessions were super- ceive any modeling from the clinician, would be given
vised by the first or fourth authors via a one-way mirror feedback on accuracy ( KR feedback) only, and should
and were video recorded for fidelity and reliability anal- listen to their productions and self-evaluate each attempt.
yses. Each session included a prepractice and a practice KR feedback initially was provided for stress pattern only,
component (Schmidt & Lee, 2005). with a correct response being perceived as relative du-
Prepractice. A random set of 10 treatment stimuli ration of about 3:2 for SW strings (i.e., the first syllable
were selected and randomly inserted into sentence-final about 50% longer than the second) and 2:3 for WS strings
position of each of 10 carrier phrases (e.g., “He bought (i.e., the second syllable about 50% longer than the first).
a____” or “Can you find my ____?”). Participants made Once success rates for stress assignment were above 50%,
attempts at producing these sentences with modeling KR feedback was given on stress assignment, fluency (i.e.,
from the clinician and detailed feedback on performance speech rate and presence of pauses), and, for M1 and M2
(i.e., KP feedback) to shape correct responses. This KP only, overall loudness level. For responses receiving KR
feedback informed the participants of the parameters of feedback, the clinician would first rate stress assignment
a correct response and prepared them for being able to (e.g., “Good emphasis”) and, when this was correct, then
adjust productions independently during the practice give feedback on fluency (e.g., “Nice and fluent”) and
when no models or KP feedback would be provided. loudness (e.g., “Not good on loudness”). To aid under-
All participants needed verbal KP feedback address- standing of this more complex KR feedback, the feedback
ing (a) relative duration of the first to second syllable sheet with the words “emphasis,” “fluency,” and “loud-
in the target string, to realize SW or WS stress pat- ness” (described above) was placed on the table. Rest
terns; ( b) maintaining habitual speech rate, rather than breaks (i.e., a 3–5-min board game) were interspersed.
slowing speech rate; and (c) avoiding pauses between As is typical in clinical practice, all KP and KR feed-
syllables in the string, between the carrier phrase and back was based on the clinician’s perceptual judgments
the string, or between presentation of the stimulus and of stress assignment, syllable duration and speech rate,
onset of the response. Participants M1 and M2 required vocal intensity, pitch variation, and speech natural-
KP feedback to discourage increasing overall loudness ness. Perceptual judgments of stress assignment corre-
for the syllable string. M2 required KP feedback to dis- late highly with acoustic measures of stress assignment
courage exaggerating pitch variation across the syllable (Davis et al., 2000; also see Results section below).
string. He also required rhythmic tapping by the cli-
nician and himself, as well as visual aids, to shape ap-
propriate relative durations of syllables without markedly
Equipment
slowed speech rate. The visual aids included short and long All baselines and experimental probes were recorded
blocks arranged on the table. To aid the participants in in a quiet room using a Marantz PMD670 PC Card Re-
processing these three types of feedback, a feedback sheet corder at 48 kHz. An Audio-Technica ATM75 cardioid
was made with the words emphasis, fluency, and loudness. headset microphone was placed 5 cm from the mouth.

1232 Journal of Speech, Language, and Hearing Research • Vol. 53 • 1227–1245 • October 2010
 Complimentary Author PDF: Not for Broad Dissemination

Dependent Measures The DUR measures were divided by number of pho-

nemes to normalize the values across the nonword and
Acoustic Measures real word data sets because real words were not con-
Acoustic measurements of syllable and /or vowel trolled for number of phonemes in each syllable, as shown
duration (ms), peak vocal intensity (dB), and peak fun- in Equation 3:
damental frequency ( F0 in Hz) were made on all non-
word and word productions in all tests. Acoustic analyses NORMDURnor ¼ ðdk þ dkþ1 Þ=ðmk þ mkþ1 Þ; ð3Þ
were done using the Praat signal-processing software where d is duration of the kth syllable, and m is the num-
(Boersma & Weenink, 2001). In nonsense syllable strings, ber of phonemes in the kth syllable.
all consonants were plosives. Syllable duration for the
first and second syllables was measured as the time The average value for real word NORMDUR at
from onset of one plosive burst to onset of the next baseline was used as a within-subject benchmark for
plosive burst. For real word stimuli, vowel duration was the desired NORMDUR for the nonsense strings at post-
measured because the words were not controlled for treatment. All nonsense strings contained four pho-
number or type of phonemes within syllables. Measure- nemes (CVCV), while four real words had four phonemes,
ment of onset and offset of vowels in the first two syl- and six contained five phonemes. Durations of the four-
lables followed the guidelines of Peterson and Lehiste phoneme and five-phoneme real words did not differ sig-
(1960), utilizing pitch and intensity contours and for- nificantly within-subject (p > .05 for all comparisons; t tests
mant trajectories generated within Praat software. Peak with Welch’s correction for unequal variances). Similarly,
intensity (dB) and peak F0 ( Hz) measures for all syl- no comparisons of the durations for SW versus WS tokens
lables were made by selecting the vowel portion and gen- for real words at baseline or for nonsense strings post-
erating the measures automatically using the Praat treatment differed significantly ( p > .05 for all compar-
software algorithms. isons). Therefore, values for SW and WS stimuli were
pooled within each real and nonsense word stimulus list.
Pairwise variability indices of lexical stress. For each
stimulus item and each acoustic measure, the pairwise
variability index ( PVI; Low, Grabe, & Nolan, 2000; see
Perceptual Measures
Equation 1 below) was calculated to determine degree of First, the percentage of responses perceived by each
asymmetry across the first two syllables of a string. PVI clinician to be correctly produced during each practice
provides a measure normalized for speech rate (or loud- session was tallied to document change in ability to
ness level, or F0 level), for more accurate comparison produce SW and WS stress contrasts on three-syllable
across participants and within participants over time. nonsense strings in sentences. Criteria for a correct re-
A positive PVI is consistent with an SW pattern (i.e., sponse were the same as those described for production
greater duration, intensity, or F0 on the first syllable), of target stress pattern (“emphasis”) and habitual speech
and a negative PVI is consistent with a WS pattern rate (“fluency”) during practice (see above). Accuracy for
(i.e., greater duration, intensity, or F0 on the second syl- stress pattern alone and for stress pattern plus speech
lable), with increasing values indicating more pronounced rate was recorded to demonstrate an accuracy/speed
contrast. A zero PVI value indicates equal stress over trade-off.
both syllables. The formula for duration is given by Second, three adults blinded to the study hypothe-
Equation 1: ses and timing of samples perceptually judged (a) the 20
productions of the treated and untreated C1V1C2V2C3V3
PVIðdurÞ ¼ 100 $ fðdk % dkþ1 Þ=½ðdk þ dkþ1 Þ=2(g; ð1Þ
strings and ( b) the 10 real words from each child’s final
where d is the duration of the kth syllable. baseline test and immediate posttreatment test. The
purpose of the latter judgments was to identify percep-
Similar formulas were used for peak intensity in dB
tible changes in production of the targeted SW and WS
[PVI(int)] and peak F0 in Hz [PVI( F0)].
nonwords and words. For each stimulus type, the sam-
Total duration of first two syllables. The average ples from each child were randomly ordered within par-
duration ( DUR) of the first two syllables of the nonsense ticipant. Samples were presented in free field from a
strings and real words was calculated (see Equation 2) to laptop computer in a quiet room. Raters were instructed
document changes in fluency (i.e., speech rate). This ad- to adjust the volume to a comfortable level, listen to
dressed whether control of lexical stress (i.e., relative each sample only once, and rate it on a 5-point equal-
duration, intensity, and pitch) was compromising speech appearing interval scale where 1 = a clear SW pattern, 2 =
rate, which can also affect speech naturalness: tending to SW, 3 = equal stress over the first two syllables,
DUR ¼ dk þ dkþ1 ; ð2Þ 4 = tending to WS, and 5 = a clear WS pattern. Order of
nonwords and words for each participant and order of
th
where d is duration of the k syllable. participants were counterbalanced across raters.

Ballard et al.: Prosody Treatment for CAS 1233

 Complimentary Author PDF: Not for Broad Dissemination

No measures of segmental accuracy or intelligibility syllable sequences from pre- to posttreatment were
were included in the study. The children were 100% in- analyzed using the Kruskal–Wallis nonparametric anal-
telligible and made few segmental errors pretreatment. ysis of variance (ANOVA) with Dunn’s multiple compar-
The consonants in the experimental stimuli were ear- isons post hoc test. Data for each participant and for each
lier developing sounds that had been mastered by all of the four stimulus sets (i.e., one treated set and three
children. untreated sets) were analyzed separately. A conserva-
tive alpha level of .01 was used to adjust for multiple
within-subject comparisons. Data for each participant
Reliability were pooled across the final two baselines and the two
Interrater agreement on the independent variable posttreatment probes (i.e., Probe 3 and the 4-week re-
(i.e., a checklist including administration of each step in tention probe) to increase statistical power for the un-
the treatment protocol, judgment of response accuracy, treated stimulus sets, which contained a small number
and provision of KR feedback during practice) was cal- of tokens.
culated on a randomly selected 10% of trials across the The DUR measures for treated and untreated strings
three participants. Interrater measures were completed were compared from pre- to posttreatment using the
live by the first and fourth authors, who viewed all ses- nonparametric Mann–Whitney U test. Again, data from
sions via video camera. While these measures were not the final two baselines and from the two retention probes
blinded, this procedure allowed any discrepancies in were pooled for each participant. This analysis tested
protocol administration to be discussed and corrected whether rate of production changed over the course of
between sessions, without compromising the overall fi- the study. Finally, the NORMDUR measures for real
delity of the treatment. Point-to-point agreement was words were compared with the NORMDUR measures of
85.1% for M1, 85.0% for F1, and 87.0% for M2. Intrarater the three nonsense string sets at pretreatment and at
agreement on the independent variable was measured posttreatment using the Kruskal–Wallis nonparametric
by having each clinician view the recording of randomly ANOVA with Dunn’s multiple comparisons test to deter-
selected trials at least 2 weeks later. For 18% of trials mine whether speech rate on the syllable strings differed
across the three participants, point-to-point agreement from habitual rate.
was 96.2% for M1, 98.0% for F1, and 95.6% for M2.
Perceptual ratings of stress production for each
Interrater reliability on the primary dependent var- nonword or word in the baseline and immediate post-
iable of syllable duration in the nonword strings was cal- treatment probes were averaged over the three listen-
culated for 12% of responses on probes across the three ers. Average within-participant ratings from pre- to
participants, and intrarater reliability was calculated on posttreatment were compared using the Mann–Whitney
8% of responses. For interrater measures on each child, U statistic to determine whether changes in stress pro-
the clinicians assigned to the other children were given duction were perceptually robust. In addition, the cor-
randomly selected samples. For intrarater measures, a cli- relation between average ratings and PVI values for
nician remeasured a set of randomly selected samples at each participant was calculated to determine the strength
least 2 weeks after original measurement. In both cases, of relationship between these perceptual and instrumental
date and original syllable duration values were concealed. measures.
Wilcoxon matched-pairs signed-ranks tests showed no
significant differences for inter- or intrarater measures
( p > .05, Pearson r = .98 and .99, respectively). The Results
average point-to-point interrater difference for syllable
duration was 1.22 ms (SE = 2.90), and the average intra- Perceptual Measures During Treatment
rater difference was 0.17 ms (SE = 0.38). Reliability was As shown in Figures 1–3, all three participants were
not calculated on the measures of peak vocal intensity perceived to improve in their ability to produce the two
and F0 for vowels, as these measures were automatically stress contrasts of SW and WS in the three-syllable treat-
generated using Praat software (Boersma & Weenink, ment stimuli over the course of the 12 sessions. From the
2001). first to the final practice session, M1 improved from 0%
to 80% correct for production of stress at habitual rate of
speech, F1 improved from 0% to 75.8% correct, and M2
Data Analysis improved from 0% to 38% correct. M2, the youngest
The perceptual measures from the practice sessions and most dysprosodic of the cases, did not reach the cri-
and the PVI measures from the baseline and experimen- terion of 80% correct over three consecutive sessions, but
tal probes were graphed for visual inspection of treat- he did perform at or above 80% correct for production of
ment, generalization, and maintenance effects. Changes the stress pattern alone in four of the last five practice
in PVI(dur), PVI(int), and PVI( F0) for SW versus WS sessions.

1234 Journal of Speech, Language, and Hearing Research • Vol. 53 • 1227–1245 • October 2010
 Complimentary Author PDF: Not for Broad Dissemination

Figure 1. Percentage of M1’s productions perceived to be correct for either stress pattern alone (open circles) or for
stress pattern and habitual speech rate (closed circles) during the 12 treatment sessions.

PVIs of Lexical Stress durational contrast in SW strings (i.e., more positive

PVI), while F1 and M2 increased the durational contrast
Treatment Effect in WS strings (i.e., more negative PVI). M1 also showed a
A positive treatment effect is represented by PVI significant difference in PVI(int) and PVI( F0) for SW and
values in experimental probes for SW and WS stim- WS nonwords at posttreatment, with some shift in PVI
uli separating, with SW stimuli having a positive PVI values for both stimulus types. F1 and M2 showed a sig-
posttreatment and WS having a negative PVI. Data for nificant posttreatment difference between SW and WS
the treated C1V1C2V2C3V3 stimuli and the untreated in PVI(int), although, for M2, the distinction was reduced
C1V1C2V2C3V3 stimuli were pooled, as performance on at 4 weeks posttreatment. F1 and M2 did not use pitch to
both sets was not significantly different for all participants contrast SW and WS posttreatment.
( Mann–Whitney U tests).
All children showed a significant treatment effect by
developing appropriate durational contrasts for SW ver- Generalization of Treatment Effects
sus WS stimuli. For the treated stimuli at baseline, no As noted above, effects of treatment on the complex
differences were found in PVI(dur), PVI(int), or PVI(F0) three-syllable strings (i.e., C1V1C2V2C3V3) generalized
across SWand WS stimuli for any participants (see Table 3 to the untreated exemplars of the same-level stimuli. Gen-
and Figures 4–6). All participants showed significant dif- eralization to less complex three-syllable strings was
ferences in PVI(dur) for SW compared to WS at posttreat- evident to some degree in all participants (see the second
ment. M1 achieved this primarily through increasing the row of charts in Figures 4–6 and Table 3). For M1, the

Figure 2. Percentage of F1’s productions perceived to be correct for either stress pattern alone (open circles) or for
stress pattern and habitual speech rate (closed circles) during the 12 treatment sessions.

Ballard et al.: Prosody Treatment for CAS 1235

 Complimentary Author PDF: Not for Broad Dissemination

Figure 3. Percentage of M2’s productions perceived to be correct for either stress pattern alone (open circles) or for
stress pattern and habitual speech rate (closed circles) during the 12 treatment sessions.

separation of the SW and WS stimuli on all PVI mea- contrasts primarily by intensity and pitch variations,
sures can been seen in the final probes in Figure 4 and with significant differences between SW and WS stimuli
with the significant difference between SWand WS stim- for PVI(int) and PVI( F0) values but not for PVI(dur),
uli on PVI(dur) and PVI(int) measures at posttreatment as indicated in Figure 4 and Table 3. With treatment
(see Table 3). A similar change from pre- to posttreat- focusing on relative durations of syllables, M1 produced
ment was seen in F1, although the focus of change was a different pattern posttreatment. That is, PVI(dur)
on the PVI(dur) and PVI( F0) measures (see Figure 5 and values were significantly different for SW and WS real
Table 3). For M2, PVI(dur) and PVI(int) for less complex word stimuli, but PVI(int) and PVI( F0) were no longer
WS stimuli improved significantly with treatment (see significantly different. F1 and M2, on the other hand,
Figure 6 and Table 3). However, duration and intensity showed a significant difference between SW and WS
contrasts between SW and WS at posttreatment were words at baseline and posttreatment for PVI(dur), as
not statistically reliable, likely due to loss of skill in the shown in Figures 5 and 6, respectively, and Table 3. For
4 weeks following treatment. M2, PVI(int) differentiated SW and WS words in base-
Generalization to more complex four-syllable strings line but not posttreatment.
was evident but more limited than the changes seen
in less complex stimuli (see the third row of charts in Retention of Treatment Effects
Figures 4–6 and Table 3). For M1, the only finding of For the statistical analysis, data from immediate post-
note is the significant improvement in differentiation of treatment probes and the 4-week retention probes were
SW and WS productions for the PVI(F0) measure at post- pooled, and so long-term retention was not analyzed
treatment. Significant change in control of relative du- separately. However, the top row of charts in Figures 4–6
ration of syllables did not reach significance due to the shows the final retention data point for the treated stim-
variable performance on the PVI(dur) measure during uli. There was no or minimal deterioration in performance on
baseline (see Figure 4). For F1, relative durations of syl- the treated items over the 4-week posttreatment period
lables in SW and WS stimuli changed from being not for M1 and F1. M2’s ability to differentiate the SW and
significantly different in baseline to being a significant WS stimuli in production was not maintained; while he
difference posttreatment (see Figure 5 and Table 3). This produced both SW and WS stimuli with equal stress (i.e.,
likely reflects limited generalization of treatment ef- PVI[dur] close to zero) in baseline, he tended to produce
fects, with PVI(dur) on more complex productions chang- all treated strings as WS in the final retention probe (i.e.,
ing but not PVI(int) or PVI( F0), as shown in Table 3. For negative PVI values for both SW and WS stimuli).
M2, there was no reliable improvement differentiation of
more complex SW and WS stimuli at posttreatment, com-
pared to baseline (see Table 3). There was an undesirable Total Duration of First Two Syllables
shift of PVI(dur) for SW stimuli from near zero at baseline Two measures were used to track changes in du-
to negative values posttreatment (see Figure 6 and Table 3). ration of the initial two syllables of the treated and
Production of real words showed a somewhat dif- untreated stimulus items: DUR and NORMDUR. Con-
ferent and unexpected pattern of results in all partici- sistent with an increase in familiarity and fluency with
pants. M1 at baseline marked the SW and WS stress the treated nonword stimuli from intensive practice, values

1236 Journal of Speech, Language, and Hearing Research • Vol. 53 • 1227–1245 • October 2010
 Complimentary Author PDF: Not for Broad Dissemination

Table 3. Results of statistical comparisons (Kruskal–Wallis test [KW] and Dunn’s multiple comparisons post hoc test) of the pairwise variability
index (PVI) for strong-weak (SW) versus weak-strong (WS) stimuli at pretreatment (Pre) and posttreatment (Post) for M1, F1, and M2.

Participant PVI(dur) PVI(int) PVI(F0)

M1
Treated KW = 45.34, p < .0001 KW = 26.22, p < .0001 KW = 29.55, p < .0001
SW vs. WS at Pre: ns; Post: p < .001a SW vs. WS at Pre: ns; Post: p < .001 SW vs. WS at Pre: ns; Post: p < .001
Pre vs. Post for SW: p < .01; WS: ns Pre vs. Post for SW: ns; WS: ns Pre vs. Post for SW: ns; WS: ns
Less complex KW = 23.42, p < .0001 KW = 17.29, p < .01 KW = 16.77, p < .001b
SW vs. WS at Pre: p < .05; Post: p < .01 SW vs. WS at Pre: ns; Post: p < .01 SW vs. WS at Pre: p < .05; Post: p < .05
Pre vs. Post for SW: ns; WS: ns Pre vs. Post for SW: ns; WS: ns Pre vs. Post for SW: ns; WS: ns
More complex KW = 13.91, p < .01 KW = 10.59, p < .05 KW = 14.86, p < .01
SW vs. WS at Pre: ns; Post: p < .05 SW vs. WS at Pre: ns; Post: p < .05 SW vs. WS at Pre: ns; Post: p < .01
Pre vs. Post for SW: ns; WS: ns Pre vs. Post for SW: ns; WS: ns Pre vs. Post for SW: ns; WS: ns
Real words KW = 26.93, p < .0001 KW = 19.64, p < .001 KW = 18.86, p < .001
SW vs. WS at Pre: p < .05; Post: p < .001 SW vs. WS at Pre: p < .01; Post: ns SW vs. WS at Pre: p < .01; Post: p < .05
Pre vs. Post for SW: ns; WS: ns Pre vs. Post for SW: ns; WS: ns Pre vs. Post for SW: ns; WS: ns
F1
Treated KW = 34.52, p < .0001 KW = 31.58, p < .0001 KW = 8.08, p < .05
SW vs. WS at Pre: ns; Post: p < .001 SW vs. WS at Pre: ns; Post: p < .001 SW vs. WS at Pre: ns; Post: ns
Pre vs. Post for SW: ns; WS: p < .01 Pre vs. Post for SW: ns; WS: p < .01 Pre vs. Post for SW: ns; WS: ns
Less complex KW = 16.86, p < .001 KW = 10.65, p < .05 KW = 18.77, p < .001
SW vs. WS at Pre: ns; Post: p < .001 SW vs. WS at Pre: ns; Post: ns SW vs. WS at Pre: ns; Post: p < .01
Pre vs. Post for SW: ns; WS: p < .01 Pre vs. Post for SW: ns; WS: p < .05 Pre vs. Post for SW: ns; WS: p < .05
More complex KW = 22.35, p < .001 KW = 10.14, p < .05 KW = 7.44, ns
SW vs. WS at Pre: ns; Post: p < .001 SW vs. WS at Pre: ns; Post: p < .05 SW vs. WS at Pre: ns; Post: p < .05
Pre vs. Post for SW: ns; WS: ns Pre vs. Post for SW: ns; WS: ns Pre vs. Post for SW: ns; WS: ns
Real words KW = 23.71, p < .0001 KW = 15.42, p < .01 KW = 4.99, ns
SW vs. WS at Pre: p < .01; Post: p < .01 SW vs. WS at Pre: ns; Post: p < .01 SW vs. WS at Pre: ns; Post: ns
Pre vs. Post for SW: ns; WS: ns Pre vs. Post for SW: ns; WS: ns Pre vs. Post for SW: ns; WS: ns
M2
Treated KW = 16.41, p < .001 KW = 25.25, p < .0001 KW = 10.76, p < .05
SW vs. WS at Pre: ns; Post: p < .01 SW vs. WS at Pre: ns; Post: p < .001 SW vs. WS at Pre: ns; Post: ns
Pre vs. Post for SW: ns; WS: p < .01 Pre vs. Post for SW: ns; WS: p < .01 Pre vs. Post for SW: ns; WS: ns
Less complex KW = 14.35, p < .01 KW = 13.01, p < .01 KW = 11.14, p < .05
SW vs. WS at Pre: ns; Post: p < .05 SW vs. WS at Pre: ns; Post: p < .05 SW vs. WS at Pre: ns; Post: ns
Pre vs. Post for SW: ns; WS: p < .01 Pre vs. Post for SW: ns; WS: p < .01 Pre vs. Post for SW: ns; WS: p < .05
More complex KW = 20.79, p < .0001 KW = 12.40, p < .01 KW = 4.84, ns
SW vs. WS at Pre: ns; Post: ns SW vs. WS at Pre: ns; Post: ns SW vs. WS at Pre: ns; Post: ns
Pre vs. Post for SW: p < .01; WS: p < .05 Pre vs. Post for SW: ns; WS: p < .05 Pre vs. Post for SW: ns; WS: ns
Real words KW = 23.78, p < .0001 KW = 27.21, p < .0001 KW = 4.35, ns
SW vs. WS at Pre: p < .01; Post: p < .01 SW vs. WS at Pre: p < .001; Post: p < .05 SW vs. WS at Pre: ns; Post: ns
Pre vs. Post for SW: ns; WS: ns Pre vs. Post for SW: ns; WS: ns Pre vs. Post for SW: ns; WS: ns

Note. PVI was calculated over the first two syllables in each string, and comparisons are presented for duration (dur), intensity (int), and fundamental
frequency (F0). Pretreatment data are pooled across the final two baselines and posttreatment data across the two retention probes for each participant.
Post hoc results show significance of comparisons between SW and WS at each time point and between time points for each lexical stress condition.
a
Effect sizes for all significant post hoc tests were large (i.e., d > 1.0). bOne of the remaining two post hoc comparisons was significant; these comparisons
are not presented here because they were not considered informative (i.e., PVI for SW at pretreatment vs. WS at posttreatment or PVI for WS at pretreatment
vs. SW at posttreatment).

of DUR decreased significantly for all participants from duration values were still longer than normal, with av-
pre- to posttreatment (see Table 4). For untreated less erage durations of 571.1 ms and 678.6 ms, respectively.
complex nonwords, DUR decreased for M1 and M2 from This equates to a speaking rate of about 3.5 syllables/s
pre- to posttreatment but remained unchanged for F1 for M1 and 2.9 syllables/s for M2. More complex words
(see Table 4). However, M1’s and M2’s posttreatment decreased in duration for M1 only, reaching an average

Ballard et al.: Prosody Treatment for CAS 1237

 Complimentary Author PDF: Not for Broad Dissemination

Figure 4. M1’s performance on baseline and experimental probes for treated and untreated stimulus sets. The dependent variables are the median
pairwise variability indices (PVIs) for duration (left panel), intensity (middle panel), and F0 (right panel). For each PVI variable, the treated
three-syllable strings are in the top chart, the untreated less complex strings are in the second chart from the top, the untreated more complex strings
are in the third chart, and the untreated real words are in the bottom chart. Filled symbols represent the strong-weak (SW) stress pattern, and
open symbols represent the weak-strong (WS) pattern. A positive PVI indicates an SW stress pattern (e.g., PVI[dur] of 40 is equivalent to a
3:2 ratio), a negative PVI indicates a WS stress pattern, and a zero PVI value indicates equal stress.

of 585.2 ms. For M2, the average durations of the more comparisons). For M1, the durations of the treated strings
complex stimuli decreased from 913.2 ms to 757.1 ms but were significantly shorter than the more complex strings
did not reach significance due to high variability on the at pretreatment ( p < .01), but no other post hoc com-
baseline tests. For real words, as expected, the summed parisons were significant. For F1 and M2, the durations
duration of the first two vowels, excluding consonants, of the treated and untreated strings did not differ from
showed no changes in duration for any participant. each other.
The normalized average duration of the first two At posttreatment, the difference in NORMDUR
syllables of the real words ( NORMDUR) at baseline across the four stimulus sets continued to be signifi-
was used as a within-subject benchmark for the desired cantly different ( M1: KW = 36.08, p < .0001; F1: KW =
duration of the first two syllables of the nonsense strings 42.92, p < .0001; M2: KW = 41.48, p < .0001), with post
at posttreatment. For all children at pretreatment, hoc tests showing that the real words continued to be sig-
NORMDUR was significantly different across stimulus nificantly shorter ( p < .01 for all comparisons) than the
sets ( M1: Kruskal–Wallis [KW] = 53.93, p < .0001; F1: treated and untreated nonsense strings. For M1 and F1,
KW = 43.94, p < .0001; M2: KW = 43.72, p < .0001), with the average NORMDUR for the treated strings was sig-
Dunn’s multiple comparisons post hoc tests showing nificantly shorter than the more complex untreated strings
that the real words were significantly shorter than the ( M1: p < .01; F1: p < .05), but the difference between
treated and untreated nonsense strings ( p < .001 for all treated and less complex untreated strings did not reach

1238 Journal of Speech, Language, and Hearing Research • Vol. 53 • 1227–1245 • October 2010
 Complimentary Author PDF: Not for Broad Dissemination

Figure 5. F1’s performance on baseline and experimental probes for treated and untreated stimulus sets. See Figure 4 for the key.

significance. For M2, NORMDUR was not significantly to 2.09; WS: 2.67 to 4.04). No pre–post statistical com-
different across treated and untreated nonsense stimu- parisons were significant for average perceptual ratings
lus sets. of real words. This is likely due to the small sample size.
However, all three children showed a tendency for stress
on SW real words to become more distinct from pre- to
Relationship Between Perceptual posttreatment (i.e., average rating moving closer to 1; see
and Instrumental Measures Table 5). There was minimal change in WS stress pro-
The average rating (out of 5) across three listeners duction on the real words. These findings are remark-
for each nonword and real word response was entered ably similar to the results of the pre- to posttreatment
into a nonparametric pre- to posttreatment paired com- statistical comparisons of PVI(dur) measures (see Table 3).
parison for each subject (see Table 5). An average rating The relationship between perceived stress pattern
near 1 indicated a clear SW pattern, a rating near 3 in- of nonwords and real words and acoustic PVI measures
dicated equal stress, and a rating near 5 indicated a clear was examined to determine whether the acoustic mea-
WS pattern. For M1, ratings of SW nonwords improved sures were capturing the perceptually based diagnostic
significantly, from an average of 2.39 to 1.24; ratings of feature of dysprosody in CAS. The analyses included
WS strings did not differ significantly from pretreat- the items that had been perceptually rated by indepen-
ment to posttreatment, being 4.11 and 4.26, respectively. dent listeners (see the Perceptual Measures subsection
For F1, only ratings of WS nonwords improved signifi- above)—that is, 20 treated and untreated C1V1C2V2C3V3
cantly, from an average of 2.45 to 4.39; ratings for SW nonwords and the 10 real words from each child’s final
strings were 2.00 at pretreatment and 1.64 at posttreat- baseline test and immediate posttreatment test (i.e.,
ment. For M2, average ratings for both SW and WS non- 40 nonwords and 20 real words per child). For each pro-
words improved significantly with treatment (SW: 2.67 duction, the average perceptual rating for the three

Ballard et al.: Prosody Treatment for CAS 1239

 Complimentary Author PDF: Not for Broad Dissemination

Figure 6. M2’s performance on baseline and experimental probes for treated and untreated stimulus sets. See Figure 4 for the key.

listeners was used. For M1, the perceptual ratings significantly correlated with PVI(dur) and PVI(int) but
were highly correlated with all PVI measures, PVI(dur): not PVI( F0), PVI(dur): r = –.80, p < .0001, 95% CI [–.89,
r = –.91, p < .0001, 95% CI [–.95, –.82]; PVI(int): r = –.51, –.65]; PVI(int): r = –.42, p < .01, 95% CI [–.65, –.12];
p < .001, 95% CI [–.71, –.23]; PVI(F0): r = –.70, p < .0001, PVI(F0): r = –.26, p > .05, 95% CI [–.53, –.06]. Similar to
95% CI [–.83, –.50]. For F1, the perceptual ratings were M1, all of M2’s PVI measures were highly correlated with

Table 4. Average summed duration (and standard deviation) of the first two syllables in the treated and untreated nonsense syllable strings and the
first two vowels, excluding consonants, in the real words compared from pretreatment to posttreatment for M1, F1, and M2.

M1 F1 M2

Pre Post Pre Post Pre Post

Treated 595.5 (81.7) 516.1 (94.6)** 636.6 (143.9) 544.2 (69.9)** 773.9 (97.7) 674.2 (97.8)***
Less complex 669.7 (72.3) 571.1 (58.8)*** 613.4 (143.1) 597.7 (63.9)ns 819.6 (125.9) 678.6 (178.9)*
More complex 708.7 (107.9) 585.2 (65.1)*** 677.4 (83.4) 674.1 (97.8)ns 913.2 (310.9) 757.1 (65.4)ns
Real words 209.7 (76.7) 205.5 (80.7)ns 171.3 (49.8) 150.7 (45.6)ns 247.0 (93.1) 250.4 (90.2)ns

Note. Pretreatment (Pre) data are pooled across the final two baselines and posttreatment (Post) data across the two retention probes for each participant.
Asterisks indicate statistically significant differences between pre- and posttreatment measures for each stimulus type. Nonparametric two-sided Mann–
Whitney U test was used. A conservative alpha level of p < .01 was used to adjust for multiple comparisons within participants. Effect sizes for all significant
comparisons were large (i.e., d > 0.82).
*p < .01. **p < .001. ***p < .0001.

1240 Journal of Speech, Language, and Hearing Research • Vol. 53 • 1227–1245 • October 2010
 Complimentary Author PDF: Not for Broad Dissemination

Table 5. Average perceptual rating (and standard deviation) of lexical stress pattern for the treated and untreated nonsense syllable strings and
the real words at pretreatment and posttreatment for M1, F1, and M2.

M1 F1 M2

Pre Post Pre Post Pre Post

Treated strings
SW 2.39 (1.03) 1.24 (0.30)** 2.00 (0.65) 1.64 (0.71)ns 2.67 (0.36) 2.09 (1.31)*
WS 4.11 (0.44) 4.26 (0.57)ns 2.46 (1.05) 4.39 (0.75)** 2.67 (0.33) 4.04 (1.10)*

Real words
SW 1.94 (0.61) 1.50 (0.59)ns 1.94 (0.57) 1.67 (0.47)ns 2.06 (0.68) 1.50 (0.18)ns
WS 4.25 (0.50) 4.25 (0.88)ns 4.12 (0.19) 4.25 (0.74)ns 3.83 (0.33) 3.92 (0.92)ns

Note. Ratings (where 1 = clear fluent SW pattern, 3 = equal stress, and 5 = clear fluent WS pattern) were provided by three listeners. Asterisks indicate
statistically significant differences between pre- and posttreatment ratings for each behavior. Nonparametric two-sided Mann–Whitney U test was used.
A conservative alpha level of p < .01 was used to adjust for multiple comparisons within participants. Effect sizes for all significant comparisons were large
(i.e., d > 1.5), with the exception of the SW treated strings for M2, which showed a medium effect size (d = 0.60).
*p < .01. **p < .001.

perceptual ratings, PVI(dur): r = –.84, p < .0001, 95% CI in CAS. They are siblings with a history of normal lan-
[–.91, –.71]; PVI(int): r = –.80, p < .0001, 95% CI [–.89, guage and intellectual development and remarkably
–.64]; PVI( F0): r = –.55, p < .001, 95% CI [–.74, –.28]. similar motor speech impairment. Previous studies of
treatment for dysprosody have included children with
diverse coexisting impairments such as language delay
Discussion (Shea & Tyler, 2001) or autism (Bellon-Harn, Harn, &
Watson, 2007). Lexical stress errors have been shown to
This study tested the efficacy of a treatment for im- differentiate children with CAS from children with other
proving production of lexical stress contrasts in three speech sound production disorders (Nijland et al., 2003;
children with CAS. It was predicted that (a) the children Shriberg et al., 1997). Furthermore, the dysprosody in CAS
would improve in their ability to produce duration con- has been reported to persist into adolescence (Velleman
trasts across syllables in treated three-syllable strings, & Shriberg, 1999), despite these children often receiving
( b) this improvement would generalize to less complex considerable intervention for segmental errors. On the
three-syllable strings, (c) effects of treatment would not other hand, the relative homogeneity of this participant
generalize to more complex four-syllable strings and pos- sample may limit generalization of the findings to the
sibly not to real words, and (d) perceptual and acoustic broader population of children with CAS and to those
measures of lexical stress would be highly correlated. The who still demonstrate reduced intelligibility due to seg-
first two hypotheses were strongly supported. There was mental errors.
some evidence of generalization to more complex stimuli,
Second, the study provides support for the hypoth-
which is encouraging, albeit counter to the third hypoth-
esis that impaired ability for learning to produce prosodic
esis. The strong relationship found between acoustic and
variations is a primary feature of CAS rather than a con-
perceptual measures of treatment effects supports both
sequence of disrupted speech sound production. Further-
use of these objective measures to quantify change in ex-
more, the study highlights the difficulty of children with
perimental studies and reliance on the faster and more
CAS in controlling temporal aspects of speech production.
economical perceptual measures in clinical practice.
These children represented a continuum of severity from
M2, the youngest with the most prosodically disrupted
speech, to M1, whose prosody was only mildly affected.
Contributions of the Study Nonetheless, the treatment task was clearly a challenge
The current study makes three contributions to our for all three participants. All required numerous trials to
understanding of the nature and treatment of CAS and achieve improvement (see Figures 1–3), and all showed
prosodic disturbances. First, it is the first study to apply some vocal and nonverbal signs of task-related stress. For
within-subject experimental designs to test the efficacy example, F1 and M2 tended to move their head upward
of a treatment for dysprosody in CAS (ASHA, 2007; Morgan on the beat of a stressed syllable, M1 made swallowing
& Vogel, 2008). The three children reported in this study and oral preparatory movements prior to beginning the
provided a unique opportunity to study prosodic deficits sentence, and M1 and M2 increased their overall vocal

Ballard et al.: Prosody Treatment for CAS 1241

 Complimentary Author PDF: Not for Broad Dissemination

intensity inappropriately during production of the stim- However, Davis and colleagues observed that, when only
uli on practice trials. These behaviors were discussed one parameter is varied to mark stress, it tended to be
with the participants within one session of their appear- duration in infants but intensity in adults. All three chil-
ance and, in all cases, dissipated within three sessions. dren here showed strong durational contrasts, as well as
Additional evidence that the children experienced intensity contrasts, to mark stress in real words.
difficulty with the treatment task was found in the anal- The corroboration of the quantitatively determined
yses of syllable durations. The normalized duration of treatment effects with data from perceptual ratings is
the first two syllables of real words was much shorter comforting for two reasons. The measures of relative du-
than the duration of the first two syllables in the non- ration, intensity, and F0 within utterance captured the
sense strings. While durations for the treated strings perception of excessive/equal/misplaced lexical stress
decreased from pre- to posttreatment, they continued to as documented by Shriberg and colleagues (Shriberg
be significantly longer than the real words. There are et al., 1997; Velleman & Shriberg, 1999) and identified in
likely two explanations for this finding. One is that the the ASHA (2007) position statement on diagnosis of CAS.
strings were not practiced sufficiently to reach the fluency Our findings support the use of these measures to ob-
level of real speech. Second, while the strings were prac- jectively quantify lexical stress patterns in CAS and to
ticed in carrier sentences, no meaning was attached to measure change with intervention. In addition, the high
them. If novel but real words had been selected for treat- correlations between perceptual ratings and the PVI
ment, it is possible that they would have been integrated measures support the continued reliance on perceptual
into the mental lexicon for use in daily communication, judgments in daily clinical practice as valid indicators
which might have facilitated independent rehearsal or of impairment and treatment-related change. However,
use and thus more fluent or faster production. interrater reliability measures will continue to prove use-
It is important to note that M2 showed poorer per- ful in safeguarding against problems such as perceptual
formance than M1 and F1 during treatment (see Figure 3) drift (Kent, 1996).
and did not retain the treatment effects at 4 weeks One problem with the current study lies in the chil-
posttreatment (see Figure 6). This may be related to the dren’s ability to differentiate SW and WS stress con-
overall greater severity of his dysprosody, suggesting trasts, through manipulation of duration and intensity,
that he required more treatment sessions to maximize in highly familiar real words during baseline. The chil-
and stabilize performance. It is also possible that there dren were referred to the study with parental report of
was a clinician effect, as each child had a different cli- dysprosody, and clinicians in the study corroborated this.
nician. This was minimized by having the supervising It is possible that the acoustic measures used here were
clinicians involved with all children and ensuring that not sensitive to more subtle residual disruptions in lexical
interrater reliability on the independent variable was stress on familiar and well-learned words or that some
high (i.e., perceptual judgment of responses and provi- other measure or measures might have better captured
sion of KP and KR feedback during treatment session). the dysprosody. Nonetheless, the children demonstrated
Future studies should control for this effect. considerable difficulty mastering the experimental stim-
The third contribution of the current study is in un- uli. Future studies might include measures of intersylla-
derstanding the interplay of duration, intensity, and F0 bic pauses or coarticulation. PVI measures of real word
in signaling stress contrasts and their relationship to the production in typically developing children would aid
perception of lexical stress patterns. During treatment, interpretation, along with information on speed of re-
practice was directed toward controlling only the dura- sponse to training on novel words.
tional contrast. Despite this, important changes were A related point that begs attention is the use of
observed in variation of intensity and F0 contrasts, con- nonword stimuli in treatment. While the three children
sistent with the well-established finding that these were able to differentiate SW and WS real words during
three features are functionally linked. Davis et al. (2000) baseline, it is reasonable to propose that acquisition of
reported that adults most often varied just intensity plus this contrast would have taken longer than normal given
frequency with first-syllable stress but varied vowel du- the protracted rate of acquisition seen during the treat-
ration and/or intensity and frequency in almost all strings ment. Evidence from studies of typically developing chil-
with second-syllable stress. Their adult data set included dren supports this assumption (Davis et al., 2000; Pollock
both real word and nonword disyllables. Here, we chose et al., 1993). Davis et al. (2000) reported that infants as
to have the children focus on syllable duration to mark young as 12 months can manipulate vowel duration, in-
first- and second-syllable stress, and, for the nonsense tensity, and F0 to produce alternating stress patterns
strings, syllable rather than vowel duration was measured. across syllable strings in prelinguistic babble. Pollock
As such, the dominance of duration variation to mark et al. (1993) showed that 3-year-old children successfully
stress is likely to be a bias introduced by our methodology. control vowel duration, intensity, and pitch to mark SW

1242 Journal of Speech, Language, and Hearing Research • Vol. 53 • 1227–1245 • October 2010
 Complimentary Author PDF: Not for Broad Dissemination

and WS contrasts on imitated C1V1C2V2 nonwords within Signorelli for assistance with treatment delivery and data
a single testing session. Therefore, the choice of nonwords analysis. We are indebted to the family for their support and
here simulated learning of unfamiliar words. It also al- willingness to be involved in the project.
lowed for careful control of stimulus characteristics in the
experimental setting and rapid, highly reliable measure-
References
ment of syllable durations. In working with younger
children, however, novel real words might be a more American Speech-Language-Hearing Association.
functional type of stimulus. This also raises the issue that (2007). Childhood apraxia of speech [Position statement].
Available from www.asha.org/policy.
no formal and reliable tests of processing and production
of lexical stress are currently available to guide diagnosis Austermann Hula, S., Robin, D. A., Maas, E., Ballard, K. J.,
& Schmidt, R. A. (2008). Effects of feedback frequency and
and intervention for dysprosody in Australian English-
timing on acquisition, retention, and transfer of speech
speaking children. skills in acquired apraxia of speech. Journal of Speech,
Language, and Hearing Research, 51, 1088–1113.
Ballard, K. J., Maas, E., & Robin, D. A. (2007). Treating
Future Directions voicing control in apraxia of speech with variable practice.
Aphasiology, 21, 1195–1217.
The current study motivates further development of
the treatment approach. The nonword stimuli were highly Bellon-Harn, M. L., Harn, W. E., & Watson, G. D. (2007).
Targeting prosody in an eight-year-old child with high-
constrained and were not assigned meaning, although
functioning autism during an interactive approach to
they were presented as nouns in the carrier sentences therapy. Child Language Teaching and Therapy, 23,
during practice. More research is required to determine 157–179.
whether best outcomes are obtained with real or non- Boersma, P., & Weenink, D. (2001). Praat 3.9.15 [Computer
sense words (see Gierut & Morrisette, 2010). In addition, software]. Amsterdam, the Netherlands: Institute of Pho-
measures of connected speech could be included to ex- netic Sciences.
plore generalization of treatment effects to more func- Cumley, G. D., & Swanson, S. (1999). Augmentative and
tional speech tasks. alternative communication options for children with devel-
opmental apraxia of speech: Three case studies. Augmenta-
At times the clinicians reported difficulty with pro- tive and Alternative Communication, 15, 110–125.
viding rapid, confident judgments of response accuracy
Davis, B. L., MacNeilage, P. F., Matyear, C. L., & Powell,
during practice. An automated instrumental procedure J. K. (2000). Prosodic correlates of stress in babbling: An
would assist this, removing potential error and bias acoustical study. Child Development, 71, 1258–1270.
associated with perceptual judgments. However, intra- Dodd, B., Hua, Z., Crosbie, S., Holm, A., & Ozanne, A.
and interrater reliability and the correlation between (2002). Diagnostic Evaluation of Articulation and Phonology.
perceptual and acoustic measures were high, suggesting Oxford, England: Pearson Assessment.
instrumental measures are not necessary for clinical Duffy, J. R. (2005). Motor speech disorders: Substrates,
practice. differential diagnosis, and management. St. Louis, MO:
Mosby-Yearbook.
In conclusion, this approach shows promise for treat-
ing production of lexical stress contrasts in CAS. Al- Forrest, K. (2003). Diagnostic criteria of developmental
apraxia of speech used by clinical speech-language pathol-
though lexical stress is produced through variations in
ogists. American Journal of Speech-Language Pathology, 12,
syllable duration and peak intensity and fundamental 376–380.
frequency, participants may only need to focus on one
Gathercole, S., & Baddeley, A. (1996). The Children’s Test of
variable to achieve change across all three. This sim- Nonword Repetition. London, England: Harcourt Assessment.
plifies the therapy task for children. Structuring the in-
Gierut, J. A., & Morrisette, M. L. (2010). Phonological
tervention according to the PML approach probably learning and lexicality of treatment stimuli. Clinical
stimulated strong retention and generalization effects; Linguistics & Phonetics, 24, 122–140.
however, this was not specifically tested here. Further Goldman, R., & Fristoe, M. (2000). Goldman Fristoe Test of
research is required to test this intervention with a Articulation— Second Edition. Bloomington, MN: Pearson
larger sample of children with CAS and measure effects Assessment.
when combined with intervention for segmental accu- Hall, P. K., Jordan, L., & Robin, D. A. (2007). Developmental
racy in younger children. apraxia of speech: Theory and clinical practice (2nd ed.).
Austin, TX: Pro-Ed.
Hargrove, P., Anderson, A., & Jones, J. (2010). A critical
Acknowledgments review of interventions targeting prosody. International
Journal of Speech-Language Pathology, 11, 298–304.
The study was supported by University of Sydney grants Kager, R. (2007). Feet and metrical stress. In P. de Lacy ( Ed.),
to Kirrie J. Ballard and Patricia McCabe. We wish to thank The Cambridge handbook of phonology ( pp. 195–227).
Samantha Warhurst, Lauren van Rees, and Monique Cambridge, England: Cambridge University Press.

Ballard et al.: Prosody Treatment for CAS 1243

 Complimentary Author PDF: Not for Broad Dissemination

Kent, R. D. (1996). Hearing and believing: Some limits to Pollock, K. E., Brammer, D. M., & Hageman, C. F. (1993).
auditory-perceptual assessment of speech and voice dis- An acoustic analysis of young children’s productions of word
orders. American Journal of Speech-Language Pathology, stress. Journal of Phonetics, 21, 183–203.
5(3), 7–23. Powell, T. W. (1996). Stimulability considerations in the
Kleim, J. A., & Jones, T. A. (2008). Principles of experience- phonological treatment of a child with a persistent disorder
dependent neural plasticity: Implications for rehabilitation of speech-sound production. Journal of Communication
after brain damage. Journal of Speech, Language, and Disorders, 29, 315–333.
Hearing Research, 51, S225–S239. Robin, D. A., Maas, E., Sandberg, Y., & Schmidt, R. A.
Knock, T. R., Ballard, K. J., Robin, D. A., & Schmidt, R. (2007). Motor control and learning and childhood apraxia
(2000). Influence of order of stimulus presentation on speech of speech. In P. K. Hall, L. Jordan, & D. A. Robin ( Eds.),
motor learning: A principled approach to treatment for Developmental apraxia of speech: Theory and clinical
apraxia of speech. Aphasiology, 14, 653–668. practice (2nd ed., pp. 67–86). Austin, TX: Pro-Ed.
Low, E. L., Grabe, E., & Nolan, F. (2000). Quantitative Schmidt, R. A. (1975). Schema theory of discrete motor skill
characterizations of speech rhythm: Syllable-timing in learning. Psychological Review, 82, 225–260.
Singapore English. Language and Speech, 43, 377–401. Schmidt, R., & Lee, T. (2005). Motor control and learning:
Ludlow, C. L., Hoit, J., Kent, R., Ramig, L. O., Shrivastav, A behavioral emphasis (4th ed.). Champaign, IL: Human
R., Strand, E., . . . Sapienza, C. M. (2008). Translating Kinetics.
principles of neural plasticity into research on speech motor Schneider, S. L., & Frens, R. A. (2005). Training four-
control recovery and rehabilitation. Journal of Speech, syllable CV patterns in individuals with acquired apraxia
Language, and Hearing Research, 51, S240–S258. of speech: Theoretical implications. Aphasiology, 19,
Maas, E., Barlow, J., Robin, D. A., & Shapiro, L. P. 451–471.
(2002). Treatment of sound errors in aphasia and apraxia Semel, E., Wiig, E. H., & Secord, W. A. (2006). Clinical
of speech: Effects of phonological complexity. Aphasiology, Evaluation of Language Fundamentals, Fourth Edition:
16, 609–622. Australian. Marrickville, New South Wales, Australia:
Maas, E., Robin, D. A., Austermann-Hula, S. N., Freedman, Harcourt Assessment.
S. E., Wulf, W., Ballard, K. J., & Schmidt, R. A. (2008). Shea, R. L., & Tyler, A. A. (2001). The effectiveness of a
Principles of motor learning in treatment of motor speech prosodic intervention on children’s metrical patterns. Child
disorders. American Journal of Speech-Language Pathology, Language Teaching and Therapy, 17, 55–76.
17, 277–298.
Shriberg, L. D., Aram, D., & Kwiatkowski, J. (1997).
McNeil, M. R., Robin, D. A., & Schmidt, R. A. (1997). Developmental apraxia of speech: III. A subtype marked
Apraxia of speech: Definition, differentiation, and treat- by inappropriate stress. Journal of Speech, Language, and
ment. In M. R. McNeil ( Ed.), Clinical management of Hearing Research, 40, 313–337.
sensorimotor speech disorders ( pp. 311–344). New York,
NY: Thieme. Skinder, A., Strand, E. A., & Mignerey, M. (1999).
Perceptual and acoustic analysis of lexical and sentential
Morgan, A. T., & Vogel, A. P. (2008). Intervention for child- stress in children with developmental apraxia of speech.
hood apraxia of speech. Cochrane Database of Systematic Journal of Medical Speech-Language Pathology, 7, 133–144.
Reviews, Issue 3, Article No. CD006278.
Strand, E. A., Stoeckel, R., & Baas, B. (2006). Treatment
Moriarty, B., & Gillon, G. (2006). Phonological awareness of severe childhood apraxia of speech: A treatment efficacy
intervention for children with childhood apraxia of speech. study. Journal of Medical Speech-Language Pathology, 14,
International Journal of Language & Communication 297–307.
Disorders, 41, 713–734.
Velleman, S. L., & Shriberg, L. D. (1999). Metrical analysis
Munson, B., Bjorum, E. M., & Windsor, J. (2003). Acoustic of the speech of children with suspected developmental
and perceptual correlates of stress in nonwords produced by apraxia of speech. Journal of Speech, Language, and
children with suspected developmental apraxia of speech Hearing Research, 42, 1444–1460.
and children with phonological disorder. Journal of Speech,
Language, and Hearing Research, 46, 189–202. Williams, P., & Stephens, H. (2004). Nuffield Centre Dyspraxia
Programme. Windsor, United Kingdom: The Miracle Factory.
Nijland, L., Maassen, B., Van Der Meulen, S., Gabreëls,
F., Kraaimaat, F. W., & Schreuder, R. (2003). Planning
syllables in children with developmental apraxia of speech. Received July 1, 2009
Clinical Linguistics & Phonetics, 17, 1–24.
Accepted February 18, 2010
Pearson Education. (2008). Clinical Evaluation of Language
Fundamentals, Fourth Edition: Technical report: Correla- DOI: 10.1044/1092-4388(2010/09-0130)
tions between the CELF-4 and WISC-4 Integrated. Retrieved Contact author: Kirrie J. Ballard, Speech Pathology,
from http://pearsonassess.com /NR/rdonlyres/D9546406- Faculty of Health Sciences, University of Sydney,
0147-4B51-A283-ECD252694FC1 /0/CELF4_WISC4_ P.O. Box 170, Lidcombe, New South Wales 1825,
TechReport.pdf. Australia. E-mail: [email protected].
Peterson, G. E., & Lehiste, I. (1960). Duration of syllable
nuclei in English. The Journal of the Acoustical Society
of America, 32, 693–703.

1244 Journal of Speech, Language, and Hearing Research • Vol. 53 • 1227–1245 • October 2010
 Complimentary Author PDF: Not for Broad Dissemination

Appendix. Stimulus list.

Treated Untreated Untreated Untreated
three-syllable three-syllable less complex more complex Untreated Carrier sentences
strings strings strings strings real words used in treatment

baguti gatibu bigiti gubitaba crocodile (SW) I saw a

bitagu tabigu bitigi tagibutu cucumber (SW) Can you find my
bugita bugati bataga tibagatu hamburger (SW) He bought a
tabugi bigatu bugutu batugibi kangaroo (SW) Here’s the new
tigabu gubati tigibi tigabubu motorbike (SW) She has a big
tubiga butiga tabaga gatibugu pattycake (SW) There’s my
gatubi gabitu tugubu gubatigi computer ( WS) It ’s going to
gibatu tibagu gibiti gabituga echidna ( WS) It ’s a purple
gutiba batugi gataba gutibati spaghetti ( WS) I want a
gubita butagi gutubu bigutagi toboggan ( WS) I went to the

Note. Examples are presented rather than the complete stimulus set. SW = strong-weak stress pattern; WS = weak-strong stress pattern.
SW real words all have the SWS stress pattern, although, for some, primary stress is on the final syllable.

Ballard et al.: Prosody Treatment for CAS 1245