Professional Documents
Culture Documents
Feedback Frequency in Treatment For PDF
Feedback Frequency in Treatment For PDF
Feedback Frequency in Treatment For PDF
Research Article
Purpose: To examine the role of feedback frequency in Results: Findings were mixed, with 2 children showing an
treatment for childhood apraxia of speech (CAS). Reducing advantage for low-frequency feedback, 1 child showing a
the frequency of feedback enhances motor learning, and small advantage for high-frequency feedback, and 1 child
recently, such feedback frequency reductions have been showing no clear improvement in either condition.
recommended for the treatment of CAS. However, no published Conclusion: These findings suggest that reducing the
studies have explicitly compared different feedback frequencies frequency of feedback may be beneficial for some children
in this population. with CAS, although this may vary with the child’s age or
Method: Using an alternating treatments single-subject severity of apraxia. Caution is warranted in extrapolating from
design with multiple baselines across behaviors, retention and the nonspeech motor learning literature to speech treatment
transfer of learning were compared following high-frequency for CAS. Finally, this study contributes another replication
feedback and low-frequency feedback in 4 children with CAS. to the literature on the efficacy of integral stimulation treatment
Feedback frequency was manipulated in the context of an for children with CAS.
integral stimulation treatment. Changes in perceptual accuracy
were quantified with effect sizes and were compared across Key Words: childhood apraxia of speech, treatment,
conditions. motor learning, feedback, speech disorders
L
earning a new motor skill or improving an existing we manipulated the amount of feedback provided during
motor skill typically requires feedback regarding the treatment and examined the effects on retention and transfer
accuracy and performance of the movement during of learning.
practice. Such feedback enables the learner to determine
which aspects of the movement were successful, which were
Speech Motor Learning in CAS
not, and how the motor plan can be modified for success on
a future attempt. For children with childhood apraxia of The current consensus in the literature is that CAS re-
speech (CAS), learning the movements of speech is thought flects an impairment of speech motor programming and
to be particularly difficult (e.g., American Speech-Language- planning, which is evidenced by difficulties moving from
Hearing Association [ASHA], 2007). As such, feedback one speech sound to the next, inconsistent errors on vowels
may be especially critical for children with CAS, and a close and consonants, and prosodic problems (e.g., ASHA, 2007;
examination of different aspects of feedback implemented Grigos & Kolenda, 2010; Nijland et al., 2002; Skinder,
during treatment may contribute to improving speech motor Strand, & Mignerey, 1999; Terband, Maassen, van Lieshout,
learning in this population. The present study was part of & Nijland, 2011). In addition, it has been suggested that
a larger research program that systematically investigated children with CAS also have difficulties with speech motor
the effects of various conditions of practice in treatment for learning, based on their often noted slow, limited prog-
CAS (e.g., Maas & Farinella, 2012). In the current study, ress in treatment (e.g., Ballard, Robin, McCabe, & McDonald,
2010; Campbell, 1999; Shriberg, Aram, & Kwiatkowski,
1997; Strand, Stoeckel, & Baas, 2006) and on recent com-
a putational modeling simulations (e.g., Terband & Maassen,
University of Arizona, Tucson
b 2010).
Panacea Therapeutics, Tucson, AZ
c Although there are few published controlled treatment
Northern Arizona University, Flagstaff
efficacy studies that focus on CAS, the clinical observations
Correspondence to Edwin Maas: emaas@arizona.edu of slow and limited progress in treatment have motivated a
Editor: Carol Scheffner Hammer search for ways to optimize speech motor learning in chil-
Associate Editor: Ken Bleile dren with CAS. The literature on motor skill learning pro-
Received August 31, 2011 vides some potentially relevant ideas in this regard, in that
Revision received January 9, 2012 some conditions of practice and feedback have been found
Accepted March 9, 2012 optimal in promoting motor learning (sometimes referred to
DOI: 10.1044/1058-0360(2012/11-0119) as “principles of motor learning”; see Maas et al., 2008, for
American Journal of Speech-Language Pathology • Vol. 21 • 239–257 • August 2012 • A American Speech-Language-Hearing Association 239
review). Various authors have advocated the use of these items was negligible in all cases. These findings suggest that
conditions of practice and feedback in CAS treatment (e.g., although practice factors derived from the motor learning
Hall, 2000; Iuzzini & Forrest, 2010; McCauley & Strand, literature may influence speech motor learning in children
1999; Robin, Maas, Sandberg, & Schmidt, 2007), and sev- with CAS, the direction of those influences may be different.
eral studies have incorporated these conditions into CAS In other words, caution is warranted in extrapolating practice
treatment (e.g., Ballard et al., 2010; Edeal & Gildersleeve- principles from the motor learning literature to CAS treatment.
Neumann, 2011; Iuzzini & Forrest, 2010; Maas & Farinella, Thus, although the motor learning literature can provide
2012; Strand & Debertine, 2000; Strand, Stoeckel, & Baas, guidance for efforts to improve speech motor learning
2006). However, only two studies to date have explicitly outcomes in children with CAS, the few direct empirical
compared the effects of different practice conditions in CAS investigations of various practice factors in this population
treatment to determine whether the same advantages apply have yielded somewhat mixed results. As such, the con-
to this population (Edeal & Gildersleeve-Neumann, 2011; tinued direct study of various practice factors in treatment
Maas & Farinella, 2012). for CAS is necessary to determine optimal practice con-
Edeal and Gildersleeve-Neumann (2011) examined the ditions for children with CAS. The present study represents
role of practice amount, arguably the most powerful factor in a continuation along these lines by examining the role of
learning (Schmidt & Lee, 2005), in an integral stimulation feedback frequency in CAS treatment.
treatment for two children with CAS. Integral stimulation
treatment was originally developed for adults with apraxia of
speech (Rosenbek, Lemme, Ahern, Harris, & Wertz, 1973) Feedback Frequency Effects in Motor Learning
and involves auditory and visual models provided by the Models of motor control and learning assume that pro-
clinician (“listen to me, watch me, and say what I say”) and duction of a skilled movement involves exploiting knowl-
principles of motor learning (see Edeal & Gildersleeve- edge of the relations between various sources of information
Neumann, 2011, and Strand & Skinder, 1999, for excellent about a movement, such as motor commands, sensory con-
reviews of the integral stimulation treatment literature for sequences of the movement, and movement outcome (e.g.,
CAS). Edeal and Gildersleeve-Neumann defined practice Bullock, Grossberg, & Guenther, 1993; Schmidt, 1975;
amount in terms of the number of practice trials and used an Willingham, 1998; Wolpert, Ghahramani, & Flanagan,
alternating treatments single-subject design in which differ- 2001). Feedback is considered critical to allow the learner
ent sound targets were assigned to different practice con- to determine these relations and figure out which combina-
ditions (100–150 vs. 30–40 trials per session). They observed tions of motor commands produce the desired outcome in
improvements in speech sound accuracy for both practice a given situation. Intrinsic feedback refers to information
conditions, but retention and transfer were greater for speech about a movement that is inherent in performance of the
sounds that were practiced 100–150 times per session than motor act. For example, in throwing a dart, the visual system
for speech sounds that were practiced 30–40 times per ses- provides information about where the dart landed, and the
sion. These findings are consistent with the motor learning somatosensory system provides information about the grip
literature (e.g., Shea, Kohl, & Indermill, 1990; see Maas force and arm excursion used. In many situations, learners
et al., 2008, for review). can use intrinsic feedback to improve their performance, but
In an integral stimulation treatment with four children there are situations in which intrinsic feedback is not suf-
with CAS, Maas and Farinella (2012) investigated another ficient or optimal. For example, a learner may not have
promising variable, namely, practice schedule. Practice access to all relevant intrinsic information (e.g., in learning a
schedule refers to the order in which different targets are novel dance routine, the learner may not have access to, or be
practiced, and a random practice schedule has been found to able to focus on, information about arm movements when
enhance motor learning compared to a blocked practice focusing on foot movements), or a learner may not have a
schedule (in which different targets are practiced in separate frame of reference that is calibrated to some standard of
blocks). Maas and Farinella compared random and blocked correctness (e.g., the learner may not know how to interpret
practice in an alternating treatments design with multiple the scores on a dart board). In such situations, externally
baselines across behaviors. Treatment was provided in two provided (augmented) feedback from an instructor may be
phases, each lasting 4 weeks (È12 hr of treatment per phase) necessary to improve the skill (Wulf & Shea, 2004).
and separated by a 2-week maintenance phase following Considerable research effort has been devoted to finding
each treatment period (same design as present study). Indi- ways to optimize augmented feedback in the learning of
vidually tailored treatment targets and probe items were motor skills (see Maas et al., 2008; Wulf & Shea, 2004,
selected for each child depending on his or her particular for reviews). One feedback variable known to affect motor
difficulties, and treatment involved the dynamic temporal learning is feedback frequency, which refers to how often
and tactile cuing (DTTC) approach (Strand, Stoeckel, & learners receive augmented feedback on their motor perfor-
Baas, 2006). The findings were mixed: Three of the four mance during practice. A typical finding is that high-frequency
children showed a clear response to treatment (in both feedback (e.g., feedback after every practice trial) enhances
conditions) and retention following treatment in terms of practice performance but degrades motor learning, as com-
overall utterance accuracy, and two of those demonstrated pared to reduced frequency feedback (e.g., feedback only
a slight net advantage for blocked over random practice, after half of all practice trials) (e.g., Bruechert, Lai, & Shea,
whereas the third child showed a clear and consistent advan- 2003; Nicholson & Schmidt, 1991; Winstein & Schmidt,
tage for random over blocked practice. Transfer to untreated 1990).
CAS001 Female 8;4 Moderate-severe None Moderate expressive delay; GFTA–2 SS: <40
low average receptive skills (<1st percentile)
PPVT–4 SS: 97 (42nd percentile) PCC: 53%
CELF–4 Core SS: 76 (5th percentile) PVC: 46%
Concepts & Following Directions = 9/10
Word Structure = 4/10
Recalling Sentences = 4/10
Formulated Sentences =7/10
CAS002 Male 5;4 Severe Very mild unilateral Expressive not testable; receptive GFTA–2 SS: <40
upper motor neuron language within normal limits (<1st percentile)
PPVT–4 SS: 92 (30th percentile) PCC: 42%
PLS–4 SS: 95 (37th percentile) PVC: 11%
CAS005 Female 7;3 Moderate-severe Moderate Severe expressive delay; GFTA–2 SS: 62
spastic-flaccid mild-moderate receptive delay (<1st percentile)
PPVT–4 SS: 79 (8th percentile) PCC: 55%
CELF–4 Core SS: 46 (< 0.1st percentile) PVC: 34%
Concepts & Following Directions = 4/10
Word Structure = 1/10
Recalling Sentences = 1/10
Formulated Sentences = 1/10
CAS012 Male 6;4 Moderate None Severe expressive delay; moderate GFTA–2 SS: 84
to severe receptive delay (12th percentile)
CELF–4 Core SS: 46 (<0.1st percentile) PCC: 59%
Concepts & Following Directions = 4/10 PVC: 62%
Word Structure = 1/10
Recalling Sentences = 1/10
Formulated Sentences = 1/10
Sentence Structure = 3/10
Note. SS = standard score; PCC = percentage of consonants correct; PVC = percentage of vowels correct; PPVT–4 = Peabody Picture
Vocabulary Test, Fourth Edition (Dunn & Dunn, 2007); CELF–4 = Clinical Evaluation of Language Fundamentals, Fourth Edition (Semel, Wiig, &
Secord, 2003); GFTA–2 = Goldman-Fristoe Test of Articulation 2 (Goldman & Fristoe, 2000); PLS–4 = Auditory Comprehension subtest of the
Preschool Language Scale, Fourth Edition (Zimmerman, Steiner, & Pond, 2002).
her PVC was 46% (particular difficulties with diphthongs PVC of 11% (difficulties with most vowels and diphthongs,
and rhotic vowels). A structural–functional oral mechanism except nonrhotic central vowels and /u/). A mild breathy/
exam (Duffy, 2005) revealed no neuromuscular deficits harsh voice quality was noted, and a structural–functional
suggestive of dysarthria. oral mechanism exam (Duffy, 2005) revealed very mild left
CAS001’s expressive language as assessed with the facial asymmetry, suggesting a possible mild unilateral upper
CELF–4 revealed a moderate delay. She demonstrated motor neuron dysarthria. Spontaneous utterances and ver-
difficulty with correct production of various morphological balizations were characterized by adequate respiratory func-
markers (e.g., possessive pronouns, regular and irregular past tion for purposes of speech breathing and by grossly normal
tense). Use of complex sentence structure was noted on loudness, pitch, and rate.
occasion (e.g., The girl who has a soccer ball was third). Her CAS002’s expressive language could not be formally
receptive language was evaluated using the CELF–4 and the assessed due to his markedly severe CAS, but language
PPVT–4 and was determined to be in the low average range. sample analysis revealed evidence of expressive language
CAS002. CAS002 was a non-Hispanic White boy, age delay. There was little evidence for correct use of age-
5;4, with severe CAS and normal receptive language. Ex- appropriate morphological markers (e.g., absence of third
pressive language was not formally tested due to his severely person singular, copula omission). When asked to describe
reduced intelligibility, but informal observation suggested and tell a story about a picture book, CAS002 verbalized
a severe expressive language delay. His CAS was charac- with high frequency and clearly was “telling” the story;
terized by inconsistent errors on vowels and consonants, however, the majority of his utterances were completely
difficulty achieving initial articulatory configurations, trial unintelligible. Occasionally, an intelligible word was noted
and error groping behaviors, inconsistent productions following (e.g., jumping), indicating that he likely had the concept of
repetition of stimulus items, intermittent hypernasality, and certain grammatical markers (i.e., verb + ing). His pragmatic
regular use of a stereotypical nonword (e.g., dunch). His language skills and play schemes were age appropriate.
speech was facilitated by verbal, visual, and tactile cues. Receptive language as assessed using the PPVT–4 and the
Performance on a word repetition list revealed a PCC of Auditory Comprehension subtest of the PLS–4 was deter-
42% (particular difficulty with fricatives and liquids) and a mined to be in the low average range. Difficulties were noted
CAS001 HFF SwS phrases in the chair; make it quick; eat a cake; wSSw phrases without asking; the cute monkey; begin
time to go;comb your hair reading; I play soccer; escape summer
LFF wSS phrases before school; a big dog; avoid heat; SwwS phrases tell me a joke; give it away; finish the
his new bike; the small cup work; under my coat; never be late
Control SSw phrases (n = 5); wSwS phrases (n = 5); 5-syllable phrases (n = 15); 4–5-syllable words (n = 10); 1-syllable words (n = 13)
CAS002 HFF 2-syllable words tiger, balloon, circle Initial clusters broom, smoke, flame
LFF Final clusters mask, point, barn 3-syllable words bicycle, tomato, radio
Control 4-syllable words; fricative CVC words; affricate CVC words; plosive CVC words; liquid CVC words; sCC clusters (n = 10 all)
CAS005 HFF wSS phrases I don’t know; before school; a big dog; SwwS phrases give it away; never be late; thanks for
his new bike; my pink shirt your help; after the storm; tell me a joke
LFF SwS phrases this is fun; time to go; not today; wSSw phrases a hot summer; the cute monkey; without
brush your teeth; take a break asking; before Friday; begin reading
Control SSw phrases (n = 5); wSwS phrases (n = 5); 5-syllable phrases (n = 15); 4–5-syllable words (n = 10); 1-syllable words (n = 13)
CAS012 HFF Sww words bicycle, pyramid, astronaut wSwS words America, photographer, emergency
LFF wSw words flamingo, volcano, coyote SwSw words Arizona, caterpillar, helicopter
Control multisyllabic words (n = 21); multisyllabic phrases (n = 28); initial clusters (CCVC) (n = 10)
Note. HFF = high-frequency feedback, LFF = low-frequency feedback, S = strong, w = weak, C = consonant, V = vowel.
words with final consonant clusters (e.g., point) in the LFF (e.g., Arizona). In addition to three transfer items per set,
condition. Phase 2 targets consisted of three words with we included 59 untreated control items, for a total list of
initial clusters (e.g., broom) in the HFF condition and three 3- 83 items.
syllable words (e.g., bicycle) in the LFF condition. For each
of these four treatment sets, seven untreated words
with similar structures were generated as transfer sets. Procedure
With 60 unrelated control words, the total list consisted Treatment and probes were administered by a trained grad-
of 100 words. uate student clinician under direct supervision of the third
CAS005. Given her difficulties with multisyllabic utter- author, who is a certified and licensed SLP with extensive
ances, targets for CAS005 included phrases with different experience in the treatment of speech disorders in children.
stress patterns. Phase 1 focused on 3-syllable phrases. The Each child was treated by the same clinician throughout
five targets in the HFF condition had a wSS stress pattern the study.
(e.g., my pink shirt), and the five targets in the LFF condition Probe procedures. The data for this study were based on
had an SwS stress pattern (e.g., take a break). In Phase 2, productions on a probe task. The probe task was a repetition
treatment focused on 4-syllable phrases: five targets with task in which the clinician produced each item at a normal
an SwwS stress pattern in the HFF condition (e.g., never rate without any other cues. No feedback was provided
be late) and five targets with a wSSw stress pattern in the during the probe task. Treatment targets, transfer items, and
LFF condition (e.g., a hot summer). Five similar, untreated control items for each child were combined into a single list
phrases were selected for each treatment set (transfer), and were presented in one of four random orders. Probes
and 48 different untreated items served as controls. The were administered three to four times before any treatment
12 treated items from the previous treatment study (Maas started (baseline) and then once per week thereafter. Probes
& Farinella, 2012) were also included (but will not be were always administered at the beginning of a session
reported here), for a list of 100 items. and were audio- and video-recorded for subsequent data
CAS012. Given his noted difficulties with prosody, analysis and reliability scoring.
targets for CAS012 included multisyllabic words. Phase 1 Treatment procedures. We elected to study the effects of
targeted 3-syllable words. The HFF condition targets con- feedback frequency in the context of the DTTC approach
sisted of three words with an Sww stress pattern (e.g., (Strand, Stoeckel, & Baas, 2006) because this treatment
pyramid ), and the LFF condition targets consisted of three has replicated efficacy support (Maas & Farinella, 2012;
words with a wSw stress pattern (e.g., coyote). In Phase 2, Strand & Debertine, 2000; Strand, Stoeckel, & Baas, 2006)
six 4-syllable words were selected as targets: three HFF con- and because it incorporates principles of motor learning,
dition targets with a wSwS stress pattern (e.g., America) allowing us to systematically vary one of these while
and three LFF condition targets with an SwSw stress pattern keeping the treatment otherwise the same. Furthermore, by
Note. S = strong (stressed), w = weak (unstressed), FB = feedback; Tx = treated items. Treatment periods are delineated with
vertical lines.
treatment showed greater accuracy for the LFF condition on untreated items. Inspection of the bottom graph in Figure 1
again. Effect size could not be calculated for the HFF con- reveals that this effect was driven primarily by improvements
dition due to zero variance in both pre- and posttreatment on 3-syllable phrases with SSw stress patterns (ES = 5.93,
data; for the LFF condition, effect size calculation showed 22% gain) and, to a lesser degree, by the 4- and 5-syllable
ES = 2.28. The mean level difference was larger for the LFF phrases (ES = 1.93, 3% gain).
condition (24% gain) than for the HFF condition (10% gain). When combining data across both phases, the LFF con-
Transfer was negligible for the LFF condition (ES = 0.99, dition shows an effect size of ES = 2.40 (30% average gain);
7% gain) and small for the HFF condition (ES = 1.18, 8% no combined HFF condition effect size could be com-
gain). The combined weighted effect size for control items in puted (14% average gain). No transfer is evident for either
Phase 2 indicated ES = 1.69 (5% gain), suggesting improvement condition, and control items show ES = 1.20 (6% gain).
Thus, the LFF condition consistently leads to greater changes (HFF: ES = 0.88, 4% gain; LFF: ES = 0.38, 1% gain), nor
on treated items than the HFF condition, and the effect sizes did control items improve (combined weighted ES = 0.70,
are larger than for the control items. 4% gain; liquids only ES = 1.64; 6% gain).
In Phase 2, a similar pattern emerged: a stable baseline,
followed by an increase in accuracy toward the end of the
CAS002 phase, with a larger increase for the HFF condition and a
For CAS002, a stable baseline was established before decrease in accuracy after treatment ended. Effect size cal-
Phase 1 treatment. As can be seen in Figure 2, an increase culations indicated treatment effects in both conditions, with
in accuracy for treated items emerged toward the end of the a relative advantage for HFF (ES = 1.91, 17% gain) com-
phase, with a larger increase for HFF items than for LFF pared to LFF (ES = 1.11, 6% gain). No transfer was evi-
items. However, these gains were not retained after cessation dent in either condition (0% change); control items also
of treatment, as evinced by the effect sizes (HFF: ES = 0.77, remained unchanged (combined weighted ES = 0.37, 0%
6% gain; LFF: ES = 0.58, 4% gain). No transfer was evident gain; liquids only ES = 1.48; 8% gain).
Taken together, CAS002 showed a small treatment effect items, except for HFF items in Phase 2 (ES = –0.51,
for HFF items (ES = 1.32, 12% gain) but not for LFF items –2% gain; 0% gain for other treated items in both con-
(ES = 0.83, 5% gain), with no transfer (ES not computable ditions and both phases). The increase in performance on
for both conditions; HFF 2% gain; LFF 1% gain) and no LFF transfer items following Phase 2 treatment (ES = 1.08)
change in control items (ES = 0.53, 2% gain; liquids only was minimal (5% gain). Some improvement was noted
ES = 1.56, 7% gain). for untreated control items in Phase 1 (ES = 1.75, 6% gain);
this effect was driven entirely by improvement on sim-
pler one-syllable words (ES = 2.09, 22% gain). In Phase 2,
CAS005 there was no change in combined control items (ES = 0.96,
CAS005 demonstrated a stable baseline for both con- 5% gain), although 3-syllable SSw phrases did improve
ditions in both phases, with no apparent change throughout (ES = 2.15, 11% gain) following treatment on 4-syllable
the study. Effect sizes could not be computed for treated phrases.
Combined across phases, there was no improvement or (ES = 1.04, 24% gain) but not for the LFF condition (ES =
transfer in either condition, and a slight improvement on 0.83, 20% gain). There was no transfer for either condition
control items (ES = 1.36, 6% gain). (HFF: ES = 0.21, 4% gain; LFF: ES = 0.28, 7% gain), and
no change for control items (ES = 0.38, 5% gain).
The baseline for Phase 2 targets was stable, and there was
CAS012 a relatively small response to treatment in the HFF condi-
For CAS012, the Phase 1 baseline was more variable than tion that was not maintained after treatment (ES not com-
for the other participants, but was relatively low and with- putable, 0% gain). For the LFF condition, there was no
out an improving trend. Following initiation of treatment, an change during treatment, but performance following treat-
increase in accuracy on treated items was evident for both ment was consistently above baseline levels, which was
conditions, with a more sustained gain for HFF. Effect sizes corroborated by the effect size calculation (ES = 4.42, 22%
indicated a small treatment effect for the HFF condition gain). Regarding transfer, a small improvement was noted
Appendix A
Treatment protocol
Clinician: “Watch me, listen carefully, and repeat after me,” and models target
Child responds
Clinician waits 2–3 seconds
If child’s response is correct:
: FB trial: Provide FB – Knowledge of Results (KR) and Knowledge of Performance (KP)
: No FB trial: Present next target
If child’s response is incorrect:
: FB trial:
: Provide FB (KR and KP)
: Slow, simultaneous production (up to 2 times); tactile cues as needed
: Fade simultaneous cue (mouth the words)
: Immediate repetition
: Wait 2–3 seconds and provide FB (KR and KP)
: No FB trial:
: “Let’s do it slowly together”
: Slow simultaneous production (up to 2 times); tactile cues as needed
: Fade simultaneous cue (mouth the words)
: Immediate repetition
: Wait 2–3 seconds, then “Now let’s do another one”
As child improves, increase delay between clinician model and child’s response (i.e., delayed repetition instead of
immediate repetition) and eventually elicit targets without clinician model.
Note. FB = feedback.
Appendix C
Treatment fidelity measures
Child Sessions checked Condition Duration (min.) # Trials # Attempts per trial FB delay (sec.) %FB % Adherence