JSLHR

Research Article

Influence of Language Load on Speech

Motor Skill in Children With Specific
Language Impairment
Meredith Saletta,a Lisa Goffman,b
Caitlin Ward,a and Jacob Olesona

Purpose: Children with specific language impairment (SLI) assessed phonetic accuracy, speech movement variability,
show particular deficits in the generation of sequenced and duration.
action: the quintessential procedural task. Practiced imitation Results: Children with SLI produced more variable
of a sequence may become rote and require reduced articulatory movements than peers with typical
procedural memory. This study explored whether speech development in the high load condition. The groups
motor deficits in children with SLI occur generally or only in converged in the low load condition. Children with
conditions of high linguistic load, whether speech motor SLI continued to show increased articulatory stability
deficits diminish with practice, and whether it is beneficial to over 3 practice sessions. Both groups produced
incorporate conditions of high load to understand speech generated sentences with increased duration and
production. variability compared with repeated sentences.
Method: Children with SLI and typical development Conclusions: Linguistic demands influence speech motor
participated in a syntactic priming task during which they production. Children with SLI show reduced speech motor
generated sentences (high linguistic load) and, then, practiced performance in tasks that require language generation but
repeating a sentence (low load) across 3 sessions. We not when task demands are reduced in rote practice.

S
pecific language impairment (SLI) has classically speech motor deficits may be broadly distributed and general
been defined as a disorder that is constrained to or may be specific to particular aspects of performance.
language, particularly to morphosyntax (Leonard, The latter option may involve demands related to sequen-
2014). However, over the last several years, it has become tial movement (Goffman, 1999; Hsu & Bishop, 2014;
evident that children diagnosed with SLI experience Lum, Gelgic, & Conti-Ramsden, 2010; Tomblin, Mainela-
other deficits related to nonlinguistic cognitive processing Arnold, & Zhang, 2007) or cognitive–linguistic processing
(Ellis Weismer, Plante, Jones, & Tomblin, 2005; Miller (Goffman, 2004; Heisler, Goffman, & Younger, 2010;
et al., 2006) and to action (Bishop, 2002; Hill, 2001; Vuolo, Maner, Smith, & Grayson, 2000). Furthermore, although
Goffman, & Zelaznik, 2017). In the speech motor domain, children with SLI do not always have overt speech motor
children with SLI show difficulties with the production of deficits, previous research has shown that they commonly
multiword phrases (Brumbach & Goffman, 2014) and weak– demonstrate more variable articulation across multiple
strong stress patterns (Goffman, 1999, 2004). Motor and productions of a sentence (Brumbach & Goffman, 2014). It
is the objective of the present work to determine whether
a
children with SLI demonstrate a deficit in speech motor
The University of Iowa, Iowa City skill that broadly influences performance or that is con-
b
Callier Center for Communication Disorders, Behavioral and Brain
strained to particular components. For example, higher lin-
Sciences, University of Texas at Dallas
guistic load tasks may be impaired, and lower linguistic load
Note that data were collected while Lisa Goffman was with the
or practiced tasks may be unaffected. In the context of the
Department of Speech, Language, and Hearing Sciences, Purdue
University, West Lafayette, IN.
current study, we define linguistic load on a continuum, with
tasks with increased formulation demands (such as word
Correspondence to Meredith Saletta: [email protected]
and sentence generation) as relatively high load and those
Editor-in-Chief: Sean Redmond
with lower formulation demands (such as word and sen-
Editor: Jan de Jong
tence imitation) as relatively low load.
Received February 16, 2017
Revision received July 5, 2017
Accepted November 16, 2017 Disclosure: The authors have declared that no competing interests existed at the time
https://doi.org/10.1044/2017_JSLHR-L-17-0066 of publication.

Journal of Speech, Language, and Hearing Research • 1–15 • Copyright © 2018 American Speech-Language-Hearing Association 1
Downloaded From: https://jslhr.pubs.asha.org/ by a University College London User on 03/12/2018
Terms of Use: https://pubs.asha.org/ss/rights_and_permissions.aspx
 Analytic Approaches to the Assessment 1999). Quantifying utterance duration can provide insight
of Speech Motor Skill into the influence of processing load on speech produc-
tion but, perhaps, in a different way from those revealed
Articulatory Variability by direct assessment of articulatory movement stability.
Speech motor skill may be measured via articulatory
kinematics. Kinematic measures provide a direct assess-
ment of the movements of the articulatory structures that Influence of Load on Articulatory
are used for speech production. One central measure is Variability and Duration
articulatory variability. The spatiotemporal index (STI) is
one measure that has been used to analyze articulatory In both adults and children, imitation and repetition
variability across repeated utterances (Smith, Goffman, paradigms have been used to assess how relatively high
Zelaznik, Ying, & McGillem, 1995). Younger children levels of language processing may influence the lower level
demonstrate more variable oral speech movements than adults components of speech production. Findings on the basis
(Goffman & Smith, 1999; Green, Moore, Higashikawa, & of imitation paradigms have provided evidence that multi-
Steeve, 2000; Nip & Green, 2013; Smith & Goffman, 1998), ple linguistic levels influence speech production.
and adults who stutter or who have Parkinson’s disease
demonstrate more variable oral speech movements than Articulatory Variability
healthy adults (Kleinow & Smith, 2000; Walsh & Smith, In the production of sentences, Maner, Smith, and
2011). Most relevant to the current study, children with Grayson (2000) showed that young children used increased
SLI demonstrate more variable oral speech movements articulatory variability when imitating utterances that
than children with typical development (TD; Brumbach & were embedded in longer and more complex sentence frames,
Goffman, 2014; Goffman, 1999). However, there is a dual as compared with imitating the same phrase in isolation.
nature of high variability. For example, Green and Nip Kleinow and Smith (2000) asked adults who stutter to
(2010) discuss how young children may demonstrate tran- produce phrases in isolation and then the same phrases
sient episodes of increased variability during times of rapid embedded in sentences with low and high syntactic com-
development or the emergence of new speech skills, pre- plexity (defined as conjoined compared with embedded
sumably because the child is in a period of reorganization clauses that were matched for length). In their study, adults
in pursuit of the best coordinative approach. In these who stutter produced utterances with higher variability
cases, increased variability in development may actually than normally fluent adults and were particularly suscepti-
be a positive or adaptive strategy. ble to increases in syntactic complexity.
In the production of lexical stress, Goffman (1999,
Speech Duration 2004) explored children’s variability and modulation
Analyses of duration have also been used to infer ar- of articulatory movement (i.e., whether weak syllables
ticulatory skills in young children (classically studied by are produced with relatively short and small amplitude
Eguchi & Hirsch, 1969). In the acoustic domain, utter- and/or variable articulatory movements), both in content
ance durations are often used as a proxy for motor skill (Goffman, 1999; Goffman & Malin, 1999) and function
or processing load. By measuring duration, researchers can (Goffman, 2004) words. Articulatory movement measures
determine whether it takes longer for a particular age group revealed production differences as a function of linguistic
or population to produce given nonwords, words, or sen- context (or load), with content words produced differently
tences, with duration differences used to infer processing from function words and iambs produced differently
load. Sevald and Dell (1994) provide a model for the ways from trochees. Surprisingly, iambs were produced with
in which processing and planning affect utterance duration. more articulatory stability than trochees, suggesting that in-
Previous findings have demonstrated that children are creased load may lead to increased rather than decreased
slower than adults (Grigos & Patel, 2007; Smith & Goffman, stability. Finally, when considering lexical–semantic load,
1998; Smith & Zelaznik, 2004; Walsh & Smith, 2002) and Heisler, Goffman, and Younger (2010) and Heisler and
that individuals with speech and/or language disorders are Goffman (2016) demonstrated that adding a lexical repre-
slower than those with typical speech and language (Strand sentation (visual or functional) in a novel word–learning
& McNeil, 1996). Durations may change when task de- task resulted in decreased articulatory variability in children
mands, such as those related to syntactic complexity or with TD and children with SLI.
length, increase (Brumbach & Goffman, 2014; Dromey &
Benson, 2003; Maner, Smith, & Grayson, 2000; Sadagopan Duration
& Smith, 2008). However, duration and articulatory vari- Utterance durations are often used as a proxy for
ability are not necessarily correlated and thus may reveal processing load. In young children, Goffman (2004) found
different underlying production processes (Brumbach & that function words were produced with shorter dura-
Goffman, 2014; Goffman & Smith, 1998; Maner, Smith, & tions than weak syllables of content words. Children also
Grayson, 2000). Most relevant to the current study, chil- use duration to distinguish between declaratives and in-
dren with SLI are slower at speech production, at least in terrogatives (Grigos & Patel, 2007). Adults, but not chil-
some tasks, compared with children with TD (Goffman, dren, reduce durations when producing sentences that are

2 Journal of Speech, Language, and Hearing Research • 1–15

Downloaded From: https://jslhr.pubs.asha.org/ by a University College London User on 03/12/2018

Terms of Use: https://pubs.asha.org/ss/rights_and_permissions.aspx
 embedded in a longer utterance rather than in isolation In summary, in both adults and children, imitation
(Maner, Smith, & Grayson, 2000). Interestingly, in these paradigms have predominantly been used to assess how
cases, durations were shorter when complexity was higher. relatively high levels of language processing may influence
In adults, movement durations are longer for bilin- the lower level components of speech production. Find-
gual talkers who are less proficient in their second language ings on the basis of imitation paradigms have provided
(Chakraborty, Goffman, & Smith, 2008). Strand and evidence that multiple linguistic levels influence speech
McNeil (1996) reported longer segment durations in long production. A few studies have incorporated language
utterances compared with short utterances in speakers with production tasks that require increased processing but have
apraxia. Dromey and Benson (2003) demonstrated that not generally applied these nonimitative tasks to studies
conditions of cognitive distraction also influence partici- of speech motor control. One next step in this line of research
pants’ speaking durations. The common theme across these is to include procedures that involve sentence generation as
studies is that utterance durations may vary in response to opposed to strictly imitation, in order to assess how lan-
conditions of different linguistic load. It is important to guage processing load influences speech movement vari-
note that these data were all obtained in the context of imi- ability and duration.
tation or repetition paradigms.
The Roles of Procedural and Declarative Processing
Nonimitated Tasks in Speech Motor Production
A small group of studies have relied on elicited or In 2005, Ullman and Pierpont proposed the proce-
spontaneous speech and language production to investi- dural deficit hypothesis (PDH). They suggested that the
gate early development. To assess the development of pattern of speech and nonspeech impairments found in
the voicing contrast, Grigos, Saxman, and Gordon (2005) children with SLI can be captured by strengths in declara-
elicited spontaneous productions by training 20-month- tive learning and deficits in procedural learning (Ullman &
old children to say “papa” when viewing a male puppet Pierpont, 2005; Ullman & Pullman, 2015).
and “baba” when viewing a toy sheep. Similarly, Grigos These individuals’ relatively robust abilities in de-
and Patel (2007) examined temporal and spatial changes clarative processing translate into strengths in producing
in prosodic control using contextual scenarios with pup- linguistic forms that are chunked. It is suggested that vo-
pets representing the target words, “Bob,” “Pop,” “bot” cabulary is less affected in children with SLI and that they
(short form of robot), and “pot.” Children had opportuni- also tend to make fewer errors when producing the irregular
ties to tell the experimenter “Show Bob a bot” and “Show past tense because, in English, these verbs rely on rote
Pop a pot” or ask her to “Show Bob a bot?” or “Show memorization, which is more heavily weighted on declar-
Pop a pot?” These studies reveal that nonimitative tasks ative mechanisms (Ullman & Pierpont, 2005). Tasks in-
that require increased load may be used to elicit speech in volving repeated imitations allow for chunking and may
young children. therefore tap into declarative learning.
Nip and Green (2013) explored the speaking rate In contrast, children with SLI demonstrate impair-
and articulatory kinematics of children and young adults ments in producing the hierarchically related sequences
producing speech tasks that varied in demand. These related to procedural learning. This type of processing is
included narrative retell, sentence repetition, and a diado- associated with the learning of new skills involving motor
chokinetic task. Their findings indicate that task demands learning or cognition and the control of established motor
influenced speaking rate (a faster rate for tasks with lower and cognitive skills. Acquiring a skill via implicit learning
language demands) and that speaking rate increased mechanisms invokes the rule-governed aspects of language
gradually between 4 and 13 years of age. Nip and Green’s (phonology, morphology, and syntax; Nicolson & Fawcett,
results are consistent with the hypothesis that older chil- 2007). Tasks involving this type of production include
dren experience a decrease in the time needed to formulate generating implicit sequences (including linguistic sequences)
an utterance and that motor practice may decrease plan- or organizing grammatical structures, which require more
ning time. implicit and procedural weighting rather than repeated,
Finally, Brumbach and Goffman (2014) introduced memorized chunking. Procedural memory is implicit be-
a unique paradigm involving structurally primed sentences. cause the learning that occurs is not available for inten-
Children with SLI or TD listened to a talker describe a tional access and is also gradual in that it occurs over
video using either a preposition or a particle (e.g., “jump multiple trials. Procedural learning deficits are also related
over the box” while watching a girl jump over a box). They to nonspeech motor impairments (gross and fine motor) in
then had to generate a sentence using the same structure individuals with SLI, particularly in those tasks requiring
but different lexical items (e.g., “tip over the block”). Speech hierarchical and sequential demands, active formulation,
movement variability and production duration were mea- and/or complex sequences of movement (Tomblin, Mainela-
sured in this priming task, and children with SLI showed Arnold, & Zhang, 2007).
more articulatory variability than their typical peers. How- Thus, overall, the PDH differentiates rote learning
ever, both groups showed similar utterance durations in from new language generation. The learning, which is pres-
this task. ent in the practiced recitation of a nursery rhyme, involves

Saletta et al.: Language Load and Speech Production in SLI 3

Downloaded From: https://jslhr.pubs.asha.org/ by a University College London User on 03/12/2018
Terms of Use: https://pubs.asha.org/ss/rights_and_permissions.aspx
 relatively declarative elements, in contrast to the learning structural oral motor screening (Robbins & Klee, 1987).
that is present in the reorganization and sequencing of new Children were recruited via advertisements and referrals
language production and that involves relatively procedural from the general community of a small city in the midwest
elements. United States. The children with SLI participated in a 4-week
Our paradigm in the current study aligns with ele- summer program that included both speech/language inter-
ments of the PDH. We set task demands such that the first vention and participation in research protocols, including
task involved relatively difficult reorganization, sequenc- the one that informed the current study.
ing, and higher formulation demands. In this condition, Children participated in standardized cognitive,
children with SLI generated aspects of sentences via scaf- language, speech, and motor testing. All children were
folded priming: a task that aligns with components of required to have normal nonverbal intelligence as mea-
procedural processing. Next, we manipulated those task sured by the Columbia Mental Maturity Scale–Third
demands such that the second task involved easier rote Edition (Burgemeister, Blum, & Lorge, 1972), which was
recitation and lower formulation demands: a task that designed to screen young children’s general nonverbal
aligns with components of declarative processing. Children cognition. Additionally, all children were required to score
practiced sentence imitation over time, presumably in- in the minimal-to-no symptoms of autism spectrum disor-
creasing rote components of the task. Through this manip- der range on the Childhood Autism Rating Scale–Second
ulation, we can assess whether children with SLI show a Edition (Schopler, Reichler, & Renner, 2002; a measure
broad and general speech motor deficit or one that dimin- designed to identify children with autism on the basis
ishes with rote and relatively declarative learning that of direct observation) and per parent report have no his-
occurs with practice and over time. tory of neurological dysfunction (such as epilepsy or head
Our goal in this study is to examine the group dif- injury).
ferences in articulatory variability and utterance duration The children with SLI met exclusionary criteria
between children with SLI and their typical peers as they based on Leonard (2014). Children who met these criteria
transitioned between these two tasks. Although these tasks demonstrated significantly impaired language abilities, as
are not the same as those aspects of processing described indicated by a standard score of 87 or below (Greenslade,
in the PDH, they were constructed to mirror those differ- Plante, & Vance, 2009) on the Structured Photographic
ences. To better understand this phenomenon, we need to Expressive Language Test–Preschool 2, or Third Edition
compare tasks composed of sentence imitation paradigms (Dawson et al., 2003; Dawson et al., 2005): a measure of
(as reported in most of the studies cited above) to parallel children’s production of morphology and syntax in a pic-
tasks composed of sentence generation paradigms. To as- ture description task. This score was inclusionary for the
sess the role of practice, it is important to study produc- children with SLI; all other measures were free to vary.
tion over time. The average Structured Photographic Expressive Language
Test scores are quite low, consistent with a diagnosis of
Hypotheses SLI (Greenslade, Plante, & Vance, 2009).
Receptive vocabulary was measured by the Peabody
The primary prediction of this study is that children Picture Vocabulary Test–Fourth Edition (Dunn & Dunn,
with SLI will show higher degrees of speech motor vari- 2007), and expressive vocabulary was measured by the
ability and longer speech movement durations than their Expressive Vocabulary Test–Second Edition (Williams,
typical peers, with differences even greater when linguistic 2007). Segmental accuracy, including consonants cor-
load is increased. That is, children with SLI will be espe- rectly produced, was measured by the Bankson-Bernthal
cially vulnerable to a condition of increased linguistic load Test of Phonology (Bankson & Bernthal, 1990). The two
in which they generate rather than imitate or repeat a sen- groups differed in their performance on all of the stan-
tence. A second prediction is that imitation practice will dardized language tests. Receptive scores from the Clinical
reduce differences in articulatory variability and duration Evaluation of Language Fundamentals Preschool–Second
in both children with TD and children with SLI (with group Edition (Semel, Wiig, & Secord, 2004) were also available
differences diminishing), suggesting that practice facilitates for the children with SLI.
motor learning and that learning more closely associated In addition, fine and gross motor skills were mea-
with declarative mechanisms is not affected in children with sured by the Movement Assessment Battery for Children–
SLI. Second Edition (Henderson, Sugden, & Barnett, 2007).
This test assesses three areas: manual dexterity, ball skills
Method (aiming and catching), and static and dynamic balance, and
may be used to identify children who demonstrate impair-
Participants ments in motor function. The two groups did not differ in
Two groups of children, matched for age, partici- their performance on the Movement Assessment Battery
pated: 18 children with SLI (10 boys, eight girls), ages 4;0– for Children–Second Edition total score or any of the three
7;1 (years;months), and 18 children with TD (seven boys, subtests. See Table 1 for a summary of these behavioral
11 girls), ages 4;4–6;2. All participants were monolingual measures. Because of family schedules, not every child com-
English speakers and passed a hearing screening and a pleted every measure.

4 Journal of Speech, Language, and Hearing Research • 1–15

Downloaded From: https://jslhr.pubs.asha.org/ by a University College London User on 03/12/2018

Terms of Use: https://pubs.asha.org/ss/rights_and_permissions.aspx
 Table 1. Scores on the measures of standardized testing.

Measure SLI, M (SD) TD, M (SD) Significance

Age 5.58 (0.90) 5.16 (0.47) t(34) = 1.74, p = .09

Father’s SES 5.38 (1.12) 6.28 (0.89) t(29) = 2.47, p = .02
Mother’s SES 4.94 (1.14) 6.11 (1.13) t(33) = 3.04, p = .005
CMMS 106.44 (10.91) 116.78 (9.75) t(32) = 2.91, p = .006
SPELT 78.50 (8.36) 113.22 (10.29) t(34) = 11.11, p < .001
PPVT 100.56 (11.73) 118.44 (10.92) t(34) = 4.74, p < .001
EVT 97.82 (17.61) 116.67 (11.14) t(33) = 3.81, p < .001
BBTOP consonant 70.67 (9.04) 100.94 (8.38) t(34) = 10.42, p < .001
BBTOP word 69.13 (9.03) 100.28 (10.80) t(31) = 8.87, p < .001
MABC total score 9.88 (3.57) 10.67 (2.40) t(33) = .77, p = .45
MABC manual dexterity 9.12 (2.20) 9.94 (2.34) t(33) = 1.07, p = .29
MABC aiming and catching 11.59 (2.62) 9.83 (2.87) t(33) = 1.88, p = .07
MABC balance 10.00 (3.76) 11.94 (2.92) t(33) = 1.71, p = .10
CELF Preschool-2 receptive 91.71 (12.45) Not administered N/A

Note. SLI = specific language impairment; TD = typical development; SES = socioeconomic status; CMMS = Columbia Mental Maturity
Scale; SPELT = Structured Photographic Expressive Language Test; PPVT = Peabody Picture Vocabulary Test; EVT = Expressive Vocabulary
Test; BBTOP = Bankson-Bernthal Test of Phonology; MABC = Movement Assessment Battery for Children; CELF Preschool-2 = Clinical
Evaluation of Language Fundamentals Preschool–Second Edition.

Equipment active formulation, indicating that it is more heavily

weighted on the hierarchical demands associated with
Three-dimensional kinematic data were collected at
procedural learning, whereas the repetition task is more
250 samples/second. A three-camera motion capture sys-
heavily weighted on what are often viewed as declarative
tem (Optotrak 3020 or 3D Investigator, both Northern
mechanisms. The sentences in these tasks were constructed
Digital Inc.) was used. Small infrared light–emitting diodes
using words that are familiar to young children, as indi-
(IREDs) were attached to each participant’s upper lip,
cated by their inclusion in the MacArthur–Bates Commu-
lower lip, and a lightweight splint under the chin at mid-
nicative Development Inventories (Fenson et al., 2007).
line (used to approximate jaw movement). In addition,
An additional constraint was that the sentences needed to
five other IREDs were used to create a three-dimensional
be heavily loaded with labial consonants to facilitate the
head coordinate system in order to subtract head motion
analysis of lip movement.
artifact (Smith, Johnson, McGillem, & Goffman, 2000).
A time-locked acoustic signal was collected at 16,000
samples/second in order to confirm that movement re- Structured Syntactic Priming Task
cords aligned with target productions. High-quality au- Children viewed 32 pairs of photographs, each
dio and video recordings were also obtained for analyses depicting an actor, a verb, and a patient. For the first
of production accuracy. item in each pair, the children listened to a female voice
describing the picture. Their task was then to generate a
sentence that included the same syntactic structure but
Procedure and Session Structure different lexical items to describe the second photograph
Children participated in three experimental sessions in each pair. For example, the primed sentence may be
on three different days. Sessions included the collection of “Mouse washes the baby,” and the child’s target sentence
kinematic data followed by behavioral assessments. The may be “Mom pats the puppy” or “Monkey pushes the
first session contained a highly scaffolded syntactic priming cow.” Two of the target sentences, “Mom pats the puppy”
task that was designed to increase processing load. At the and “Mom pats the baby,” were elicited 10 times over the
end of the first session and then over two sessions that course of the experiment; these two sentences were sub-
followed, children engaged in a sentence repetition task. jected to analysis. An additional eight filler sentences with
See Figure 1 for a representation of the structure of these the same syntactic structure (such as “Mouse washes the
sessions. Of these two tasks, the priming task carried a baby” and “Monkey pushes the cow”) were quasirandomly
higher processing load because it required the child to gen- (no consecutive presentation of the same sentence) distrib-
erate a variety of sentences in response to picture cues. uted throughout the experimental block. These fillers served
In comparison, the repetition task carried a lower process- to increase the processing load. See Table 2 for the lists of
ing load because it required only that the child repeat a target and filler sentences. It is important to note that the
single sentence multiple times rather than generate multi- children heard and produced both the analyzed target
ple different sentences. In this way, the task manipulations sentences and the fillers.
were designed to mirror elements of procedural versus Initially, to verify that the child knew the target verb,
declarative processing. The generation task required more he or she was exposed to the verb “pat” in the children’s book

Saletta et al.: Language Load and Speech Production in SLI 5

Downloaded From: https://jslhr.pubs.asha.org/ by a University College London User on 03/12/2018
Terms of Use: https://pubs.asha.org/ss/rights_and_permissions.aspx
 Figure 1. Session structure.

“Pat the Bunny” (Kunhardt, 2001). Next, to assure famil- sentence “Mom pats the puppy” 10 times consecutively.
iarity with the subjects and objects included in the experi- This is similar to prior tasks used to assess speech motor
mental sentences, photographs of these were presented and variability in children (Smith & Goffman, 1998; Smith &
labeled. Finally, the child completed three practice trials Zelaznik, 2004).
demonstrating his or her understanding of the priming par-
adigm before beginning the experimental task. The sen-
tences and their accompanying illustrations were presented Data Processing
via PowerPoint at a comfortable loudness level. If a child The kinematic data from the target sentences “Mom
produced errors in target sentences (such as substituting pats the puppy” and “Mom pats the baby” were processed
“dog” for “puppy” or producing a syntactic error), the ex- in customized MATLAB routines (The Mathworks, 2009).
aminer would remind the child of the correct production and When processing the kinematic data, some productions
encourage him or her to reattempt the sentence. During the could not be used, including those that contained extra or
first experimental session, the children engaged in the scaf- missing syllables or phonemes, interruptions in the speech
folded priming task. Imitated sentences were not analyzed. signal, extra opening and closing of the lips, or IREDs
missing from the cameras’ view. Incorrect but consistent
productions were analyzed if they outnumbered correct
Sentence Repetition Task productions, that is, if the child said “Mom petted the
At the end of the first and during the second and puppy” eight times and “Mom pats the puppy” twice, the
third sessions, the child was instructed to repeat the target eight consistent productions were used.

Table 2. Target and filler sentences.

Calculating Articulatory Variability
Target sentences Filler sentences The STI for lip aperture was used to measure vari-
ability of articulatory movements across multiple produc-
Mom pats the puppy. Mom pushes the bunny.
tions of a sentence. In this case, the STI quantifies the
Mom pats the baby. Mom washes the plate.
Mouse pushes the puppy. movement of three effectors (upper lip, lower lip, and jaw)
Mouse washes the bird. as they interact to control the opening and closing of the
Mouse pats the flower. mouth during speech (Smith & Zelaznik, 2004). To calcu-
Monkey pats the cat. late the STI, the onsets and offsets of each target sen-
Monkey washes the baby.
Monkey pushes the cow. tence (e.g., the opening movement following the first [m] in
“mom” and the opening movement following the second [p]

6 Journal of Speech, Language, and Hearing Research • 1–15

Downloaded From: https://jslhr.pubs.asha.org/ by a University College London User on 03/12/2018

Terms of Use: https://pubs.asha.org/ss/rights_and_permissions.aspx
 in “puppy”) were extracted from lower lip movement on to note that increased articulatory variability may index
the basis of peak velocity. We then took the lower lip sig- flexibility in learning and thus may be viewed as a positive.
nals and subtracted them point-by-point from the upper Increased articulatory pattern stability also suggests that
lip signals, giving us the lip aperture values. Movement a motor pattern has been acquired (e.g., the acquisition of a
onsets and offsets were selected by visually inspecting the tennis swing). Both components of variability are relevant
velocity record for local minima. An algorithm then estab- for this study: Higher variability may indicate flexibility and
lished the minimum value, determining the point at which learning, and lower variability, reliance on an established
velocity crossed 0 within a 100-ms window of the point se- motor pattern.
lected by the experimenter. We measured the movement
within that entire time window. The movement trajectories
were then linearly time normalized by setting each ex- Calculating Duration
tracted record to a time base of 1,000 points and using a In addition to the STI, we measured the mean dura-
cubic spline algorithm to interpolate between points. They tions of each production. Duration was assessed from the
were amplitude normalized by setting the mean to 0 and the same kinematic records as those used to calculate the STI.
SD to 1. The rationale for time and amplitude normaliza-
tion was to rule out differences in rate and loudness; the
goal was to assess spatiotemporal patterning. After nor- Statistical Analyses
malizing the data, standard deviations were computed at The research questions were answered using linear
2% intervals in relative time across the 10 records and mixed-effects models via the lmer function requiring
then summed. The sum of the 50 SDs is the STI; a higher the lme4, lmerTest, and lsmeans packages in R (Bates,
value reflects greater movement variability (Goffman, 1999; Maechler, Bolker, & Walker, 2015; Kuznetsova, Brockhoff,
Goffman, Gerken, & Lucchesi, 2007; Goffman & Smith, & Christensen, 2016; Lenth, 2016). We examined the
1999; Smith & Goffman, 1998; Smith et al., 1995; Smith outcome variables of speech stability and duration sepa-
& Zelaznik, 2004). A maximum of 10 productions were ana- rately. Fixed effects in the model were group (SLI, TD)
lyzed, but if errors were present, a minimum of five consis- and task (primed/puppy, primed/baby, first repetition, sec-
tent productions were used to calculate the STI. An example ond repetition, and third repetition). A random intercept
of the STI calculation is presented in Figure 2. It is important was included in the model to account for within-subject

Figure 2. Extracted movement sequences from a child with SLI’s productions of the sentence, “Mom pats the puppy.” The left column represents
productions elicited by priming; the right column represents repeated productions. The top two panels represent the raw movement records.
The middle two panels represent the same records, now time normalized and amplitude normalized. The bottom two panels represent the
standard deviations of the normalized records, the sums of which compose the spatiotemporal index (STI) values for lip aperture, or the lip
aperture index (LA index). SLI = specific language impairment.

Saletta et al.: Language Load and Speech Production in SLI 7

Downloaded From: https://jslhr.pubs.asha.org/ by a University College London User on 03/12/2018
Terms of Use: https://pubs.asha.org/ss/rights_and_permissions.aspx
 correlation arising from the within-subject task factor. Analysis 2. Articulatory Variability
The model for duration included different variances by as an Index of Speech Motor Control
task. The covariance structure used in each model was
chosen based on the smallest Akaike information criterion Effects of Group and Task
value. The mixed-effects model allows us to test for overall Tests for the main effects and interactions from the
group effect, overall task effect, the interaction between linear mixed-effects regression model were completed using
group and task, and test for pairwise differences of interest Type III F tests with a Satterthwaite approximation for
between the two primed conditions and the three repeti- degrees of freedom. Model results for STI show that there
tions, while appropriately controlling for repeated observa- was not a significant interaction between group and task,
tions and unequal group variances all within one unifying F(4, 136) = 0.907, p = .46, but there was a significant main
model. A follow-up analysis was performed to find the re- effect of group, F(1, 34) = 6.159, p = .02. This indicates
lationship between STI and duration, after adjusting for that children with SLI produced more variable articulatory
group and task, using a similar mixed-effects model. Our pri- movements than their peers with TD in both a condition
mary tests of interest are the pairwise comparisons between of relatively high formulation demands and a condition
the two primed conditions and the three repetitions. To of relatively low formulation demands. Because the
control overall Type I error resulting from performing mul- priming task has an additional sentence that was not
tiple tests, we use the false discovery rate (FDR; Benjamini given to the children during the repetition sessions, to
& Hochberg, 1995). FDR-adjusted p values control the compare the effect of task, we must look at the pairwise
number of false discoveries (Type I errors) so that .05 im- comparisons of primed/puppy versus the three repeated
plies that 5% of significant tests will result in false positives. puppy sentences. Because there was not a significant
The FDR-adjusted p value is less conservative than a interaction between group and task, we can analyze the
Bonferroni adjustment. The alpha level was set to .05. differences between the tasks by averaging over the two
groups. The effect of task is highly significant in all three
comparisons, p < .0001 and FDR < .0001 (see Table 4).
This is convincing evidence that the higher processing load
Results obligated by the priming condition led to increased articu-
Analysis 1. Errors in Sentence Production latory variability for all children, regardless of group (see
Figure 3).
In the structured syntactic priming task, children in
both groups produced errors that we divided into two
types: word selection errors (e.g., substituting “mommy” Priming Task
for “mom”) and syntactic errors (e.g., substituting “pat” In order to incorporate both primed sentences and
for “pats”). The majority of the word choice errors were to evaluate the influence of SLI on sentence production that
appropriate; however, children were corrected because, requires higher formulation demands, we examined the
for the kinematic analysis, it is necessary that labial conso- pairwise comparisons between group and sentence ( puppy
nants and a consistent number of syllables be included. vs. baby; see Table 4). The effect of group was significant
Syntactic errors were generally true errors, as in omitting only for puppy, p = .01, FDR = .05, meaning that there is
the regular third-person singular morpheme. These data evidence that children with SLI were more variable than
are presented in Table 3. Results indicate that the children children with TD when asked to produce the sentence
with SLI produced more syntactic errors but that both “Mom pats the puppy.” There was no statistically signifi-
groups produced a similar number of word choice errors. cant effect of sentence for either group, p = .42, FDR = .61,
As expected on the basis of their diagnosis, children with and p = .12, FDR = .21, for children with TD and SLI,
SLI demonstrated significant difficulty with accurately respectively.
producing the third-person singular form. The numerous
syntactic errors or nonerrorful substitutions made by both Repetition Task
groups (and by the SLI group in particular) suggest that We were also interested in whether children with
the task was sufficiently challenging and successfully indexed SLI differed from their peers with TD in a task with lower
components of procedural learning. formulation demands that was practiced over time. To do

Table 3. Children’s word choice errors and syntactic errors while completing the priming task.

Group Puppy: nontarget word choice Puppy: syntactic error Baby: nontarget word choice Baby: syntactic error

SLI 47.93% (34.38) 70.48% (36.44) 42.34% (41.56) 73.24% (39.81)

TD 56.95% (37.65) 16.77% (31.44) 49.10% (42.65) 15.85% (32.14)
F(1, 47) = 0.76, F(1, 47) = 28.62, F(1, 47) = 0.31, F(1, 47) = 28.64,
p = .39 p < .001 p = .58 p < .001

Note. SLI = specific language impairment; TD = typical development.

8 Journal of Speech, Language, and Hearing Research • 1–15

Downloaded From: https://jslhr.pubs.asha.org/ by a University College London User on 03/12/2018

Terms of Use: https://pubs.asha.org/ss/rights_and_permissions.aspx
 Table 4. Pairwise comparisons of interest for speech stability.

Contrast Estimate SE df t ratio p value FDR

PrimePuppy – Rep1 8.9278 1.0045 136 8.888 < .0001 < .0001
PrimePuppy – Rep2 10.7400 1.0045 136 10.692 < .0001 < .0001
PrimePuppy – Rep3 11.1486 1.0045 136 11.099 < .0001 < .0001
TD, PrimeBaby – SLI, PrimeBaby −3.1811 1.6822 128 −1.891 0.0609 0.1218
TD, PrimePuppy – SLI, PrimePuppy −4.2722 1.6822 128 −2.540 0.0123 0.0492
TD, PrimeBaby – TD, PrimePuppy −1.1522 1.4205 136 −0.811 0.4187 0.6090
SLI, PrimeBaby – SLI, PrimePuppy −2.2433 1.4205 136 −1.579 0.1166 0.2073
TD, Rep1 – SLI, Rep1 −3.5767 1.6822 128 −2.126 0.0354 0.0944
TD, Rep2 – SLI, Rep2 −1.5133 1.6822 128 −0.900 0.3700 0.5920
TD, Rep3 – SLI, Rep3 −1.1361 1.6822 128 −0.675 0.5007 0.6162
TD, Rep1 – TD, Rep2 0.7806 1.4205 136 0.549 0.5836 0.6669
SLI, Rep1 – SLI, Rep2 2.8439 1.4205 136 2.002 0.0473 0.1081
TD, Rep1 – TD, Rep3 1.0006 1.4205 136 0.704 0.4824 0.6162
SLI, Rep1 – SLI, Rep3 3.4411 1.4205 136 2.422 0.0167 0.0536
TD, Rep2 – TD, Rep3 0.2200 1.4205 136 0.155 0.8772 0.8772
SLI, Rep2 – SLI, Rep3 0.5972 1.4205 136 0.420 0.6748 0.7198

Note. FDR = false discovery rate; SLI = specific language impairment; TD = typical development.

this, we again observed the pairwise comparisons between children with SLI between the first and third sessions,
group and repetition session (see Table 4). There was not although the size is small, p = .02, FDR = .05. This indi-
a significant effect of group for any of the three repetitions, cates that, in conditions of low formulation demands,
p = .04, .37, .50, and FDR = .09, .59, .62, for Sessions 1–3, children with SLI show similar articulatory variability to
respectively. The effect of session was significant only for their peers with TD. Importantly, with imitation practice,

Figure 3. Speech stability by group and task. SLI = specific language impairment; STI = spatiotemporal index; TD = typical development.

Saletta et al.: Language Load and Speech Production in SLI 9

Downloaded From: https://jslhr.pubs.asha.org/ by a University College London User on 03/12/2018
Terms of Use: https://pubs.asha.org/ss/rights_and_permissions.aspx
 children with SLI demonstrate a decrease in articulatory repeating a target sentence can also be analyzed by looking
variability. at the pairwise comparisons between group and repeti-
tion session (see Table 5). There was not a significant
effect of group for any of the three repetitions, p = .87,
Analysis 3. Movement Duration as an Index .82, .45, and FDR = .92, .92, .79, for Sessions 1–3, respec-
of Language Processing tively. There were statistically significant differences in
Effects of Group and Task speech duration between the first and third sessions and
The model results for duration show that there was between the second and third sessions, only for children
not a significant interaction between group and task, with SLI, p = .002, FDR = .01, and p = .01, FDR = .04,
F(4, 136) = 2.15, p = .08, or a significant main effect of respectively. This indicates that, with practice, children
group, F(1, 34) = 0.51, p = .48. Because there was not a sig- with SLI produce sentences with shorter durations.
nificant interaction between group and task, we can analyze
the differences in duration between the tasks by averaging Analysis 4. Relationship Between STI and Duration
over the two groups. The difference between tasks was sig-
The mixed model shows that duration is highly sig-
nificant at all three sessions; p < .0001 and FDR < .0001
nificant in the model for STI, F(1, 166.88) = 29.53, p <
for all three comparisons (see Table 5). This is convincing
.0001. The slope estimate indicates that, after adjusting
evidence that the higher formulation demands obligated
for relationship with group and task, an increase in duration
by the priming condition led to increased sentence duration
of 1-s increases the STI by 6.57, on average.
for all children, regardless of group (see Figure 4).

Priming Task Discussion

To investigate how children with SLI differed from In this study, we examined the impact of manipulat-
their peers with TD in sentence production in a context ing language load on the speech production of two groups
with high formulation demands, we analyzed the pairwise of children. Children with SLI or TD produced the same
comparisons between group and sentence (see Table 5). sentence (“Mom pats the puppy” ) in two tasks. In one task,
There was no statistically significant effect of group for they generated two target sentences 10 times in the context
either sentence, p = .84, FDR = .92, and p = .18, FDR = of highly scaffolded priming (higher formulation demands,
.37, for primed/baby and primed/puppy, respectively. There aligning with elements of procedural processing). In the
was also no statistically significant effect of sentence for second task, they repeated one of the sentences 10 times
either group, p = .74, FDR = .92, and p = .08, FDR = .21, sequentially (lower formulation demands, aligning with
for children with TD and SLI, respectively. This indicates elements of declarative processing) across three practice
that neither children with SLI nor TD had a difference in sessions. The goal was to evaluate whether there is a gener-
speech duration when primed for different sentences. alized and global speech motor deficit in children with
SLI or one that is associated with specific task demands
Repetition Task (Goffman, 1999; Heisler, Goffman, & Younger, 2010;
The differences in children with SLI and their peers Hsu & Bishop, 2014; Lum, Gelgic, & Conti-Ramsden,
with TD in the task with lower formulation demands of 2010; Tomblin, Mainela-Arnold, & Zhang, 2007). We also

Table 5. Pairwise comparisons of interest for speech duration.

Contrast Estimate SE df t ratio p value FDR

PrimePuppy – Rep1 0.4572 0.0690 136 6.629 < .0001 < .0001
PrimePuppy – Rep2 0.4861 0.0733 136 6.636 < .0001 < .0001
PrimePuppy – Rep3 0.5117 0.0725 136 7.061 < .0001 < .0001
TD, PrimeBaby – SLI, PrimeBaby −0.0339 0.1620 34 −0.209 .8356 .9244
TD, PrimePuppy – SLI, PrimePuppy −0.2456 0.1805 34 −1.361 .1826 .3651
TD, PrimeBaby – TD, PrimePuppy 0.0333 0.1003 136 0.332 .7402 .9244
SLI, PrimeBaby – SLI, PrimePuppy −0.1783 0.1003 136 −1.777 .0778 .2073
TD, Rep1 – SLI, Rep1 −0.0167 0.0985 34 −0.169 .8666 .9244
TD, Rep2 – SLI, Rep2 0.0200 0.0881 34 0.227 .8219 .9244
TD, Rep3 – SLI, Rep3 0.0689 0.0894 34 0.771 .4463 .7933
TD, Rep1 – TD, Rep2 0.0106 0.0304 136 0.348 .7286 .9244
SLI, Rep1 – SLI, Rep2 0.0472 0.0304 136 1.555 .1222 .2792
TD, Rep1 – TD, Rep3 0.0117 0.0304 136 0.384 .7015 .9244
SLI, Rep1 – SLI, Rep3 0.0972 0.0304 136 3.201 .0017 .0068
TD, Rep2 – TD, Rep3 0.0011 0.0194 136 0.057 .9545 .9545
SLI, Rep2 – SLI, Rep3 0.0500 0.0194 136 2.575 .0111 .0355

Note. FDR = false discovery rate; SLI = specific language impairment; TD = typical development.

10 Journal of Speech, Language, and Hearing Research • 1–15

Downloaded From: https://jslhr.pubs.asha.org/ by a University College London User on 03/12/2018

Terms of Use: https://pubs.asha.org/ss/rights_and_permissions.aspx
 Figure 4. Speech duration by group and task. SLI = specific language impairment; TD = typical development.

inquired whether practice in imitating the sentence (that the interaction term was not statistically significant because
is, as the task transitions between an emphasis on high planned analyses of interest were the pairwise compari-
formulation demands to a task comprising lower formula- sons, we evaluated all of the pairwise comparisons while
tion demands) would cause the two groups to converge. adjusting the p value with a suitable Type I error adjust-
We address each of our predictions below. ment, FDR. As predicted, the mean variability score was
significantly different between SLI and TD in the priming
condition, but the group means were not significantly
The Speech Motor Deficit in Children With SLI different in the repetition condition. In addition, practice
First, we predicted that children with SLI would facilitated motor learning in children with SLI.
demonstrate higher speech movement variability and lon- Our findings indicate that articulatory variability
ger utterance durations. Our findings indicate that the and utterance duration decreased by Session 3 of practice
SLI group produced more variable speech movements in for children with SLI, indicating motor learning with prac-
both tasks, but that there were no group differences in tice even in children with potential deficiencies in speech
duration. We also found that both groups had more vari- motor learning. The TD group did not show this differ-
able movements and longer utterance durations in the con- ence. This indicates that, as the demands of the experi-
dition of high formulation demands than in the condition mental task diminished, the two groups converged, and
of low formulation demands. This indicated that the re- children with SLI no longer showed differences from their
duced demands associated with imitation facilitated stable peers with TD.
and relatively fast articulatory movements in both groups. These discoveries contribute to characterization of
Second, we predicted that group differences would the motor deficit in children with SLI. Our comparison of
be higher when formulation demands were higher, in that speech production in conditions of high formulation de-
children with SLI would be especially susceptible to this mands versus conditions of motor practice/rote recitation
manipulation. Specifically, both groups would demonstrate revealed that children with SLI do not have a generalized
higher variability in the condition of high formulation speech motor deficit. Actually, the differences between
demands, but with motor practice and lower formulation speech motor production in children with SLI versus chil-
demands, group differences would diminish. Even though dren with TD appeared predominately in the condition

Saletta et al.: Language Load and Speech Production in SLI 11

Downloaded From: https://jslhr.pubs.asha.org/ by a University College London User on 03/12/2018
Terms of Use: https://pubs.asha.org/ss/rights_and_permissions.aspx
 of high linguistic load. Our findings here are novel, as they being weighted toward processing associated with proce-
are the first to show that motor skills are more variable dural learning (organizing and sequencing) to processing
under generation demands versus repetition demands. associated with declarative learning (rote recitation or
Third, we found that speech movement stability memorization). This discovery is consistent with previous
showed group effects, whereas utterance duration did not, studies exploring multiple repetitions of the sentence,
in that there were group differences for stability but not “Buy Bobby a puppy” (Smith & Goffman, 1998; Walsh
duration. It therefore appears that duration captures differ- & Smith, 2002). A production task that is normally associ-
ent components of processing from articulatory variability. ated with procedural learning becomes more associated
These findings suggest that motor learning is not always with declarative learning as the utterance is repeated again
the same as other aspects of processing. Rather, these two and again over time. These results speak to the clinical
speech production processes differ: The first is related to observation that a child with SLI may be able to recite a
movement implementation, and the second is related to nursery rhyme from memory even though he or she is un-
broader dimensions of language processing. Yet, the mixed able to construct a sentence containing those same words. It
model showed that these two measures were correlated, in also may imply that increased articulatory stability may
that more stable STI values corresponded to shorter utter- reflect less generative and flexible learning, in that children
ance durations. In sum, both children with SLI and their with SLI converge to the variability level of their peers in
peers with TD produce longer and more variable move- a task that is rote, rather than languagelike.
ments in a high load retrieval condition. However, only
speech motor stability shows group differences, with chil-
dren with SLI more variable than their peers in the high The Importance of Including Higher Level
retrieval condition. Motor aspects of speech production ap- Tasks in Speech Production Studies
pear to be implicated in SLI, but only as related to the
stability of movement implementation in a relatively high Our third significant discovery in this study relates
load task. to the ways in which children’s language production is
studied, in that there may be substantial differences be-
tween requiring imitation and requiring generation. Most
The Change in Processing Demands previous studies of speech motor control have not tapped
Impacts Speech Production into these types of language demands because the methods
In previous studies, children with SLI were hypothe- used to investigate language in children with SLI were
sized to show relative strengths in declarative learning based on imitation tasks. Researchers have found that
and relative difficulties in procedural learning (Ullman & these imitation paradigms reveal how language influences
Pierpont, 2005; Ullman & Pullman, 2015). This inherent speech production (operationalized by measuring articu-
deficit can be observed via tasks demanding generation of latory stability and duration). However, in the current
sentences with correct English syntax and morphology. study, we have discovered that imitation paradigms do
In the current study, the task that we propose aligned most not fully capture the influences of language load. Both
closely with procedural processing was our priming task, groups of children were highly sensitive to these differences
requiring understanding of English underlying grammar, in task demands, and the higher linguistic load was associ-
online production of a grammatical structure, the creative ated with increased speech motor variability and longer
variable lexical items, and the correct organization and utterance duration. We therefore conclude that increasing
sequencing of elements within the sentence. In contrast, linguistic load via sentence generation demands, rather
individuals with SLI experience relative strengths in the use than relying solely on imitation paradigms, may provide
of declarative processing: the type of learning that can be important and innovative information regarding speech
consciously and explicitly recalled and that involves remem- production. Furthermore, the inferences that can be made
bering idiosyncratic representations and linguistic map- based on imitation are limited. Requiring children to gener-
pings memorized in the “mental lexicon.” As a compensatory ate an utterance leads to substantial increases in spatio-
strategy, individuals with SLI may memorize complex se- temporal variability. The inclusion of increased sentence
quences and forms rather than attempt to generate them generation influences speech production and may be impor-
anew during each attempt. When a word or concept be- tant to incorporate into the study of SLI. Indeed, articula-
comes learned in a way that is rote, as in the memorization tory variability likely indexes increased flexibility in the
of verbs with irregular past tense marking, the word or generation of an utterance. Clearly, manipulation of lin-
concept switches from procedural awareness to declarative guistic and motoric difficulty can make an important dif-
awareness. In the current study, the task that we propose ference in the results of a given study.
aligned most closely to declarative learning was our repeti-
tion task.
Therefore, our second major finding in this study Clinical Implications
was that the change in processing demands impacts speech There are two significant clinical implications for
production in children with SLI. Following our priming our findings: It is crucial to incorporate both higher formu-
task with our repetition task changed the demands from lation demands and motor practice in therapy designs.

12 Journal of Speech, Language, and Hearing Research • 1–15

Downloaded From: https://jslhr.pubs.asha.org/ by a University College London User on 03/12/2018

Terms of Use: https://pubs.asha.org/ss/rights_and_permissions.aspx
 First, it may be important for clinicians to include Limitations and Future Directions
increased linguistic load in their therapy because that is ac-
One consideration is how the findings of increased
tually where the child’s difficulty may lie. This is in con-
articulatory variability should be interpreted. Tradition-
trast to traditional therapy, which usually begins with the
ally, gross-motor or fine-motor variability is seen to be
child imitating the production of the clinician. Speech-
indicative of an adaptive process that actually facilitates
language pathologists (SLPs) classically utilize a hierarchy
learning. For example, in infants’ reaching trajectories,
of difficulty that begins with imitation. That is, children
early variability in the body’s posture, reach positions
with language impairments are instructed to imitate utter-
and speeds, and the freedom of the other arm actually sup-
ances produced by the SLP or instructor as the first com-
port learning (Thelen & Spencer, 1998). However, this
ponent of therapy. They then go on to use procedures,
concept has not been addressed in studies of speech motor
such as script fading, in which the SLP gradually encour-
processes, even though it is generally assumed that greater
ages the child to produce sentences with an increasing num-
stability is more positive than greater variability (Saletta,
ber of empty spaces (i.e., beginning with “I like to play
2015). Future work should address this question directly.
video ___,” then proceeding to “I like to play ___” and then
Additionally, this study is limited by the deficiency
“I like ___,” and so forth, until the client can produce the
of external validity. That is, repeated motor practice is not
script independently; Paul & Norbury, 2012). However,
an element of natural language learning (except in, e.g.,
it may be the case that scaffolded retrieval tasks are more
the use of music and poetry), and sentences are not re-
effective at tapping into the deficit, which is specific to chil-
peated verbatim over and over in the course of typical de-
dren with SLI. This concept is consistent with the works
velopment. Future works should increase the clinical
of Karpicke and Blunt (2011) and Karpicke and Roediger
feasibility of this paradigm by incorporating activities that
(2008) who have explored the role of retrieval in learning.
more closely mirror natural development and learning.
Similarly, our findings suggest that eliciting productions via
Another important next step in this line of work is to
conditions of higher linguistic load may actually facilitate
conduct an additional priming/high linguistic load task af-
deeper learning, and the increased articulatory variability
ter the three repetition/low linguistic load tasks. This will
observed reflects this learning.
speak to the issue of whether the priming exercise changes
This idea has been confirmed by studies indicating
as a result of previous exposure to the repetition exercise.
that increased complexity in teaching facilitates learning
Also, further work is necessary to ascertain how imita-
and generalization. For example, Guo, Van Horne, and
tion paradigms tap into language processing. Finally, fu-
Tomblin (2011) discuss how sentence complexity interacts
ture research in child language intervention should
with children’s knowledge of grammar. Returning to the
continue to explore the role of sentence generation/produc-
current study, requiring the child to either generate or
tion tasks rather than relying exclusively on activities in-
repeat an utterance may have a large impact on his or her
volving direct imitation or repetition.
performance and, most importantly, the success of his or
her learning. It is therefore crucial to focus more carefully
on the demands of the task that the child is asked to Conclusions
perform and to make sure that this variable is controlled In conclusion, our data indicate that, when investi-
across the therapy program. gating the connection between speech and language, it is
The incorporation of motor practice (generally essential to include tasks that tap into high levels of lin-
speaking, consisting of a variety of motor exercises) is also guistic processing. All children, with TD or SLI, show high
a crucial element of speech and language therapy. Specifi- levels of movement variability and long durations in sen-
cally, motor-based treatment approaches elicit target tence production tasks of relatively high load. However,
speech behaviors from the client by teaching the correct group differences only emerge in a high load condition that
production of an individual sound; in this process, repeti- presumably requires active retrieval. It is in this task that
tion is crucial (Bernthal, Bankson, & Flipsen, 2013). Typi- children with SLI show a distinct pattern of speech motor
cal speech sound production involves both accurate and learning, as they demonstrate great variability (but not ut-
efficient sound production at the motor level and the use of terance duration) compared with their peers with TD. In
speech sounds in accordance with language rules (Bernthal the articulatory kinematic domain, in which speech motor
et al., 2013). Similarly, grammatical intervention often relies deficits are explicitly assessed, children with SLI become
upon repetitions and recasts or imitations of a given struc- closer to the performance of their peers with TD in imita-
ture (Fey, Long, & Finestack, 2003). Clearly, therefore, tion conditions of lower linguistic load. These data extend
motor practice coupled with language training should be upon the theoretical account of the PDH and the nature of
an important component of speech and language interven- the (speech) motor deficit in SLI and speak to important
tion. Despite the fact that children with SLI do not always methodological considerations for future studies.
have speech motor, gross motor, or fine motor deficits,
our results indicate that motor practice still has relevance in
a clinical setting because it may impact language process- Acknowledgments
ing. This is a novel conclusion that we can draw from our This project was supported by the National Institute on
data. Deafness and Other Communication Disorders Grant R01

Saletta et al.: Language Load and Speech Production in SLI 13

Downloaded From: https://jslhr.pubs.asha.org/ by a University College London User on 03/12/2018
Terms of Use: https://pubs.asha.org/ss/rights_and_permissions.aspx
 DC04826 (awarded to Lisa Goffman). The authors would like to deficits. Journal of Speech, Language, and Hearing Research,
thank Mitch Barna, Sara Benham, Janna Berlin, Barbara Brown, 42, 1499–1516. https://doi.org/10.1044/jslhr.4206.1499
Lakin Brown, Allison Gladfelter, Amanda Steeb, and Janet Vuolo Goffman, L. (2004). Kinematic differentiation of prosodic catego-
for their help with multiple components of this project. ries in normal and disordered language development. Journal
of Speech, Language, and Hearing Research, 47, 1088–1102.
https://doi.org/1092-4388/04/4705-1088
References Goffman, L., Gerken, L., & Lucchesi, J. (2007). Relations between
Bankson, N. W., & Bernthal, J. E. (1990). Bankson-Bernthal Test segmental and motor variability in prosodically complex non-
of Phonology. Austin, TX: Pro-Ed. word sequences. Journal of Speech, Language, and Hearing Re-
Bates, D., Maechler, M., Bolker, B., & Walker, S. (2015). Fit- search, 50(2), 444–458.
ting linear mixed-effects models using lme4. Journal of Goffman, L., & Malin, C. (1999). Metrical effects on speech move-
Statistical Software, 67(1), 1–48. https://doi.org/10.18637/jss. ments in children and adults. Journal of Speech, Language,
v067.i01 and Hearing Research, 42, 1003–1015.
Benjamini, Y., & Hochberg, Y. (1995). Controlling the false Goffman, L., & Smith, A. (1998). Stability and patterning of
discovery rate: A practical and powerful approach to multiple speech movement sequences in children and adults. Journal
testing. Journal of the Royal Statistical Society, Series B of Speech, Language, and Hearing Research, 41, 18–30. https://
(Methodological), 289–300. doi.org/ 1092-4388/98/4101-0018
Bernthal, J. E., Bankson, N., & Flipsen, P. (2013). Articulation Goffman, L., & Smith, A. (1999). Development and phonetic dif-
and phonological disorders: Speech sound disorders in children ferentiation of speech movement patterns. Journal of Experi-
(7th ed.). Upper Saddle River, NJ: Pearson. mental Psychology: Human Perception and Performance, 25(3),
Bishop, D. V. M. (2002). Motor immaturity and specific speech 649–660.
and language impairment: Evidence for a common genetic basis. Green, J. R., Moore, C. A., Higashikawa, M., & Steeve, R. W. (2000).
American Journal of Medical Genetics, 114, 56–63. The physiologic development of speech motor control: Lip and
Brumbach, A. C. D., & Goffman, L. (2014). Interaction of lan- jaw coordination. Journal of Speech, Language, and Hearing
guage processing and motor skill in children with specific lan- Research, 43, 239–255. https://doi.org/1092-4388/00/4301-0239
guage impairment. Journal of Speech, Language, and Hearing Green, J. R., & Nip, I. S. B. (2010). Some organization principles
Research, 57(1), 158–171. https://doi.org/10.1044/1092-4388 in early speech development. In B. Maassen & P. V. Lieshout
(2013/12-0215) (Eds.), Speech motor control: New developments in basic and
Burgemeister, B., Blum, L., & Lorge, I. (1972). Columbia Mental applied research (pp. 171–187). Oxford, United Kingdom:
Maturity Scale–Third Edition. New York, NY: Harcourt Brace Oxford University Press.
Jovanovich. Greenslade, K. J., Plante, E., & Vance, R. (2009). The diagnostic
Chakraborty, R., Goffman, L., & Smith, A. (2008). Physiological accuracy and construct validity of the Structured Photographic
indices of bilingualism: Oral-motor coordination and speech Expressive Language Test, Preschool, Second Edition. Lan-
rate in Bengali–English speakers. Journal of Speech, Language, guage, Speech, and Hearing Services in Schools, 40(2), 150–160.
and Hearing Research, 51(2), 321–332. https://doi.org/10.1044/ https://doi.org/10.1044/0161-1461(2008/07-0049)
1092-4388(2008/024) Grigos, M. I., & Patel, R. (2007). Articulator movement associ-
Dawson, J. I., Stout, C., Eyer, J. A., Tattersall, P., Fonkalsrud, J., ated with the development of prosodic control in children. Jour-
& Croley, K. (2003). Structured Photographic Expressive Lan- nal of Speech, Language, and Hearing Research, 50, 119–130.
guage Test–Third Edition. DeKalb, IL: Janelle Publications. https://doi.org/1092-4388/07/5001-0119
Dawson, J. I., Stout, C., Eyer, J. A., Tattersall, P., Fonkalsrud, J., Grigos, M. I., Saxman, J. H., & Gordon, A. M. (2005). Speech
& Croley, K. (2005). Structured Photographic Expressive Lan- motor development during acquisition of the voicing con-
guage Test–Preschool 2. DeKalb, IL: Janelle Publications. trast. Journal of Speech, Language, and Hearing Research, 48,
Dromey, C., & Benson, A. (2003). Effects of concurrent motor, 739–752. https://doi.org/1092-4388/05/4804-0739
linguistic, or cognitive tasks on speech motor performance. Guo, L. Y., Van Horne, A. J. O., & Tomblin, J. B. (2011). The
Journal of Speech, Language, and Hearing Research, 46, role of developmental levels in examining the effect of subject
1234–1246. https://doi.org/1092-4388/03/4605-1234 types on the production of auxiliary is in young English-
Dunn, L. M., & Dunn, D. M. (2007). Peabody Picture Vocabulary speaking children. Journal of Speech, Language, and Hearing
Test–Fourth Edition. San Antonio, TX: Pearson. Research, 54, 1658–1666. https://doi.org/10.1044/1092-4388
Eguchi, S., & Hirsch, I. J. (1969). Development of speech sounds (2011/10-0140)
in children. Uppsala, Sweden: Almqvist & Wiksells. Heisler, L., & Goffman, L. (2016). The influence of phonotactic
Ellis Weismer, S., Plante, E., Jones, M., & Tomblin, J. B. (2005). probability and neighborhood density on children’s production
A functional magnetic resonance imaging investigation of of newly learned words. Language Learning and Development,
verbal working memory in adolescents with specific language 21(3), 338–356. https://doi.org/10.1080/15475441.2015.1117977
impairment. Journal of Speech, Language, and Hearing Re- Heisler, L., Goffman, L., & Younger, B. (2010). Lexical and artic-
search, 48, 405–425. https://doi.org/1092-4388/05/4802-0405 ulatory interactions in children’s language production. Devel-
Fenson, L., Marchman, V. A., Thal, D. J., Dale, P. S., Reznick, opmental Science, 13(5), 722–730. https://doi.org/10.1111/
J. S., & Bates, E. (2007). MacArthur–Bates Communicative j.1467-7687.2009.00930.x
Development Inventories. Baltimore, MD: Brookes. Henderson, S. E., Sugden, D. A., & Barnett, A. (2007). Movement
Fey, M. E., Long, S. H., & Finestack, L. H. (2003). Ten principles Assessment Battery for Children–Second Edition. San Antonio,
of grammar facilitation for children with specific language TX: Pearson.
impairments. American Journal of Speech-Language Pathology, Hill, E. L. (2001). Non-specific nature of specific language impair-
12, 3–15. ment: A review of the literature with regard to concomitant
Goffman, L. (1999). Prosodic influences on speech production motor impairments. International Journal of Language and
in children with specific language impairment and speech Communication Disorders, 36(2), 149–171.

14 Journal of Speech, Language, and Hearing Research • 1–15

Downloaded From: https://jslhr.pubs.asha.org/ by a University College London User on 03/12/2018

Terms of Use: https://pubs.asha.org/ss/rights_and_permissions.aspx
 Hsu, H. J., & Bishop, D. V. M. (2014). Sequence-specific procedural Saletta, M. (2015). Literacy transforms speech production. Fron-
learning deficits in children with specific language impairment. tiers in Psychology, 6, 1458. https://doi.org/10.3389/fpsyg.2015.
Developmental Science, 17(3), 352–365. 01458
Karpicke, J. D., & Blunt, J. R. (2011). Retrieval practice produces Schopler, E., Reichler, R. J., & Renner, B. R. (2002). The Childhood
more learning than elaborative studying with concept map- Autism Rating Scale (CARS). Los Angeles, CA: Western
ping. Science, 331(6018), 772–775. https://doi.org/10.1126/ Psychological Services.
science.1199327 Semel, E. S., Wiig, E. H., & Secord, W. A. (2004). Clinical Evalu-
Karpicke, J. D., & Roediger, H. L. (2008). The critical importance ation of Language Fundamentals Preschool–Second Edition
of retrieval for learning. Science, 319(5865), 966–968. https:// (CELF Preschool-2). San Antonio, TX: Pearson.
doi.org/10.1126/science.1152408 Sevald, C. A., & Dell, G. S. (1994). The sequential cueing effect
Kleinow, J., & Smith, A. (2000). Influences of length and syntactic in speech production. Cognition, 52, 91–127.
complexity on the speech motor stability of the fluent speech Smith, A., & Goffman, L. (1998). Stability and patterning of
of adults who stutter. Journal of Speech, Language, and Hear- speech movement sequences in children and adults. Journal of
ing Research, 43, 548–559. Speech, Language, and Hearing Research, 41, 18–30. https://
Kunhardt, D. (2001). Pat the bunny. New York, NY: Random doi.org/1092-4388/98/4101-0018
House Children’s Books. Smith, A., Goffman, L., Zelaznik, H. N., Ying, G., & McGillem, C.
Kuznetsova, A., Brockhoff, P. B., & Christensen, R. H. B. (2016). (1995). Spatiotemporal stability and patterning of speech move-
lmerTest: Tests in linear mixed effects models. R package ver- ment sequences. Experimental Bain Research, 104, 493–501.
sion 2.0-33. https://CRAN.R-project.org/package=lmerTest Smith, A., Johnson, M., McGillem, C., & Goffman, L. (2000).
Lenth, R. V. (2016). Least-squares means: The R package lsmeans. On the assessment of stability and patterning of speech move-
Journal of Statistical Software, 69(1), 1–33. https://doi.org/ ments. Journal of Speech, Language, and Hearing Research,
10.18637/jss.v069.i01 43, 277–286. https://doi.org/1092-4388/00/4301-0277
Leonard, L. B. (2014). Children with specific language impairment. Smith, A., & Zelaznik, H. N. (2004). Development of functional
Cambridge, MA: MIT Press. synergies for speech motor coordination in childhood and ado-
Lum, J. A. G., Gelgic, C., & Conti-Ramsden, G. (2010). Proce- lescence. Developmental Psychobiology, 45(1), 22–33.
dural and declarative memory in children with and without Strand, E. A., & McNeil, M. R. (1996). Effects of length and
specific language impairment. International Journal of Commu- linguistic complexity on temporal acoustic measures in
nication Disorders, 45(1), 96–107. apraxia of speech. Journal of Speech and Hearing Research,
Maner, K. J., Smith, A., & Grayson, L. (2000). Influences of utter- 39, 1018–1033.
ance length and complexity on speech motor performance in Thelen, E., & Spencer, J. P. (1998). Postural control during reach-
children and adults. Journal of Speech, Language, and Hearing ing in young infants: A dynamic systems approach. Neuro-
Research, 43, 560–573. https://doi.org/1092-4388/00/4302-0560 science and Biobehavioral Reviews, 22, 507–514.
Mathworks, Inc. (2009). MATLAB: High performance numeric Tomblin, J. B., Mainela-Arnold, E., & Zhang, X. (2007). Proce-
computation and visualization software [computer program]. dural learning in adolescents with and without specific lan-
Natick, MA: Author. guage impairment. Language Learning and Development, 3,
Miller, C. A., Leonard, L. B., Vail, R. K., Zhang, X., Tomblin, 269–293. https://doi.org/10.1080/15475440701377477
J. B., & Francis, D. J. (2006). Response time in 14-year-olds Ullman, M. T., & Pierpont, E. I. (2005). Specific language im-
with language impairment. Journal of Speech, Language, and pairment is not specific to language: The procedural deficit
Hearing Research, 49, 712–728. https://doi.org/1092-4388/06/ hypothesis. Cortex, 41, 399–433.
4904-0712 Ullman, M. T., & Pullman, M. Y. (2015). A compensatory role
Nicolson, R. I., & Fawcett, A. J. (2007). Procedural learning dif- for declarative memory in neurodevelopmental disorders. Neu-
ficulties: Reuniting the developmental disorders? TRENDS roscience and Biobehavioral Reviews, 51, 205–222. https://doi.
in Neuroscience, 30(4), 135–141. https://doi.org/10.1016/j.tins. org/10.1016/j.neubiorev.2015.01.008
2007.02.003 Vuolo, J., Goffman, L., & Zelaznik, H. N. (2017). Deficits in coor-
Nip, I. S. B., & Green, J. R. (2013). Increases in cognitive and dinative bimanual timing control in children with specific
linguistic processing primarily account for increases in speak- language impairment. Journal of Speech, Language, and Hear-
ing rate with age. Child Development, 84(4), 1324–1337. https:// ing Research, 60, 393–405. https://doi.org/10.1044/2016_
doi.org/10.1111/cdev.12052 JSLHR-L-15-0100
Paul, R., & Norbury, C. (2012). Language disorders from infancy Walsh, B., & Smith, A. (2002). Articulatory movements in adoles-
through adolescence: Listening, speaking, reading, writing, and cents: Evidence for a protracted development of speech motor
communicating (4th ed.). St. Louis, MO: Elsevier. control processes. Journal of Speech, Language, and Hearing Re-
Robbins, J., & Klee, T. (1987). Clinical assessment of oropharyn- search, 45, 1119–1133. https://doi.org/1092-4388/02/4506-1119
geal motor development in young children. Journal of Speech Walsh, B., & Smith, A. (2011). Linguistic complexity, speech pro-
and Hearing Disorders, 52, 271–277. duction, and comprehension in Parkinson’s disease: Behavioral
Sadagopan, N., & Smith, A. (2008). Developmental changes and physiological indices. Journal of Speech, Language, and
in the effects of utterance length and complexity on speech Hearing Research, 54, 787–802. https://doi.org/10.1044/1092-
movement variability. Journal of Speech, Language, and 4388(2010/09-0085)
Hearing Research, 51, 1138–1151. https://doi.org/1092-4388/08/ Williams, K. T. (2007). Expressive Vocabulary Test–Second Edi-
5105-1138 tion. San Antonio, TX: Pearson Assessments.

Saletta et al.: Language Load and Speech Production in SLI 15

Downloaded From: https://jslhr.pubs.asha.org/ by a University College London User on 03/12/2018
Terms of Use: https://pubs.asha.org/ss/rights_and_permissions.aspx