Content uploaded by Toby Macrae
Author content
All content in this area was uploaded by Toby Macrae on May 31, 2017
Content may be subject to copyright.
AJSLP
Research Article
Lexical and Phonological Variability
in Preschool Children With
Speech Sound Disorder
Toby Macrae,
a
Ann A. Tyler,
b
and Kerry E. Lewis
c
Purpose: The authors of this study examined relationships
between measures of word and speech error variability and
between these and other speech and language measures
in preschool children with speech sound disorder (SSD).
Method: In this correlational study, 18 preschool children
with SSD, age-appropriate receptive vocabulary, and normal
oral motor functioning and hearing were assessed across
2 sessions. Experimental measures included word and
speech error variability, receptive vocabulary, nonword
repetition (NWR), and expressive language. Pearson
product–moment correlation coefficients were calculated
among the experimental measures.
Results: The correlation between word and speech error
variability was slight and nonsignificant. The correlation
between word variability and receptive vocabulary was
moderate and negative, although nonsignificant. High word
variability was associated with small receptive vocabularies.
The correlations between speech error variability and NWR
and between speech error variability and the mean length of
children’s utterances were moderate and negative, although
both were nonsignificant. High speech error variability was
associated with poor NWR and language scores.
Conclusion: High word variability may reflect unstable lexical
representations, whereas high speech error variability may
reflect indistinct phonological representations. Preschool
children with SSD who show abnormally high levels of
different types of speech variability may require slightly
different approaches to intervention.
Key Words: speech sound disorder, speech variability,
word variability, error variability, inconsistency
Historically, speech variability has received much
less attention than speech accuracy in studies of
typical and disordered speech development in
young children. Studying speech variability in addition to
accuracy is clinically important for several reasons. First,
although a relationship clearly exists between speech sound
accuracy and speech intelligibility, the correlation is modest
(r= .42; Shriberg & Kwiatkowski, 1982). Shriberg and
Kwiatkowski (1982) concluded that “speech intelligibility
reflects a complex of factors in addition to articulation pro-
ficiency”(p. 264). In adult speakers with dysarthria, increased
variability in the production of vowels and syllables is as-
sociated with decreased speech intelligibility and increased
severity of impairment (Kim, Hasegawa-Johnson, & Perlman,
2010; Ziegler, Hartmann, & Hoole, 1993). Young children
with speech sound disorders (SSD) become more intelligible
as listeners become familiar with their speech error patterns
(Shriberg & Kwiatkowski, 1982). Children with variable
speech are likely to be less intelligible than children with con-
sistent speech because of the unpredictability in their speech
productions (Holm, Crosbie, & Dodd, 2005).
Speech variability is also clinically important because
word variability (one type of speech variability) is a core feature
of certain subtypes of SSD in children—namely, childhood
apraxia of speech (CAS; American Speech-Language-Hearing
Association [ASHA], 2007) and inconsistent disorder (Dodd,
2005). Word variability refers to variability in repeated pro-
ductions of the same word. For example, a child may produce
the target word cat as [kæt], [dæt], and [dæ] on three different
occasions in conversation during a play session. Children with
these subtypes of SSD may require unique approaches to
intervention. For example, children with CAS may benefit
from an approach based on motor learning theory, such as the
Nuffield Centre Dyspraxia Programme (3rd ed.; see Williams
& Stephens, 2010). Children with inconsistent disorder may
benefit from an approach that focuses on consistent produc-
tion of a core set of vocabulary items (e.g., core vocabulary
intervention; Dodd, Holm, Crosbie, & McIntosh, 2010).
a
The Florida State University, Tallahassee
b
Western Michigan University, Kalamazoo
c
University of Nevada, Reno
Correspondence to Toby Macrae: toby.macrae@cci.fsu.edu
Editor: Carol Scheffner Hammer
Associate Editor: Lynn Williams
Received April 11, 2012
Revision received August 6, 2012
Accepted May 5, 2013
DOI: 10.1044/1058-0360(2013/12-0037)
Disclosure: The authors have declared that no competing interests existed at the
time of publication.
American Journal of Speech-Language Pathology •Vol. 23 •27–35 •February 2014 •AAmerican Speech-Language-Hearing Association 27
Downloaded From: http://ajslp.pubs.asha.org/ by Florida State University, Toby Macrae on 03/03/2014
The third reason for the clinical importance of speech
variability is that it may relate to prognosis in treatment
for children with SSD. Forrest and colleagues (Forrest,
Dinnsen, & Elbert, 1997; Forrest, Elbert, & Dinnsen, 2000)
studied the effect of speech error variability (another type of
speech variability) on speech sound learning and general-
ization in children with SSD. Speech error variability refers to
variability in speech sounds that are produced in place of
misarticulated targets. Targets with a consistent speech error
are those that are misarticulated and for which only one
speech sound is produced in their place. For example, a child
may produce the target sound / ^/ as [ts] in chair ([tsɛr]),
church ([tsɝts]), and patch ([pæts]). Targets with variable
speech errors are those that are misarticulated and for which
several speech sounds are produced in their place. For
example, a child may produce the target /s/ as [t] in saw ([t]),
as [d] in sun ([dÃn]), and as an omission in bus ([bÃ]). Children
receiving treatment for a target produced with a consistent
error showed greater learning and generalization of the
target than children receiving treatment for a target pro-
duced with variable errors (Forrest et al., 1997, 2000).
Studying speech variability is also theoretically im-
portant. Dynamic systems theory is a theory of action that
was formulated to account for real-time changes in motor
behavior in infants (Thelen & Bates, 2003). This theory was
extended to account for longer term changes in motor de-
velopment. A dynamic systems view of development “con-
siders the origins and functions of variability as absolutely
central for understanding change”(Thelen & Smith, 1994,
p. 145). Variability is associated with transitions between
developmental stages and is viewed as a “potential driving
force of development and a potential indicator of ongoing
processes”(van Geert & van Dijk, 2002, p. 341). When lying
on their backs, for example, newborn infants perform highly
coordinated alternating leg kicks. At about 1 month of age,
coordination between the legs becomes highly variable. This
variability leads to new forms of coordination between the legs,
for example, simultaneous kicking (Thelen & Smith, 1994).
Thelen and Smith (1994) suggested that infants must
free themselves of the stable patterns of the newborn period
before they can assemble new patterns of coordination.
The variability present during developmental transitions pro-
vides infants with a wide array of coordinative possibilities.
In other words, variability in motor development allows
infants to explore new patterns of motor behavior. Variability
is characteristic of the development of other biological and
psychological systems, including speech-language development
(van Geert & van Dijk, 2002). Dynamic systems theory has
potential in explaining the variability that is seen in speech-
language development. Peaks in variability in speech produc-
tion may presage developmental change, just as variability has
been shown to precede real-time changes in motor behaviors
and longer term changes in other areas of development.
Word Variability
Sosa and Stoel-Gammon (2006) studied whole-word
variability in four children between the ages of 1 and 2 years.
They used Ingram’s (2002) measure of variability, the pro-
portion of whole-word variation, which is calculated by
dividing the number of different phonetic forms of a word by
the total number of productions of that word in a speech
sample. The authors found that variability fluctuated through-
out the duration of the study and peaked when children
had acquired È150–200 words and when two-word combi-
nations were first observed. Sosa and Stoel-Gammon sug-
gested that this peak in variability reflected a reorganization
of the linguistic system that included a transition from
underlying holistic to phonemic representations as well as the
analysis and combination of individual words. Underlying
representations are “part of the mental lexicon that stores
the information needed to recognize and produce words”
(Stoel-Gammon, 2011, p. 17).
Sosa and Stoel-Gammon’s (2006) interpretation is con-
sistent with Metsala and Walley’s (1998) lexical restructur-
ing model (LRM). According to this model, infants’early
lexical representations are holistic in nature as there is simply
no need to represent words in a more detailed manner. As
children’s vocabularies grow, however, the increasing simi-
larity among words in the lexicon creates pressure to form
more fine-grained, phonemic representations to allow for
accurate word recognition and production, and this con-
tinues into middle childhood (Metsala & Walley, 1998).
Metsala and Walley (1998) summarized findings from
the speech perception literature to support their model. First,
phonemic perception shows a protracted course of devel-
opment in children and may continue into young adulthood.
Children’s ability to identify and discriminate phonemes
improves as their vocabularies grow, and this is thought to
reflect more clearly defined phonemic categories. Second,
adults are more sensitive than children to the phonemic com-
position of words in word recognition tasks and are more
adept at identifying a spoken word when only part of the
word is presented to them. Children need to perceive the
entire word if they are to access its underlying representation.
Although word variability has been shown to peak
during developmental change, studies point to a general
trend of decreasing word variability throughout both typical
and disordered speech development (Burt, Holm, & Dodd,
1999; Holm, Crosbie, & Dodd, 2007; Iuzzini & Forrest,
2011). Burt et al. (1999) found a significant negative corre-
lation between age and word variability in children with
typical speech development, ages 3;10 (years;months) to
4;10. These authors measured variability in the production
of the same words across different linguistic contexts (e.g.,
imitation, spontaneous production in response to picture
prompts, and spontaneous production in connected speech).
Holm et al. (2007) observed a significant decrease in word
variability with increasing age in children with typical speech
development, ages 3;0 to 6;11. These authors measured
variability using Dodd’s (1995) Inconsistency Assessment.
This assessment requires children to name 25 colored pictures
on three separate occasions within a session. The youngest
group, ages 3;0–3;5, showed significantly more variability in
their word productions than all other groups. An average of
13% of target words was produced variably by children in the
28 American Journal of Speech-Language Pathology •Vol. 23 •27–35 •February 2014
Downloaded From: http://ajslp.pubs.asha.org/ by Florida State University, Toby Macrae on 03/03/2014
youngest group. Iuzzini and Forrest (2011) found decreasing
word variability with increasing age in children with typical
and disordered speech development, ages 3;0 to 5;8. In this
study, variability reflected the proportion of target words that
were produced variably across three or more productions,
although it is unclear what linguistic context was used.
Within an overall trend of decreasing word variability
in children with typical speech development, peaks may
presage developmental change. In children with SSD, how-
ever, persistent variability, or variability that is higher than
for other children with SSD, may be characteristic of a
subgroup with a unique underlying deficit. In children with
SSD, word variability is most often associated with CAS.
CAS is characterized by three key features, one of which is
variability. Specifically, CAS is associated with “(a) incon-
sistent errors on consonants and vowels in repeated pro-
ductions of syllables or words, (b) lengthened and disrupted
coarticulatory transitions between sounds and syllables, and
(c) inappropriate prosody, especially in the realization of
lexical or phrasal stress”(ASHA, 2007, p. 2). CAS is thought
to reflect an impairment in speech output, specifically,
“planning and/or programming spatiotemporal parameters
of movement sequences”(ASHA, 2007, p. 1). This motor
planning and/or programming deficit may be responsible for
the high levels of word variability that are seen in children
with CAS (Marquardt, Jacks, & Davis, 2004).
Dodd (2005) proposed a classification system of chil-
dren with SSD that includes subgroups characterized ac-
cording to word variability. According to Dodd, children
who use developmental and nondevelopmental phonolog-
ical processes and show variable productions of less than
10 of the 25 target words on Dodd’s (1995) Inconsistency
Assessment are identified as having consistent disorder, and
children who show variable productions of 10 or more of the
25 target words are identified as having inconsistent disorder.
Whereas a diagnosis of CAS is based on other features in
addition to word variability, a diagnosis of inconsistent
disorder is based entirely on word variability.
Dodd and colleagues (Bradford & Dodd, 1994; Holm,
Farrier, & Dodd, 2008) conducted several studies in an
attempt to identify unique deficits in children with in-
consistent disorder. Children in this subgroup showed
age-appropriate oral motor functioning but deficits in
spontaneously producing, imitating, and spelling nonwords
and real words. In the most recent study, Holm et al. (2008)
pointed to a “linguistic breakdown rather than a motor
planning and implementation disorder”(p. 312). Specifi-
cally, the authors proposed a deficit in selecting and assembling
phonemes to form words, which they referred to as pho-
nological assembly, to account for the deficits and highly
variable word productions that are seen in children with
inconsistent disorder. More research is required to deter-
mine if high word variability in children with SSD other
than CAS reflects an underlying deficit and, if so, what this
deficit might be.
Relationships between measures of word variability
and other measures of speech and language abilities may
point to underlying deficits. Several studies that have
examined the relationship between word variability and
vocabulary knowledge have yielded mixed results. Sosa and
Stoel-Gammon (2006) found a negligible relationship be-
tween word variability and expressive vocabulary in children
with typical speech development. Holm et al. (2008) found
that children in their inconsistent disorder subgroup had
receptive vocabulary scores that were no different than those
of children in the other subgroups. More recently, however,
Sosa and Stoel-Gammon (2012) found a high negative
correlation (r=–.81) between word variability and expres-
sive vocabulary. The conflicting findings across Sosa and
Stoel-Gammon’s two studies may be due to their different
measures of word variability. Sosa and Stoel-Gammon (2006)
calculated overall word variability as a ratio of the number
of different phonetic forms of all words produced two or
more times to the total number of productions of these words.
This measure does not account for participants attempting
different numbers of target words. For example, a child might
produce three target words four times each, with the exact
same phonetic form for each production of each word. This
yields a word variability ratio of 3:12, which equals .25.
Another child might produce six target words two times each,
also with the exact same phonetic form for each production
of each word. This yields a word variability ratio of 6:12,
which equals .50. Both of these children are completely con-
sistent in their word productions, yet they receive quite dif-
ferent variability scores. Sosa and Stoel-Gammon (2012)
controlled for different numbers of target words across par-
ticipants by calculating the mean number of different phonetic
forms per target word. This measure, therefore, would seem
to be a more valid reflection of word variability than the
measure that was employed in the earlier study. The relation-
ship between word variability and vocabulary knowledge
requires further exploration. Negative correlations may point
to unstable lexical representations in children with high word
variability.
Speech Error Variability
Studies examining the relationship between speech
error variability and phonological change in children with
SSD provide some insight into the nature of the deficit
underlying this type of variability. Recall that targets with a
consistent speech error have shown greater gains following
treatment than targets with variable speech errors (Forrest
et al., 1997, 2000). Variable speech errors may reflect a lack
of categorical representation for a target sound, and “the
development of categories is a prerequisite to phonological
learning”(Forrest et al., 1997, p. 74). A later study, how-
ever, revealed that speech error variability calculated for all
sounds produced in error predicted phonological change
such that children with the most variable substitutes showed
the greatest gains (Tyler, Lewis, & Welch, 2003). This re-
lationship is in the opposite direction to what would be ex-
pected based on Forrest and colleagues’studies and may be
due to methodological differences across the studies. First,
Tyler et al. (2003) included children with SSD and expressive
language impairments, whereas only some of Forrest and
Macrae et al.: Lexical and Phonological Variability in Preschoolers 29
Downloaded From: http://ajslp.pubs.asha.org/ by Florida State University, Toby Macrae on 03/03/2014
colleagues’participants had documented SSD, and all had
age-appropriate expressive language abilities. Second, Tyler
et al. measured speech error variability and phonological
change across the entire phonological system. Speech error
variability was measured using the Error Consistency Index
(ECI; Tyler et al., 2003), which reflects the total number of
different speech errors for all consonant targets in a single-
word elicitation task. In contrast, Forrest and colleagues
measured speech error variability and phonological change
for individual targeted sounds. Third, participants in the
Forrest et al. studies received treatment for one singleton
fricative that was omitted from the phonetic inventory and
for which participants produced variable speech errors.
In contrast, participants in the Tyler et al. study received
treatment for anywhere between four and eight singleton
or cluster targets. When selecting treatment targets, these
authors did not control for the variability of speech errors.
These methodological differences make it difficult to draw
conclusions about the effect of speech error variability on
phonological change in children with SSD and are likely
responsible for the seemingly conflicting findings across the
studies. More research is needed to study speech error var-
iability in children with SSD.
Significant correlational relationships may elucidate
the deficits underlying high speech error variability in chil-
dren with SSD. Along these lines, Preston and Koenig (2011)
examined relationships among measures of word, speech
error, and phonetic variability in school-age children with
residual speech errors. Speech error variability was highly
correlated with word variability and moderately correlated
with two measures of phonetic variability (i.e., word du-
ration variability and voice onset time variability). Although
variability in vowel formants, another measure of phonetic
variability, was moderately correlated with word and speech
error variability, the correlations were negative. That is,
children with high variability in their vowel formants showed
low word and speech error variability. Low or negligible
correlations were observed among acoustic measures of
phonetic variability and among all other measures of vari-
ability. One explanation given for the lack of convergence
among transcription-based and acoustic measures of vari-
ability was that transcription-based measures may reflect
phonological-level processing, whereas acoustic measures
may reflect speech motor functioning. Iuzzini and Forrest
(2011) examined relationships among measures of word and
speech error variability in children with and without SSD,
including those with CAS. Like Preston and Koenig, Iuzzini
and Forrest found high correlations between measures of
word and speech error variability.
Although nonword repetition (NWR) has been iden-
tified as a potential clinical marker of specific language im-
pairment (SLI; Bishop, North, & Donlan, 1996; Dollaghan
& Campbell, 1998), it may also provide insight into the
deficits underlying high speech error variability. According
to Munson, Kurtz, and Windsor (2005), NWR is a complex
task involving several cognitive processes, including “per-
ceiving and discriminating the acoustic signal, matching
the signal with phonological representations in memory,
planning the articulatory movements required to replicate
the nonword, and executing the response”(p. 1,033). With
regard to phonological representations, Gathercole (2006)
suggested that NWR “is influenced by the quality and per-
sistence of the phonological representations that are char-
acteristic of an individualIand by prior factors affecting
the initial construction of the phonological representation”
(p. 519). In order for speakers to be able to combine pho-
nemes into unfamiliar strings during NWR tasks, therefore,
they must possess, among other things, distinct phonological
representations. Elbro, Borstrom, and Petersen (1998) re-
ferred to distinctness as “the relative distance between a
phonological representation and its neighbors”(p. 40) and
suggested that indistinct representations may be more easily
confused with their neighbors than distinct representations.
Rispens and Baker (2012) found that the distinctness of
children’s phonological representations was highly correlat-
ed with NWR in 5-year-old children with typical language
development. In their study, both phonological short-term
memory and phonological representations predicted signif-
icant amounts of variance in NWR in 5- to 8-year-olds with
typical language, with phonological representations predict-
ing a larger amount of variance (31%) than phonological
short-term memory (10%). Sosa and Stoel-Gammon (2012)
pointed out that no study has examined speech variability
and NWR in the same group of children. Research is re-
quired to determine the relationship between speech error
variability and NWR in children with SSD. Negative cor-
relations may point to indistinct phonological representa-
tions in children with high speech error variability.
Purpose
The purpose of the present study was to determine if
word variability and phonological-level variability reflect
unique underlying processes and, if so, what these processes
might be. First, we examined the relationship between word
and speech error variability. According to the LRM, the
development of distinct phonological representations occurs
only after children have acquired sufficient lexical representa-
tions. Based on the proposal that high word variability reflects
unstable lexical representations, and high phonological vari-
ability reflects indistinct phonological representations, peaks in
phonological variability would be expected to follow peaks in
word variability. Therefore, we hypothesized that word and
speech error variability would not be correlated. Second, we
examined relationships between measures of speech variability
and other speech and language measures—namely, receptive
vocabulary, the standard score from a standardized test of
expressive and receptive language, mean length of utterance
(MLU), and NWR—to provide insight into what each measure
of speech variability reflects. Although some speech variability
may reflect developmental change, high levels of variability
in children with SSD may be indicative of an underlying deficit.
Negative correlations would support this. We hypothesized
that word variability would be negatively correlated with re-
ceptive vocabulary, and speech error variability would be
negatively correlated with NWR performance.
30 American Journal of Speech-Language Pathology •Vol. 23 •27–35 •February 2014
Downloaded From: http://ajslp.pubs.asha.org/ by Florida State University, Toby Macrae on 03/03/2014
Method
Participants
Eighteen children (13 boys and 5 girls), ages 3;6–5;5
(M
age
= 4;8), participated in the study. All participants met
the following inclusion criteria: (a) presence of SSD, as
confirmed by a score at least 1 SD below the mean on the
Bankson–Bernthal Test of Phonology (BBTOP; Bankson &
Bernthal, 1990); (b) no existing diagnosis of CAS; (c) age-
appropriate receptive vocabulary, as confirmed by a score
>1SD below the mean on the Peabody Picture Vocabulary
Test—III (PPVT–III; Dunn, Dunn, & Williams, 1997);
(d) normal oral motor functioning, as confirmed by a score
>1SD below the age-appropriate mean on at least one
subtest of the Oral and Speech Motor Control Protocol
(Robbins & Klee, 1987); and (e) normal hearing, as con-
firmed by positive responses to 1000-Hz, 2000-Hz, and 4000-Hz
stimuli presented at 25 dB HL in audiometric testing.
Participants’descriptive statistics for the inclusion
criteria are presented in Table 1. Potential participants were
recruited from early childhood programs in the Washoe
County School District, Reno, Nevada, and from the speech
and hearing clinic at the University of Nevada, Reno (UNR).
Letters of invitation were given to parents of children who
were receiving treatment for SSD or waiting to receive an
assessment for suspected SSD.
Experimental Measures
All participants received a 2-hr assessment in order to
determine if they met the inclusion criteria and to obtain
results for the experimental measures. These included
severity of SSD, as measured by percentage of consonants
correct (PCC), word and speech error variability, NWR, and
language abilities. All testing was conducted by the first
author in the UNR speech and hearing clinic in a quiet room,
across two sessions. Sessions were audio-recorded using a
Sony ICD-P320 digital audio recorder at a sampling rate of
8 kHz.
Word variability. Word variability was calculated using
a story retell. Pancakes for Breakfast (dePaola, 1978) is a
wordless book with repetitive themes. A story script based
on the book and designed for the purposes of this study was
first read to each participant in its entirety at the very start of
the assessment session. The story was again read to each
participant, in six sections corresponding to six broad events
in the story, toward the end of the session. At the end of
each section, the examiner instructed the participant to retell
that part of the story. The story script has 67 content words
that are repeated at least once by the examiner. Content
words produced three or more times by the participant were
transcribed using International Phonetic Alphabet (IPA)
broad transcription and were included in the analysis. All
function words were excluded. Additionally, only words
with CVC or more complex syllable structures were included.
All words with more basic structures (i.e., V, CV, and VC)
were excluded.
Word variability was calculated from the story retell
using target variability (TV). TV reflects the proportion of
target words with variable productions. It was calculated
by dividing the number of content words produced three or
more times with variable productions by the total number of
content words produced three or more times. TV is the inverse
of Marquardt et al.’s (2004) target stability measure, which
reflects the proportion of target words for which all tokens are
produced alike. Only consonant productions were considered
when determining whether phonetic forms of a word were
the same or different.
Speech error variability. Speech error variability was
calculated using the ECI. The ECI was based on the re-
sponses from the BBTOP and 20 additional words that were
selected to ensure that each of the 23 consonants was sam-
pled three times each in initial and final word positions.
Participants’responses were transcribed using IPA broad
transcription. For each target consonant, the total number of
different substitutions/omissions, regardless of word posi-
tion, was summed. For example, /f/ occurred word initially
in the words fish, fire, and feather, and word finally in the
words knife, leaf, and chief. If a child produced [pIS] for fish,
[faIr] for fire, [wɛdə] for feather, [naIs] for knife, [lip] for leaf,
and [^if] for chief, the number of different sound substitu-
tions/omissions for /f/ across initial and final word positions
was three. This process was repeated for each of the other
23 consonants. The number of different speech errors was
summed across all 23 consonants to yield the ECI for each
participant.
NWR. NWR was measured using the Syllable Repe-
tition Task (SRT; Shriberg et al., 2009), which was designed
for use with children with SSD. The SRT consists of 18 multi-
syllabic nonwords containing CV sequences: eight 2-syllable
nonwords (CVCV), six 3-syllable nonwords (CVCVCV), and
four 4-syllable nonwords (CVCVCVCV). The nonwords
contain only early developing consonants (/b, d, m, n/) and
one vowel (/A/). As per Shriberg et al. (2009), scoring for the
SRT was based on the PCC.
Language. Three measures of language were obtained.
The first was the Core Language score from the Clinical
Evaluation of Language Fundamentals—Preschool, Second
Edition (CELF–P2; Semel, Wiig, & Secord, 2004). This
is a standardized summary score based on participants’
Table 1. Participant summary data for inclusion criteria.
Measure MSD Range
BBTOP standard score 67.89 3.68 65.00–76.00
PPVT–III standard score 106.72 10.65 86.00–126.00
OSMCP TSS 23.78 0.55 22.00–24.00
OSMCP TFS 106.50 3.43 22.00–24.00
OSMCP PRR 1.51 0.38 1.00–2.33
Note. BBTOP = Bankson–Bernthal Test of Phonology (Bankson &
Bernthal; 1990); PPVT–III = Peabody Picture Vocabulary Test—III
(Dunn et al., 1997); OSMCP = Oral and Speech Motor Control
Protocol (Robbins & Klee, 1987); TSS = total structural score;
TFS = total functional score; PRR = polysyllabic repetition rate.
Macrae et al.: Lexical and Phonological Variability in Preschoolers 31
Downloaded From: http://ajslp.pubs.asha.org/ by Florida State University, Toby Macrae on 03/03/2014
performance on the Sentence Structure, Word Structure, and
Expressive Vocabulary subtests of the CELF–P2. The second
measure, MLU
m
, reflects the average length in morphemes
of participants’spoken utterances from the story retell. Par-
ticipants’retell samples were transcribed and MLU
m
was
calculated using the Systematic Analysis of Language Tran-
scripts program (SALT; Miller & Chapman, 2000). The third
measure of language was the standard score from the
PPVT—III, the test of receptive vocabulary. Descriptive
statistics for the experimental measures are provided in
Table 2.
Reliability
Inter- and intrarater reliability were assessed for the
following: (a) identification of the participants’consonant
productions in single words using IPA broad transcription,
the basis of the majority of the experimental tasks, including
the story retell, the BBTOP, the 20 additional words, and
the SRT; and (b) transcription of the participants’spoken
utterances from the story retell and subsequent calculation
of MLU
m
. For IPA transcription, audio recordings of
BBTOP samples from two participants (11%) were selected
at random. Interrater reliability was calculated as the percent
agreement between the first author, who was the original
transcriber, and a trained research assistant, who was a
graduate student in communication science and disorders,
in the identification of consonants in the BBTOP responses.
Interrater reliability was 87.25%. Intrarater reliability was
calculated as the percent agreement between the first author’s
original transcription and the repeated transcription.
Intrarater reliability was 95.05%.
For the calculation of MLU
m
, audio recordings of the
story retells from two participants, selected at random, were
retranscribed and the MLU
m
was recalculated using the
SALT program. Interrater reliability was calculated as the
mean absolute difference between the first research assis-
tant’s original MLU
m
and the second research assistant’s
recalculated MLU
m
across the two participants. The mean
absolute difference was 0.22, which represents 4.9% of the
original mean MLU
m
that was calculated across the two
participants. Intrarater reliability was calculated as the mean
absolute difference between the first research assistant’s
original and recalculated MLU
m
across the two participants.
The mean absolute difference was 0.51, which represents
11.3% of the original mean MLU
m
that was calculated across
the two participants.
Results
The data in Table 2 reveal some important participant
characteristics. First, the participants presented with SSD
ranging in severity from mild–moderate to severe, with a
mean PCC representing a moderate–severe rating (Shriberg
& Kwiatkowski, 1982). Although PCC scores were fairly
evenly distributed among mild–moderate (n= 5), moderate-
severe (n= 7), and severe (n= 6) ratings, no participants
presented with mild SSD. Second, the majority of partici-
pants presented with age-appropriate language abilities.
All participants scored > 1 SD below the mean on the
PPVT–III (see Table 1); most participants (14 of 18 par-
ticipants, 78%) also scored above this cutoff on the CELF–P2.
Third, participants’TV scores ranged from .15 to .79, with a
mean of .41. That is, 15% of the least variable child’starget
words and 79% of the most variable child’s target words were
produced with variable productions. This represents a wide
range of word variability across the participants. The number
of target words included in the word variability analysis ranged
from9to26(M= 20), and the total number of word pro-
ductions across all targets ranged from 43 to 146 (M=97),
across the participants.
Pearson product–moment correlation coefficients were
calculated between the variability measures and between
these and the other speech and language measures. The sig-
nificance level was set at p< .006, based on a Bonferroni
correction for multiple (nine) comparisons. Three of the six
experimental measures rely, to some degree, on the accuracy
of speech sound production, including the measures of word
and speech error variability (i.e., TV and ECI). For two
words or two sounds to be considered variable, at least one of
the productions must be in error. Furthermore, the SRT is
scored according to the PCC. In an effort to control for the
effect of severity of SSD on these measures, partial corre-
lations controlling for PCC were calculated. The correla-
tion between TV and the ECI was slight (Guilford, 1956)
and nonsignificant. The correlation between TV and the
PPVT–III standard score was moderate, negative, and sig-
nificant (r=–.45, p= .034). High word variability was asso-
ciated with small receptive vocabularies. The correlations
between the ECI and SRT (r=–.44, p= .040) and between
the ECI and MLU
m
(r=–.43, p= .042) were moderate,
negative, and significant. The correlation between the ECI
and the Core Language score from the CELF–P2 was neg-
ative and approached moderate in size and significance
(r=–.39, p= .063). All correlations were nonsignificant.
High error variability was associated with poor SRT per-
formance and low language scores.
Table 2. Participant summary data for the experimental measures.
Measure MSDRange
PCC 56.63 12.99 35.4–78.3
TV .41 .20 .15–.79
ECI 21.94 7.97 12–38
SRT 74.11 18.13 28–96
CELF–P2 95.61 14.25 65–119
MLU
m
5.00 1.69 2.10–7.74
Note. PCC = percentage of consonants correct; TV = target
variability; ECI = Error Consistency Index (Tyler et al., 2003); SRT = PCC
on the Syllable Repetition Task (Shriberg et al., 2009); CELF–P2 =
core Language score from the Clinical Evaluation of Language
Fundamentals—Preschool, Second Edition (Semel et al., 2004);
MLU
m
= mean length of utterance in morphemes.
32 American Journal of Speech-Language Pathology •Vol. 23 •27–35 •February 2014
Downloaded From: http://ajslp.pubs.asha.org/ by Florida State University, Toby Macrae on 03/03/2014
Discussion
Relationship Between Word and
Speech Error Variability
The first aim of this study was to examine the rela-
tionship between word and speech error variability in young
children with SSD. The LRM proposes that infants’bur-
geoning lexicons precipitate increases in the level of detail
they store in their underlying lexical representations. Accord-
ing to this model, the development of distinct phonological
representations follows the acquisition of a sufficient number
of lexical representations. Therefore, one would expect peaks
in phonological-level variability to follow peaks in word
variability. Although this study measured variability at only
one point in time, the slight nonsignificant correlation between
word and speech error variability is in line with the assumption
that these two different types of speech variability reflect
unique underlying processes and would show different devel-
opmental progressions, thereby supporting the LRM.
The slight correlation is in contrast to Preston and
Koenig’s (2011) and Iuzzini and Forrest’s (2011) findings of
high correlations between word variability and speech error
variability. These differences may be due to the different
populations of interest: preschoolers with SSD in the present
study; school-age children with residual speech errors in
Preston and Koenig, whose variability may have been more
closely related to speech motor abilities than linguistic abil-
ities; and preschoolers with and without SSD, including
those with CAS, in Iuzzini and Forrest. In the children with
CAS, at least, their high word variability was also likely
related to their speech motor abilities. Preston and Koenig
also suggested that different measures of variability may
reflect unique underlying processes: “Transcriptional mea-
sures may capture a phonological level of representation/
processing, whereas at least some of the acoustic measuresImay
better reflect phonetic or motoric processes”(p. 181). Preston
and Koenig’s transcriptional measures included word and
phonological variability. It is suggested that word variability
reflects lexical-level processing, whereas speech error vari-
ability reflects phonological-level processing. Specifically, high
word variability reflects unstable lexical representations,
whereas high speech error variability reflects indistinct pho-
nological representations. The findings from the present study
that support these proposals will now be discussed.
Relationships Between Variability and
Speech and Language Measures
The second aim of this study was to examine the
relationships between measures of speech variability and
other speech and language measures. Word variability was
negatively correlated with receptive vocabulary, indicating
that children with high word variability had small receptive
vocabularies. This is consistent with Sosa and Stoel-Gammon’s
(2012) finding of a negative correlation and supports Holm
et al.’s (2008) suggestion that children with high word var-
iability have a linguistic deficit. These findings contribute to
our understanding of the role of word variability in speech
development. Sosa and Stoel-Gammon (2006) found that,
in children with typical development, peaks in word var-
iability presaged developmental change, which is in line with
dynamic systems theory. The results of the present study
suggest that, in preschoolers with SSD, high levels of word
variability reflect an underlying deficit characterized by
unstable lexical representations, which may be due to small
vocabularies. Unstable lexical representations may be
responsible for Holm et al.’s proposed deficit in phonological
assembly in children with high word variability. Presumably,
the more children use words and the more their vocabularies
grow, the more stable their underlying representations will
become.
In our study, speech error variability was negatively
correlated with SRT; that is, children with the most variable
speech errors showed the poorest SRT performance. This
supports the proposal that variable speech errors, like poor
SRT performance, are an overt manifestation of indistinct
phonological representations (Forrest et al., 1997, 2000; Sosa
& Stoel-Gammon, 2012; Tyler & Lewis, 2005). Sosa and
Stoel-Gammon (2012) proposed that production variability
and SRT may both assess “the degree of abstract phonemic
knowledge”(p. 605). Forrest et al. (1997) suggested that
children with a consistent error for a target sound may lack the
ability to produce the sound, but they understand that the
sound must be produced consistently in different contexts
and word positions. That is, they understand the cate-
gorical nature of the sound. Categorical representation
of speech sounds develops with increasing age (Coady, Evans,
Mainela-Arnold, & Kluender, 2007; Hazan & Barrett, 2000;
Liker & Gibbon, 2008; Mayo, Scobbie, Hewlett, & Waters,
2003; Nittrouer, 2002; Nittrouer & Miller, 1997; Tyler
& Saxman, 1991). Children with SSD and high speech
error variability may have delayed development of these
representations.
Speech error variability was also negatively correlated
with MLU
m
and the Core Language score from the CELF–
P2. That is, children with the most variable speech errors
showed the poorest language abilities. Broomfield and Dodd
(2004) also found the poorest language skills in children
with the most variable speech output, although their study
focused on word variability and not speech error variability.
As alluded to by Preston and Koenig (2011), the relationship
between language skills and variable output may be bidirec-
tional: Variable output may hinder a child’s ability to map
linguistic information to a spoken production. The converse
may also be true: Unstable linguistic representations may lead
to variable output.
Limitations
There are several limitations associated with the pre-
sent study. First, the two measures of speech variability
showed high variability among the participants. For exam-
ple, for word variability, the standard deviation was approx-
imately half of the mean value. This high variability is likely
due to the small sample size and may be at least partially
responsible for the nonsignificant correlations. Given the
Macrae et al.: Lexical and Phonological Variability in Preschoolers 33
Downloaded From: http://ajslp.pubs.asha.org/ by Florida State University, Toby Macrae on 03/03/2014
small sample size and the nonsignificant correlations, the
findings from the present study should be considered pre-
liminary. Second, the participants presented with a wide range
of ages (3;6–5;5). It is not unreasonable to assume that any
relationship or lack of relationship seen between different
measures of speech variability and between these measures
and other measures of speech and language may be dependent
on the age of the speaker. Recall that Preston and Koenig
(2011) found high correlations between measures of word and
speech error variability in their participants. There was no
such relationship in the present study’s participants, who were
younger than those in Preston and Koenig. In order to obtain
the clearest picture of lexical- and phonological-level proces-
sing at any point in development, a sufficiently narrow view of
development is necessary. The wide range of ages of the
participants also may have been partially responsible for the
nonsignificant correlations. Third, although the omission of
vowels from calculations of word variability is not without
precedent (Ferguson & Farwell, 1975; Ingram, 2002; Leonard,
Rowan, Morris, & Fey, 1982; Sosa & Stoel-Gammon, 2006,
2012), it is considered a limitation in the present study. In-
cluding vowels in the calculations would lead to more sensitive
measures of word variability, particularly for children with
vowel difficulties. This is especially important when studying
children with suspected or diagnosed CAS given that such a
diagnosis is based, in part, on “inconsistent errors on conso-
nants and vowels”(ASHA, 2007, p. 2). Future research in
word variability in children should include vowels in varia-
bility calculations.
Clinical Implications
In preschool children with SSD, it would appear
that high levels of speech variability reflect underlying
linguistic deficits. These findings have important clinical
implications. First, preschool children with SSD and high
word variability should be monitored closely to ensure that
they are acquiring vocabulary knowledge at a rate that is
typical for their age. Clinicians should aggressively target
vocabulary expansion in children with SSD and vocabulary
deficits. Second, according to Metsala and Walley’s (1998)
LRM, a delay in the segmental restructuring of lexical
representations can lead to reading disabilities in some
children. Given the present study’s finding of an inverse
relationship between speech error variability and SRT
performance, it would seem that children with high speech
error variability are at risk for reading disabilities because of
indistinct phonological representations. Clinicians should
monitor these children for potential phonemic awareness and
reading deficits and remediate these when they are present.
Identifying underlying deficits in children with SSD will
allow us to design more focused interventions that target
these deficits. In children with SSD and high speech varia-
bility, interventions that address not just the accuracy of
production but also its consistency may increase the stability
of their underlying representations and lead to the greatest
treatment gains.
Acknowledgments
This research is based on data from a dissertation that was
submitted by the first author in partial fulfillment of the require-
ments for the degree of Doctor of Philosophy in Speech Pathology
from the University of Nevada, Reno. Parts of this research were
presented at the November 2009 annual convention of the American
Speech-Language-Hearing Association in New Orleans, Louisiana.
References
American Speech-Language-Hearing Association. (2007). Childhood
apraxia of speech [Position statement]. Available from www.
asha.org/policy
Bankson, N., & Bernthal, J. E. (1990). Bankson–Bernthal Test of
Phonology. Austin, TX: Pro-Ed.
Bishop, D. V. M., North, T., & Donlan, C. (1996). Nonword rep-
etition as a behavioural marker for inherited language impair-
ment: Evidence from a twin study. Journal of Child Psychology
and Psychiatry, 37, 391–403.
Bradford, A., & Dodd, B. (1994). The motor planning abilities of
phonologically disordered children. European Journal of Dis-
orders of Communication, 29, 349–369.
Broomfield, J., & Dodd, B. (2004). The nature of referred subtypes
of primary speech disability. Child Language Teaching and
Therapy, 20, 135–151.
Burt, L., Holm, A., & Dodd, B. (1999). Phonological awareness skills
of 4-year-old British children: An assessment and developmental
data. International Journal of Language and Communication
Disorders, 34, 311–335.
Coady, J. A., Evans, J. L., Mainela-Arnold, E., & Kluender, K. R.
(2007). Children with specific language impairment perceive
speech most categorically when tokens are natural and mean-
ingful. Journal of Speech, Language, and Hearing Research,
50, 41–57.
dePaola, T. (1978). Pancakes for breakfast. Orlando, FL: Harcourt.
Dodd, B. (1995). Differential diagnosis and treatment of children with
speech disorder. London, United Kingdom: Whurr.
Dodd, B. (2005). Differential diagnosis and treatment of children with
speech disorder (2nd ed.). London, United Kingdom: Whurr.
Dodd, B., Holm, A., Crosbie, S., & McIntosh, B. (2010). Core
vocabulary intervention. In A. Williams, S. McLeod, & R. J.
McCauley (Eds.), Interventions for speech sound disorders in
children (pp. 117–136). Baltimore, MD: Brookes.
Dollaghan, C., & Campbell, T. F. (1998). Nonword repetition and
child language impairment. Journal of Speech, Language, and
Hearing Research, 41, 1136–1146.
Dunn, L., Dunn, L., & Williams, K. T. (1997). Peabody Picture
Vocabulary Test—III. Circle Pines, MN: AGS.
Elbro, C., Borstrom, I., & Petersen, D. K. (1998). Predicting dyslexia
from kindergarten: The importance of distinctness of phono-
logical representations of lexical items. Reading Research
Quarterly, 33, 36–60.
Ferguson, C. A., & Farwell, C. B. (1975). Words and sounds in early
language acquisition. Language, 51, 419–439.
Forrest, K., Dinnsen, D. A., & Elbert, M. (1997). Impact of sub-
stitution patterns on phonological learning by misarticulating
children. Clinical Linguistics & Phonetics, 11, 63–76.
Forrest, K., Elbert, M., & Dinnsen, D. A. (2000). The effect of
substitution patterns on phonological treatment outcomes.
Clinical Linguistics & Phonetics, 14, 519–531.
Gathercole, S. E. (2006). Nonword repetition and word learning:
The nature of the relationship. Applied Psycholinguistics, 27,
513–543.
34 American Journal of Speech-Language Pathology •Vol. 23 •27–35 •February 2014
Downloaded From: http://ajslp.pubs.asha.org/ by Florida State University, Toby Macrae on 03/03/2014
Guilford, J. P. (1956). Fundamental statistics in psychology and
education. New York, NY: McGraw–Hill.
Hazan, V., & Barrett, S. (2000). The development of phonemic
categorization in children aged 6–12. Journal of Phonetics, 28,
377–396.
Holm, A., Crosbie, S., & Dodd, B. (2005). Treating inconsistent
speech disorders. In B. Dodd (Ed.), Differential diagnosis and
treatment of children with speech disorder (pp. 182–201). London,
United Kingdom: Whurr.
Holm, A., Crosbie, S., & Dodd, B. (2007). Differentiating normal
variability from inconsistency in children’s speech: Normative
data. International Journal of Language and Communication
Disorders, 42, 467–486.
Holm, A., Farrier, F., & Dodd, B. (2008). Phonological awareness,
reading accuracy and spelling ability of children with inconsis-
tent phonological disorder. International Journal of Language
and Communication Disorders, 43, 300–322.
Ingram, D. (2002). The measurement of whole-word productions.
Journal of Child Language, 29, 713–733.
Iuzzini, J., & Forrest, K. (2011, November). Lexical and segmental
inconsistency in disordered and typical developing speech. Poster
session presented at the annual convention of the American
Speech-Language-Hearing Association, San Diego, CA.
Kim, H., Hasegawa-Johnson, M., & Perlman, A. (2010). Vowel
contrast and speech intelligibility in dysarthria. Folia Phoniatrica
et Logopaedica, 63, 187–194.
Leonard, L. B., Rowan, L. E., Morris, B., & Fey, M. E. (1982). Intra-
word phonological variability in young children. Journal of
Child Language, 9, 55–69.
Liker, M., & Gibbon, F. E. (2008). Tongue palate contact patterns of
velar stops in normal adult English speakers. Clinical Linguistics
& Phonetics, 22, 137–148.
Marquardt, T. P., Jacks, A., & Davis, B. L. (2004). Token-to-token
variability in developmental apraxia of speech: Three longitu-
dinal case studies. Clinical Linguistics & Phonetics, 18, 127–144.
Mayo, C., Scobbie, J. M., Hewlett, N., & Waters, D. (2003). The
influence of phonemic awareness development on acoustic cue
weighting strategies in children’s speech perception. Journal
of Speech, Language, and Hearing Research, 46, 1184–1196.
Metsala, J. L., & Walley, A. C. (1998). Spoken vocabulary growth
and the segmental restructuring of lexical representations:
Precursors to phonemic awareness and early reading ability. In
J. L. Metsala & L. C. Ehri (Eds.), Word recognition in beginning
literacy (pp. 89–120). Mahwah, NJ: Erlbaum.
Miller, J. F., & Chapman, R. S. (2000). Systematic Analysis of
Language Transcripts (Version 6.1) [Computer software].
Madison, WI: Waisman Center.
Munson, B., Kurtz, B. A., & Windsor, J. (2005). The influence of
vocabulary size, phonotactic probability, and wordlikeness
on nonword repetitions of children with and without specific
language impairment. Journal of Speech, Language, and Hearing
Research, 48, 1033–1047.
Nittrouer, S. (2002). From ear to cortex: A perspective on what
clinicians need to understand about speech perception and
language processing. Language, Speech, and Hearing Services in
Schools, 33, 237–252.
Nittrouer, S., & Miller, M. E. (1997). Predicting developmental
shifts in perceptual weighting schemes. The Journal of the
Acoustical Society of America, 101, 2253–2266.
Preston, J. L., & Koenig, L. L. (2011). Phonetic variability in
residual speech sound disorders. Topics in Language Disorders,
31, 168–184.
Rispens, J., & Baker, A. (2012). Nonword repetition: The relative
contributions of phonological short-term memory and phono-
logical representations in children with language and reading
impairment. Journal of Speech, Language, and Hearing
Research, 55, 683–694.
Robbins, J., & Klee, T. (1987). Clinical assessment of oropharyngeal
motor development in young children. Journal of Speech and
Hearing Disorders, 52, 271–277.
Semel, E., Wiig, E., & Secord, W. (2004). Clinical Evaluation of
Language Fundamentals—Preschool, Second Edition. San
Antonio, TX: The Psychological Corporation.
Shriberg, L. D., & Kwiatkowski, J. (1982). Phonological disorders:
III. A procedure for assessing severity of involvement. Journal of
Speech and Hearing Disorders, 47, 256–270.
Shriberg, L. D., Lohmeier, H. L., Campbell, T. F., Dollaghan, C. A.,
Green, J. R., & Moore, C. A. (2009). A nonword repetition task
for speakers with misarticulations: The Syllable Repetition Task
(SRT). Journal of Speech, Language, and Hearing Research,
52, 1189–1212.
Sosa, A. V., & Stoel-Gammon, C. (2006). Patterns of intra-word
phonological variability during the second year of life. Journal of
Child Language, 33, 31–50.
Sosa, A. V., & Stoel-Gammon, C. (2012). Lexical and phonological
effects in early word production. Journal of Speech, Language,
and Hearing Research, 55, 596–608.
Stoel-Gammon, C. (2011). Relationships between lexical and
phonological development in young children. Journal of Child
Language, 38, 1–34.
Thelen, E., & Bates, E. (2003). Connectionism and dynamic systems:
Are they really different? Developmental Science, 6, 378–391.
Thelen, E., & Smith, L. B. (1994). A dynamic systems approach to the
development of cognition and action. Cambridge, MA: MIT Press.
Tyler, A. A., & Lewis, K. E. (2005). Relationships among consistency/
variability and other phonological measures over time. Topics
in Language Disorders, 25, 243–253.
Tyler, A. A., Lewis, K. E., & Welch, C. M. (2003). Predictors of
phonological change following intervention. American Journal
of Speech-Language Pathology, 12, 289–298.
Tyler, A. A., & Saxman, J. H. (1991). Initial voicing contrast
acquisition in normal and phonologically disordered children.
Applied Psycholinguistics, 12, 453–479.
van Geert, P., & van Dijk, M. (2002). Focus on variability: New
tools to study intra-individual variability in developmental data.
Infant Behavior & Development, 25, 340–374.
Williams, P., & Stephens, H. (2010). The Nuffield Centre Dyspraxia
Programme. In A. Williams, S. McLeod, & R. J. McCauley
(Eds.), Interventions for speech sound disorders in children
(pp. 159–177). Baltimore, MD: Brookes.
Ziegler, W., Hartmann, E., & Hoole, P. (1993). Syllabic timing in
dysarthria. Journal of Speech and Hearing Research, 36, 683–693.
Macrae et al.: Lexical and Phonological Variability in Preschoolers 35
Downloaded From: http://ajslp.pubs.asha.org/ by Florida State University, Toby Macrae on 03/03/2014
