jonathan preston dissertationspeechproductionlab.syr.edu/_pdfs/jonathan preston dissertation.pdf ·...
Post on 22-Jul-2020
11 Views
Preview:
TRANSCRIPT
ABSTRACT
Some children with speech sound disorders (SSD) have difficulty with literacy-
related skills. In particular, they often have trouble with phonological processing, which
is a robust predictor of early literacy. This study investigates the phonological processing
abilities of preschoolers with SSD and uses a regression model to evaluate the degree to
which these abilities can be concurrently predicted by types of speech sound errors.
Forty-three English-speaking preschoolers (ages four to five) with SSD of
unknown origin participated in an assessment of phonological processing skills and
speech sound production. Productions elicited on a 125-item picture naming task were
phonetically transcribed, and errors were coded in two ways: (1) according to Percent
Consonants Correct (PCC), which weights all consonant errors equally, and (2) according
to a three-category system: typical sound changes, atypical sound changes, and
distortions. Phonological awareness (PA) was assessed via rhyme matching, onset
(initial consonant) matching, onset segmentation and matching, and blending.
Phonological memory was assessed using a syllable repetition task. Children also rapidly
named pictures of monosyllabic and disyllabic words.
Results showed that performance on a PA composite score could be predicted, in
part, by vocabulary and age (about 33%). Atypical sound changes were found to account
for additional variance in PA (another 6%), but distortions and typical errors did not
account for significant variance in PA. Thus, use of more atypical sound changes was
associated with poorer performance on PA tasks. When the same consonant errors were
classified using PCC, speech sound errors were not found to predict significant variance
in PA. Atypical sound changes also significantly predicted variance in phonological
memory (about 31%) and rapid naming (about 10%) tasks beyond what had already been
predicted by vocabulary and age.
The results support the notion that poorer performance on phonological
processing tasks is associated with lower receptive vocabularies and production of more
atypical speech sound changes. Results are interpreted in the context of the accuracy of
phonological representations. Thus, atypical sound changes are seen as reflecting poorly
specified internal representations of the sound features of words.
Phonological Processing and Speech Production
in Preschoolers with Speech Sound Disorders
By
Jonathan Preston
B.S. Elmira College
M.S. Syracuse University
DISSERTATION
Submitted in partial fulfillment of the requirements for the degree of
Doctor of Philosophy in Speech-Language Pathology
Department of Communication Sciences and Disorders
Syracuse University
August, 2008
Approved: _______________________ Professor Mary Louise Edwards Date: ____________________
Copyright 2008 Jonathan Preston
All rights reserved
ACKNOWLEDGEMENTS
Thanks to the families that participated in this research, to the clinicians who
referred children, and to my colleagues and friends in the field who offered
encouragement and intellectual support. Thanks in particular to my advisor, Dr. Mary
Louise Edwards, for her support. I am appreciative of comments and feedback from my
committee members, Dr. Raymond Colton, Dr. Linda Milosky, Dr. Benita Blachman, and
Dr. Annette Jenner-Matthews. I also would like to thank Renail Richards for assisting
with reliability, and Dr. Lawrence Shriberg for providing the Power Point stimuli for the
syllable repetition task. In addition, Dr. Beth Prieve’s flexibility was important in
making this project happen.
This research was supported in part by the 2007 American Speech-Language-
Hearing Foundation grant in Early Child Language awarded to the author.
v
TABLE OF CONTENTS
CHAPTERS:
I : INTRODUCTION......................................................................................................... 1
II : METHODS ................................................................................................................ 37
III : RESULTS ................................................................................................................. 70
IV : DISCUSSION........................................................................................................... 87
REFERENCES ............................................................................................................... 105
FIGURES
Figure 1: Theoretical framework for the study.................................................................. 5
Figure 2: Flow chart of procedures with number of participants..................................... 47
Figure 3: Examples of PA stimuli.................................................................................... 60
Figure 4: Scatterplots of speech sound production error types and phonological
awareness Principal Component ............................................................................... 77
Figure 5: Observed PA Principal Component scores and PA scores predicted by the
regression (age, vocabulary, atypical sound changes) for the 43 children with SSD80
vi
TABLES
Table 1: Summary of speech sound error types and their suspected reflection of
underlying phonological representations .................................................................. 26
Table 2: Inclusionary criteria for the study...................................................................... 45
Table 3: Descriptive statistics for the 43 preschoolers who participated in Part II and
were used in the final analysis .................................................................................. 46
Table 4: Summary of speech sound (in)accuracy for 43 preschoolers with SSD............ 71
Table 5: Pearson’s correlation coefficients (r) of speech sound error types.................... 72
Table 6: Summary of the performance of 43 children on the phonological processing
tasks........................................................................................................................... 73
Table 7: Pearson correlation coefficients (r) for the phonological awareness tasks for 43
children with speech sound disorders ....................................................................... 74
Table 8: Principal Component Analysis summary derived from the four Phonological
Awareness tasks ........................................................................................................ 75
Table 9: Hierarchical regression used to predict PA Principal Component .................... 78
Table 10: Regression using PCC as the speech production variable to predict PA......... 81
Table 11: Regression explaining variance in Phonological Memory (Syllable Repetition
Task) ......................................................................................................................... 83
Table 12: Regression explaining variance in Rapid Naming (average Z scores of two
Rapid Naming tasks) ................................................................................................ 85
vii
APPENDICES
Appendix A: Transcription Rules and Coding Sound Changes..................................... 125
Appendix B: Errors with Interacting Sound Changes: Which is Preferred? ................ 146
Appendix C: Words Used on the Picture Naming Task ................................................ 150
Appendix D: Phonological Awareness Tasks................................................................ 151
Appendix E: Syllable Repetition Task (from Shriberg et al, 2006)............................... 155
Appendix F: Rapid Naming Task .................................................................................. 156
Appendix G: Complete Correlation Matrix ................................................................... 157
Appendix H: Measurement Issues ................................................................................. 159
Appendix I: Regression Diagnostics.............................................................................. 163
Appendix J: Caveats and Limitations: The Role of Children’s Experiences................ 166
Appendix K: Speech Perception .................................................................................... 168
1
I : INTRODUCTION
Literacy problems are a significant international concern, with as much as 15-
20% of the world’s population having some sort of reading difficulty (International
Dyslexia Association, 2000). Early identification of such problems is essential so that
early intervention can take place. Fortunately, it is now possible to identify skills in
preschool that are good predictors of later literacy. This study will focus on preschoolers
with speech sound disorders (SSD), who are known to be at risk for preliteracy and
literacy problems (particularly phonological processing). Exactly how SSDs are related
to preliteracy deficits is unclear. Therefore, to aid in the identification of early preliteracy
problems, this study will explore the relationship between the specific types of speech
sound errors produced by preschoolers with SSD and phonological processing skills,
known to predict early literacy.
Phonological processing, which is the ability to process speech sound
information, is related to both speech production and literacy development (e.g.,
Stackhouse & Wells, 1997). Because phonological processing skills do not necessarily
rely on alphabet knowledge, it is possible to assess these skills in preschool children
(prior to formal literacy instruction). Phonological processing has been discussed as
including three domains: phonological awareness (PA), phonological memory (PM),
and phonological retrieval (as assessed by rapid naming, RN) (e.g., Wagner &
Torgesen, 1987). Children with SSD have been reported to have weaknesses in each of
these domains (e.g., Leitao et al., 1997). The degree to which variability in speech sound
production is related to variability in each of these three components of phonological
processing has not been thoroughly explored. This study addresses that issue.
2
The term speech sound disorder (SSD) will be used to refer to children who have
clinically significant difficulties producing or using the speech sounds of their native
language for their age and dialect groups (cf. NIDCD, 2006). Other reports have referred
to these children as having ‘articulation’ or ‘phonological’ disorders and/or delays (e.g.,
Dodd, 1995; e.g., Gibbon, 1999). The current investigation will limit the definition to
include children whose primary deficits are in speech communication, and who have no
known oral structural problems (e.g., cleft palate) or developmental disorders (e.g.,
cerebral palsy). Approximately 8-9% of young children are diagnosed with a SSD
(NIDCD, 2006); thus, the problem affects millions of children.
There is emerging evidence that children who begin kindergarten with a SSD and
poor phonological awareness are at particular risk for later literacy problems (e.g.,
Nathan et al., 2004); thus, early identification of these problems is crucial. The specific
relationship between speech sound production patterns and phonological processing in
children with SSD, however, remains unclear. Previous investigations have used
measures of speech sound production (e.g., Percent Consonants Correct) that may not be
sensitive to the nature of the errors a child makes. Therefore, the current study will
examine the relationship between types of speech sound errors, quantified in a more
precise manner than in previous investigations, and each domain of phonological
processing in preschoolers with SSD. The primary focus will be the relationship between
speech sound production and phonological awareness, with exploratory analyses
examining the concurrent relationship between speech production and the other two
domains of phonological processing, phonological memory and phonological
retrieval/rapid naming.
3
The goals of this research are (1) to confirm previous assertions (which largely
lack empirical support) regarding the strength of the relationship between various types
of speech sound errors and measures of phonological processing in children with SSD;
(2) to improve our understanding of how specific types of speech sound changes account
for unique variance in phonological processing. The clinical contributions of the study
include identifying speech production characteristics that may be indicative of risk for
early literacy problems.
4
Review of the Literature
The concept of phonological representations will be reviewed first, as
phonological representations have been discussed as an underlying contributor to
performance on phonological processing tasks as well as speech sound production. The
literature concerning the relationship between phonological awareness (PA) and literacy
development will be reviewed briefly to highlight the importance of being able to identify
potential indicators of phonological processing difficulty. Also, the known connection
between SSD and PA will be outlined, and limitations in our current understanding of
this relationship will be addressed. The quantification of speech sound errors will be
discussed, along with the justification for a more specific measurement system that could
advance our understanding of the relationship between speech sound errors and
phonological processing. Finally, literature related to two other domains of phonological
processing, phonological memory and phonological retrieval/rapid naming, will be
reviewed; the relationship between SSD and these two domains will be investigated by
exploratory analyses.
Figure 1 (similar to a model by Rvachew & Grawburg, 2006) was adapted for the
current study to explicate the relationship between speech sound accuracy and
phonological processing, and to show the presumed relationship of each to phonological
representations. The literature review will use this figure as a guide in discussing the
relationships among the concepts of interest.
5
PHONOLOGICAL PROCESSING Phonological Awareness
Phonological Memory
Phonological Retrieval
Figure 1: Theoretical framework for the study
Notes: The link between speech sound accuracy and phonological processing (heavy dotted line) remains unclear and will be examined here. The link between phonological processing and literacy is well established (not shown here). The variables in jagged boxes (age and vocabulary) are control variables that have been discussed as being associated with the accuracy of phonological representations. The shading of Speech Sound Accuracy indicates that there may be varying degrees of (in)accuracy of speech sound production. No significant relationship is generally reported between receptive vocabulary and speech production in preschoolers (Bishop & Adams, 1990; Rvachew & Grawburg, 2006); hence Figure 1 does not include a link between vocabulary and speech sound accuracy. Concurrent relationships will be explored in this study, not causality.
PHONOLOGICAL
REPRESENTATIONS
Age Vocabulary
SPEECH SOUND
ACCURACY
Atypical Sound Changes
Typical Sound Changes
Distortions
6
Phonological Representations
Phonological representations are stored (internal) representations in the mental
lexicon that contain the phonological (speech-sound related) features of words (Edwards,
1995; Pascoe et al., 2006; Rvachew, 2006; Stackhouse & Wells, 1997). These
representations may include the constituent phonemes and phoneme combinations of
words, and possibly the associated phonetic specifications of the segments, such as
acoustic or motoric features (e.g., Shuster, 1998). Because these representations are
internal, they cannot be directly measured. Therefore, researchers rely on measurable
behaviors to make inferences about phonological representations. While some theorists
hold that there are “input” representations and “output” representations (see Edwards,
1995 for a review), empirical data provide support for a strong relationship between the
two (Foy & Mann, 2001; Shuster, 1998; Sutherland & Gillon, 2005). As most current
models rely on a single underlying representation (Baddeley, 2003; Rvachew &
Grawburg, 2006; Stackhouse & Wells, 1997), this is the view assumed in the current
study. As in other studies (Elbro et al., 1998; Rvachew & Grawburg, 2006; Rvachew et
al., 2003; Sénéchal et al., 2004; Strange & Broen, 1981), the current investigation will
use speech sound production as one way of inferring the accuracy of phonological
representations (see below).
It is generally assumed that, as children get older, phonological representations
develop and improve (i.e., become more adult-like) (Nathan et al., 2004; Sutherland &
Gillon, 2005). Therefore, age must be taken into account when considering a child’s
phonological representations. However, not all children develop more accurate
phonological representations at the same rate or with the same precision. Thus, some
7
children may have more accurate or “stronger” phonological representations than others
(Rvachew & Grawburg, 2006; Rvachew et al., 2003; Snowling, 2000; Stackhouse, 1997;
Swan & Goswami, 1997a).
Weaknesses in the accuracy (or “strength”) of phonological representations have
been discussed as a basis for both impaired speech and poor phonological processing
(and, by extension, poor literacy skills) (Elbro et al., 1998; Larivee & Catts, 1999;
Rvachew, 2007; Sutherland & Gillon, 2005; Swan & Goswami, 1997a). For example,
Senechal, Ouellette and Young (2004) suggest that "the quality of phonemic
[phonological] representations may be reflected in children's expressive phonology or
articulation" (p. 243). Similarly, Swan and Goswami (1997a) make the claim that weak
phonological representations contribute to the poor phonological awareness skills of
children with literacy problems. If a relationship is found between speech sound
accuracy and phonological processing, this would provide support for the notion that
phonological representations are an underlying factor in both speech sound production
and phonological processing skills.
The current study continues a line of research investigating the phonological
deficit hypothesis, in which phonological processing is causally related to literacy skills
(Snowling, 2000; Wagner & Torgesen, 1987). Accurate or precise phonological
representations are considered to be important for the development of phonological
processing skills (Fowler, 1991; Snowling, 2000). In fact, weaknesses in phonological
representations have been discussed as the causal factor in poor performance on
phonological processing tasks by children with preliteracy and literacy problems (Swan
& Goswami, 1997a, 1997b). It has been presumed that children with inaccurate
8
phonological representations will have difficulty with tasks that require them to utilize
those representations, such as comparing initial phonemes in words, or comparing
rhymes.
Phonological Representations, Phonological Awareness, and Literacy
Phonological awareness (PA) refers to awareness of spoken units of speech, such
as syllables and rhyming words (see explanations below). It includes phonemic
awareness, which is the awareness of individual sounds (Report of the National Reading
Panel, 2000). Converging evidence suggests that PA skills are related to spelling, reading
decoding, reading comprehension, and reading fluency, both concurrently and
longitudinally (Bradley & Bryant, 1983; Catts et al., 2001; National-Reading-Panel,
2000; Phillips & Torgesen, 2006; Snow et al., 1998; Wagner & Torgesen, 1987).
Additionally, PA is a primary area of focus in this study because there is evidence that
explicit instruction in PA can have positive benefits for literacy development in children
both with and without SSD (Ball & Blachman, 1991; Bradley & Bryant, 1983; Gillon,
2000, 2005; Tangel & Blachman, 1992).
Syllables are units of speech that must include a nucleus (typically a vowel), with
optional consonants preceding the nucleus (the “onset”) and/or following the nucleus (the
“coda”). Developmentally, awareness of syllables precedes awareness of rhyme (vowel
plus coda), which precedes awareness of phonemes (individual consonants or vowels)
(Liberman et al., 1974; Stackhouse, 1997). Hence, children become aware of smaller and
smaller units of speech. While phonological awareness in preschoolers may help to
predict later literacy, it is awareness of speech at the phoneme level (phonemic
9
awareness) that is most critical in learning to read and spell (Bradley & Bryant, 1983).
This is because, in an alphabetic system such as English and many other written
languages, letters represent phonemes, not rhymes or syllables. Preschoolers, who are the
focus of this study, are often at the stage of learning to (a) identify and produce rhymes (a
vowel plus coda, e.g, the “at” in hat), (b) identify and produce initial phonemes (e.g., the
“h” in hat), and (c) blend units spoken separately to form words (e.g., blend “h” and “at”
to form hat) (Bird et al., 1995; Catts, 1991; Gillon, 2000; Rvachew et al., 2003;
Stackhouse, 1997; Stackhouse & Wells, 1997).
Many studies have examined variables that relate to phonological awareness.
These include receptive vocabulary (McDowell et al., 2007; Rvachew & Grawburg,
2006), expressive vocabulary (Elbro et al., 1998), letter naming (Elbro et al., 1998),
socioeconomic status (McDowell et al., 2007; Nittrouer & Burton, 2005), and speech
perception (Rvachew & Grawburg, 2006), etc. The present study will control for age and
receptive vocabulary, the variables that have been most commonly discussed as relating
to the development of phonological representations.
The Role of Age and Vocabulary in PA Development
As depicted in Figure 1, phonological processing skills (including PA) are related
to age and vocabulary. As children get older and their vocabulary skills increase, they
have a larger internal ‘dataset’ from which to make inferences about phonological
features of words. Therefore, phonological representations are thought to become more
accurate (or precise) as vocabulary skills develop and as children get older (Metsala,
1999; Walley et al., 2003). It is also believed that as children become more attuned to
10
smaller phonological features of words, performance on PA tasks improves (Fowler,
1991; Liberman et al., 1974; Snowling, 2000).
Age and PA. Longitudinal studies have reported growth in PA skills and literacy
as children age (e.g., Caravolas et al., 2001; Nathan et al., 2004). In a cross-sectional
study, Chafouleas et al. (1997) reported that age can account for as much as 60% of the
variance in PA from kindergarten to second grade, providing evidence for rapid
developmental growth in PA skills. This growth in PA and (pre-)literacy at a young age
is often discussed as a function of more mature phonological representations (Fowler,
1991; Nathan et al., 2004; Swan & Goswami, 1997a). Thus, age is one important factor
to consider when assessing PA.
Vocabulary and PA. There is also a strong relationship between PA/literacy
development and language skills in young children. For example, language impairment
negatively impacts literacy development (Aram et al., 1984; Bishop & Adams, 1990;
Bishop & Clarkson, 2003; Catts, 1993, 1997; Catts et al., 1994; Kamhi & Catts, 1986;
Kamhi et al., 1988; Nathan et al., 2004), and for children with and without speech and
language impairments, vocabulary has proven to be the most robust language measure
when predicting PA. That is, vocabulary is reported to account for approximately 25-
30% of the variance in PA in preschool and young school-age children (Bishop &
Adams, 1990; Elbro et al., 1998; Rvachew, 2006; Rvachew & Grawburg, 2006; Rvachew
et al., 2004). In the present study, the primary interest is in receptive vocabulary, in part
because speech sound impairments influence the ability to reliably interpret a child’s
spoken vocabulary.
There is empirical evidence that vocabulary and PA skills are positively
11
correlated (Rvachew, 2006; Rvachew & Grawburg, 2006; Swanson et al., 2003). For
example, Metsala (1999) found that larger receptive vocabularies in children, as
measured by the Peabody Picture Vocabulary Test-Revised (Dunn & Dunn, 1981), were
correlated with better performance on phonological processing tasks (including blending,
initial phoneme isolation, and rhyming). Vocabulary and PA were related even when the
influence of age was controlled. She attributes this phenomenon to underlying
representations that are more adult-like in their features and their organization. That is,
children who know more words are thought to have more accurately defined
phonological representations, because they must keep words separate from similar-
sounding words.
Phonological representations are also related to speech sound production
(Edwards et al., 2004; Hodson & Edwards, 1997; Shuster, 1998; Stackhouse & Wells,
1997). Therefore, there is reason to believe that phonological processing skills also relate
to SSD. One important question is whether speech production can predict variance in
PA, and whether it can predict variance in PA above and beyond the known contribution
of receptive vocabulary and age.
Suspected Causes of Poor Phonological Representations
There is much speculation as to why phonological representations would be weak
in some children, including many children with SSD and literacy difficulties. One
possibility is that genetic factors, or a combination of genetic and environmental factors,
play a role in speech and literacy difficulties (Lewis et al., 2002; Lewis et al., 2006;
Raitano et al., 2004; Shriberg et al., 2005). Also, speech perception skills (including
12
temporal order judgment, phoneme discrimination, phoneme boundary identification, and
amplitude envelope rise time) have been found to relate to phonological processing and
literacy in several studies of children with different levels of reading skill (Lieberman et
al., 1985; Mody et al., 1997; Richardson et al., 2004; Savage et al., 2005; Sénéchal et al.,
2004; Watson & Miller, 1993) and children with SSD (Bridgeman & Snowling, 1988;
Jamieson & Rvachew, 1992; Ohde & Sharf, 1988; Rvachew, 1994; Rvachew &
Grawburg, 2006; Rvachew et al., 2004; Rvachew et al., 2003; Rvachew et al., 1999;
Sharf et al., 1988). (Appendix K provides further discussion of this topic.) The current
study is continuing a research line that presumes that phonological representations may
be impaired, but does not attempt to explain why they are weak in some children with
poor PA and/or speech sound production problems.
Regardless of the mechanism responsible for weak phonological representations,
it remains clear that performance on phonological processing tasks varies widely in
young children, including those with SSD. The current study seeks to determine if a new
measure of speech sound accuracy can provide additional explanation for variance in
phonological processing, because both speech production and phonological processing
are presumed to rely on phonological representations. This new measure could help to
provide a clinical indicator of PA skills in young children with SSD.
Phonological Awareness in Children with Speech Sound Disorders
Children with SSD, as a group, have been found to have poor PA. Therefore,
they are generally considered at risk for later literacy problems. For example, Lewis and
Freebairn (1992) compared preschoolers, school-age children, adolescents, and adults
13
with histories of SSD to age-matched peers without such histories on a variety of PA
tasks. Significant differences were found between the groups at all age levels, suggesting
that a history of SSD constitutes risk for PA/literacy problems. However, because this
was a retrospective study, specific speech sound production characteristics were not
considered when evaluating PA/literacy outcomes.
Raitano et al. (2004) found that five to six year olds with SSD performed below
age-matched controls on a PA factor score which included rhyme, elision (segment
deletion), blending, and sound matching. Bird et al. (1995) also found that five to seven
year olds with SSD performed below controls on measures of rhyme, initial consonant
matching, initial consonant segmentation and matching, and nonword reading and
spelling, regardless of whether or not they had concomitant language impairments.
In studies examining the effects of PA intervention for children with SSD, Gillon
(2000, 2005) found large group differences between children with SSD and controls on
measures of PA prior to intervention. Specifically, she found that children with SSD who
were receiving intervention that did not include a PA component had a slower rate of
literacy skill acquisition compared to typically developing control children without SSD.
However, children with SSD who received PA intervention improved PA skills at a rate
similar to typically developing control children.
Leitao et al. (1997) reported that six year olds with SSD performed below
typically developing children on PA measures such as elision (deletion of sounds),
blending, segmentation, and invented spelling. They found that some (but not all)
children with SSD perform below the range of typically developing children. Although
no statistical analyses were performed to address the issue, they suggested that the
14
children with SSD performed differently based on the types of speech sound errors that
they exhibited (see below for further discussion). This is one of the few attempts that has
been made to (qualitatively) relate the variability in PA to types of speech sound errors.
One exception to the above findings of low PA performance by children with
SSD was provided by Catts (1993), who reported that a group of 15 kindergarten children
who had ‘articulation impairments’ performed as well as typically developing children on
several early reading measures when assessed in second grade. However, these children
were identified based only on the number of errors on a widely used articulation test, the
Goldman Fristoe Test of Articulation (Goldman & Fristoe, 1986). No further speech
analysis was reported. This greatly limits the ability to interpret the speech sound
characteristics of the sample. This also highlights the possibility that some children with
SSD may perform within normal limits on phonological processing and literacy
measures. Again, the within-group variance among children with SSD has yet to be
thoroughly explained.
In summary, there is evidence that children with SSD, as a group, often have
below-average PA, putting them at risk for later literacy problems. However, this is not
the case for every child with SSD, and the variability in PA skills in this population is
largely unexplained. Of interest in the present study is whether this variability can be
partly explained by the relative occurrence of the different types of speech sound errors
the child exhibits.
Measuring Speech Sound Errors
There is no universally accepted way of quantifying the accuracy of speech sound
15
production. Several different methods of measuring speech sound errors have been used.
Research evaluating the relationship between phonological processing and speech sound
disorders, quantified by the total number of errors or standard scores on a standardized
articulation test, have yielded mixed results (Catts, 1993; Larivee & Catts, 1999;
Rvachew & Grawburg, 2006). Drawbacks to the use of standardized tests include (a) the
speech sample is often small (under 60 words), (b) the sample often includes just one
occurrence of each sound in each word position, and (c) all types of errors are equally
weighted (e.g., speech sound distortions may be counted the same as phoneme
substitutions or omissions or unusual sound changes).
One way to measure general speech sound accuracy is to (statistically) combine
multiple measures of speech sound production to approximate or estimate “speech sound
production” as a global construct. For example, Nathan et al. (2004) explored preschool
speech sound production and its relationship with early literacy skills using path analysis.
Speech sound production was measured as percent consonants correct (PCC) derived
from naming 20 pictures and also repeating several real words and nonsense words. This
composite of speech production was not a significant predictor of PA and literacy skills
over the next two years. However, limitations of this study include the small speech
sample, the use of PCC to measure speech sound accuracy (see below), the use of a
repetition task (i.e., a phonological memory task) to evaluate speech sound accuracy, and
the fact that different types of speech sound errors were not considered.
Rvachew and Grawburg (2006) used structural equation modeling to examine
whether PA could be predicted from speech production, estimated by PCC in connected
speech and scores on the Goldman-Fristoe Test of Articulation-2 (Goldman & Fristoe,
16
2000). They found that a model without a link between PA and speech production
(estimated by PCC and GFTA-2 scores) was preferred to a model that used speech sound
production to predict PA. Thus, the speech sound production-PA relationship was not
confirmed using a global estimate of speech production. However, this study failed to
evaluate the types of speech sound errors, which is argued to be an important difference
in speech production between children.
McDowell et al. (2007) used the GFTA-2 along with a measure of nonsense word
repetition to estimate speech sound accuracy in 700 children between the ages of two and
five. PA was measured by rhyming tasks, blending tasks, and elision (sound deletion)
tasks. The combined GFTA-2 and nonword repetition measure was found to account for
significant variance (5%) in PA beyond receptive vocabulary. However, limitations of
this study include the use of a small speech sample, and the use of a phonological
memory task (nonword repetition) to assess speech sound accuracy. It is also unclear
how many of these children had a speech sound disorder. Importantly, this study also did
not evaluate the types of speech sound errors made by the children.
Frequency of Speech Sound Errors: Percent Consonants Correct (PCC)
PCC is a widely used method of assessing severity of speech sound disorders
(Shriberg et al., 1997a; Shriberg & Kwiatkowski, 1982). In this calculation, the number
of correct consonants in a sample is divided by the number of attempted consonants. All
consonant errors are therefore equally weighted. Although PCC in conversational speech
is said to be related to severity of speech production problems (Shriberg & Kwiatkowski,
1982), it may not be the best measure for evaluating the relationship between speech
17
sound accuracy and phonological processing. That is, while it captures the frequency of
consonant errors, it does not distinguish between types of errors (distortions,
substitutions, omissions).
In some instances, PCC based on a picture naming task has been found to predict
PA and early literacy. For instance, Bishop and Adams (1990) reported that speech
production measured by PCC at age five-and-a-half predicted later reading accuracy and
spelling (known to be related to PA), although the contribution of speech sound errors to
follow-up prediction of reading was relatively modest (PCC at four-and-a-half years
explained 5.4% of the variance in reading accuracy at eight years of age, beyond
vocabulary and IQ). Bird et al. (1995) found PCC in a picture naming task to contribute
to predicting later literacy difficulty among five to seven year olds with SSD. Larivee
and Catts (1999) also found PCC in multisyllabic words at the end of kindergarten to
predict reading in first grade. The variance in reading ability explained by PCC
overlapped with the variance in reading that was explained by PA; the authors
hypothesized that this is evidence that PCC in multisyllabic words taps similar skills to
PA, specifically the quality of phonological representations.
In contrast, Gillon (2005) found no significant correlation between PCC in
conversation and several measures of PA (rhyme oddity, phoneme matching, letter
recognition, alliteration, syllable segmentation, letter-sound knowledge, phoneme
isolation) during five assessment periods between three and six years of age.
Additionally, Rvachew and Grawburg (2006) found that PCC in conversation was not
related to PA in a large study of 95 preschoolers with SSD.
In conclusion, the results of studies that have used PCC to predict PA and/or
18
literacy are mixed. One limitation is that PCC weights all speech sound errors the same,
regardless of the type of error1. Thus, PCC does not capture differences between speech
sound patterns in children. Therefore, a new procedure for measuring speech sound
errors will be used. It is hypothesized that this procedure will be more sensitive to PA
problems than the standard PCC measure, in part because it takes into consideration the
presumed relationship between the type of error and phonological representations. It is
hypothesized that errors representing relatively weak phonological representations will
make a significant contribution to the variance in phonological processing, while errors
representing minor deviations from a target (and presumably more accurate phonological
representations) will not make a significant contribution to the variance in phonological
processing.
Types of Speech Sound Errors
Difficulty in learning to produce speech sounds correctly can be manifested in a
variety of types of speech sound errors. However, not all errors are necessarily
equivalent, as would appear to be the case when using PCC or raw number of errors on a
standardized test. It is possible to consider speech sound errors and error patterns
differently, as has been done by researchers and clinicians since the 1970s (Edwards &
Shriberg, 1983; Ingram, 1976; Khan, 1982). Thus, the current study will categorize
speech sound errors according to typical and atypical sound changes, often referred to as
phonological processes. In this type of analysis, errors are analyzed in terms of place,
1 One alternative to PCC would be to use a revised measure (PCC-R) (Shriberg et al., 1997a), which considers phoneme omissions and substitutions as errors, but ignores distortions. However, this would not capture differences in sound changes involving one feature (e.g., [t] for /k/) and sound changes involving two or more features (e.g., [d] for /k/), nor would it differentiate between typical sound changes (e.g., [t] for /k/) and atypical sound changes (e.g., [s] for /k/). This distinction is further discussed later.
19
manner, voicing, and syllable structure. Such sound changes have been used in the
literature for many years to describe speech sound errors produced in typically
developing children and those with SSD, but previous investigations have often used
these sound changes to describe the types of errors individual children (or small groups)
make. There have been relatively few attempts to use such errors patterns to
quantitatively describe children’s speech sound accuracy.
In this study, each speech sound error exhibited by each child will be classified
according to the types of individual (component) changes involved: distortions, typical
sound changes, and atypical sound changes. It is hypothesized that the types of sound
changes represent different degrees of similarity between a target representation for a
phoneme and the child’s actual production.
Distortion Errors. Errors that are typically referred to as distortions involve
productions that are in the correct phoneme category, but are produced without phonetic
precision or accuracy. Distortions, which reflect a slight alteration in the production of a
sound (such as a slight problem with tongue shape or placement), are prevalent in the
speech of young children with typically developing speech as well as those with SSD
(Shriberg & Kwiatkowski, 1994; Smit et al., 1990). For example, the voiceless alveolar
fricative /s/ in “Sue” could be produced with the tongue blade or tip too close to the teeth,
resulting in a dentalized production of /s/, transcribed as [sʝu]. Such a production would
still be recognized as belonging to the /s/ phoneme category. It has been suggested that
distortions (e.g., dentalized or lateralized /s/, labialized /r/) may represent a breakdown in
motoric processes (Dodd, 1995; Dworkin, 1980; Fletcher et al., 1961; Hall, 1989;
20
Shriberg et al., 2005). It is hypothesized that, because such motor differences are not
likely to be related to phonological representations, distortion errors will not be closely
related to phonological processing. In fact, Shriberg (1997) states, “Unlike phoneme
deletions and phoneme substitutions, phoneme distortions have not been associated with
deficits in the phonological skills underlying reading, writing, and other verbal skills” (p.
107).
One investigation that empirically evaluated the relationship between distortions
and phonological awareness in children with SSD was by Rvachew et al. (2007). These
authors found no significant group differences between four to five year olds with
normally developing PA and those with delayed PA in the number of distortions
produced on the Goldman-Fristoe Test of Articulation-2 (Goldman & Fristoe, 2000).
Preston and Edwards (2007) also reported that speech sound distortions, when counted as
errors, reduce the correlation between speech sound errors and phonological awareness in
adolescents. Thus, it is hypothesized that distortions are not indicative of weak
phonological representations and therefore will not be related to phonological processing.
Phonemic Sound Changes (Typical and Atypical). Phonemic sound changes, in
which the target phoneme is not produced, may be considered less accurate productions
than distortions. Phonemic sound changes include substitutions, in which a different
phoneme is produced. For example, cat /kæt/ could be produced as [tæt] (k�t).
Patterns of omissions may also be observed; for example [kæ] for /kæt/ (t � Ø).
Some of these sound changes may be “atypical;” that is, they are found rarely, if at all, in
normal development (Dodd, 1995, 2005; Dodd & Iacano, 1989; Dodd et al., 1989;
Edwards & Shriberg, 1983; Ingram, 1976; Leitao & Fletcher, 2004). Therefore, atypical
21
sound changes are thought of as being less accurate than typical sound changes and are
hypothesized to be related to weak phonological representations.
‘Typical’ Sound Changes. Typical sound changes represent systematic
substitutions or omissions that affect a class of sounds (e.g., velars or fricatives) or a
sound sequence (e.g., /s/ plus stop clusters) (Edwards & Shriberg, 1983). For example,
children with typically developing speech as well as children with SSD may produce the
name “Sue” (/su/) as [tu], replacing the fricative /s/ with the presumably easier stop [t],
with which it shares several features (a pattern often called “stopping of fricatives”). It is
also possible for a child to produce errors that involve more than one feature change at a
time. Such changes may be considered “interacting” or “overlapping” (Edwards &
Shriberg, 1983). For example “Sue” could be produced as [du] by stopping the fricative
and adding voicing. This more complex two-feature change would not be captured using
Percent Consonants Correct, because in both [tu] and [du], the one consonant that is
assessed (/s/) is produced incorrectly, thus both productions are counted the same.
Children with SSD may continue to use these typical phonemic sound changes beyond
the ages at which they should have been outgrown (Edwards & Shriberg, 1983). The
continued use of these sound changes may reflect a delay in learning linguistic ‘rules’ for
speech production, which could also be reflected in other phonological abilities such as
phonological awareness. That is, frequent use of these typical sound errors may reflect a
delay in phonological development for both speech production and phonological
processing.
‘Atypical’ Sound Changes: Some speech sound errors exhibited by children with
22
SSD represent sound changes that are found rarely, if at all, in typical phonological
development. For example, children with SSD may delete the initial consonant in a
word, producing “Sue” as [u] (Dodd & Iacano, 1989), or they may replace the /s/ with a
sound produced further back in the mouth, as in [gu] for Sue. Such errors have been
characterized as unusual, deviant, atypical, nondevelopmental, or different from those of
normally developing English-speaking children (Dodd, 2005; Dodd & Iacano, 1989;
Dodd et al., 1989; Edwards & Shriberg, 1983; Ingram, 1976; Klein & Spector, 1985;
Leonard, 1985; Lowe, 1994). However, there is no complete list of typical and atypical
changes, and there are some sound changes that are less clear-cut. Other changes are
uncommon, but still phonetically plausible. In this study, an effort was made to define
atypical errors based on existing literature. Definitions of typical and atypical sound
changes as adapted for this study are found in Appendix A, along with examples.
One of the goals of this study was to investigate the hypothesis that atypical sound
changes may represent a greater degree of phonological impairment than other sound
changes. According to Dodd and Iacano (1989), “A child who follows the normal course
of development, albeit slowly, is less linguistically impaired than a child who produces
(atypical) errors” (p. 334). They also suggest that, “The use of (atypical) processes
reflects a linguistic deficit, i.e., an impaired ability to abstract the rules governing
phonology” (p. 335). If this is the case, then atypical sound changes should be more
strongly related to poor PA than typical sound changes.
Indeed, there is some evidence that atypical phonemic sound changes may be
associated with poorer PA outcomes. For example, Dodd et al. (1989) grouped children
based on the nature of their speech error patterns. The authors reported that preschoolers
23
who consistently used atypical sound changes had an impaired ability to detect whether a
word was phonologically ‘legal’ (e.g., /zmebi/ is not phonologically legal, because it
violates rules of English that prohibit initial consonant sequences such as /zm/). While
this study generally lends support to the notion that atypical errors may reflect a poorer
understanding of phonological rules, examination of the data indicates that the children
who were grouped as having atypical speech errors had more errors overall than children
who were in the group that used primarily typical errors. Additionally, vocabulary and
age were not considered when the groups were compared. Hence, it is unclear whether
atypical sound changes, more typical sound errors, more distortions, or other factors (e.g.,
vocabulary or age differences) were indicative of low performance on the phonological
processing task. This is a common problem that is encountered when subgroups of
children are compared.
Leitao et al. (1997) compared typically developing, speech impaired, language
impaired, and speech and language impaired six year olds on several measures of
phonological processing. The authors noticed a range in the data, with a possible trend
for a bimodal distribution on phonological awareness tasks among six year olds with
speech impairment (i.e., possibly two separate subpopulations). They noted that children
who frequently used atypical sound changes performed more poorly on PA tasks than
those who frequently used typical sound changes. In a follow-up study, Leitao and
Fletcher (2004) examined two cohorts of children with SSD at age six, and followed
them prospectively until ages 12-13. They discovered that children in the group that used
more atypical sound changes when they were young (i.e., had 10% or more of their sound
changes classified as atypical) performed significantly more poorly on phonological
24
awareness and literacy measures at follow-up than children who had few atypical
phonemic sound changes (less than 10%). However, there were only seven children in
each group, making it difficult to generalize findings.
Among the studies most relevant to the current project is the work done by
Rvachew, Chiang, and Evans (2007). They made an attempt to elucidate the relationship
between PA and speech sound errors by analyzing children’s consonant errors on the
GFTA-2. The participants were 58 children with SSD ages four to five, divided into two
groups: those with and without PA problems. The groups were compared on the types of
speech sound errors they produced. Errors were classified as distortions, typical syllable
structure errors (e.g., final consonant deletion), typical segmental errors (e.g., /s/ � [t]),
atypical syllable structure errors (e.g., initial consonant deletion), and atypical segmental
errors (e.g., t � [k]). When in preschool, the only significant group difference was that
the children with PA problems produced more typical syllable structure changes. When
in kindergarten, the only significant difference was that children with PA problems
produced more atypical segmental errors.
A limitation in the existing research has been the attempt to categorize children
into discrete groups when, in fact, the variable(s) on which they were classified are
continuous. For example, to evaluate the relationship between PA and speech sound
errors, Rvachew et al. (2007) used a grouping variable to divide children according to
their score above vs. below a cut point (one standard deviation below the mean of a group
of control children) on a PA task. Dodd (1995) recommended using qualitative
judgments for grouping children based on the presence/absence of atypical errors (as well
25
as the consistency of those errors). Related to this, Leitao and Fletcher (2004) grouped
children based on percentage of atypical phonemic sound changes. This may have
resulted in assigning children who had more speech sound errors overall into the
‘atypical’ group (i.e., the children who had at least 10% of their sound changes defined as
‘atypical’ may have also used more typical sound changes, so the relative contribution of
atypical sound changes remains in question). Hence, it would be necessary to control for
the use of all other sound changes when examining the effects of atypical sound changes.
An analysis that predicts phonological processing from the relative occurrence of
different types of speech sound errors has the advantage of being able to examine the
separate influences of these errors, and it does not rely on grouping definitions to predict
variance.
Given the above descriptions of speech sound errors, Table 1 summarizes how
speech sound error types are thought to relate to underlying phonological representations.
Measurement System for Quantifying Sound Changes in the Present Study
In the current study, a summary of each child’s speech will include a score within
each of the following categories, determined through narrow phonetic transcription.
Appendix A provides definitions and examples of the types of sound changes and
examples to show how the sound changes are quantified.
26
Table 1: Summary of speech sound error types and their suspected reflection of underlying phonological representations
Error Type Proposed Reflection of
Phonological Representations
Proposed Statistical Relationship with
Phonological Processing Distortions
Relatively accurate, because
phonemically correct. Closest to the
adult form.
Weakest
Typical Sound
Changes
Moderately accurate; phonetically
motivated and found in the speech of
many typically developing children
Moderate
Atypical Sound
Changes
Poorly represented; uncommon and
relatively far from the adult form;
not phonetically plausible
Strongest
1) Distortions Per Consonant: The number of consonants distorted divided by
the total number of consonants attempted. Sound changes that are dialectally acceptable
(e.g., partial devoicing of voiced final consonants) are not considered errors.
2) Typical Sound Changes Per Consonant: The number of typical sound
changes divided by the total number of consonants attempted (an adaptation of the
Process Density Index described by Edwards, 1992, and the Relative Influence on
Unintelligibility by Dodd & Iacano, 1989).
3) Atypical Sound Changes Per Consonant: The number of atypical sound
changes divided by the total number or consonants attempted (based on the Relative
Influence on Unintelligibility by Dodd & Iacano, 1989). Whenever possible, these
atypical sound changes are identified based on previous research; they are outlined in
Appendix A.
27
The main advantages of this system are as follows. Both the types of sound
changes and their frequency can be specifically defined using this three-category system.
It should be noted that the current classification system also captures sound changes that
co-occur on the same phoneme (i.e., are ‘interacting’ or ‘overlapping’) (Edwards &
Shriberg, 1983). That is, if the word ‘cap’ /kæp/ is produced as [dæp], two sound
changes affect the initial phoneme (Velar Fronting [k �t] and Initial Voicing [t�d]).
Both of the constituent (component) changes of this error are counted in the present
analysis, whereas only one error would be counted using PCC.
In addition, it is important to note that a particular sound error may require coding
in more than one category. That is, a child’s production of a phoneme may be comprised
of more than one type of sound change. For example, if zipper /zǺpǪ/ is said as [sʝǺpǪ],
both an atypical error (devoicing of the /z/ to [s] in word-initial position) and a distortion
(dentalization) occur.
Speech Samples
Spontaneous (i.e., non-imitated) speech production samples are considered to
provide good evidence of what a child is independently capable of producing. Speech
samples taken from conversational speech, although useful for evaluating severity in a
clinical setting, would be inadequate for purposes of this study. This is because such
samples may fail to elicit a variety of syllable structures and phonemes, may be
confounded by morphosyntactic and pragmatic elements, and inherently provide different
samples from different children (Campbell & Shriberg, 1982; Paul & Shriberg, 1982).
Thus, a picture naming task that controls the speech sounds and word structures sampled
28
would be the most representative and equivalent across children. Additionally, a naming
task minimizes the complications associated with glossing a child’s conversational
speech (i.e., determining what the child intended to say), which may be difficult if the
child is hard to understand.
Because sound changes may affect both syllable/word structure (e.g., Final
Consonant Deletion, Consonant Cluster Reduction) and individual phoneme production
(e.g., Velar Fronting, Stopping), extensive samples containing a variety of syllable
structures and phonemes in different word positions are needed. Larivee and Catts
(1999) reported that the production of multisyllabic words is more sensitive to the
prediction of reading than is the production of single-syllable words, so a complete
sample would also include several multisyllabic words. These are not extensively
sampled in many standardized articulation tests. In addition, all consonants should be
sampled more than once across multiple word positions, to be certain that there are ample
opportunities for observing any of the child’s error patterns. Therefore, this study utilizes
a 125 item picture naming task adapted from earlier research (Wolk et al., 1993) to meet
the requirements outlined above.
Exploratory Analyses
Phonological awareness has been discussed as one component of phonological
processing. Two other areas that have received attention in the literature (and that are
also related to reading ability), as described earlier, are phonological memory and
phonological retrieval/rapid naming. Both of these skills are also thought to rely, in part,
on the accuracy of a child’s phonological representations. Although children with SSD
29
have been found to perform more poorly than typically developing peers on phonological
memory and rapid naming tasks, there is a significant lack of research addressing the
relationship between types of speech sound errors and performance on these tasks in
preschoolers. Therefore, exploratory analyses will address this issue. The resulting data
could aid in the interpretation of the main findings and could provide insights into
directions for future research.
First Exploratory Analysis: Phonological Memory
Phonological memory (PM) is the ability to retain phonological information in
short-term memory. It has been argued that this ability is essential for children who are
learning to read and spell (Brady, 1991; Metsala, 1999; Wagner & Torgesen, 1987). For
example, children who are attempting to sound out (decode) a printed word with which
they are unfamiliar often rehearse the sounds associated with the letters, either overtly or
covertly. Once they reach the end of the word, they must recall all of those sounds. In
fact, phonological memory has been found to be related to literacy skills, including
reading and spelling accuracy, and may be weak in poor readers (Elbro et al., 1998;
Griffiths & Snowling, 2002; Kamhi et al., 1988; Wagner & Torgesen, 1987).
While phonological memory skills have been discussed as being related to
phonological representations (Metsala, 1999), phonological memory tasks are not
intended to draw upon stored phonological representations of words. Instead, they rely
on temporary retention of phonological information. Similar to other domains of
phonological processing, phonological memory has been found to be related to a child’s
age and receptive vocabulary skills (e.g., Edwards et al., 2004; Metsala, 1999; Munson et
30
al., 2005).
Two common ways of assessing phonological memory involve repetition of
numbers and repetition of nonsense words (nonwords). Number repetition assesses an
individual’s ability to immediately recall sequences of random numbers (e.g., 7, 4, 9;
Elbro et al., 1998). This may not be appropriate for preschool children, as some children
may have significantly more familiarity with numerical concepts than others, and because
a semantic component is involved in the task. Therefore, number repetition is not used in
this study.
Nonword repetition skills have been found to separate good and poor readers, and
to relate to literacy skills such as decoding of nonwords and spelling (e.g., Griffiths &
Snowling, 2002; Kamhi et al., 1988; Lewis et al., 2004). Performance on nonword
repetition tasks has been shown to be related to age (Metsala, 1999; Roy & Chiat, 2004),
as well as receptive vocabulary ability (Edwards et al., 2004; Metsala, 1999; Munson et
al., 2005). Because nonword repetition is more appropriate for preschoolers than number
repetition, the present study will utilizes a nonword repetition task.
Phonological Memory in Children with SSD. It has been argued that a child’s
ability to hold speech sound information in memory should be related to speech
production development (Brady, 1991; Locke & Scott, 1979) and, in fact, weaknesses in
phonological memory have been reported for children with SSD compared to their
typically-developing peers (e.g., Munson et al., 2005; Preston & Edwards, 2007).
Several limitations exist with nonword repetition tasks for children with SSD.
Nonword repetition requires the ability to recall phonological input, establish a temporary
31
representation, plan the motor movements of the articulators necessary for the sound
sequence, and execute those motor movements. That is, the ability to accurately repeat
nonwords not only requires the ability to recall phonemes, but also the ability to perform
complex motor movements, which may be influenced by adjacent phonemes (i.e.,
coarticulation effects). Hence, it is unclear which of these processes are disrupted in
children with SSD. Therefore, it might be beneficial to use a “purer” task to assess
phonological memory (i.e., one that simplifies motor demands and coarticulatory effects).
In such a task, children would be required to repeat simple syllables that are likely to be
within their repertoire of production abilities (e.g., /ma/, /da/, /ba/). Thus, a task is used
in this study that assesses phonological memory in children with SSD without some of
the complications associated with previous nonword repetition tasks (Shriberg et al.,
2006). This will help to determine whether the ability to remember speech sounds is
problematic for children with SSD. It is possible that difficulty recalling phonological
information could be related to the ability to form (store) phonological representations
such that children who have trouble retaining phonological information in short-term
memory (as evidenced by poor nonword repetition abilities) might be expected to have
trouble forming accurate long-term phonological representations. It is therefore
hypothesized that performance on the nonword repetition task will be related to the
accuracy of phonological representations, as indicated by the use of atypical speech
sound changes. That is, atypical sound changes will be more strongly related to nonword
repetition than will lower-level errors (i.e., distortions).
To date, only one investigation with adolescents has investigated the potential
relationship between speech sound accuracy and phonological memory. Preston and
32
Edwards (2007) found a significant relationship (r = 0.65) between percent of consonant
errors and nonword repetition. However, this relationship was found in adolescents, not
preschoolers, and the speech error analysis and phonological memory task differed from
the current study. The present study will examine the contributions of different types of
speech sound errors to variance in phonological memory in preschoolers with SSD.
Second Exploratory Analysis: Rapid Naming
Phonological retrieval is often assessed using rapid naming tasks. Rapid naming
(RN) tasks require children to name a series of pictures/objects/ letters/ numbers as
rapidly as possible. These tasks are frequently used to assess the ability to retrieve
phonological information quickly. RN has been found to predict literacy skills, both
concurrently and longitudinally (Allor, 2002; Catts et al., 2001; Kirby et al., 2003;
Schatschneider et al., 2004). RN tasks have also been reported to separate good and poor
readers (Denckla & Rudel, 1976), and to separate children at higher risk of reading
problems from those at lower risk (Cardoso-Martins & Pennington, 2004). Similar to
other domains of phonological processing, age is also a significant predictor of
performance on RN tasks (Troia et al., 1996). However, RN and vocabulary tend not to
be highly correlated, as reported in a recent meta-analysis of school-age children (r =
0.26) (Swanson et al., 2003).
It has been argued that slow performance on naming tasks is due to poor
phonological skills (Denckla & Rudel, 1976; Kirby et al., 2003; Raitano et al., 2004;
Stringer et al., 2004; Swan & Goswami, 1997b; Troia et al., 1996). That is, when children
are slow to name pictures, objects, numbers, or letters, the deficit may be because of poor
33
access to the phonological features of the word. Hence, the ability to quickly access
phonological representations could impact the speed of naming. However, it is still
unclear whether naming speed is a function of poorly stored representations and/or poor
retrieval of phonological information.
Debate exists as to whether RN should be considered a phonologically-based task
(Wolf & Bowers, 1999), but its predictive value in literacy development is well
documented. Therefore, predicting performance on RN in preschool children could
provide some insight into processes that underlie literacy development. However, it
should be noted that RN tasks have not been frequently used with preschoolers, hence,
the exploratory nature of the RN component of the study.
Rapid Naming in Children with SSD. A small body of research suggests that
children with SSD may perform more slowly on RN tasks than their typically developing
peers (Leitao et al., 1997; Preston & Edwards, 2006). However, the research is limited
with this population, and the underlying reason for this difference in naming speed is
unclear. Recent research indicates that rapid naming of phonologically complex words
may be more challenging for adolescents with SSD than for their normally speaking
peers; however, no group difference was observed on naming of monosyllabic stimuli
(Preston & Edwards, 2006). Leitao et al. (1997) found that six year olds with SSD (as
well as those with language impairments) performed below typically developing peers on
several rapid naming tasks: letters, numbers, objects, and colors. Catts (1993) also
reported that, for children with speech and language impairment, rapid naming of animals
in kindergarten was moderately correlated with word reading in second grade.
34
Unfortunately, none of these studies examined specific speech production errors relative
to rapid naming, and none used this task with preschoolers.
One contradiction to the above evidence has been research reported by Raitano
et al. (2004), who found no difference between five to six year olds with SSD and control
participants on a rapid naming factor score which included naming of colors and objects.
However, if syllable length plays a role, as suggested by Preston and Edwards (2006),
then words of more than one syllable should be included in the stimuli. This limitation
will be addressed in the present study by using both a monosyllabic and a disyllabic RN
task.
Because Rapid Naming is thought to rely on rapid access to phonological
representations, it is hypothesized that the speech error measurement system based on the
presumed accuracy of phonological representations will significantly predict variance in
RN.
Primary Goals of the Study and Hypotheses
The importance of understanding how phonological processing skills vary in
children with SSD has been described. Because types of sound changes are presumed to
reflect the accuracy of phonological representations, it is possible that types of speech
sound changes can explain variance in phonological processing. This hypothesis will be
tested, and a new speech error classification system will be compared to the commonly
used Percent Consonants Correct (PCC).
Thus, while previous studies have examined how the frequency of speech errors
relates to phonological awareness using PCC, this study is unique because it evaluates
35
both the frequency and the types of sound changes that are involved in children’s speech
errors. The following hypotheses are investigated in the current study:
Summary
To reiterate, phonological processing has been defined to include phonological
awareness, phonological memory, and rapid naming. Phonological processing skills are
Hypothesis 1: Phonological awareness (PA) will be related to (correlated with)
speech sound error types in preschoolers with SSD, according to the proposed
accuracy of phonological representations.
Hypothesis 2: Types of speech sound errors thought to reflect weak phonological
representations will predict variance in PA above and beyond receptive vocabulary
and age in preschoolers with SSD.
Hypothesis 3: An analysis that characterizes sound changes according to the relative
accuracy of phonological representations will provide a better explanation of the
variance in PA than an analysis that considers all consonant errors to be equal (PCC).
(Exploratory) Hypothesis 4: A speech production analysis that considers three types
of sound changes will predict variance in phonological memory beyond the
contribution of age and receptive vocabulary.
(Exploratory) Hypothesis 5: A speech production analysis that considers three types
of sound changes will predict variance in rapid naming beyond the contribution of age
and receptive vocabulary.
36
important in predicting literacy development. The goal of the present investigation is to
determine whether speech sound accuracy can predict concurrent performance on
phonological processing tasks in children with SSD. Two procedures for analyzing
speech will be compared: (1) Percent Consonants Correct (PCC), and (2) an analysis that
represents both the frequency and type of speech sound errors.
37
II : METHODS
This study was approved by the Institutional Review Board at Syracuse
University. General results (e.g., test scores) were made available to the parents of
children who participated at the end of each session. Children were given books for their
participation, and parents were financially compensated for their time.
Participants
Children sought for the study were preschoolers, ages four to five years, with
speech sound disorders (SSD) of unknown cause (i.e., functional or idiopathic). No
attempt was made to include or exclude children based on the type of SSD (articulation
or phonological disorder, suspected childhood apraxia of speech, deviant or delayed
speech sound production, etc.) because of the lack of agreed-upon criteria for such
diagnoses. Children who were eligible for the study met the following criteria (described
below in more detail):
1. Diagnosed by a speech-language pathologist with a SSD (articulation/
phonological disorder, suspected childhood apraxia of speech)
2. Primary language and dialect was General American English
3. Had no known developmental, neurological, or oral structural difficulties (such as
mental retardation, cerebral palsy, pervasive developmental disorder/autism, cleft palate,
permanent hearing loss, etc.) that might cause the SSD. A history of ear infections was
acceptable.
4. Four or five years old and had not yet begun kindergarten.
5. Did not have a moderate or severe receptive language delay. Mild receptive
38
language delay was acceptable. Children were not excluded from the study for
expressive language concerns.
Recruiting
The primary method of recruiting relied on referrals from speech-language
clinicians in Upstate New York. These professionals were contacted via email addresses
that were publicly available on the internet (American Speech-Language-Hearing
Association, New York State Speech-Language-Hearing Association, local agency web
sites), presentations to local agencies, personal contacts, advertisements in professional
newsletters, and direct mailings to agencies and preschools. A description about the
study went out to these professionals, indicating the need for children who met the
criteria listed above. Flyers were made available to these clinicians to pass along to
parents of children who might qualify. In addition to clinical referrals, announcements
were made available to the public in newspapers, the Syracuse University SUNews list
serve, the Gebbie Speech-Language-Hearing Clinic, the Gebbie Clinic web site, and
posters in local preschools. Parents then directly contacted the researcher if they were
interested in obtaining further information.
Parent Phone Interview
Once the parent contacted the researcher about the study, a phone interview was
conducted to confirm that the child was of the appropriate age and that the child had
difficulty with speech sound production. Most of the children (all but two) were in
39
speech therapy2. Once the study procedures were explained to the parent, they were
asked if they wished to participate. All parents of children who met the criteria indicated
that they wished to participate, so a screening session was scheduled (n = 53). Two of
these parents contacted the researcher and scheduled the Part I Screening session, but
cancelled the session and did not reschedule. Additionally, two parents contacted the
researcher about participating, but the children were not of the appropriate age for the
study, so they were not included.
All parents reported that their child had no known permanent hearing loss or
developmental disabilities that might cause a SSD (such as cleft palate, autism, cerebral
palsy). Parents also confirmed that none of the children were exposed to a
parent/guardian who spoke a language other than English at home, and all parents
reported that the adults in the home were speakers of General American English. This
was also informally confirmed in the home visit.
Part I: Screening
A screening was first conducted to determine eligibility for the study. Informed
consent was obtained from one parent prior to the screening, and children provided oral
assent for participation. Screenings took place either at the child’s home (n = 49) or at a
quiet room used for child research at Syracuse University (n = 2), based on parents’
preferences. Parents were allowed to observe if they wished.
2 One child (P47) was not in speech-language therapy but the parent expressed concerns about the child’s articulation. The child was seen for the study but was later excluded because of high articulation score. A second child (P40) was not in therapy, although parents indicated that he did qualify for services. He achieved low speech sound production scores on the GFTA-2 and was included in the study.
40
A case history form was completed by parents during this session. Parents
provided additional detail about the child’s developmental history (medical, social,
educational, speech/language) and family background. Socioeconomic status (SES),
collected for descriptive purposes, was measured by the number of years of parental
education, similar to other studies. This variable has been found to relate both to the
prevalence of SSD and to PA skills (cf. Campbell et al., 2003; cf. Catts et al., 2001;
Nittrouer & Burton, 2005).
The entire screening protocol for Part I was pilot tested with two typically
developing preschoolers, and portions of the protocol were pilot tested with two other
children. This was done to obtain a time estimate of the length of the sessions, and to
familiarize the examiner with the administration procedures for the tests. It took between
40-65 minutes to administer all the screening tasks for Part I.
Task Order for Part I
Four tasks were randomly ordered: Goldman-Fristoe Test of Articulation-2
(GFTA-2) (Goldman & Fristoe, 2000); Concepts and Following Directions subtest and
the Sentence Structure subtest of the Clinical Evaluation of Language Fundamentals:
Preschool-2 (CELF:P-2) (Wiig et al., 2004); Peabody Picture Vocabulary Test-4 (PPVT-
4) (Dunn & Dunn, 2007); Pattern Construction subtest of the Differential Ability Scales
(DAS) (Elliott, 1990). An oral mechanism screening, devised for this study, was either
the fourth or fifth task. This was late in the session so that rapport had been established,
in case some children might find it embarrassing to make movements with their mouth.
41
In addition, fatigue should have very little effect on the pass/fail outcome of the oral
mechanism screening. Participants were offered a short break after two or three tasks.
Speech Sound Production
The Sounds-in-Words subtest of the GFTA-2 (Goldman & Fristoe, 2000) was
chosen to screen speech because it is a commonly used test for the clinical diagnosis of
speech sound disorders in preschoolers. This requires children to name pictures on 34
plates eliciting 53 target words, with online judgment of the accuracy of production of 61
consonants in initial, medial, and final position and consonant clusters. To qualify for the
study, children had to achieve a standard score below 90 on this test. This task was audio
recorded following procedures described below, but the task was scored online (live) in
order to determine eligibility at the time of the screening. Audio recordings were
consulted only if the child’s score was 80 or above because the misidentification of a few
errors would impact eligibility. According to the manual, the median test-retest
reliability for phonemes in the initial, medial, and final positions of words is 98%
agreement. Median inter-rater agreement for the presence of errors is 93% in the initial
position, and 90% in the medial and final positions. The alpha reliabilities for the age
groups in this study range from 0.94-0.96. Approximately 10 communities from Upstate
New York are represented in the standardization sample.
An informal oral peripheral screening was also used to confirm that there were no
gross structural or functional problems contributing to the SSD. This involved having the
child imitate the examiner’s mouth movements: close lips, purse lips, smile, elevate
tongue, protrude tongue, and lateralize tongue. Oral structures were also observed for
42
abnormalities (teeth, hard and soft palate, lips, face). All participants demonstrated
adequate structural/functional integrity of the oral peripheral mechanism.
Language
Because the children would be required to participate in phonological processing
tasks, participants were required to demonstrate adequate receptive language skills. This
was operationally defined as achieving scores not lower than one and one-third SD below
the mean on at least two of three receptive language tasks: the Peabody Picture
Vocabulary Test-IV (PPVT-4), the Concepts and Following Directions subtest of the
Clinical Evaluation of Language Fundamentals: Preschool-2 (CELF:P-2), or the Sentence
Structure subtest of the CELF:P-2. This was believed to be a reasonable means of not
excluding children who might have subtle receptive language difficulties, but who would
still be likely to follow directions and understand vocabulary well enough to participate
in research tasks. Expressive language was not formally evaluated, because none of the
experimental tasks required more than single word responses, and because the theoretical
justification for the study did not rely on a child’s expressive language skills.
Two subtests of the CELF: P-2 (Wiig et al., 2004) were used to screen receptive
language skills. The Concepts and Following Directions subtest requires children to
follow verbal directions by pointing to pictures of animals, usually in a specified order.
For example, “Point to the big dog, then point to the little monkey.” Items increase in
length and complexity. There are 22 items, and testing is discontinued after five
consecutive errors. The manual indicates that test-retest correlation is 0.83 for 4 year
43
olds and 0.88 for 5 year olds. Coefficient alpha for children in the age range seen here
were 0.78-0.85, and split-half reliability is reported to be 0.87-0.94.
The CELF: P-2 Sentence Structure subtest requires children to point to a colored
picture (from a field of 4) that accurately depicts a scene corresponding to the examiner’s
description. For example, “Point to The girl who is standing in the front of the line is
wearing a backpack,” and the distracter pictures typically show slight variations, such as
a girl in the back of the line with a backpack, the second girl in line wearing a backpack,
etc. There are 22 items, and testing is discontinued after five consecutive errors. Test-
retest correlation is reported to be 0.85 for 4 year olds and 0.79 for 5 year olds.
Coefficient alphas for children in the age range seen here were 0.78-0.83, and split-half
reliability is reported to be 0.81-0.85.
The PPVT-4 (Dunn & Dunn, 2007) measures single word receptive vocabulary by
requiring children to point to a colored picture (from a field of four) that corresponds
with the single word spoken by the examiner. Items increase in complexity, and the
testing continues until a ceiling is reached. Earlier versions of this instrument have been
used in several studies to estimate receptive vocabulary skills in children with SSD
(Rvachew, 2006; Rvachew & Grawburg, 2006). The newest version of this instrument
was updated, in part, to improve reliability in preschoolers. Test-retest reliability for the
age groups in this study range from r of 0.91 - 0.94. Split-half reliability for the children
ages 4;0-5;6 range from 0.94 - 0.96, and coefficient alphas are 0.96 - 0.97. About 10
facilities from upstate New York are represented in the standardization sample.
44
Nonverbal Cognition
The Pattern Construction subtest of the Differential Ability Scales (DAS, Elliott,
1990) was used as a brief screening of nonverbal intelligence. Children are shown
pictures of patterns of yellow and black squares. They then try to manipulate and arrange
the blocks to replicate the patterns shown in the picture. Both speed and accuracy of the
pattern construction are considered in scoring. Of the nonverbal subtests in the DAS, the
Pattern Construction subtest was chosen because it is a relatively efficient means of
estimating nonverbal cognition (i.e., it can be scored live, and it has the highest
correlation of all nonverbal subtests of the DAS with the Nonverbal Ability Composite).
Children were included if they achieved a T score above 37. Test-retest correlations for
the ages in this study are r = 0.62 - 0.73, and internal reliability is 0.82-0.90.
Table 2 shows a summary of the tasks from Part I, along with the criteria for
inclusion in the study.
Participants Included in Part II
Fifty-one children participated in Part I (screening), and the 44 who met the
criteria described above were invited to participate in Part II. Because one parent
scheduled and then canceled the Part II session, a total of 43 children participated in the
experimental tasks. Time between Part I and Part II ranged from 0-27 days, with an
average of 10 days between sessions. Table 3 summarizes the performance on the Part I
tasks for the 43 children who participated in Part II. The seven who did not qualify are
excluded, as is the one who chose not to participate (see also Figure 2).
45
Table 2: Inclusionary criteria for the study
Speech • Diagnosed with a speech sound disorder
• Standard Score of <90 on GFTA-2
• Exposed to General American English as the primary dialect, as
reported by parents and observed in the screening
• Speech disorder not a result of permanent hearing loss or
developmental disability, as reported by parents
• No obvious oral structural or functional problems
Receptive
Language
(met at least two
of three criteria)
• PPVT-4 Standard Score >80
• CELF:P-2 Sentence Structure Scaled Score >6
• CELF:P-2 Concepts & Following Directions Scaled Score >6
Nonverbal
Cognition
• Differential Ability Scales: Pattern Construction subtest
T score >37
Other • No known developmental disabilities, as reported by the parent
The 43 participants in Part II included 34 males and 9 females, a 3.78:1 gender
ratio. This ratio is not statistically different than the 2.75:1 male: female ratio reported
for children with SSD by Shriberg (1994) (χ2 [1] =0.724, p = 0.395). All participants
were Caucasian except for one female who was adopted from Asia. The average reported
maternal education level was 16 years of formal schooling, or the equivalent of four years
of college. The average paternal education level was 15 years of formal schooling, or
about three years of college. It is evident from Table 3 that some of the participants had
46
relatively high vocabulary skills compared to the standardization sample of the PPVT-4,
as well as relatively high nonverbal cognition, as measured by the DAS Pattern
Construction subtest. Both the PPVT-4 and DAS Pattern Construction subtest were
significantly above the expected mean based on one-sample t-tests (p’s <0.01). Possible
explanations for this include referral bias, as this project relied on SLPs to distribute
information about the study, and self-selection bias, with families from higher
socioeconomic homes perhaps being more likely to participate.
Table 3: Descriptive statistics for the 43 preschoolers who participated in Part II and were included in the final analysis
Mean SD Range
Age at Part II In months 54.7 5.4 48-69
Standard Score (mean 100, SD 15) 71.1 11.7 49-89 GFTA-2 Sounds in
Words Subtest Percentile 8.3 5.6 0.5-23
T score (mean 50, SD 10) 57.2 7.8 43-70 DAS Pattern
Construction Percentile 71.4 23.1 24-98
Sentence Structure Scaled Score (mean 10, SD 3)
10.9 2.4 6-15
CELF:P-2 Concepts & Following Directions Scaled Score
10.5 2.5 4-15
Standard Score (mean 100, SD 15) 112.4 12.3 84-145 PPVT-4
Percentile 73.8 21.6 14-99
Mother 16.0 2.3 12-21 Years of Parental
Education Father 15.3 2.9 9-22
47
Figure 2: Flow chart of procedures with number of participants
Qualified through Phone Interview & Scheduled Part I Screening (n=53)
Participated in Part I Screening (n=51)
• GFTA-2
• CELF:P-2 Sentence Structure
• CELF:P-2 Concepts & Following
Directions
• PPVT-IV
• DAS Pattern Construction subtest
• Oral –Peripheral Exam
Did Not Participate in Part II (n=8)
• GFTA-2 > 90 (n=4)
• Did not complete one or
more tasks
(noncompliant; n=2)
• Language and nonverbal
cognitive scores too low
(n=1)
• Qualified but cancelled
Part II (n=1)
Participated in Part II Experimental
Tasks (n=43)
• Hearing Screening
• Four Phonological Awareness
Tasks
• Picture Naming Task
• Syllable Repetition Task
• Rapid Naming Tasks
Did Not Participate in Part I Screening (session cancelled, n=2)
48
Part II: Experimental Tasks
Part II was conducted at Syracuse University for nine of 43 children; the
remainder were seen at their homes. Part II took between 75-125 minutes, and was split
into two sessions if the child showed significant signs of fatigue or was distracted.
Children were offered frequent breaks throughout Part II. Task order was pseudo-
randomized, with tasks being administered in the following way:
1. Hearing Screening
2. Introduce PA pictures: naming of/familiarization with 96 target words
3. Randomly chosen PA task
4. Randomly chosen PA task
5. Picture Naming task (for speech sample)
6. Randomly chosen PA task
7. Randomly chosen PA task
8. Rapid Naming or Syllable Repetition Task
9. Rapid Naming or Syllable Repetition Task
This order was chosen because it was essential that children be familiarized with
the phonological awareness (PA) task pictures before being exposed to them in the
experimental tasks. All four PA tasks required use of the laptop and were similar in
format (e.g., nonverbal response to stimuli); so these four tasks were split into groups of
two, with the picture naming task between. Because the final two tasks were exploratory,
they were completed at the end in case there was insufficient time to complete them.
Only one child failed to complete one of the exploratory tasks.
49
Hearing
Hearing was screened using a portable MAICO MA 27 audiometer. Behavioral
responses were required (i.e., raising the hand when the tone is presented). Following a
training/familiarization at about 65 dB, pure tones were presented at 20 dBSPL at 1000,
2000, and 4000 Hz (ASHA Audiologic Assessment Panel 1996, 1997). If tested at home,
failure to respond at 20 dB was followed by presentation of the same frequency at 25 dB,
and a response at this level was accepted as a pass due to presumed ambient noise levels
in the home. Forty-one participants passed the screening. One participant (P38) was not
screened because the audiometer was not available. One participant (P46) passed in the
left ear but did not pass the screening in the right ear (right ear threshold of 30 dB at 1000
Hz, passed at 25 dB at 2000 Hz, threshold of 35 dB at 4000 Hz.). He was kept in the
study because there was no history of permanent hearing loss, and he did not appear to be
an outlier in the dataset. Because this participant had a cold at the time of testing, failure
to respond may have been due to otitis media. Care was taken so that all recorded stimuli
were presented to this participant at a loudness level that he indicated was adequate. All
analyses were repeated without this participant in the dataset, and the conclusions were
unchanged.
Speech Assessment
Recording Procedure
All tasks requiring verbal responses (Picture Naming Task, Syllable Repetition
Task, Rapid Naming Tasks) were audio recorded. Two digital recorders were used, so as
to have a backup recording if one device failed: (a) Zoom H4 Handy Recorder with two
50
studio quality X/Y pattern condenser microphones set to record as digital WAV files at
24-bit quantization and 48 kHz sampling rate; (b) Olympus WS-331M digital voice
recorder with built-in stereo microphone, recorded on extra-high-quality stereo mode
with no low-cut filter. This device saved as Windows Media Audio (WMA) sound files
with a 44.1 kHz sampling rate. For later review of audio files, the WMA files were
converted to WAV files so that they could be reviewed in the Praat (reference?) acoustic
analysis software program. The clearest of the two recordings (usually the Olympus
device) was used for transcription/analysis. For one participant (P35), the digital audio
equipment was not brought to Part II; therefore, a cassette recording was made and this
was later digitized.
Speech Sample
A 125 word picture naming task (PNT) adapted from Wolk, Edwards and Conture
(1993) was used to assess all consonants in nearly every position in which they occur in
words (initial, medial, and final). All vowels of General American English were included
at least twice, as well as numerous consonant clusters/blends and multisyllabic words
(see Appendix C). The entire sample consisted of 480 consonants, although this total was
adjusted when necessary (e.g., if the child did not produce a particular word). Scripted
prompts were used to elicit the target word if the child mislabeled a picture. For
example, for the target splash, some children said, “Jumping into the pool,” so the
examiner said, “He jumped into the pool and it made a big ____”. If a child failed to
respond with the target word after several attempts at eliciting it, a delayed imitative
response was allowed. That is, a model was provided by the examiner, followed by a
51
comment, then the child was again prompted to produced the word (e.g., “He made a big
splash. See? There’s water going everywhere. He made a big ___.”)
For half of the children, the PNT was administered in order from item 1 to item
125. For the other half of the children, the PNT was administered in reverse (i.e., from
item 125 to item 1).
The picture naming task was piloted with two typically developing children, ages
four and five, and one seven year-old with a SSD. To the extent possible, pictures that
were mislabeled by these children were replaced with newer or more explicit pictures to
elicit the target words.
Transcription
Children’s responses on the picture naming task were narrowly phonetically
transcribed by the author. Praat software was used to play the digital files in free-field in
a quiet room. Time between the initial assessment and the first phonetic transcriptions
varied from one day to approximately 4 months, depending on the participant. To ensure
accuracy of the transcriptions, audio files were reviewed by the author a minimum of
three times for each participant. Transcriptions were entered directly into the Logical
International Phonetic Programs software (LIPP, Oller & Delgado, 2001). For detailed
phonetic variations, the author used the diacritics in this software program, and
supplemented with the use of a nonspecific diacritic for clinical distortions (e.g.,
derhoticized /r/, lateralized /s/). Hence, any phoneme that had this “distortion” diacritic
was counted as incorrect using PCC, and classified as a distortion using the three-
category system devised for this study. The transcriber who completed reliability
52
listened to the sound files using AKG K 240 headphones, and wrote out her detailed
transcriptions rather than using LIPP (reliability details are provided later).
If the child spoke a word more than once, the clearest recording of the two
renditions was used; if both were clear, the first was chosen. When there was overlay
with another speaker or there was background noise covering a portion of the word, the
child was given credit for producing those overlaid sounds correctly. If a child added
morphological endings, those were not analyzed (e.g., if a child said “toys” instead of
“toy,” the plural was not scored). Further detail regarding transcription rules and
procedures is included in Appendix A.
Types of Speech Errors
Using these transcriptions, two consonant analysis schemes were compared to see
if either was better able to predict variance in phonological processing:
1) Percent Consonants Correct (PCC) was calculated from the picture naming
task, with all consonant errors being weighted the same (i.e., substitutions, omissions, and
distortions). Each consonant was therefore judged to be correct or incorrect.
2) Three types of speech sound changes: Distortions per consonant, Typical
Sound Changes per consonant, and Atypical Sound Changes per consonant were
calculated from the narrow transcription of the child’s productions on the picture naming
task.
Note that speech errors for both analyses were computed by hand, rather than by
computer, to allow for dialectal variations (e.g., partial devoicing, affrication of /tw, dw,
tr, dr/ clusters, glottal stop replacement for final /t/, etc.) and for interacting sound
53
changes. This is because the LIPP program is limited in its ability to accurately code
some of these sound changes (White, 1997). Initial coding of speech sound errors was
completed by the author at the same time of the transcription. However, because it was
necessary to refine some of the sound change definitions (Appendix A) as the study
progressed, each participant’s phonetic transcriptions were reviewed a minimum of three
times to ensure accuracy and consistency of error coding.
Typical and atypical sound changes were defined based on previous research.
Changes in place of articulation, manner of articulation, voicing, and syllable structure
that are commonly found in the speech sound development of children have been
generally well described (Edwards & Shriberg, 1983; Ingram, 1976; Khan, 1982). In
addition, there has been a moderate amount of discussion about what constitutes atypical
or unusual sound changes. However, some sound changes have not been discussed
adequately or the definitions are not fully agreed upon. For the present study, atypical
sound changes were defined based on prior research, to the extent possible, but some
definitions had to be refined to be sufficiently explicit (see Appendix A). A relatively
conservative approach to defining sound changes as atypical was used. When there was
lack of agreement in the literature, a general rule of phonetic plausibility was adopted.
Thus, if a consonant sound change occurred that was potentially due to phonetic context,
word position, or the influence of other consonants in the word, it was not considered
atypical. Appendix A and Appendix B provide further detail about the coding of sound
errors. To give a common example, velarization (or backing) of alveolar stops (e.g.,
d�g) has often been considered atypical (e.g., Dodd & Iacano, 1989) because typically
developing children generally replace back sounds with front sounds. Given the
54
definitions developed for this study, this sound change would be considered atypical only
if it could not be accounted for by a typical sound change, such as velar assimilation.
Thus, /d/� [g] in the word “dinosaur” would be considered atypical because there are no
other velars in the word to trigger this change. However, if /d/� [g] occurred in the word
“pudding,” it would be accounted for by the typical error of velar assimilation. (That is,
/d/ assimilates to the velar feature of the /ŋ/.)
Phonological Awareness
While some PA tasks require spoken responses, this may confound results when
assessing PA in children whose speech is often hard to understand (Sutherland & Gillon,
2005). Therefore, PA tasks that were selected for this study met the following criteria:
(1) no spoken response was required; (2) the task has been shown to be related to later
literacy development; and (3) the task was age-appropriate. PA assessment tasks and
protocols were therefore based on prior research (see below).
PA Stimuli Preparation and Presentation
Ninety-six words (that were different from the picture naming task) were selected
for use in the PA tasks. All 96 words were monosyllabic, and most were made up of
CVC syllables (e.g., dog), with a few being CV (e.g., shoe) or CCVC (e.g., spoon).
Words were chosen based on their phonological features (consonant and vowel
components) and picturability/interpretability by four year olds. Most words were nouns,
but there were two verbs (run, tap) and one adjective (red).
To limit the number of items with which the children had to be familiar, each
55
word was used either two or three times, but no word was used more than twice in a
given task, and never twice as the target response. For example, “coat” appeared once as
a distracter item in the Onset Matching task, once as a correct target in the Rhyme
Matching task, and once as a correct target in the Blending task. The stimuli for the four
PA tasks are listed in Appendix D.
Audio stimuli and instructions for the PA tasks were recorded by an adult male
(the author) using a Sure WH22 head mounted microphone fed into a Rolls MX 54s Pro
Mixer Plus in a double-walled soundproof booth. The signal was recorded at 44 kHz
sampling rate on a Dell Inspiron 8600 laptop in Praat v. 4.2.19. Stimuli were stored as
WAV files. They were presented to the children using the same computer, and were
imported into Microsoft Power Point. Audio stimuli were paired with visual stimuli,
which were clip art pictures taken from a variety of sources (e.g., Microsoft Word,
Google Images, and other internet sources). An external speaker was used to amplify the
audio signal in environments where the internal speakers of the laptop were judged to be
insufficient.
PA tasks were pilot tested with one typically developing four year old, two
typically developing five year olds, and a seven year old with a SSD. As with the picture
naming task, if some of the children had difficulty identifying the pictures, different
pictures were selected. Approximately five of the 96 pictures were replaced with newer
clipart in order to better represent the target words.
Familiarization
Before any of the PA tasks were administered, children were familiarized with
56
the 96 target words to be used in the experimental PA tasks. Children were shown the
pictures on the laptop. Instructions were, “I am going to show you some pictures on the
computer. Tell me the names of the pictures that you see.” The examiner controlled the
rate of presentation of the pictures (i.e., they were not time-controlled by the software).
If a child was unfamiliar with the picture or provided the wrong label, a spoken model
was provided, the child was asked to imitate the word, and then another model was
provided. For example, when shown a picture of hen, if a child said “rooster,” the correct
label was provided (e.g., “That’s a picture of a hen. Can I hear you try that word? Good.
That’s hen.).
General Procedure for PA Tasks
Children sat on the floor or at a table in front of the laptop. For each task, three
or four pictures appeared together on the computer screen. These were arranged in a
random configuration on the screen, so that the correct response picture was not
consistently in the same position. Because three PA tasks used a field of four choices
from which the child could select a response, the screen was divided into four quadrants.
For the Blending task, three picture choices were arranged in a row. Figure 3 (shown
after all tasks are described) provides examples of the visual layout for each of the PA
tasks.
Because all of the PA tasks required nonverbal responses to audio/visual stimuli,
children were given a 12-inch “magic wand” with a soft end that was used as a pointer.
They used this to lightly touch the computer screen to indicate their response. Some
children chose to provide a verbal response, but they were encouraged to point as well
57
because verbal responses could be unintelligible. The recorded audio stimuli were
played only once, unless the child failed to respond (e.g., if distracted) or requested
repetition. The examiner pointed to the pictures on the screen as they were named. If a
child changed his/her response, the final response was scored. There were five training
items for all of the PA tasks, with feedback and instruction provided if the child
responded incorrectly. All responses were noted online by the examiner.
The first three PA tasks described below were adapted from Bird et al. (1995).
These include rhyme matching, onset matching, and onset segmentation and matching.
All three tasks have been used with preschoolers with SSD to predict early literacy skills
(Rvachew, 2006; Rvachew & Grawburg, 2006). The tasks were adapted to be presented
with recorded audio stimuli and clip art pictures on a laptop in PowerPoint (instead of
using puppets, as in the original research). Additionally, target and distracter items were
modified to control phonological similarity of distracter items to the targets, as described
below. The stimuli used for all PA tasks are in Appendix D.
Rhyme Matching. The rhyme matching task included 16 experimental items,
with four blocks of four rhymes (i.e, four items that rhyme with the names Dan, Doug,
Pete, Ned). For each trial, four pictures appeared on the computer screen at once, the
correct picture and three “distracters.” Each block was introduced by presentation of a
photo of a person paired with audio recording. For example, “This is Dan. Dan likes
things that rhyme with his name. Help Dan find things that rhyme with his name.” The
name was repeated during each item: “Which one rhymes with Dan? spoon, cap, mouse,
pan. Which one rhymes with Dan?” (child points). For each item in the Rhyme
Matching, one of the distracters had the same vowel as the target (here, /æ/ in cap), one
58
had the same final consonant (here, /n/ in spoon), and one had no phonemes in common
with the target (here, mouse). A picture of the person whose name was to be rhymed
always appeared in the upper-left hand portion of the screen (here, a picture of Dan).
Five training items were provided with corrective feedback as necessary. Audio stimuli
for each trial were recorded, and the examiner controlled when each item was presented.
Onset Segmentation and Matching. A similar paradigm was used for the Onset
Segmentation and Matching task. When presented with a field of four pictures, children
were instructed to find a word that “begins like” a particular name. For example, “Which
one begins like Tom? Pin, juice, tie, door. Which one begins like Tom?” (child points).
Five training items were provided with corrective feedback. Prior to the training items,
the children were shown a slide with examples of correct responses, such as “Time and
turtle begin like Tom. Now let’s find some more.” One of the distracter items always
began with a phoneme that children frequently produce as a substitute for the target
phoneme. For example, all of the matching items for Tom included a correct target
beginning with /t/, but also a “foil” beginning with /d/ (e.g., door). There were five
experimental items that begin with /t/ (to match Tom), and five that begin with /s/ (to
match Sam).
Onset Matching. The Onset Matching task required children to find a word from
a field of four that began with a given sound. Unlike the Onset Segmentation and
Matching task where children had to determine the initial sound of a word before
matching it, in the Onset Matching task, children were given the phoneme they had to
listen for. For example, “Which one begins with /p/? Deer, kite, bug, pin. Which one
begins with /p/?” Five training items were provided (three with /r/, two with /m/) to
59
familiarize the child with the task. The experimental items included five where the child
had to choose a word beginning with /p/, and five beginning with /tȓ/ (ch). As with the
Onset Segmentation and Matching, one foil or distracter item in the Onset Matching
began with a phoneme that children often produce as a substitute for the target. Thus, all
/p/ matching items had a foil beginning with /b/, and all of the /tȓ/ (ch) items had a foil
beginning with /ȓ/ (sh). The remaining two distracters began with phonemes that are
less similar to the target (e.g., /d/ in deer differs from /p/ in both place and manner of
articulation).
Blending. To assess onset-rhyme and C-V-C phoneme blending (or synthesis), a
task was adapted from previous research (Larivee & Catts, 1999). Children were
presented with a set of three pictures on the computer screen (e.g., fan, fish, dish), and
listened to a recorded presentation of the target word spoken in segments (e.g., /f--Ǻ--ȓ/).
There was approximately 1.0 second between phonemes. The child pointed to the picture
to indicate a response. To introduce the task, children were shown a picture of a monster
and told, “This monster says things in a funny way. He says words in pieces. See if you
can guess what he is saying.” For each item, a carrier phrase spoken by a female (“Point
to the one that you hear”) preceded the segments, spoken by the monster (a male).
Twelve experimental items for the Blending task were presented in a game-like
format in PowerPoint. The first six items required onset-rhyme blending (i.e., initial
consonant [onset], then vowel-consonant pair [rhyme]). The last six required blending of
individual phonemes (consonant, then vowel, then consonant). All targets were CVC
words. Three training items with corrective feedback were presented before the six
60
Figure 3: Examples of PA stimuli
Rhyme: Which one rhymes with Dan? Cat, fan, run, bike. Which one rhymes with Dan?
Onset Segmentation & Matching: Which one begins like Tom? Pin, juice, tie, door? Which one begins like Tom?
61
Onset Matching: Which one begins with /p/? deer, kite, bug, pin. Which one begins with /p/?
Blending: Female: Point to the one that you hear:
Male: /m/m/m/m --- aaaaȚs/Țs/Țs/Țs/
62
onset-rhyme blending items (e.g., /f--Ǻȓ/). Two additional training items with corrective
feedback were used before the six C-V-C blending items (e.g., /f--Ǻ--ȓ/). Both distracter
words had phonological similarity to the target: one foil began with the same phoneme
as the target (e.g., fan begins like fish), and one foil had the same vowel and/or final
consonant as the target (e.g., dish has the same rhyme as fish).
Exploratory Analyses
Phonological Memory
Phonological memory is an additional phonological processing domain that is
discussed as being related to phonological representations, speech sound disorders, and
literacy. In this study, phonological memory was assessed by the Syllable Repetition
Task, which was developed for children with poor intelligibility (Shriberg et al., 2006).
Shriberg et al. (2006) report data from 99 children confirming that the four consonants
used in this task are in the phonetic inventories of children with SSD, that scores on the
syllable repetition task correlate moderately with other nonword repetition tasks, and that
the scores on this task met distributional requirements for parametric statistical analysis.
Stimuli (provided by the first author of the original work) were two to four syllables in
length, and were spoken by an adult female. For each item, the stimuli were produced
with no pause between syllables. Four early developing consonants (Shriberg et al.,
1994; Smit et al., 1990) were presented in combinations of CV syllables using the /a/
vowel (/ba, ma, da, na/). The stimuli therefore limited the articulatory demands of the
task. The children were directed to imitate the examiner’s productions of the individual
syllables prior to beginning the syllable repetition task to be certain that they could
63
produce the sounds. All participants were able to imitate the single syllables.
The recorded audio stimuli for this task were spoken by an adult female. They
were presented in free field via Power Point on a laptop, using external speakers to
amplify the signal if necessary. When the stimuli provided for this task were presented,
the letters (e.g., “bada”) appeared on the screen in conjunction with the auditory stimuli;
therefore, the children were instructed to turn away and/or close their eyes so they could
not see the laptop screen. As in Shriberg et al. (2006), items were repeated if the child
failed to respond or requested a repetition. The instructions were as follows: “You are
going to hear the computer speak some funny words. Just say exactly what you hear. If
the computer says /ba/, you say____. What if the computer says /da/? How about /na/?
What about /ma/? Good. Now listen to the lady on the computer say these silly words,
and say exactly what she says.” Similar to Shriberg et al. (2006), if the child failed to
respond within several seconds of the presentation of a stimulus, or requested a repetition
(e.g., “what?”), the stimulus was played again. Appendix E lists the stimulus items for
this task.
Audio recordings of each child’s productions were reviewed using Praat software
and were phonetically transcribed by the author. Scoring procedures followed those
outlined by Shriberg et al. (2006), and additional detail about scoring procedures was
provided by the first author of the original work. Each consonant was scored as correct
or incorrect, ignoring distortion errors. If the wrong number of syllables was produced,
the child’s productions were aligned with the target response to provide the highest score
for that item. For example, if the child produced [bada] for the target [mabada], the child
was given credit for two of three consonants in the word ([ba] and [da]). There were 50
64
total consonants in the stimuli, and percent consonants correct-revised (PCC-R, Shriberg
et al., 1997a) was calculated from this task for each child.
Phonological Retrieval/Rapid Naming
A third domain of phonological processing is phonological retrieval. As in other
studies, Rapid Naming (RN) tasks were used to assess retrieval of phonological forms.
Because there is reason to believe that speed of recall may be influenced by the number
of syllables (Preston & Edwards, 2006), two RN tasks were used: one with
monosyllables and one with disyllables. For both tasks, stimuli were presented as color
pictures on 8 ½ ” x 13” legal-sized paper. The pictures were arranged in five rows of six
pictures (30 pictures). The two tasks (monosyllables and disyllables) were presented
consecutively, but the order of the two RN tasks was randomly chosen for each
participant. Appendix F has the RN stimuli.
Children were first familiarized with the Rapid Naming paradigm and were
briefly trained. The children named each of the four pictures (fish, cat, ball, book) and
were given corrective feedback if they mislabeled them. Then, the children were shown
the training page of the four color pictures repeated five times (20 pictures for the
training). They were told, “We are going to have a race to see how fast you can talk.”
The instructions were, “You are going to name all of the pictures on this page as fast as
you can. Start at the top and go through each row until you come to the end. Watch me
do it first.” During this familiarization trial, the examiner first modeled by naming the
pictures quickly from left to right, then asked the child to do so. The examiner followed
along by pointing to keep the child on the correct picture and to continue in a left-to-right
65
fashion.
The instructions were repeated for the two experimental RN trials. The children
were first familiarized with the individual pictures, and corrective feedback was given if a
picture was mislabeled. The Monosyllabic RN task included colored pictures of the
items dog, chair, hat, boat, and fire. Each picture appeared six times, for a total of 30
pictures (adapted from Torgesen & Wagner, 1997). The order of the five pictures
differed in each of the six iterations.
The Disyllabic RN task involved rapid naming of colored pictures of five two-
syllable items: money, apple, finger, pencil, and table. Items were taken from the 3-4
year items from the PPVT-III and Expressive Vocabulary Test, as well as Carroll,
Davies, and Richman (1971). All were two-syllable words with a trochaic (strong-weak)
stress pattern. Each picture appeared six times, for a total of 30 pictures. The order of
the five pictures differed from one iteration to the next.
Digital sound files were used to score both of the Rapid Naming tasks. Acoustic
waveforms were marked by the author using Praat, timing from the beginning of energy
onset of the first word to the end of energy offset of the final word. If the child became
distracted or went off task during the rapid naming (e.g., made a comment, asked a
question, laughed), the duration of the off-task behavior was removed from the total
naming time by subtracting the time from the beginning of the off-task behavior to the
beginning of the naming of the next picture. This had to be done for five participants on
the monosyllable RN task and nine participants on the disyllable RN task. For statistical
analysis, the average z-score of the two RN tasks was used to summarize the construct of
phonological retrieval.
66
Reliability of Measures
Speech Production Reliability
Analyzing the data derived from the picture naming task involved a two-step
process. Each speech sample was narrowly phonetically transcribed; then the
transcriptions were reviewed and coded for errors according to the scheme developed for
this study (Appendix A). Therefore, reliability was obtained for both steps. A transcriber
with more than 30 years of experience with phonetic transcription of children’s speech
completed reliability for these speech production measures.
The first reliability measure evaluated the reliability of the error coding scheme.
The reliability judge reviewed the author’s narrow transcription of a randomly-selected
sample of at least 20 words from each participant. She used the error coding system to
classify each speech sound error (Appendix A). Word-by-word agreement was
computed, scoring “agree” if the initial rater and the reliability judge completely agreed
on the number of distortions, typical sound changes, and atypical sound changes in the
word. Disagreements were reviewed and were used to further refine definitions of error
patterns in Appendix A. Given the phonetic transcription of a child’s speech, the two
judges completely agreed on speech error coding of all of the sound changes in 834 of
903 words (92.4% of words; range 62-100% agreement on words from individual
participants). Following adjustments to the coding system, 41 of those words that the two
judges disagreed upon were independently coded a second time. Agreement was reached
on 83% (34/41) of these words on which the judges had disagreed.
67
For the second reliability measure, the reliability judge independently transcribed
25 consecutive words of the 125 word speech sample from the picture naming task (20%)
for 30 of the participants. This represents 14% of all words that were transcribed. The
starting point for the 25 consecutive words was randomly chosen for each participant.
The reliability judge then coded her transcriptions based on the definitions in Appendix
A. This was a “worst case scenario” measure because differences in phonetic
transcription could inherently result in different coding of speech sound changes. These
25 word samples ranged from 90-100 consonants, depending on the specific words
transcribed. For each 25 word sample, the number of distortion errors per consonant,
typical sound changes per consonant, and atypical sound changes per consonant was
computed. For these 30 participants, the mean (absolute) difference between the
reliability judge’s estimate and the original estimate for the 25 word sample was 2.7
atypical sound changes per consonant (SD 3.0, range 0-9.3), 3.6 typical sound changes
per consonant (SD 4.7, range 0-11.2), and 2.9 distortions per consonant (SD 2.8, range 0-
8.0). The concordance correlation coefficient3 was 0.73 for atypical errors, 0.94 for
typical errors, and 0.73 for distortions.
Syllable Repetition Task Reliability
For 15 participants, productions elicited on the syllable repetition task (SRT) were
independently transcribed by a trained research assistant, an undergraduate senior
majoring in Communication Sciences and Disorders who had taken a course in applied
phonetics, in which she learned phonetic transcription. Reliability was computed for 15
3 The concordance correlation coefficient (Lin, 1989) is similar to a Pearson’s r but it provides an estimate of the departure of two ratings from exact agreement (i.e., 45o line, or when both axes are an identical scale). Hence, it is a more conservative estimate of agreement than a Pearson’s r.
68
participants by making correct/incorrect judgments on each consonant produced and
computing a percent consonants correct (PCC) for the 50 consonants. SRT scores
obtained by the reliability judge were within +/- 6% of the original estimate for all
participants (mean difference 0.27%). There was no statistically significant difference
between the original score and the score obtained by the reliability judge (t = 0.33, p =
0.744), and the two scores were very highly correlated (r = 0.978, p< 0.001; concordance
correlation coefficient 0.978).
Rapid Naming Reliability
For 14 participants, the durations for each of the two Rapid Naming (RN) tasks
were independently re-timed by a trained research assistant using waveforms in Praat, as
described above. The duration estimates between the two judges were very highly
correlated for both the RN monosyllable task (r = 0.996; p < 0.001) and the RN disyllable
task (r = 1.00; p < 0.001). The mean difference between the two judges in timing the RN
monosyllable task was .04 sec (range of absolute differences 0.00 – 3.56 sec). The mean
difference between the two judges in timing the RN disyllable task was .01 sec (range of
absolute differences 0.00-0.65 sec). Paired t-tests revealed no statistically significant
differences in the durations measured by the original measurement and the reliability
judge for the RN monosyllable task (t = 0.11; p = 0.916) or the RN disyllable task (t =
0.14; p = 0.889). To ensure accuracy of the data, it was determined that a discrepancy in
duration estimate of greater than +/- 0.5 sec would prompt a re-timing of the RN task.
This was done for two participants on the RN monosyllable tasks and one participant on
the RN disyllable tasks. In all three cases, the source of disagreement involved
69
measuring the duration of off-task behavior. The re-timing always agreed with one of the
duration measures (the original or that of the reliability judge), so the retiming was used
in the final data analysis.
Data Analysis
Statistics were computed using SPSS version 15.0 (SPSS, 2006). A correlational
design was used to examine the concurrent relationship between measures of speech
sound accuracy and phonological processing in children with SSD. Hierarchical multiple
regression was used to evaluate the proportion of variance in phonological awareness that
could be explained by speech sound errors. For all regressions, an alpha level of 0.05
was used as a guide for statistical significance testing. The study was designed to be able
to predict variance in PA by detecting a change in R2 (or ∆R2) of about 0.10 with power
of approximately 0.80. See Appendix H for a discussion of observed power.
70
III : RESULTS
Summary of Speech Sound Production
A primary goal of the present study was to evaluate appropriate methods of
quantifying speech sound errors in children with SSD and to determine how those errors
relate to phonological awareness (PA) skills. Speech sound accuracy scores were based
on phonetic transcriptions of each child’s consonant productions from the 125-item
picture naming task. Percent Consonants Correct (PCC) was calculated for each child,
with any phonemic change (substitution or omission) or clinical distortion being
considered an error. Hence, each consonant was judged as correct or incorrect. The PCC
scores for children in this study are shown in Table 4. Although normative data are not
available for PCC in picture naming samples, they have been reported in connected
speech samples. The mean in the present study is significantly lower than data reported
elsewhere from connected speech samples in normally developing children and are near
values reported for conversational samples from children with SSD (Campbell et al.,
2007; Shriberg et al., 1997a; Shriberg & Kwiatkowski, 1982). The mean PCC is 4%
lower than that reported by Bird and Bishop (1995) in a picture naming task with children
with SSD who were, on average, 16 months older than the participants in this study.
Wolk (1990) reported PCC on a similar picture naming task for 14 phonologically
disordered children ages 4;2-5;11 (half of whom also stuttered); the mean PCC of the
present study is 5.8% below the mean reported in that study. Therefore, the PCC scores
appear to reflect a range of speech sound (in)accuracy and are consistent with values
expected for children with speech sound disorders.
71
Table 4: Summary of speech sound (in)accuracy for 43 preschoolers with SSD
Mean SD Range
Percent Consonants Correct (PCC) 48.45 11.44 16.29-69.17
Distortions per Consonant 0.047 .036 0.00-0.156
Typical Sound Changes per Cons. 0.453 0.128 0.236-0.819
Atypical Sound Changes per Cons. 0.073 0.044 0.015-0.249
From those same picture naming speech samples, all sound changes were also
analyzed based on the three-category system described earlier: distortions, typical sound
changes, and atypical sound changes. Many sound errors required more than one sound
change (i.e., interactions) to explain the child’s production (e.g., catch /kætȓ/ � [dætȓ]
requires both Velar Fronting and Initial Voicing to explain a single error /k/ � [d]).
Descriptive data are shown in Table 4. As expected, children produce significantly more
typical sound changes per consonant than atypical sound changes per consonant.
Distortions were produced relatively infrequently, as reported in other studies (Gruber,
1999). However, all children were found to produce at least some atypical sound changes.
Higher values on the Typical Sound Changes per Consonant, Atypical Sound
Changes per Consonant, and Distortions per Consonant indicate more errors and,
therefore, less accurate speech production, while higher PCC values are indicative of
greater speech sound accuracy (correct consonants). Therefore, one would expect PCC to
be (negatively) related to these error types.4
4 As described earlier, PCC is not simply a linear combination of the three error types, as PCC does not take into account the components/features of sound changes.
72
Table 5 reports correlations among the three categories of speech sound errors.
Similar correlation matrices are not available from other studies, as this is the first study
to numerically quantify all consonant errors according to this three-category system;
however, these values are not unexpected. Typical and atypical sound changes were
positively correlated (r = 0.344, p < 0.05), suggesting that children who have more
atypical sound changes also have more typical sound changes. Distortions were
negatively correlated with typical sound changes (r = -0.440, p <0.01). This is in accord
with literature on speech development that has suggested that children may progress from
making phonemic errors (substitutions and omissions) to distortion errors as their speech
sound accuracy improves (Gruber, 1999). Hence, more distortions are associated with
fewer typical sound changes.
Table 5: Pearson’s correlation coefficients (r) of speech sound error types
PCC
Typical Sound Changes Per Consonant
Distortions Per
Consonant Typical Changes Per Consonant
-0.924(**)
Distortions Per Consonant
0.302(*) -0.440(**)
Atypical Changes Per Consonant
-0.600(**) 0.344(*) -0.183
** Correlation is significant at the 0.01 level (2-tailed). * Correlation is significant at the 0.05 level (2-tailed).
Summary of Phonological Awareness
Table 6 summarizes the group performance of the preschoolers with SSD on the
phonological processing tasks, which includes the four PA tasks: Rhyme Matching,
73
Onset Matching, Onset Segmentation and Matching, and Blending. As expected, there
was a broad range in performance on the PA tasks among these 43 children with SSD.
Therefore, there is interest in explaining this variability of PA skills.
Table 6: Summary of the performance of 43 children on the phonological processing tasks
Task Mean SD Range
Rhyme Matching (out of 16 items) 6.8 3.4 2-14
Onset Matching (out of 10 items) 4.5 2.6 0-10
Onset Segmentation & Matching (out of 10 items) 3.5 2.2 0-10
Blending (out of 12 items) 7.0 2.6 2-12
Syllable Repetition (PCC-R) 69.5 15.5 24-94
Monosyllable Rapid Naming (in seconds )* 42.0 14.0 25.6-79.7
Disyllable Rapid Naming (in seconds) 49.1 16.5 24.6-112.5
*One participant did not complete the Monosyllable Rapid Naming
There was no evidence of floor or ceiling effects, indicating the appropriateness
of these tasks for detecting differences in PA skills. This provides support for the use of
these tasks with this age group, and indicates that they may be sensitive to differences in
PA skills. The means and standard deviations are generally in agreement (within +/- 1
items correct) with those reported in other studies that used similar tasks with 4 to 6 year
olds with SSD (Bird et al., 1995; Rvachew & Grawburg, 2006). All variables were
normally distributed based on Kolmogorov-Smirnov tests for normality (all p’s > 0.15)
and visual inspection of histograms.
74
As shown in Table 7 and as anticipated from other studies, significant positive
correlations were found among the phonological awareness variables. That is, children
who performed relatively well on a given PA task were likely to perform relatively well
on the other tasks. A more complete correlation matrix is available in Appendix G.
Table 7: Pearson correlation coefficients (r) for the phonological awareness tasks for 43 children with speech sound disorders
Onset Matching
Onset Segmentation & Matching
Blending
Rhyme .621(**) .508(**) .356(*)
Onset Matching .637(**) .401(**)
Onset Segmentation & Matching .490(**)
** Correlation is significant at the 0.01 level (2-tailed). * Correlation is significant at the 0.05 level (2-tailed)
A Phonological Awareness (PA) composite score was calculated by using a
Principal Component Analysis to summarize the four PA tasks. This is a multivariate
technique used to derive a linear combination of several variables while retaining the
maximum possible variance. Each child therefore ends up with a single composite score
for PA (with a mean of 0 and SD of 1). For these data, the principal component derived
from the four PA tasks retained 63% of the variance of the tasks. The factor
loading/Pearson’s correlation coefficient of each PA task with the overall phonological
awareness principal component is summarized in Table 8, along with the communality
(or the proportion of variance of a variable that is retained in the principal component).
75
It can be seen in Table 7 and Table 8 that the Blending task has the lowest
correlations with the other three PA tasks, and also has the lowest correlation with the
overall PA composite. Blending, therefore, may require somewhat different skills or
have demands that differ from the other tasks (e.g., memory and synthesis, see
Discussion). It is also possible that performance on this task was more variable because
it could be more highly influenced by guessing; that is, each item had only three picture
choices, compared to the other PA tasks (Rhyme, Onset Matching, Onset Segmentation
& Matching) which had four. Hence, 33% correct on the Blending task was equivalent to
random guessing, whereas 25% correct on the other tasks was equivalent to random
guessing (see Appendix H: Measurement Issues for further discussion).
Table 8: Principal Component Analysis summary derived from the four Phonological Awareness tasks
Task
Correlation with Principal
Component
Communality*
Rhyme 0.789 0.622
Onset Matching 0.854 0.729
Onset Segmentation & Matching 0.841 0.706
Blending
Total
0.681 0.463
0.63
* The communality is the proportion of variance of that task that is retained in the PA Principal Component.
76
Hypothesis 1
Hypothesis 1 was that a relationship would be found between PA and speech
sound error types. It was believed that PA would be most strongly predicted by the
speech error category that represented the weakest phonological representations (Atypical
Sound Changes per Consonant). Correlational analyses and visual inspection of
scatterplots between the PA composite and speech production variables (Figure 4)
showed that there was very little relationship between PA and distortions (bottom plot of
Figure 4), and very little relationship between PA and typical sound changes (middle plot
of Figure 4); both of these correlations were not statistically significant (p’s > 0.05).
However, a significant relationship was found between PA and Atypical Sound Changes
per Consonant (r = -0.362, p = 0.009; top plot of Figure 4). That is, atypical sound
changes predicted about 13% of the variance in PA. As anticipated, the negative
correlation indicates that children with more atypical sound changes performed more
poorly on the PA tasks. This supports the hypothesis that atypical speech errors are
related to poor PA, presumably because both reflect weak phonological representations.
Hypothesis 2
As reported earlier, PA skills are often related to vocabulary and age, and this
study examined the extent to which PA variance can be predicted by speech sound errors
when vocabulary and age are taken into account. To address this question, hierarchical
multiple regression was used (Table 9). See Appendix I for regression diagnostics5.
5 Briefly, there were no significant violations of the assumptions of normal distribution of the variables or the residuals; the interaction terms did not account for significant variance in the model; there were no cases with standardized residuals more than 2.0 SD from the mean; tolerance statistics were high, indicating that multicollinearity is not a significant concern.
77
Figure 4: Scatterplots of speech sound production error types and phonological awareness composite (principal component)
Typ
ical
Ch
ang
es
Per
Co
nso
nan
t
Aty
pic
al C
han
ges
P
er C
on
son
ant
78
Table 9: Hierarchical regression used to predict PA Principal Component
Model
Var. Method b
(SE) β Sig. F df p R2 Adj R2
1 PPVT4 Enter .044
(.010)
.545 .000 11.5 2, 40 .000 .365 .333
Age Enter .058
(.023)
.314 .017
∆F df p ∆∆∆∆R2
∆∆∆∆ Adj R2
2 PPVT4
.040
(.010)
.487 .000 4.8 1, 39 .033 .070 .059
Age
.060
(.022)
.322 .011
ATYP Stepwise -6.149
(2.799)
-.217 .033
Not in the Equation Sig.
Distortions .557
Typical Errors .247
Total R2 = 0.435
Total Adjusted R2=0.392
Notes: b = Unstandardized coefficient (an estimate of the change in the PA Principal Component score for each 1-unit change in that variable; Keith, 2006); SE = standard error of the regression coefficient; β = standardized coefficient (coefficient when all variables are expressed in standardized [z-score] form; SPSS, 2006); R2 = the variance explained in PA by the variables in the model (Keith, 2006); Adj. R2 = Adjusted R2 (an attempt to correct R2 to more closely reflect the fit of the model to the population; SPSS, 2006); PPVT4 = Standard score of the Peabody Picture Vocabulary Test-4; ATYP = Atypical sound changes per consonant
As in other studies, receptive vocabulary was correlated with PA, and in this
study vocabulary accounted for about 27% of the variance in the PA composite (r =
0.517, p < 0.001). The bivariate relationship between age and the PA composite was not
79
statistically significant (r = 0.264, p > 0.05), possibly due to the restricted age range of
the participants in this study (21 months). However, when included in a model with other
variables, age is a significant predictor of the variance in PA (see below).
In the first step of the regression, age (in months) and receptive vocabulary
(PPVT-4 standard score) were used to predict PA. These two variables accounted for
about 33.3% of the variance in PA (F [2, 40] = 11.5, p < 0.001, R2 = 0.365, Adjusted R2
= 0.333), and both are statistically significant predictors of PA (p <0.05). In the second
step, the three speech production variables were tested in the model using stepwise entry
(adding variables to the equation based on those with the smallest probability of F, if the
probability is small enough)6. Atypical Sound Changes per Consonant was the only
significant speech production variable selected into the equation, and vocabulary and age
remained significant predictors of PA as well. The new model accounted for additional
variance in PA (∆R2 = 0.070, ∆ adjusted R2 = 0.059, p = 0.033). Therefore, Hypothesis
2 was confirmed: atypical sound changes predicted approximately 5.9% of the variance
in PA beyond what was already accounted for by vocabulary and age. Figure 5 displays
a scatterplot of the observed PA values (those achieved by the participants) and the PA
values predicted by the regression (age, receptive vocabulary, and atypical errors).
6 Stepwise entry was chosen for the second step of the regression (as opposed to forcing the three speech variables into the equation together) because it was presumed that some of the variables would not be related to PA. Therefore, only those speech production variables that contribute to the prediction of PA variance would be chosen. That is, the goal is to determine if certain variables are more robust predictors of PA, not to determine if all three speech variables together are robust predictors.
80
Figure 5: Observed PA Principal Component scores and PA scores predicted by the regression (age, vocabulary, atypical sound changes) for the 43 children with SSD
Hypothesis 3
Because it was found that atypical sound changes predict significant variance in
PA beyond variance explained by vocabulary and age, it is of interest to determine
whether a similar result would be found using PCC (which does not distinguish between
types of incorrect productions) as the speech sound accuracy variable (Hypothesis 3).
The bivariate correlation between PCC and the PA composite was not statistically
significant (r = 0.222, p = 0.153). However, a similar hierarchical multiple regression
was performed to predict PA, with PCC forced to enter in the second step, after
vocabulary and age. The results (Table 10) indicate that PCC does not explain any
variance in PA beyond receptive vocabulary and age in these 43 children with SSD (∆R2
= 0.000, p = 0.923). Therefore, the speech analysis based on presumed reflection of
phonological representations appears to provide a better explanation for the relationship
between PA and speech sound production than does the analysis using PCC.
81
Table 10: Regression using PCC as the speech production variable to predict PA
Model
Var. Method b
(SE) β Sig. F df p R2 Adj R2
1 PPVT4 Enter .044
(.010)
.545 .000 11.5 2, 40 .000 .365 .333
Age Enter .058
(.023)
.314 .017
∆∆∆∆F df p ∆∆∆∆R2 ∆∆∆∆
Adj R2
2 PPVT4 .044
(.011)
.540 .000 .009 1, 39 .932 .000 -.017
Age .058
(.024)
.314 .011
PCC Enter .001
(.012)
.013 .923
Notes: b = Unstandardized coefficient (an estimate of the change in the PA Principal Component for each 1-unit change in that variable; Keith, 2006); SE = standard error of the regression coefficient; β = standardized coefficient (coefficient when all variables are expressed in standardized [z-score] form; SPSS, 2006); R2 = the variance explained in PA by the variables in the model (Keith, 2006); Adj. R2 = Adjusted R2 (an attempt to correct R2 to more closely reflect the fit of the model to the population; SPSS, 2006); PPVT4 = Standard score of the Peabody Picture Vocabulary Test-4.
Exploratory Hypotheses
This study also investigated how phonological memory and phonological
retrieval/rapid naming skills are related to speech sound errors. The above regressions
were therefore repeated to explain variance in phonological memory (with scores on the
Syllable Repetition Task as the dependent variable; Hypothesis 4) and to explain variance
in rapid naming (with the average Z score on the two rapid naming tasks as the dependent
variable; Hypothesis 5).
82
(Exploratory) Hypothesis 4
Syllable Repetition Task was strongly correlated with Atypical Sound changes per
Consonant (r = -0.611, p <0.001), weakly correlated with Typical Sound Changes per
Consonants (r = -0.340, p = 0.026), and not significantly correlated with Distortions per
Consonant (r = 0.031, p = 0.842). The regression analysis then evaluated whether
variance in phonological memory (as assessed by the Syllable Repetition Task) could be
explained by types of speech sound errors (Hypothesis 4). Results of the regression are
shown in Table 11. The initial model, which includes receptive vocabulary and age, does
not predict a statistically significant amount of variance in phonological memory (F [2,
40] = 2.25, p = 0.118; R2 = 0.101; adjusted R2 = 0.056). This is somewhat unexpected,
as it is in contrast to other studies that have shown nonword repetition to correlate with
vocabulary skills and age (e.g., Edwards et al., 2004; Metsala, 1999). When the three
speech variables are added in the next step, the overall regression model becomes
significant (F [1, 39] = 22.6, p <0.001,R2 =0.409, Adjusted R2 = 0.364). Atypical Sound
Changes per Consonant becomes the only significant predictor of variance in
phonological memory (p < 0.001). Atypical changes explain about 30.8% of the unique
variance in the PA composite (∆ R2 = 0.308, ∆adjusted R2 = 0.308) and age and
vocabulary remain nonsignificant predictors. Typical Sound Changes per Consonant and
Distortions per Consonant are not selected for entry by the stepwise method (p > 0.05).
Therefore, as was the case with PA, atypical speech errors explain a significant amount of
the variance (30.8%) in phonological memory. This confirms Hypothesis 4.
83
Table 11: Regression explaining variance in Phonological Memory (Syllable Repetition Task)
Model
Var. Method b
(SE) β Sig. F df p R2 Adj R2
1 PPVT4 Enter .387
(.190)
.307 .048 2.25 2, 40 .118 .101 .056
Age Enter -.175
(.431)
-.061 .686
∆∆∆∆F df p ∆∆∆∆R2 ∆ ∆ ∆ ∆
Adj R2
2 PPVT4
.232
(.160)
.184 .154 20.3 1, 39 .000 .308 .308
Age
-.124
(.354)
-.043 .792
ATYP Stepwise -199.6
(55.3)
-.569 .000
Not in the Equation Sig.
Distortions .416
Typical Errors .507
Total R2 =0.409
Total Adjusted R2 =0.364
Notes: b = Unstandardized coefficient (an estimate of the change in the syllable repetition task for each 1-unit change in that variable; Keith, 2006); SE = standard error of the regression coefficient; β = standardized coefficient (coefficient when all variables are expressed in standardized [z-score] form; SPSS, 2006); R2 = the variance explained in the syllable repetition task by the variables in the model (Keith, 2006); Adj. R2 = Adjusted R2 (an attempt to correct R2 to more closely reflect the fit of the model to the population; SPSS, 2006); PPVT4 = Standard score of the Peabody Picture Vocabulary Test-4; ATYP = Atypical sound changes per consonant
84
(Exploratory) Hypothesis 5
The two RN tasks, which were moderately correlated (r = 0.55, p < 0.001), were
combined using average z-scores. A final hierarchical regression was run to predict
variance in RN, and the results are shown in Table 12. Age and receptive vocabulary did
not predict significant variance in the RN composite (F [2, 39] = 2.15, p = 0.130,
Adjusted R2 =0.053). Next, the three speech production variables (Distortions per
Consonant, Typical Sound Changes per Consonant, Atypical Sound Changes per
Consonant) were tested in the regression model, using stepwise entry, to determine if
types of speech sound production errors could explain performance in rapid naming.
Note that the bivariate correlations between RN and all three speech production variables
were nonsignificant, but they were entered into the equation to address the theoretical
question and to determine if any unique variance in RN could be explained. The overall
regression model that includes atypical sound changes, age, and receptive vocabulary is
significant (F [1, 38] = 7.71, p = 0.026; R2 = 0.213; adjusted R2 = 0.151). Atypical
Sound Changes Per Consonant was the only speech production variable that explained
significant variance in the RN composite (p = 0.024), and the result was an increase in
adjusted R2 of 9.9%. When Atypical Sound Changes per Consonant is added, receptive
vocabulary becomes a significant predictor of RN as well, and the model accounts for
about 15.2% of the variance in Rapid Naming. Therefore, atypical sound changes predict
significant variance in RN beyond age and vocabulary, confirming Hypothesis 5.
85
Table 12: Regression explaining variance in Rapid Naming (average z scores of two Rapid Naming tasks) 7
Model
Var. Method B
(SE) β Sig. F df p R2 Adj R2
1 PPVT4 Enter -.021
(.011)
-.291 .066 2.15 2, 39 .130 .099 .053
Age Enter -.027
(.025)
-.168 .279
∆∆∆∆F df p ∆∆∆∆R2 ∆ ∆ ∆ ∆
Adj R2
2 PPVT4
-.027
(.011)
-.378 .016 5.56 1, 38 .026 .114 .099
Age
-.028
(.024)
-.170 .248
ATYP Stepwise -6.898
(2.924)
-.350 .024
Not in the Equation Sig.
Distortions .783
Typical Errors .719
Total R2 =0.213
Total Adjusted R2 =0.152
Notes: b = Unstandardized coefficient (an estimate of the change in the rapid naming average z-score for each 1-unit change in that variable; Keith, 2006); SE = standard error of the regression coefficient; β = standardized coefficient (coefficient when all variables are expressed in standardized [z-score] form; SPSS, 2006); R2 = the variance explained in rapid naming by the variables in the model (Keith, 2006); Adj. R2 = Adjusted R2 (an attempt to correct R2 to more closely reflect the fit of the model to the population; SPSS, 2006); PPVT4 = Standard score of the Peabody Picture Vocabulary Test-4; ATYP = Atypical sound changes per consonant
7 Because one participant (P20) did not complete the Monosyllable Rapid Naming task, her data are not included in the regression reported. However, the regression was run again with her included (using the Z score from the disyllabic Rapid Naming task that she did complete) and the conclusions were unchanged.
86
Summary
The results presented above generally confirm the five hypotheses. Atypical
sound changes are significantly correlated with PA, but other speech sound error types
(distortions and typical sound changes) are not significantly related to PA. Atypical
sound changes account for a significant amount of variance in PA (about 5.9%) above
and beyond the variance explained by receptive vocabulary and age. The model that
categorizes speech errors into three types (based on the presumed accuracy of
phonological representations) was found to be better in explaining variance in PA,
compared to the model which considers all consonant errors equally (PCC). Finally,
atypical sound changes help to explain significant variance in phonological memory and
phonological retrieval/rapid naming skills in children with speech sound disorders. In all
of the models, more atypical errors are associated with poorer performance on
phonological processing tasks.
87
IV : DISCUSSION
In this study, the relationship between speech sound errors and three domains of
phonological processing (phonological awareness, phonological retrieval/rapid naming,
and phonological memory) was assessed. Phonological processing skills, which are often
reported to be weak in poor readers, are also weak in some children with SSD. The fact
that there is often a wide range of performance on phonological processing tasks by
children with SSD was confirmed in the present study.
The variability in PA found in this study could be accounted for, in part, by
vocabulary skills and age (about 33%). Both of these factors have been discussed in the
past as contributing to the development of PA and also to the accuracy of phonological
representations. However, as with prior studies, there remained much unexplained
variance in the performance of the children in this study on phonological awareness tasks.
One additional consideration, therefore, was that speech sound production, which is also
thought to rely, in part, on phonological representations, could predict performance on
phonological processing tasks. That is, certain types of speech sound errors may be
indicative of poorly specified or inaccurate phonological representations, and therefore
these errors may be related to a child’s performance on PA tasks. The results of this
study confirmed that prediction: a measure thought to reflect poorly specified
phonological representations in speech sound production, the number of atypical sound
changes per consonant, was found to account for significant variance in PA. The
variance accounted for was above and beyond any variance explained by vocabulary and
age. However, no additional variance was explained in PA when Percent Consonants
88
Correct (PCC) was used to measure speech sound accuracy, suggesting that PCC may not
be sensitive to variation in PA skills.
It was also found that variance in two other phonological processing domains,
phonological memory and phonological retrieval , could be explained in part by atypical
sound changes. Therefore, models intended to account for phonological processing
performance in children with SSD can be informed by these findings. Further discussion
is provided below regarding types of consonant errors, each domain of phonological
processing (phonological awareness, phonological memory, and phonological retrieval),
and the presumed theoretical link to phonological representations.
Consonant Error Types
Percent Consonants Correct (PCC), which weights all speech sound errors
equally, was not found to be significantly correlated with the PA composite score, and it
did not account for any variance in PA beyond age and receptive vocabulary. This
finding is generally consistent with prior research and supports Hypothesis 3. It suggests
that PCC may not be a sensitive indicator of the relationship between speech sound errors
and PA. However, as discussed in Appendix H, larger samples would be required to have
adequate power to reject the use of PCC in predicting variance in PA.
One of the unique features of the current study is that it attempts to provide a
more complete explanation of the component feature changes involved in children’s
speech sound errors than has been done in the past. Whereas PCC simply considers all
speech sound errors as equal, the current study calls upon phonetically-motivated
explanations of how those errors could be derived. That is, the three-category system
89
was designed to more fully account for all of the features of a child’s errors.
As expected, distortion errors, which involve phonetic (within-phoneme) changes,
were unrelated to performance on any of the phonological processing tasks. This is in
line with previous findings and supports the notion that these phonetic variations are not
indicative of poorly specified phonological representations (Preston & Edwards, 2007;
Rvachew et al., 2007; Shriberg et al., 2005).
Typical sound changes were also found not to be correlated with phonological
processing skills in this study. While it was predicted that these errors might have a
moderate correlation with phonological processing skills, the correlations were low and
not statistically significant. Thus, it appears that the occurrence of typical sound changes
provides little information regarding a child’s phonological processing skills.
In contrast, atypical sound changes were found to account for significant variance
in all three domains of phonological processing. The primary analysis was intended to
predict variance in phonological awareness (PA); atypical sound changes predicted about
13% of the variance in PA, and about 5.9% of the unique variance in PA when
controlling for age and receptive vocabulary. While this is not necessarily a robust
explanation of the variance in PA skills, it may be indicative of a shared phonological
deficit, speculated here to be weak underlying phonological representations. That is,
children with SSD who use unusual sounds changes to produce words may also have
trouble attending to the sound features of words in tasks such as rhyming, initial
consonant matching, and blending. Children who use more of these atypical sound
changes also tend to be less accurate in syllable repetition and slower on rapid naming
tasks. Thus, the fact that this measure accounts for significant variance in three separate
90
domains of phonological processing provides support for this type of analysis of a child’s
speech sound errors.
To avoid overestimation of what was considered “atypical,” the classification
scheme developed for this study was relatively conservative. The decision was made a
priori to score errors in a manner that would give children the most possible “credit” (i.e.,
counting the smallest number of errors possible, and considering typical errors rather than
atypical errors when alternate accounts were possible, as described in Appendix B).
However, even with this conservative estimate, all participants were found to have at
least a few occurrences of atypical errors in their speech samples. It is acknowledged that
defining atypical errors differently could result in different findings, and other ways of
analyzing speech sound errors could yield different results. For example, the
investigation by Rvachew et al. (2007) reported no significant relationship between
atypical sound changes and phonological processing in preschoolers with SSD. One
explanation might be differences in statistical techniques (e.g., the regression used in the
current investigation, vs. the t-tests used by Rvachew et al. to compare two groups of
children with “normal” and “delayed” PA). An alternate explanation is that the speech
error coding scheme developed for this study was more fine-grained and took phonetic
plausibility and the effects of nearby sounds into consideration when trying to logically
account for errors. For example, Rvachew et al. considered /d/ � [g] to be atypical
regardless of phonetic context, whereas the current investigation counted that change as
typical if it could be accounted for by the typical sound change of velar assimilation. For
example, if pudding is produced as [pȚgǺŋ], /d/ �[g] is accounted for by velar
assimilation, with the /d/ taking on the “back” feature of the /ŋ/. Thus, the notions of
91
phonetic plausibility and component sound changes that were applied in this study were
less apparent in the work by Rvachew et al. (2007).
Phonological Awareness
As expected, variance in phonological awareness (PA) could be predicted, in part,
by vocabulary and age in preschoolers with SSD. Also as hypothesized, additional
variance in PA was explained by atypical sound changes, a speech production variable
thought to be associated with weak phonological representations. The ability to develop
accurate or refined phonological representations for PA tasks (to the point that they may
be used for comparing and contrasting initial consonants and rhymes of words) had a
modest (but significant) negative relationship with the production of atypical sound
changes. The primary impact of low PA is likely to be on early decoding and spelling.
That is, if children do not have clearly defined representations for the essential sound
features of words, they may have difficulty using phonological information for sounding
out (decoding) words and spelling. It could be speculated that imprecise phonological
representations might additionally hinder the ability to associate an orthographic symbol
with a phoneme. However, this hypothesis was not tested, as orthographic knowledge
and sound-symbol associations were not assessed. Appendix J provides further
discussion of this issue.
There is mounting evidence that children who enter kindergarten with a SSD and
weak PA skills are at particular risk for early literacy problems (Bird et al., 1995; Nathan
et al., 2004). Thus, the results of this study could have diagnostic significance. Early
identification of PA problems is essential for early intervention to take place. Clinically,
92
PA assessments are not routine for all children with SSD. This study provides support
for the notion that children with numerous atypical sound changes will be at an elevated
risk for PA problems and should therefore be assessed in that domain.
Phonological Retrieval/Rapid Naming
The finding that atypical sound changes account for 4.6% of the unique variance
in rapid naming (beyond the contribution of age and vocabulary) provides tentative
support for the notion that problems with quickly retrieving phonological representations
are related to the production of more atypical sound changes. The implications are that
children who produce more atypical speech sound errors may be at added risk for literacy
difficulties, as rapid naming tasks have been found to relate to reading fluency, spelling,
and decoding (Allor, 2002; Kamhi et al., 1988; Kirby et al., 2003; Wolf & Bowers, 1999;
Wolf et al., 2002).
It should be noted that rapid naming tasks are not frequently used with
preschoolers, and the predictive validity of the specific rapid naming tasks used in this
study has not been investigated. As reported earlier, some of the preschoolers found it
difficult to attend (uninterrupted) to a series of 30 pictures. Hence, other processes
beyond phonological retrieval are clearly involved in rapid naming (attention, visual
recognition, inhibition of recently retrieved words, etc., Wolf & Bowers, 1999). These
other processes may account for the relatively weak association between speech sound
errors and rapid naming in the present study.
As expected, it took most participants longer to rapidly name 30 disyllabic words
than 30 monosyllabic words. The data from this study could now be compared to the
93
naming abilities of preschoolers without SSD to determine if naming of disyllabic words
takes disproportionately longer for children with SSD (as noted for adolescents in Preston
& Edwards, 2006).
Phonological Memory
One of the interesting relationships found in the exploratory analyses in this study
was that atypical sound changes showed a relatively strong correlation (r = -0.611) with
the measure of phonological memory used here and contributed a relatively large
proportion of variance to phonological memory beyond age and vocabulary (30.8%).
Because performance on repetition tasks has been strongly tied to language and literacy
performance (Brady, 1991; Dollaghan & Campbell, 1998; Metsala, 1999; Munson et al.,
2005), this finding has significant implications for identification of children at risk for
literacy problems. As explained below (Speculations on Phonological Representations),
it is possible that poor phonological memory is causally connected with both atypical
sound changes and poorly specified phonological representations. The predictive value
of this syllable repetition task in children with SSD should be investigated to determine if
growth in speech sound accuracy and/or PA development over time could be predicted by
performance on this task.
In this study, the ability to repeat syllables was unrelated to age and receptive
vocabulary. This in contrast to previous studies that evaluated nonword repetition in
preschoolers (Edwards et al., 2004; Metsala, 1999; Roy & Chiat, 2004). It is unclear why
this might be, although one possible explanation could be the restricted age (4;0-5;9) and
PPVT-4 standard scores (>84) in this study. An additional possibility is that previous
94
reports of nonword repetition have utilized stimuli that require more complex articulatory
demands. Those stimuli may have been more sensitive to age than the stimuli in this
study, which utilized only four early developing consonants paired with the same vowel.
Clinical Implications
As expected, there were several children with SSD who performed quite well on
the phonological processing tasks. Thus, as reported in prior research, not all children
with SSD are necessarily at risk for literacy problems. Based on the results of this study,
the use of atypical sound changes can be considered an indicator of weak phonological
processing skills and, in particular, poor phonological memory. It could be argued, then,
that intervention or treatment that focuses on speech sound production and phonological
processing should be implemented for children who exhibit atypical sound changes (cf.
Gillon, 2005), perhaps targeting atypical errors. There are few studies investigating
treatment of children with atypical sound changes, but those that exist suggest that these
errors can be improved with standard phonological treatment techniques, such as minimal
pair intervention and facilitating contexts (Dodd & Iacano, 1989; Leonard & Brown,
1984; Stringfellow & McLeod, 1994).
Given the relatively strong relationship between atypical sound changes and
phonological memory, we might speculate about whether treatment directed at the
improvement of phonological memory would have any impact on speech sound
production. That is, children who can better recall from working memory the
phonological features they just heard might be better able to store accurate phonological
representations. However, there is a lack of research addressing the question of whether
95
phonological memory can be improved by intervention in individuals with poor nonword
repetition skills. Although phoneme-based interventions have shown some success in
improving phonological memory in aphasics (Kendall et al., 2008), the training of
working phonological memory in children with SSD seems to be an area in need of
further investigation.
Caveats and Limitations
Interpretation of Results
Several caveats related to the findings of the current study should be noted. For
example, given the sample size (n = 43), the confidence interval around the model R2 in
the prediction of PA is relatively large (R2 = 0.435, 95% CI = 0.233 - 0.636). Replication
of these results in other samples will help to clarify the effect size and the strength of the
relationship between PA and types of speech sound errors. Additionally, in the analysis
of behavioral data, there remains debate as to how best to interpret the “size” of R2
change (Keith, 2006). The amount of variance in PA that is explained by adding
Atypical Sound Changes per Consonant to the equation is relatively modest (∆R2
adjusted = 0.059). Thus, in comparison to the other variables (particularly receptive
vocabulary), this does not appear to be a large effect. However, because a significant
amount of additional variance can be accounted for by adding Atypical Sound Changes
and because there is theoretical reason to include this variable (i.e., it is thought to be
indicative of weak phonological representations), this suggests that the model is useful in
explaining variance in PA for children with SSD.
96
Nevertheless, to keep the findings of the current study in perspective, it is
important to consider the relative strength of the relationships. The bivariate correlations,
as well as the size of the standardized coefficients in the regression, suggest that PA is
more strongly related to receptive vocabulary than to speech sound errors. This may be
because vocabulary is thought to be causally related to the development of accurate
phonological representations. That is, increases in vocabulary size might result in
refinement of phonological representations. In contrast, some speech sound errors might
be considered a result of inaccurate phonological representations. This remains a matter
of theoretical speculation, as there is no direct way to evaluate the (in)accuracy of a
child’s phonological representations.
Additionally, the results should be interpreted within the scope of the participant
characteristics (e.g., primarily middle class, monolingual English-speaking children with
idiopathic SSD). Thus, the results may not be applicable to all children with
phonological processing difficulties. For example, many children have problems with
PA but do not have speech sound production difficulties. Therefore, it is unlikely that
any additional variance in the phonological processing skills could be explained by
atypical errors in children without SSD, as they, by definition, rarely (if ever) exhibit
atypical sound changes.
Caveats on Speech Sound Errors
As discussed earlier, debate exists concerning how to categorize speech sound
errors, and particularly which errors should be considered “atypical.” While attempts
were made to consult the literature regarding such sound changes, relevant literature was
97
sometimes absent or contradictory. Thus, other researchers might reach slightly different
conclusions about which sound changes should be considered atypical. However, the
error coding system was intended to be comprehensive and replicable and was based on
extant literature and the notion of phonetic plausibility. Therefore, is believed to be a
valid system for categorizing consonant errors.
The method of quantifying speech sound accuracy in this study, while
theoretically motivated, is not the only method for analyzing speech sound production
skills. Other transcription-based methods exist for examining speech production output,
although they have not consistently revealed a relationship between speech production
and phonological processing. For example, phonological processing skills have been
found to be unrelated to phonological features (e.g., sonorant, labial, nasal, etc.; Rvachew
et al., 2007) and some standardized tests of speech sound accuracy (e.g., Larivee & Catts,
1999). Therefore, it is possible that transcription-based methods might not be highly
sensitive to subtleties in speech sound production that relate to phonological processing
and/or phonological representations. Future studies could implement instrumental
analysis of phonetic output, including segmental and suprasegmental analysis. For
example, possible speech-related predictors of phonological processing skills might
include subtle acoustic features such as voice onset time (Tyler et al., 1990) and vowel
formants (cf. Elbro et al., 1998), or prosodic characteristics such as lexical stress
(Shriberg et al., 2003a; Shriberg et al., 2003b) and speaking rate (Smith et al., 2006).
Vowels. This study did not analyze vowel production errors, in part because
vowel accuracy is generally thought to develop earlier than consonant accuracy (Lowe,
1994, but see Pollock, 1991) and because vowel errors are less often discussed as a
98
characteristic of speech sound disorders. One study with Danish children suggested that
the quality of vowel productions may be related to later literacy achievement in
kindergarteners (Elbro et al., 1998). However, because the current study attempted to
remedy some of the limitations associated with the Percent Consonants Correct measure,
vowel errors were not analyzed here. Vowels were transcribed and vowel errors did
occur. Thus, data are available for exploratory analyses in the future. A measure that is
capable of describing and adequately weighting both consonant and vowel errors might
be the most comprehensive measure of phonological output accuracy (such a measure is
currently being developed by the author).
Subgroups. Although the current study examined types of errors and their
frequency, it did not quantify the consistency with which errors occur. There has been
some discussion that the consistency of errors across multiple attempts at the production
of a word may be indicative of childhood apraxia of speech (e.g., producing "elephant"
four different ways on four different attempts, Dodd, 1995). Children with a diagnosis of
childhood apraxia of speech (CAS) have been found to have difficulty with phonological
processing (Dodd, 1995; Lewis et al., 2004). However, this study did not take different
subgroups of children with SSD of unknown origin into consideration. Therefore,
children with suspected CAS were included but were not looked at separately.8
Several classification systems exist for SSD, based on suspected etiology
(Shriberg et al., 1997b), speech error patterns (Bradford & Dodd, 1996; Crary, 1984;
Dodd, 2005; Gibbon, 1999), concomitant speech disorders (Wolk et al., 1993), or
8 Approximately half of the parents or referring clinicians of participants in this study indicated that CAS was diagnosed or suspected, which is well above prevalence data for CAS. However, this is consistent with the notion that CAS definitions are broad and that the disorder is often clinically over-diagnosed (American-Speech-Language-Hearing-Association, 2007).
99
concomitant language disorders (Bird et al., 1995; Bishop & Adams, 1990; Leitao et al.,
1997), etc., but there is poor consensus on how to differentially diagnose particular
subtypes of SSD. Because none of these classification systems have robust empirical
support, they were not utilized here. Although many of the children in this study
probably fell into one or more of the subgroups described in the literature, looking at
subgroups was not the focus of this study. Moreover, it is unclear how the results found
here would be influenced by particular subgroups, as there is no clear description of the
use of different types of sound errors by subgroups of children with SSD.
Speculations on Phonological Representations
As previously discussed, the results reported above may be accounted for in part
by the accuracy of phonological representations. Phonological representations are
thought to develop with vocabulary and age and to rely on a child’s ability to extract
and/or infer linguistically meaningful sound patterns in the speech signal. As vocabulary
skills increase, children develop a broader variety of words from which to draw
inferences about the essential phonological features of words (Fowler, 1991; Metsala,
1999). These inferences are thought to help children recognize underlying contrasts (e.g.,
voiced-voiceless, nasal-nonnasal, etc) and to recognize sound patterns and appropriate
sound combinations in the adult language. Phonological representations might then
become more specified and closer to the adult target as the child has more experience
with a word and with similar-sounding words (Fowler, 1991). When children’s ability to
extract salient phonological features of words (and use them to form phonological
100
representations) is weak, their ability to recognize some of the salient phonological
features (such as rhymes or initial consonants) might be weak as well.
Some children with SSD, especially those with lower vocabularies and those who
produce relatively more atypical sound changes, were found in this study to have greater
difficulty on the PA tasks which required them to focus on the linguistically meaningful
sound patterns of the speech signal (i.e., identify rhyme, initial consonants, etc.). For
these children, salient features in speech sound production may be poorly represented,
resulting in unusual productions of words (e.g., deletions of initial consonants, strong
syllables, or unmarked members of consonant clusters).
The current investigation provides support for the notion that phonological
processing and types of speech sound errors are linked in preschoolers with SSD (Figure
1). The assumed link, though not directly tested here, is poorly specified (“weak”)
phonological representations (cf. Swan & Goswami, 1997a). It remains unclear exactly
why these representations are weak. Some possible explanations include:
(a) Speech perception problem: poor detection or encoding of the phonetic
features of the rapidly-changing speech signal (e.g., poor speech perception) may result
in insufficient, incomplete, or inaccurately perceived information to store in phonological
representations. Appendix K provides further detail related to this issue.
(b) Storage problem: Some children with SSD may perceive speech signals
accurately but have difficulty making the appropriate inferences about the phonetic or
phonological components of words to store them correctly.
(c) Phonological rehearsal problem. Articulatory rehearsal (e.g., the
“phonological loop”) which involves the ability to subvocally repeat/rehearse verbal
101
input to keep it in temporary storage for a longer time, has been discussed in models of
working memory (Baddeley, 2003; Baddeley et al., 1998). There is evidence that this
rehearsal may be impaired in some children with SSD (Bishop et al., 1990; Locke &
Scott, 1979). Given the strength of the relationship between the phonological memory
task and atypical sound changes found here, one hypothesis is that a limitation in the
ability to accurately rehearse phonological information to form temporary phonological
representations results in a lack of information available to form accurate long-term
representations. That is, there may be a (covert) rehearsal mechanism that is impaired in
some children with SSD, and this may impact the ability to accurately retain speech-
related information in working memory as well as form a long-term representation.
Hence, the phonological memory/rehearsal deficit would be viewed as a causal factor in
weak phonological representations. This would have implications for the ability to learn
and store new phonological forms (cf. Sutherland & Gillon, 2005).
Although the premise discussed thus far has been that phonological
representations influence speech sound production, it is possible that a bidirectional
relationship exists between phonological representations and speech sound production.
This would mean that producing speech sounds correctly in words could reinforce adult-
like representations, whereas producing speech sounds incorrectly could inhibit the
development of adult-like phonological representation (Bishop et al., 1990; Nicolson et
al., 2001). Such a view could be considered in accord with the Motor Theory of Speech
Perception (Galantucci, Foweler, & Turvey, 2006).
102
Ideally, longitudinal follow-up of the children in this study would be helpful in
determining if the tasks and speech production measures used here are able to predict
long-term growth of phonological processing and literacy skills. For example, as
kindergarteners, invented spelling tasks (which have often been discussed as relating to
PA development in somewhat older children) could provide insight into these children’s
early knowledge of phonemes and phoneme-grapheme correspondence (e.g., Ball &
Blachman, 1991). Additionally, it would be of interest to determine if periodic
assessments would reveal concurrent reduction of atypical sound changes and
improvement in phonological processing skills, both presumably due to the refinement of
phonological representations.
SIGNIFICANCE AND CONCLUSIONS
This study evaluated the relationship between phonological processing and
speech sound errors in children with speech sound disorders (SSD), while addressing
some of the limitations of previous studies. The relative influence of different types of
speech sound errors had not been well-explored in a systematic fashion. This study
appears to be the first to address within-group variability in phonological processing
through a measurement system that separates all consonant errors based on both types
and frequency. A three-category scheme for coding speech sound errors was developed
that accounted for the component features and changes in the children’s sound errors, and
it was found that atypical sound changes were better predictors of phonological
processing than distortions and typical sound changes. Some of the limitations of
previous studies, which formed discrete groups based on one measure of speech
103
production to predict variance in phonological processing, were addressed by using a
regression technique. Additionally, the study explored phonological processing and
speech sound production in preschoolers with SSD, a population that has not previously
been studied in this way.
This study has both clinical and theoretical importance, as it has helped to
advance our understanding of how certain types of speech sound errors relate to specific
phonological processing domains known to be related to early literacy. Atypical sound
changes were found to predict unique variance in three phonological processing domains,
whereas distortions and typical sound changes were not. Poorly specified phonological
representations have been discussed as the link between phonological processing
difficulties and some speech sound errors.
This research suggests that more frequent use of atypical sound changes is related
to greater risk of preliteracy problems (to the extent that they are tapped by these tasks) in
children with SSD. This research provides important evidence in light of the critical age
hypothesis for literacy development, which suggests that children who enter kindergarten
with speech sound production problems and PA problems are at significant risk for
literacy problems (Bird & Bishop, 1992; Nathan et al., 2004). Thus, it would be prudent
for clinicians to consider the specific types of speech sound errors that reflect relatively
greater risk for phonological processing (and, by extension, literacy) when evaluating and
treating preschool children. Children with SSD who exhibit frequent atypical sound
changes would be appropriate candidates for further evaluation of phonological
processing. Therefore, it is hoped that this research may help to further our
104
understanding of which children are at particular risk for preliteracy and literacy
problems so that early intervention can be implemented.
105
REFERENCES
Allor, J. H. (2002). The relationships of phonemic awareness and rapid naming to reading
development. Learning Disability Quarterly, 25, 47-57.
American-Speech-Language-Hearing-Association. (2007). Childhood Apraxia of Speech
[Technical Report]. available from www.asha.org/policy.
Aram, D. M., Ekelman, B. L., & Nation, J. E. (1984). Preschoolers with language
disorders: 10 years later. Journal of Speech & Hearing Research, 27(2), 232-244.
ASHA Audiologic Assessment Panel 1996. (1997). Guidelines for audiological
screening. Rockville, MD: Author.
Baddeley, A. (2003). Working memory and language: An overview. Journal of
Communication Disorders, 36, 189-208.
Baddeley, A., Gathercole, S., & Papagno, C. (1998). The phonological loop as a language
learning device. Psychological Review, 105(1), 158-173.
Baguley, T. (2004). Understanding statistical power in the context of applied research.
Applied Ergonomics, 35, 73-80.
Ball, E. W., & Blachman, B. A. (1991). Does phoneme awareness training in
kindergarten make a difference in early word recognition and developmental
spelling. Reading Research Quarterly, 26(1), 49-66.
Bird, J., & Bishop, D. (1992). Perception and awareness of phonemes in phonologically
impaired children. European Journal of Disorders of Communication, 27, 289-
311.
106
Bird, J., Bishop, D. V. M., & Freeman, N. H. (1995). Phonological awareness and
literacy development in children with expressive phonological impairments.
Journal of Speech and Hearing Research, 38, 446-462.
Bishop, D. V., Brown, B. B., & Robson, J. (1990). The relationship between phoneme
discrimination, speech production, and language comprehension in cerebral-
palsied individuals. Journal of Speech and Hearing Research, 33(2), 210-219.
Bishop, D. V. M., & Adams, C. (1990). A prospective study of the relationship between
specific language impairment, phonological disorders and reading retardation.
Journal of Child Psychology and Psychiatry, 31(7), 1027-1050.
Bishop, D. V. M., & Clarkson, B. W. (2003). Written language as a window into residual
language deficits: A study of children with persistent and residual speech and
language impairments. Cortex, 39(2), 215-237.
Bradford, A., & Dodd, B. (1996). Do all speech-disordered children have motor deficits?
Clinical Linguistics and Phonetics, 10(2), 77-101.
Bradley, L., & Bryant, P. E. (1983). Categorizing sounds and learning to read - a causal
connection. Nature, 301, 419-421.
Brady, S. A. (1991). The role of memory in reading disability. In S. Brady & D.
Shankweiler (Eds.), Phonological processes in literacy: A tribute to Isabelle Y.
Liberman (pp. 129-151). Hillsdale, NJ: Lawrence Erlbaum Ass.
Bridgeman, E., & Snowling, M. (1988). The perception of phoneme sequence: A
comparison of dyspraxic and normal children. British Journal of Disorders of
Communication, 23, 245-252.
107
Burgess, S. R., & Lonigan, C. J. (1998). Bidirectional relations of phonological
sensitivity and prereading abilities: Evidence from a preschool sample. Journal of
Experimental Child Psychology, 70, 117-141.
Campbell, T. F., Dollaghan, C., Janosky, J. E., & Adelson, P. D. (2007). A performance
curve for assessing change in Percentage of Consonants Correct Revised (PCC-
R). Journal of Speech, Language, and Hearing Research, 50(4), 1110-1119.
Campbell, T. F., Dollaghan, C. A., Rockette, H. E., Paradise, J. L., Feldman, H. M.,
Shriberg, L. D., et al. (2003). Risk factors for speech delay of unknown origin in
3-year-old children. Child Development, 74(2), 346-357.
Campbell, T. F., & Shriberg, L. D. (1982). Associations among pragmatic functions,
linguistic stress, and natural phonological processes in speech-delayed children.
Journal of Speech and Hearing Research, 25, 547-553.
Caravolas, M., Hulme, C., & Snowling, M. J. (2001). The foundations of spelling ability:
Evidence from a 3-year longitudinal study. Journal of Memory and Language, 45,
751-774.
Cardoso-Martins, C., & Pennington, B. F. (2004). The relationship between phoneme
awareness and rapid serial naming skills and literacy acquisition: The role of
developmental period and reading ability. Scientific Studies of Reading, 8(1), 27-
52.
Carroll, J. B., Davies, P., & Richman, B. (1971). Word frequency book. Boston:
Houghton Mifflin Co.
Catts, H. W. (1991). Facilitating phonological awareness: Role of speech-language
pathologists. Language, Speech and Hearing Services in Schools, 22, 196-203.
108
Catts, H. W. (1993). The relationship between speech-language impairments and reading
disabilities. Journal of Speech and Hearing Research, 36, 948-958.
Catts, H. W. (1997). The early identification of language-based reading disabilites.
Language, Speech and Hearing Services in Schools, 28, 86-89.
Catts, H. W., Fey, M. E., Zhang, X., & Tomblin, J. B. (2001). Estimating the risk of
future reading difficulties in kindergarten children: A research-based model and
its clinical implementation. Language, Speech and Hearing Services in Schools,
32(1), 38-50.
Catts, H. W., Hu, C.-F., Larrivee, L., & Swank, L. (1994). Early identification of reading
disabilities in children with speech-language impairments. In R. V. Watkins & M.
L. Rice (Eds.), Specific language impairments in children (pp. 145-160).
Baltimore, MD: Paul H. Brookes Publishing.
Crary, M. A. (1984). Phonological characteristics of developmental verbal dyspraxia.
Seminars in Speech and Language, 5(2), 71-83.
Denckla, M. B., & Rudel, R. G. (1976). Rapid 'automatized' naming (R.A.N.): Dyslexia
differentiated from other learning disabilities. Neuropsychologia, 14, 471-479.
Dodd, B. (1995). The differential diagnosis and treatment of children with speech
disorders. San Diego: Singular Publishing Group.
Dodd, B. (Ed.). (2005). Differential diagnosis and treatment of children with speech
disorder (2 ed.). Philadelphia: Whurr Publishers.
Dodd, B., & Iacano, T. (1989). Phonological disorders in children: Changes in
phonological process use during treatment. British Journal of Disorders of
Communication, 24, 333-351.
109
Dodd, B., Leahy, J., & Hambly, G. (1989). Phonological disorders in children:
Underlying cognitive deficits. British Journal of Developmental Psychology, 7,
55-71.
Dollaghan, C., & Campbell, T. F. (1998). Nonword repetition and child language
impairment. Journal of Speech, Language, and Hearing Research, 41, 1136-1146.
Dunn, L. M., & Dunn, D. M. (2007). Peabody Picture Vocabulary Test (4 ed.).
Minneapolis, MN: Pearson.
Dunn, L. M., & Dunn, L. M. (1981). Peabody Picture Vocabulary Test-Revised. Circle
Pines, MN: American Guidance Service.
Dworkin, J. P. (1980). Characteristics of frontal lispers clustered according to severity.
Journal of Speech and Hearing Disorders, 35, 37-44.
Edwards, J., Beckman, M. E., & Munson, B. (2004). The interaction between vocabulary
size and phonotactic probability effects on children’s production accuracy and
fluency in nonword repetition. Journal of Speech, Language, and Hearing
Research, 47, 421-436.
Edwards, M. L. (1992). In support of phonological processes. Language, Speech and
Hearing Services in Schools, 23, 233-240.
Edwards, M. L. (1995). Developmental Phonology. In H. Winitz (Ed.), Human
communication and its disorders: A review (Vol. IV). Timonium, MD: York
Press.
Edwards, M. L., & Shriberg, L. D. (1983). Phonology: Applications in communicative
disorders. San Diego, CA: College-Hill Press.
110
Elbro, C., Borstrom, I., & Peterson, D. K. (1998). Predicting dyslexia from kindergarten:
The importance of distinctness of phonological representations of lexical items.
Reading Research Quarterly, 33(1), 36-60.
Elliott, C. D. (1990). Differential Ability Scales. San Antonio, TX: Psychological
Corporation/Harcourt Brace.
Fletcher, S. G., Casteel, R. L., & Bradley, D. P. (1961). Tongue-thrust swallow, speech
articulation, and age. Journal of Speech & Hearing Disorders 26, 201-208.
Fowler, A. E. (1991). How early phonological development might set the stage for
phoneme awareness. In S. A. Brady & D. P. Shankweiler (Eds.), Phonological
processes in literacy: A tribute to Isabelle Y. Liberman (pp. 97-117). Hillsdale,
NJ: Lawrence Erlbaum Associates, Publishers.
Foy, J. G., & Mann, V. (2001). Does strength of phonological representations predict
phonological awareness in preschool children? Applied Psycholinguistics, 22(3),
301-325.
Galantucci, B., Fowler, C.A. & Turvery, M.T. (2006). The motor theory of speech
perception reviewed. Psychonomic Bulletin & Review, 13 (3), 361-377.
Gibbon, F. E. (1999). Undifferentiated lingual gestures in children with
articulation/phonological disorders. Journal of Speech, Language, and Hearing
Research, 42, 382-397.
Gillon, G. T. (2000). The efficacy of phonological awareness intervention for children
with spoken language impairment. Language, Speech, & Hearing Services in
Schools, 31(2), 126-141.
111
Gillon, G. T. (2005). Facilitating phoneme awareness development in 3- and 4-year-old
children with speech impairment. Language, Speech, and Hearing Services in
Schools, 36(4), 308-324.
Goldman, R., & Fristoe, M. (1986). Goldman-Fristoe Test of Articulation. Circle Pines,
MN: American Guidance Service.
Goldman, R., & Fristoe, M. (2000). Goldman Fristoe Test of Articulation - Second Ed.
Circle Pines, MN: AGS.
Griffiths, Y. M., & Snowling, M. J. (2002). Predictors of exception word and nonword
reading in dyslexic children: The severity hypothesis. Journal of Educational
Psychology, 94(1), 34-43.
Gruber, F. A. (1999). Probability estimates and paths to consonant normalization in
children with speech delay. Journal of Speech, Language, and Hearing Research,
42, 448-459.
Hall, P. K. (1989). The occurrence of developmental apraxia of speech in a mild
articulation disorder: A case study. Journal of Communication Disorders, 22,
265-276.
Hall, S. L., & Moats, L. C. (1999). Straight talk about reading: How parents can make a
difference during the early years. Chicago, Ill: Contemporary Books.
Hauner, K. K. Y., Shriberg, L. D., Kwiatkowski, J., & Allen, C. T. (2005). A subtype of
speech delay associated with developmental psychosocial involvement. Journal of
Speech, Language, and Hearing Research, 48, 635-650.
Hesketh, A., Adams, C., Nightingale, C., & Hall, R. (2000). Phonological awareness
therapy and articulatory training approaches for children with phonological
112
disorders: A comparative outcome study. International Journal of Language and
Communication Disorders, 35(3), 337-354.
Hodson, B., & Edwards, M. L. (1997). Perspectives in applied phonology. Gaithersburg,
MD: Aspen Publishers, Inc.
Ingram, D. (1976). Phonological disability in children. New York: Elsevier.
International-Dyslexia-Association. (2000). Dyslexia Basics: Fact Sheet #62.
Jamieson, D. G., & Rvachew, S. (1992). Remediating speech production errors with
sound identification training. Journal of Speech Language Pathology and
Audiology, 16(3), 201-210.
Justice, L. M., & Pullen, P. C. (2003). Promising interventions for promoting emergent
literacy skills: Three evidence-based approaches. Topics in Early Childhood
Special Education, 23(3), 99-113.
Kamhi, A. G., & Catts, H. W. (1986). Toward an understanding of developmental
language and reading disorders. Journal of Speech & Hearing Disorders, 51(4),
337-347.
Kamhi, A. G., Catts, H. W., Mauer, D., Apel, K., & Gentry, B. F. (1988). Phonological
and spatial processing abilities in language- and reading- impaired children.
Journal of Speech and Hearing Disorders, 53, 316-327.
Kazdin, A. E. (2003). Research design in clinical psychology. Boston: Allyn and Bacon.
Keith, T. Z. (2006). Multiple regression and beyond. Boston: Pearson Education.
Kendall, D. L., Rosenbek, J. C., Heilman, K. M., Conway, T., Klenberg, K., Gonzalez
Rothi, L. J., et al. (2008). Phoneme-based rehabilitation of anomia in aphasia.
Brain and Language, 105(1), 1-17.
113
Khan, L. M. L. (1982). A review of 16 major phonological processes. Language, Speech
and Hearing Services in Schools, 13, 77-85.
Kirby, J. R., Parrilla, R. K., & Pfeiffer, S. L. (2003). Naming speed and phonological
awareness as predictors of reading development. Journal of Educational
Psychology, 95(3), 453-464.
Klein, H. B., & Spector, C. C. (1985). Effect of syllable stress and serial position on error
variability in polysyllabic productions of speech-delayed children. Journal of
Speech and Hearing Disorders, 50(4), 391-402.
Larivee, L. S., & Catts, H. W. (1999). Early reading achievement in children with
expressive phonological disorders. American Journal of Speech-Language
Pathology, 8, 118-128.
Law, J., Garrett, Z., & Nye, C. (2004). The efficacy of treatment for children with
developmental speech and language delay/disorder: A meta-analysis. Journal of
Speech, Language, & Hearing Research, 47(4), 924-943.
Leitao, S., & Fletcher, J. (2004). Literacy outcomes for students with speech impairment:
Long-term follow-up. International Journal of Language & Communication
Disorders, 39(2), 245-256.
Leitao, S., Hogben, J., & Fletcher, J. (1997). Phonological processing skills in speech and
language impaired children. European Journal of Disorders of Communication,
32, 91-111.
Leonard, L. B. (1985). Unusual and subtle phonological behavior in the speech of
phonologically disordered children. Journal of Speech and Hearing Disorders,
50(1), 4-13.
114
Leonard, L. B., & Brown, B. L. (1984). Nature and boundaries of phonologic categories:
A case study of an unusual phonologic pattern in a language-impaired child.
Journal of Speech and Hearing Disorders, 49(4), 419-428.
Leonard, L. B., & McGregor, K. K. (1991). Unusual phonological patterns and their
underlying representations: A case study. Journal of Child Language, 18, 261-
271.
Lewis, B. A., & Freebairn, L. (1992). Residual effects of preschool phonology disorders
in grade school, adolescence, and adulthood. Journal of Speech and Hearing
Research, 35, 819-831.
Lewis, B. A., Freebairn, L. A., Hansen, A. J., Iyengar, S. K., & Taylor, H. G. (2004).
School-age follow-up of children with childhood apraxia of speech. Language,
Speech, and Hearing Services in Schools, 35, 122-140.
Lewis, B. A., Freebairn, L. A., & Taylor, H. (2002). Correlates of spelling abilities in
children with early speech sound disorders. Reading and Writing, 15(3-4), 389-
407.
Lewis, B. A., Shriberg, L. D., Freebairn, L. A., Hansen, A. J., Stein, C. M., Taylor, H. G.,
et al. (2006). The genetic bases of speech sound disorders: Evidence from spoken
and written language. Journal of Speech, Language, & Hearing Research, 49(6),
1294-1312.
Liberman, I. Y., Shankweiler, D., Fischer, F. W., & Carter, B. (1974). Explicit syllable
and phoneme segmentation in the young child. Journal of Experimental Child
Psychology, 18, 201-212.
115
Lieberman, P., Meskill, R. H., Chatillon, M., & Schupack, H. (1985). Phonetic Speech
Perception Deficits in Dyslexia. Journal of Speech and Hearing Research, 28(4),
480-486.
Lin, L. I. (1989). A concordance correlation coefficient to evaluate reproducibility.
Biometrics, 45(1), 255-268.
Locke, J. L., & Scott, K. K. (1979). Phonetically mediated recall in the phonetically
disordered child. Journal of Communication Disorders, 12, 17-32.
Lowe, R. J. (1994). Phonology assessment and intervention applications in speech
pathology. Baltimore, MD: Williams & Wilkins.
McDowell, K. D., Lonigan, C. J., & Goldstein, H. (2007). Relations Among
Socioeconomic Status, Age, and Predictors of Phonological Awareness. Journal
of Speech, Language, and Hearing Research, 50(4), 1079-1092.
McGrath, L. M., Pennington, B. F., Willcutt, E. G., Boada, R., Shriberg, L. D., & Smith,
S. D. (2007). Gene x Environment interactions in speech sound disorder predict
language and preliteracy outcomes. Developmental Psychopathology, 19(4),
1047-1072.
Metsala, J. L. (1999). Young children's phonological awareness and nonword repetition
as a function of vocabulary development. Journal of Educational Psychology,
91(1), 3.
Mody, M., Studdert-Kennedy, M., & Brady, S. (1997). Speech perception deficits in poor
readers: Auditory processing or phonological coding. Journal of Experimental
Child Psychology, 64, 199-231.
116
Munson, B., Edwards, J., & Beckman, M. E. (2005). Relationships between nonword
repetition accuracy and other measures of linguistic development in children with
phonological disorders. Journal of Speech, Language, & Hearing Research,
48(1), 61-78.
Nathan, L., Stackhouse, J., Goulandris, N., & Snowling, M. J. (2004). The development
of early literacy skills among children with speech difficulties: A test of the
"critical age hypothesis." Journal of Speech, Language, and Hearing Research,
47, 377-391.
National-Reading-Panel. (2000). Teaching children to read: An evidence-based
assessment of the scientific research literature on reading and its implications for
reading instruction. Washington, DC: National Institutes of Health.
Nicolson, R. I., Fawcett, A. J., & Dean, P. (2001). Developmental dyslexia: the cerebellar
deficit hypothesis. Trends in Neurosciences, 24(9), 508-511.
NIDCD. (2006, Jan 13). Statistics on voice, speech, and language. Retrieved October
12, 2006, 2006, from http://www.nidcd.nih.gov/health/statistics/vsl.asp#2
Nittrouer, S., & Burton, L. T. (2005). The role of early language experience in the
development of speech perception and phonological processing abilities:
Evidence from 5-year-olds with histories of otitis media with effusion and low
socioeconomic status. Journal of Communication Disorders, 38, 29-63.
Ohde, R. N., & Sharf, D. J. (1988). Perceptual categorization and consistency of
synthesized /r-w/ continua by adults, normal children and /r/-misarticulating
children. Journal of Speech and Hearing Research, 31, 556-568.
117
Oller, D. K., & Delgado, R. E. (2001). Logical International Phonetic Programs (LIPP).
Miami, FL: Intelligent Hearing Systems.
Oller, D. K., & Ramsdell, H. L. (2006). A weighted reliability measure for phonetic
transcription. Journal of Speech, Language, & Hearing Research, 49(6), 1391-
1411.
Pascoe, M., Stackhouse, J., & Wells, B. (2006). Persisting speech difficulties in children.
Chichester, England: Wiley.
Paul, R., & Shriberg, L. D. (1982). Associations between phonology and syntax in speech
delayed children. Journal of Speech and Hearing Research, 25, 536-547.
Phillips, B. M., & Torgesen, J. K. (2006). Phonemic awareness and reading: Beyond the
growth of initial reading accuracy. In D. K. Dickinson & S. B. Neuman (Eds.),
Handbook of early literacy research (pp. 101-112). New York, NY: Guilford
Press.
Preston, J. L., & Edwards, M. L. (2006). Rapid naming abilities of adolescents with
residual speech sound errors, American Speech-Language-Hearing Association
Annual Convention. Miami, FL.
Preston, J. L., & Edwards, M. L. (2007). Phonological processing skills of adolescents
with residual speech sound errors. Language, Speech and Hearing Services in
Schools, 38, 297-308.
Raitano, N. A., Pennington, B. F., Tunick, R. A., Boada, R., & Shriberg, L. D. (2004).
Pre-literacy skills of subgroups of children with speech sound disorders. Journal
of Child Psychology and Psychiatry, 45(4), 821-835.
118
Richardson, U., Thomson, J. M., Scott, S. K., & Goswami, U. (2004). Auditory
processing skills and phonological representation in dyslexic children. Dyslexia:
An International Journal of Research and Practice Selected Papers from the 6th
Conference of the British Dyslexia Association, 10(3), 215-233.
Roy, P., & Chiat, S. (2004). A prosodically controlled word and nonword repetition task
for 2- to 4-year-olds: Evidence from typically developing children. Journal of
Speech, Language, and Hearing Research, 47, 223-234.
Rvachew, S. (1994). Speech perception training can facilitate sound production learning.
Journal of Speech and Hearing Research, 37, 347-357.
Rvachew, S. (2006). Longitudinal predictors of implicit phonological awareness skills.
American Journal of Speech-Language Pathology, 15(2), 165-176.
Rvachew, S. (2007). Phonological processing and reading in children with speech sound
disorders. American Journal of Speech-Language Pathology, 16(3), 260-270.
Rvachew, S., Chiang, P.-Y., & Evans, N. (2007). Characteristics of speech errors
produced by children with and without delayed phonological awareness skills.
Language, Speech and Hearing Services in Schools, 38(1), 60-71.
Rvachew, S., & Grawburg, M. (2006). Correlates of phonological awareness in
preschoolers with speech sound disorders. Journal of Speech, Language, &
Hearing Research, 49(1), 74-87.
Rvachew, S., Nowak, M., & Cloutier, G. (2004). Effect of phonemic perception training
on the speech production and phonological awareness skills of children with
expressive phonological delay. American Journal of Speech-Language Pathology,
13(3), 250-263.
119
Rvachew, S., Ohberg, A., Grawburg, M., & Heyding, J. (2003). Phonological awareness
and phonemic perception in 4-year-old children with delayed expressive
phonology skills. American Journal of Speech-Language Pathology, 12(4), 463-
471.
Rvachew, S., Rafaat, S., & Martin, M. (1999). Stimulability, speech perception skills, and
the treatment of phonological disorders. American Journal of Speech-Language
Pathology, 8, 33-43.
Savage, R. S., Frederickson, N., Goodwin, R., Patni, U., Smith, N., & Tuersley, L.
(2005). Relationships among rapid digit naming, phonological processing, motor
automaticity, and speech perception in poor, average and good readers and
spellers. Journal of Learning Disabilities, 38(1), 12-28.
Schatschneider, C., Fletcher, J. M., Francis, D. J., Carlson, C. D., & Foorman, B. R.
(2004). Kindergarten prediction of reading skills: A longitudinal comparative
analysis. Journal of Educational Psychology, 96(2), 265-282.
Sénéchal, M., Ouellette, G., & Young, L. (2004). Testing the concurrent and predictive
relations among articulation accuracy, speech perception, and phoneme
awareness. Journal of Experimental Child Psychology, 89(3), 242-269.
Sharf, D. J., Ohde, R. N., & Lehman, M. E. (1988). Relationship between the
discrimination of /w-r/ and /t-d/ continua and the identification of distorted /r/.
Journal of Speech and Hearing Research, 31(2), 193-206.
Shriberg, L. D. (1994). Five subtypes of developmental phonological disorders. Clinics in
Communication Disorders, 4(1), 38-53.
120
Shriberg, L. D. (1997). Developmental phonological disorders. In B. W. Hodson & M. L.
Edwards (Eds.), Perspectives in Applied Phonology (pp. 105-131). Gaithersburg,
MD: Aspen Publishers, Inc.
Shriberg, L. D., Austin, D., Lewis, B. A., McSweeny, J. L., & Wilson, D. L. (1997a). The
percentage of consonants correct (PCC) metric: Extensions and reliability data.
Journal of Speech, Language, and Hearing Research, 40, 708-722.
Shriberg, L. D., Campbell, T. F., Karlsson, H. B., Brown, R. L., McSweeny, J. L., &
Nadler, C. J. (2003a). A diagnostic marker for childhood apraxia of speech: The
lexical stress ratio. Clinical Linguistics & Phonetics, 17(7), 549-574.
Shriberg, L. D., Green, J. R., Campbell, T. F., McSweeny, J. L., & Scheer, A. R. (2003b).
A diagnostic marker for childhood apraxia of speech: the coefficient of variation
ratio. Clinical Linguistics and Phonetics, 17(7), 575-595.
Shriberg, L. D., & Kwiatkowski, J. (1982). Phonological disorders III: A procedure for
assessing severity of involvement. Journal of Speech and Hearing Disorders, 47,
242-256.
Shriberg, L. D., & Kwiatkowski, J. (1994). Developmental phonological disorders I: A
clinical profile. Journal of Speech, Language, and Hearing Research.
Shriberg, L. D., Kwiatkowski, J., & Gruber, F. A. (1994). Developmental phonological
disorders II: Short-term speech-sound normalization. Journal of Speech and
Hearing Research, 37.
Shriberg, L. D., Kwiatkowski, J., & Hoffmann, K. (1984). A procedure for phonetic
transcriptions by consensus. Journal of Speech & Hearing Research, 27(3), 456-
465.
121
Shriberg, L. D., Lewis, B. A., McSweeny, J. L., & Wilson, D. L. (1997b). The speech
disorders classification system (SDCS): Extensions and lifespan reference data.
Journal of Speech, Language, and Hearing Research, 40, 723-740.
Shriberg, L. D., Lewis, B. A., Tomblin, J. B., McSweeny, J. L., Karlsson, H. B., &
Scheer, A. R. (2005). Toward diagnostic and phenotype markers for genetically
transmitted speech delay. Journal of Speech, Language, & Hearing Research,
48(4), 834-852.
Shriberg, L. D., Lohmeier, H. L., McSweeny, J. L., & Kwiatkowski, J. (2006). A
nonword repetition task for speakers with misarticulations: The Syllable
Repetition Task. Poster presented at the Annual Convention of the American
Speech Language Hearing Association, Miami, FL.
Shuster, L. I. (1998). The perception of correctly and incorrectly produced /r/. Journal of
Speech, Language, and Hearing Research, 41, 941-950.
Smit, A. B., Hand, L., Freilinger, J. J., Bernthal, J. E., & Bird, A. (1990). The Iowa
articulation norms project and its Nebraska replication. Journal of Speech and
Hearing Disorders, 55(4), 779-797.
Smith, A. B., Roberts, J., Lambrecht Smith, S., Locke, J. L., & Bennett, J. (2006).
Reduced speaking rate as an early predictor of reading disability. American
Journal of Speech-Language Pathology, 15(3), 289-297.
Snow, C. E., Burns, M. S., & Griffin, P. (1998). Preventing reading difficulties in young
children. Washington, DC: National Academy Press.
Snowling, M. J. (2000). Dyslexia (Vol. 2nd). Oxford: Blackwell Publishers.
SPSS. (2006). SPSS for Windows, Rel. 15.0.
122
Stackhouse, J. (1997). Phonological awareness: Connecting speech and literacy
problems. In B. W. Hodson & M. L. Edwards (Eds.), Perspectives in applied
phonology (pp. 157-196). Gaithersburg, MD: Aspen Publishers, Inc.
Stackhouse, J., & Wells, B. (1997). Children's speech and literacy difficulties. London:
Whurr Publishers Ltd.
Strange, W., & Broen, P. A. (1981). The relationship between perception and production
of /w/, /r/, and /l/ by three-year-old children. Journal of Experimental Child
Psychology, 31, 81-102.
Stringer, R. W., Toplak, M. E., & Stanovich, K. E. (2004). Differential relationships
between RAN performance, behavior ratings, and executive function measures:
Searching for a double dissociation. Reading and Writing: An Interdisciplinary
Journal, 17, 891-914.
Stringfellow, K., & McLeod, S. (1994). Using a facilitating phonetic context to reduce an
unusual form of gliding. Language, Speech and Hearing Services in Schools, 25,
191-193.
Sutherland, D., & Gillon, G. T. (2005). Assessment of phonological representations in
children with speech impairment. Language, Speech and Hearing Services in
Schools, 36, 294-307.
Swan, D., & Goswami, U. (1997a). Phonological awareness deficits in developmental
dyslexia and the phonological representations hypothesis. Journal of
Experimental Child Psychology, 66, 18-41.
Swan, D., & Goswami, U. (1997b). Picture naming deficits in developmental dyslexia:
The phonological representations hypothesis. Brain and Language, 56, 334-353.
123
Swanson, H. L., Trainin, G., Necoechea, D. M., & Hammill, D. D. (2003). Rapid naming,
phonological awareness, and reading: A meta-analysis of the correlation
evidence. Review of Educational Research, 73(4), 407-440.
Tangel, D. M., & Blachman, B. A. (1992). Effect of phoneme awareness instruction on
kindergarten children's invented spelling. Journal of Reading Behavior, 24(2),
233-261.
Torgesen, J. K., & Wagner, R. K. (1997). Comprehensive Test of Phonological Processes
in Reading. Austin, TX: PRO-ED.
Troia, G. A., Roth, F. P., & Yeni-Komshian, G. H. (1996). Word frequency and age
effects in normally developing children's phonological processing. Journal of
Speech and Hearing Research, 39, 1099-1108.
Tyler, A. A., Edwards, M. L., & Saxman, J. H. (1990). Acoustic validation of
phonological knowledge and its relationship to treatment. Journal of Speech and
Hearing Disorders, 55(2), 251-261.
Usdan, V. L. (1978). Utilization of the "straw technique" for correction of the lateral lisp.
Language, Speech and Hearing Services in Schools, 9(1), 5-7.
Wagner, R. K., & Torgesen, J. K. (1987). The nature of phonological processing and its
causal role in the acquisition of reading skills. Psychological Bulletin, 101(2),
192-212.
Walley, A. C., Metsala, J. L., & Garlock, V. M. (2003). Spoken vocabulary growth: Its
role in the development of phoneme awareness and early reading ability. Reading
and Writing: An Interdisciplinary Journal, 16(1-2), 5-20.
124
Watson, B. U., & Miller, T. K. (1993). Auditory perception, phonological processing, and
reading ability/disability. Journal of Speech and Hearing Research, 36(4), 850-
863.
White, H. R. (1997). Comparison of computer-assisted and by-hand phonological
analyses for young children with disordered phonology. Unpublished Masters
Thesis, Syracuse University, Syracuse, NY.
Wiig, E. H., Secord, W. A., & Semel, E. (2004). Clinical Evaluation of Language
Fundamentals Preschool (2 ed.). San Antonio, TX: Harcourt/Psych Corp.
Wolf, M., & Bowers, P. G. (1999). The double-deficit hypothesis for developmental
dyslexias. Journal of Educational Psychology, 91(3), 415-438.
Wolf, M., O'Rourke, A. G., Gidiney, C., Lovett, M., Cirino, P., & Morris, R. (2002). The
second deficit: An investigation of the independence of phonological naming-
speed deficits in developmental dyslexia. Reading and Writing: An
Interdisciplinary Journal, 15, 43-72.
Wolk, L. (1990). An investigation of stuttering and phonological difficulties in young
children. Unpublished Doctoral dissertation, Syracuse University, Syracuse.
Wolk, L., Edwards, M. L., & Conture, E. G. (1993). Coexistence of stuttering and
disordered phonology in young children. Journal of Speech and Hearing
Research, 36, 907-917.
125
Appendix A: Transcription Rules and Coding Sound Changes
1) Do not count errors on morphological endings that are added (e.g., toys, shrimps).
2) Transcribe and code any errors in the second part of a two-word phrase (e.g., plant it;
this one).
3) If part of an utterance is overlaid by another speaker, do your best to transcribe
accurately. However, if uncertain, give the child as much credit as possible (i.e.,
count overlaid sound as correct).
4) Do not penalize the child for dialectally acceptable forms
a) Count t � Ȏ as correct between stressed and unstressed vowels (e.g.,
hippopotamus; spaghetti). However, t� d is a typical voicing error.
b) Allow the following alternations in the target form (i.e., don’t consider them as
errors):
i) Beige: /beȴ/ or /beȢ/ (#4)
ii) Newspaper: /nuspepǪ/ or /nuzpepǪ/ (#13)
iii) Garage: /gǩrǡȢ/ or /gǩrǡȴ/ (#70)
5) For “Leaf” allow: /lif/ or /liv/ (#72)
(/liv/ is allowed due to “back formation” from the plural leaves)
6) For “Quack” allow /kwæk/ or /kwækwæk/ (#100)
Adjust denominator (total number of consonants) accordingly
126
7) Do not penalize for affrication/palatalization of /tr, dr, tw, dw/ clusters, or if these
could be intermediate steps in a sound change (due to dialect)
a) Ex: /tri/ � [tȓri] is no error
b) Ex: /twǺnz/ � [tȓwǺnz] is no error
8) “Surface errors” should be broken down into separate sound changes, each of which
is coded as a typical sound change, an atypical sound change, and/or a distortion.
That is, an error may involve interacting changes (more than one typical, atypical,
and/or distortion errors). Code surface errors based on the smallest number of sound
changes to arrive at the child’s production. If multiple “paths” of interacting changes
are possible to account for a child’s production, choose the one with the fewest
atypical changes (see Appendix B).
9) Use the list of error pattern descriptions below to account for a child’s production of a
word. To the extent possible, try to account for all component feature changes (e.g.,
manner, place, voicing).
10) Typical and atypical changes can be interacting, as can multiple atypical changes.
a) Ex: “ladder” [mædǩ] would be gliding of /l/ � /w/ (typical) + Nasalization of
/w/ � [m] (atypical)
b) Ex: “yawn” [vǤn] would be Glide Interchange /j/ � /w/ (atypical) and Frication
of a Glide /w/ � [v] (atypical)
127
11) Glottal stops:
a) Not considered an error at the end of a word/compound word if replacing /t/,
because that is an acceptable dialectal variant
i) Ex: “squirtgun” [skwǭȤ gȜn] is not an error
ii) Ex: “basket” [bǽskǺȤ] is not an error
iii) Ex: “elephant” [ǫlǝfǺnȤ] is no error, but [ǫlǝfǺȤ]] is Nasal Cluster Reduction,
but no penalization for glottal stop replacing /t/
b) Atypical sound changes: the intrusion of a glottal stop, or a glottal stop
substitution in word-initial or intervocalic position or in place of a consonant
(other than a final /t/) in a final cluster or final position
i) Ex: “teeth” [Ȥiθ] is an atypical sound change
ii) Ex: “ladder” [lǽȤǪ] is an atypical sound change (intervocalic substitution)
iii) Ex: “chicken” [tȓǺȤkǫn] is an atypical sound change (intrusion)
iv) Ex: “ketchup” [kǽtȓǝȤ] is an atypical sound change
v) Ex: “tractor” [trǽȤtǪ] is an atypical sound change
12) Partial voicing/devoicing errors should be transcribed, but they are not considered as
distortions; they will be considered acceptable phonetic variants.
13) Assimilation errors are considered typical, whether they involve partial assimilation
(one or several features) or complete assimilation (all features) (see section below on
assimilations).
128
TYPICAL PHONEMIC SOUND CHANGES:
(based on Edwards & Shriberg, 1983; Lowe, 1994)
Typical Syllable Structure Changes
Examples
Sound Change Definition Target Word Production
Final Consonant Deletion (FCD)
Final singleton consonants are deleted in words or compound words
spoon /spun/
ice cream
/aǺs krim/
[spu]
[aǺ krim] S Cluster Reduction (SCR) - initial
/s/ is deleted in a consonant cluster in syllable-initial position
spoon /spun/
snake /snek/
[pun]
[nek]
S Cluster Reduction (SCR) - final
Either /s/ or another phoneme is deleted in word-final position
dentist
/dǫntǺst/
mailbox
/melbǤks/
[dǫntǺs] or
[dǫntǺt]
[melbǤk] or
[melbǤs] Liquid Cluster Reduction (LCR)
Deletion of /r/ or /l/ in any liquid cluster (syllable initial or final)
three
/ſri/
present
/prǫzǺnt/
skateboard
/sketbord/
black
/blæk/
sled
/slǫd/
[ſi]
[pǫzǺnt]
[sketbod]
[bæk]
[sǫd] Glide Cluster Reduction (GCR)
Deletion of /w/ or /j/ in a cluster
twins /twǺnz/
vacuum
/vǽkjum/
[tǺnz]
[vǽkum]
129
Nasal Cluster Reduction (NCR)
Deletion of either element of a nasal cluster
elephant
/ǫlǩfǺnt/
[ǫlǩfǺt] or
[ǫlǩfǺn]
Consonant Sequence Reduction (CSR) (Hodson, 1997)
Deletion of a consonant in a sequence that crosses syllable or word boundaries
helicopter
/hǫlǺkǤptǪ/
tractor
/trǽktǪ/
[hǫlǺkǤpǪ] or
[hǫlǺkǤtǪ]
[trǽtǪ] or
[trǽkǪ]
Weak Syllable Deletion
Deletion of unstressed syllable
banana
/bǩnǽnǩ/
hippopotamus
/hǺpǩpǡtǩmǺs/
[nǽnǩ]
[hǺpǩpǡmǺs]
Epenthesis (EP)
Insertion of a vowel, often a schwa, in consonant clusters. Does not include insertion of other phonemes in other positions (see atypical intrusive consonants, vowels, syllables)
black /blæk/
plate /plet/
[bǩlæk]
[pȚlet]
Segment Coalescence (Seg-COA)
Features from two adjacent phonemes combine to form a new segment that retains features of both phonemes. In our definition, segment coalescence can involve place and manner features, but not voicing.
black /blæk/
spider/spaǺdǪ/
zebra /zibrǩ/
- - - - - - - -
BUT:
plate /plet/
[væk] ([v] has
labial feature
of /b/ and
continuant
feature of /l/)
[faǺdǪ]
[zivǩ]
- - - - - - - -
[vet] is Seg-
COA + Voicing
130
Syllable Coalescence (Syl-COA)
Segments from two adjacent syllables combine, with a weak vowel (and sometimes a following sonorant) being deleted
garage
/gǩrǡȢ/
banana
/bǩnǽnǩ/
[gǡȢ]
[bǽnǩ]
Reduplication (REDUP)
Entire stressed syllable is repeated, or a simplified version of it
pudding
/pȚdǺŋ/
[pȚpȚ]
Typical Place of Articulation Changes
Examples
Sound Change Definition Target Word Production
Depalatalization/ Palatal Fronting (DEPAL)
Palatal obstruent is replaced by an alveolar
chocolate
/tȓǤklǺt/
cage /keȴ/
[tsǤklǺt]
[kedz] Velar Fronting Velar phoneme
replaced by alveolar cage /keȴ/
green /grin/
[teȴ]
[drin] Labialization (LAB)
Alveolar or interdental becomes labial. NOTE: labial replacing palatal or velar is atypical (see atypical place changes)
three /θri/
toy /tǤǺ/
scissors
/sǺzǪz/
[fri]
[pǤǺ]
[fǺzǪz] Alveolarization (ALV)
Interdental or labial consonant is replaced by alveolar consonant
thumb /θȜm/
beige /beȢ/
[sȜm]
[deȢ]
131
Typical Manner of Articulation Changes
Examples
Sound Change Definition Target Word Production
Gliding of Liquids (GL)
liquids /r, l/ become glides [w] or [j]
rabbit /ræbǺt/
leaf /lif/
[wæbǺt]
[wif] or [jif] Gliding of Fricatives (GF)
Homorganic glide replaces a fricative in intervocalic position only (if fricatives are glided in word-initial position, this is atypical)
television
/tǫlǩvǺȢǺn/
[tǫlǩwǺȢǺn] or
[tǫlǩvǺjǺn]
Stopping (ST) Fricatives or affricates become homorganic stops (i.e., same place of articulation) Stopping of palatal fricatives and affricates is just one change (ST, not ST + DEPAL) Stopping of interdentals to alveolar is one change. If resulting stop is interdental, do not count as distortion
zebra /zibrǩ/
leaf/lif/
cage /keȴ/
catch /kætȓ/
three /θri/
thimble /θǺmblʜ/
[dibrǩ]
[lip]
[ked]
[kæt]
[ti] or [tʝi]
[tǺmblʜ] or
[tʝʝǺmblʜ]
Vocalization (VOC)
Postvocalic /l, r/ and syllabic liquids are replaced by a vowel
thimble /θǺmblʜ/
spider/spaǺdǪ/
[θǺmbo]
[spaǺdǩ]
Deaffrication (DEAFF)
Affricates are replaced by homorganic fricative
cage /keȴ/
chocolate
/tȓǤklǺt/
[keȢ]
[ȓǤklǺt]
132
Affrication of fricatives (AFF)
Fricatives become homorganic affricates
washing /wǤȓǺŋ/
television
/tǫlǩvǺȢǺn/
zebra /zibrǩ/
leaf /lif/
[wǤtȓǺŋ]
[tǫlǩvǺȴǺn]
[dzibrǩ]
[lipf]
Typical Voicing Changes
Examples
Sound Change Definition Target Word Production
Initial Voicing (IV)
Voiceless obstruents become voiced before a sonorant. Note: partial voicing is not considered an error
cage /keȴ/
truck /trȜk/
[geȴ]
[drȜk]
Final Devoicing (FD)
Voiced obstruents become voiceless at the end of a word or syllable. Note: partial devoicing is not considered an error
cage /keȴ/
garage
/gǩrǡȢ/
[ketȓ ]
[gǩrǡȓ]
133
Other Typical Changes:
Examples
Sound Change Definition Target Word Production
Metathesis (MET)
Two consonants in a word exchange positions Note: if the change results in a phonotactic violation, consider it atypical
animals
/ænǺmǩlz/
BUT:
spring /sprǺŋ/
[æmǺnǩlz]
[pswǺŋ] is
atypical
because
[psw]
clusters are
not allowed
in English
134
Assimilations (ASSIM) are Typical Changes
One phoneme takes on one or more features of another phoneme in the word: velar, alveolar, labial, palatal, nasal, liquid, fricative, continuant, etc. It may involve place and/or manner, but not voicing, as defined for this study
Velar ASSIM:
Palatal ASSIM:
Nasal ASSIM:
Liquid ASSIM:
Frication ASSIM:
guitar /gǩtar/
shovel /ȓȜvǩl/
banana /bǩnǽnǩ/
ladder /lædǪ/
beige /beȢ/
[gǩkar]
[ȓȜȢǩl]
[mǩnǽnǩ]
[lælǪ]
[veȢ]
Even if assimilations involve more than one feature, they count as just one typical change Complete assimilation (in which all features assimilate) is preferred if it can reduce the total number of steps
ASSIM to place
and manner:
Complete ASSIM:
GL + Complete
ASSIM is
Preferred over GL
+ ST + Lab ASSIM
to explain initial
[p]
leaf /lif/
flag /flæg/
shrimp /ȓrǺmp/
[vif]
[glæg]
[pwǺmp]
135
ATYPICAL SOUND CHANGES: (Indicated by an asterisk *)
Atypical Syllable Structure Changes
Examples
Sound Change Definition Target Word Production
Atypical /s/ Cluster Reduction (*ASCR)
In word-initial /s/ clusters, the /s/ remains and the stop or nasal is deleted. ------------------------ This is not applied to /sl/ and /sw/ clusters
snowman
/snomæn/
school /skul/
--------------
BUT:
sled /slǫd/
swim /swǺm/
[somæn]
[sul]
-----------
[sǫd] is LCR
[sǺm] is GCR
Atypical Liquid Cluster Reduction (*ALCR; Lowe, 1994)
In liquid clusters, the liquid is retained This can also be applied to syllable-final liquid clusters
tree /tri/
plant /plænt/
twelve /twǫlv/
skateboard
/sketbord/
[ri]
[lænt]
[twǫl]
[sketbor]
Atypical Glide Cluster Reduction (*AGCR)
In stop + glide cluster, the glide is retained
twelve /twǫlv/
twin /twǺn/
[wǫlv]
[wǺn]
Initial Consonant Deletion (*ICD; Dodd & Iacano, 1989)
Word-initial singleton consonants are deleted This can be evident when both elements of an intial cluster are deleted
toy /toǺ/
leaf /lif/
plant /plænt/
[oǺ]
[if]
[ænt] is LCR
and ICD
136
Medial (intervocalic) Consonant Deletion (*MCD; Dodd & Iacano, 1989)
Intervocalic consonants are deleted
ladder /lædǪ/
scissors /sǺzǪz/
[læǪ]
[sǺǪz]
Addition of consonants, vowels or syllables (*ADD)
Individual consonants, vowels, or whole syllables are added Note: not used between members of consonant clusters (see Epenthesis)
shovel /ȓȜvǩl/
spring /sprǺŋ/
beige /beȢ/
[ȓȜvǩvǩl]
[sprǺŋk]
[beȢa]
Migration (*MIG) (Leonard & McGregor, 1991)
A consonant is moved to another part of the word
soap /sop/
[ops]
Strong Syllable Deletion (*SSD)
Syllable/ vowel with primary or secondary stress is deleted.
shovel /ȓȜvǩl/
basket /bǽskǺt/
hippopotamus
/hǺpǩpǡtǩmǺs/
[vo]
[kǺt]
[pǡtǩmǺs] is
WSD + *SSD
137
Atypical Place of Articulation Changes
Examples
Sound Change Definition Target Word Production
Glottal Replacement (*GR; Dodd & Iacano, 1989)
Glottal stop [Ȥ] replaces a consonant (except syllable final /t/, in which case no error is counted) This is also used if a glottal stop replaces /h/
leaf /lif/
chocolate
/tȓǤklǺt/
BUT:
hippo /hǺpo/
[Ȥif] or [liȤ]
[ȤǤklǺt]
[tȓǤklǺȤ] is NOT
an error
[ȤǺpo] is an
atypical error
Backing (*BACK)
A labial, dental, alveolar, or palatal is backed to a velar. Used when velar assimilation is not possible.
toy /toǺ/
banana
/bǩnænǩ/
[koǺ]
[gǩnænǩ]
Palatalization (*PAL)
A non-palatal fricative or affricate (usually, but not restricted to alveolar) becomes a palatal phoneme. Used when palatal assimilation is not possible.
scissors
/sǺzǪz/
zebra /zibrǩ/
teeth /tiθ/
BUT:
teeth /tiθ/
[ȓǺzǪz] or
[sǺȢǪz]
[Ȣibrǩ]
[tiȓ]
[tȓiθ] is
*Affrication +
*PAL
138
Atypical Labialization (*ALAB)
Velar or palatal phoneme becomes labial. Used when labial assimilation is not possible. Note that alveolar or interdentals becoming labial is typical
guitar /gǩtar/
catch /kætȓ/
[bǩtar]
[pætȓ] or
[kæp] (ST+
*LAB)
Glide Interchange (*GLINT)
Interchange between /j/ and /w/. Used when complete assimilation is not possible.
yawn /jǤn/
washing
/wǤȓǺŋ/
BUT:
vacuum cleaner
/vækjum
klinǪ/
yoyo /jojo/
[wǤn]
[jǤȓǺŋ]
[vækwum
kwinǪ] is GL +
complete
Assim.
[wowo] is
GLINT +
complete
Assim
Liquid Interchange (*LIQINT)
Interchange between /r/ and /l/
rabbit /ræbǺt/
leaf /lif/
green /grin/
[læbǺt]
[rif]
[glin]
139
Atypical Manner of Articulation Changes
Examples Sound Change
Definition
Target Word Production
Denasalization (*DENAS)
Nasal phoneme become homorganic voiced stop (Dodd & Iacano, 1989; Shriberg, 1993)
nose /noz/ [doz]
Nasalization (*NAS)
Non-nasal phoneme becomes homorganic nasal. Occurs only when nasal assimilation is not possible (cf. Shriberg, 1993)
leaf /lif/
beige /beȢ/
[nif]
[meȢ]
Fricatives Replace Stops (*FRS)
Fricative replaces homorganic stop. Only occurs when assimilation of fricatives is not possible (cf. Lowe, 1994)
toy /toǺ/
crib /krǺb/
[soǺ]
[krǺv]
Liquids Replacing Glides (*LIQ)
Glides become liquids (Stringfellow & McLeod, 1994)
you /ju/
twin /twǺn/
[lu]
[trǺn]
Tetism (*TET) (Edwards & Shriberg, 1983)
/f/ � [t] Used when assimilation to alveolar stop is not possible
feather /fǫðǪ/
leaf /lif/
[tǫðǪ]
[lit]
Atypical Gliding of Intervocalic Consonants (*AGL)
Intervocalic consonants (other than fricatives) are replaced by glides
ladder /lǽdǪ/
rabbit /rǽbǺt/
[lǽjǪ]
[rǽwǺt]
140
Atypical Stopping of Liquids or Glides (*AST)
Glide or liquid becomes a homorganic stop. Note that place of articulation changes should be counted as separate sound changes.
leaf /lif/
washing
/wǤȓǺŋ/
BUT:
yawn /jǤn/
leaf /lif/
/dif/
[bǤȓǺŋ]
[bǤn] is *AST
+ LAB
[gif] is *AST
and *BACK
Atypical Voicing Changes
Examples
Sound Change Definition Target Word Production
Initial/Prevocalic Devoicing (*IDEV)
Prevocalic obstruents becomes devoiced (Dodd & Iacano, 1989)
dog /dǤg/ [tǤg]
Final Voicing (*FV)
Postvocalic/final obstruents become voiced
hat /hæt/ [hæd]
141
Interacting Atypical Sound Changes
Atypical changes may interact with atypical changes, typical changes or distortions.
Examples
Sound Change Definition Target Word Production
Atypical + Atypical
*Nasalization + *Atypical Labialization
guitar
/gǺtar/
[mǺtar]
Atypical + Distortion
*Liquid Interchange + Distorted /r/
yellow /jǫlo/ [jǫrwo]
Atypical + Typical
*Palatalization + Initial Voicing
scissors
/sǺzǪz/
/ȢǺzǪz/
If an atypical change is repeated more than once in a word, it is coded as one atypical change plus assimilation (typical) Assimilations should be considered prior to considering atypical changes
*LIQ + ASSIM *BACKING + VELAR ASSIM (twice) For /l/ � [d], prefer ASSIM (to alveolar stop) over *AST For /b/ � [v], prefer ASSIM (to labial fricative) over *FRS
yoyo /jojo/
dentist
/dǫntǺst/
telephone
/tǫlǩfon/
thimble
/θǺmbǩl/
[lolo]
[gǫŋkǺst]
[tǫdǩfon]
[fǺmvǩl]
142
Examples of Consonant Cluster Changes: Coalescence, Assimilation,
other Changes
Do not penalize for affrication/palatalization of /tr, dr, tw, dw/ clusters, or if these could
be intermediate steps in a sound change.
a) Ex: /θri/ � [tȓi] can be coded as Stopping [tri], with no penalization for
affrication [tȓri], followed by Liquid Cluster Reduction to [tȓi] .
b) Ex: /twǺnz/ � [tȓwǺnz] no error is coded
/twǺnz/ � [ȴǺnz] is Initial Voicing + Glide Cluster Reduction
Stop + Liquid or Stop + Glide resulting in a homorganic Fricative + Liquid or Fricative + Glide is Continuant Assimilation
Continuant ASSIM:
Coalescence:
(features from
adjacent segments
combine)
ASSIM to Place of
Artic:
Gliding of Liquid +
Coalesc.
twin /twǺn/
plate /plet/
princess
/prǺnsǫs/
flag /flæg/
plate /plet/
flag /flæg/
tree /tri/
drum/drȜm/
[swǺn]
[flet]
[frǺnsǫs] or
[fwǺnsǫs]
[sæg]
[fet]
[slæg]
[fi]
[bȜm]
INTERACTIONS:
Initial Devoicing +
LCR + *Affric. of stop
Gliding + ASSIM
(continuant and labial
features)
bridge /brǺȴ/
crib /krǺb/
green /grin/
[pfǺȴ]
[fwǺb]
[vwin]
143
Gliding + ASSIM
(continuant and labial
features) + Initial
Devoicing
Gliding +
Depalatalization
Liquid Cluster Red +
Deaff
/s/ Cluster Red +
Deaffrication + Liquid
Cluster Red
bridge /brǺȴ/
drive /draǺv/
tree /tri, tȓri/
drive /ȴraǺv/
tree /tri, tȓri/
drive /ȴraǺv/
string /stȓrǺŋ/
strawberry
/strǤbǫri/
[fwǺȴ]
[fwaǺv]
[tswi]
[dzwaǺv]
[ȓi]
[ȢaǺv]
[ȓǺŋ]
[ȓǤbǫri]
144
DISTORTION ERRORS:
Only clinically significant distortions (not appropriate for the context) are considered as
distortion errors. They can be marked for any consonant (not just sibilants and liquids).
• NOTES ON DISTORTIONS:
o Partial voicing and partial devoicing are not considered to be distortion
errors
o Only one distortion is coded on a particular phoneme
Sibilant Distortions
Examples
Sound Change Target Word Production
Lateralization Although this is sometimes considered atypical, it will be considered a distortion in this study because of the suspected motoric (rather than linguistic) involvement (Usdan, 1978)
soap
/sop/
zebra
/zibrǩ/
[sʢop] or [Ǽop]
[zʢibrǩ]
Dentalization/
Interdentalization
Includes substitution of interdental phonemes for sibilants
soap /sop/
zebra
/zibrǩ/
[sʝop] or [θop]
[zʝebra] [ðibrǩ]
Other Includes salivary (wet), whistled, flat tongue position, and other/ nonspecific sibilant distortions
145
Rhotic Distortions
Examples
Sound Change Target Word Production
Derhoticization
of /r, Ǫ, ǭ/ (Shriberg, 1994)
rabbit /ræbǺt/
cracker /kræǪ/
[rʢæbǺt]
[krʢæǪʢ]
Labialization of /r/ (Shriberg, 1993)
rabbit /ræbǺt/ [rwæbǺt]
Other Other specific or nonspecific rhotic distortions are possible
Other Distortions
Note: This list is not exhaustive, but is illustrative of the types of distortions observed
Sound Change Examples
Partly Nasalized rabbit guitar
Partly Denasalized mailbox banana
Rhoticization of /w/ washing twins
Dentalization of nonsibilant alveolars
screwdriver toy dinosaur
Distortions Interacting with Other Sound Changes
Examples
Target Word Production
Depalatalization + Sibilant Distortion
shovel /ȓȜvǩl/ [θȜvǩl] or [sʝȜvǩl]
Alveolarization + Sibilant Distortion
leaf /lif/ [lisʝ]
146
Appendix B: Errors with Interacting Sound Changes: Which is Preferred?
Sometimes more than one “path” could account for a particular sound change.
Examples are shown below to show how one path was selected Most of these different
paths relate to scoring consonant clusters. Note that this does not claim that the child is
going through these steps, but just that these are phonetically plausible paths connecting
the child’s production to the corresponding adult form (called derivations).
Abbreviations are found in Appendix A. An asterisk (*) indicates an atypical sound
change.
a. No difference in total scoring. If there is no difference in the resulting score in any of
the categories, either path is deemed acceptable.
Ex: P43 word #35 “drive” /draǺv/ � [baǺf].
Path #1 Path #2
/draǺv/ GL /draǺv/ LCR
[dwaǺv ] COA [daǺv] Labial ASSIM
[baǺv] FD [baǺv] FD
[baǺf] [baǺf]
RESULT: 3 Typical RESULT: 3 Typical
147
b. One path results in more atypical sound changes than another. The selected path is
always the one with fewer atypical sound changes (so as not to penalize the child).
Ex: P35 #116 “string” /strǺŋ/ � [ȓwǺŋ]. Recall /stȓrǺŋ/ is an allowable target
Preferred Path Non-Preferred Path
/stȓrǺŋ/ /s/ CR /stȓrǺŋ/ *ASCR
[tȓrǺŋ] GL [srǺŋ] GL
[tȓwǺŋ] DEAFF [swǺŋ] *PAL
[ȓwǺŋ] [ȓwǺŋ]
RESULT: 3 Typical RESULT: 2 Atypical, 1 Typical
Ex: P43 word #36 “clown” / klaȚn/ � [baȚn].
Preferred Path Non-Preferred Path
/klaȚn/ GL /klaȚn/ LCR
[kwaȚn] COA [kaȚn] IV
[baȚn] [gaȚn] ****LAB (atypical)
[baȚn]
RESULT: 2 Typical RESULT: 2 Typical, 1 Atypical
148
Ex: P41 #20 “zebra” /zibrǩ/ � [zivǩ]
Preferred Path Non-Preferred Path
/zibrǩ/ COA /zibrǩ/ ASSIM (Cont.)
[zivǩ] [zivrǩ] LCR
[zivǩ]
RESULT: 1 Typical RESULT: 1 Typical, 1 Atypical
c. If one path results in more typical sound changes than another (but atypical
changes remain the same), chose the path with the smallest number of typical errors.
Ex: P43 word #55 “queen” /kwin/ � [bind]
Preferred Path Non-Preferred Path
/kwin/ COA /kwin/ ASSIM (labial)
[bin] *ADD [pwin] IV
[bind] [bwin] GCR
[bin] *ADD
[bind]
RESULT: 1 Typical, 1 Atypical RESULT: 3 Typical, 1 Atypical
149
Ex: P 41 word #42 “shrimp” /ȓrǺmp/ � [pwǺmp]
Preferred Path Non-Preferred Path
ȓrǺmp GL ȓrǺmp GL
[ȓwǺmp] ASSIM to /p/ [ȓwǺmp] ASSIM (Lab)
[pwǺmp] [fwǺmp] ST
[pwǺmp]
RESULT: 2 Typical, 0 Atypical RESULT: 3 Typical, 0 Atypical
150
Appendix C: Words Used on the Picture Naming Task
(adapted from Wolk, Edwards & Conture, 1993) 1. parachute 2. baby carriage 3. bathtub 4. beige 5. teeth 6. dinosaur 7. toy 8. ketchup 9. cookie 10. catch 11. guitar 12. measuring cup 13. newspaper 14. giraffe 15. fire truck 16. valentine 17. thimble 18. this 19. scissors 20. zebra 21. xylophone 22. shovel 23. hippopotamus 24. ladder 25. refrigerator 26. washing machine 27. yoyo 28. animals 29. plant 30. princess 31. black 32. brother 33. bridge 34. tractor 35. drive 36. clown 37. cracker 38. glasses 39. grasshopper 40. flag 41. french-fries 42. shrimp
43. spaghetti 44. sticker 45. smooth 46. snake 47. sleep 48. swing 49. splash 50. spread 51. strawberry 52. screwdriver 53. squirrel 54. twelve 55. queen 56. three 57. skateboard 58. ladybug 59. basket 60. chicken 61. pajamas 62. ice cream 63. banana 64. telephone 65. television 66. toothbrush 67. dishwasher 68. cage 69. cowboy 70. garage 71. mailbox 72. leaf 73. nose 74. chocolate 75. jump rope 76. jelly 77. feather 78. vacuum cleaner 79. thank you 80. thirsty 81. there 82. sandwich 83. zipper 84. shampoo
85. helicopter 86. library 87. rabbit 88. window 89. yawn 90. elephant 91. plate 92. present 93. blanket 94. breathe 95. tree house 96. twins 97. pudding 98. dragon 99. crib 100. quack 101. glove 102. green 103. flower 104. frog 105. throw 106. shrunk 107. spider 108. stamp 109. school bus 110. smoke 111. snowman 112. slide 113. swimming pool 114. splinter 115. spring 116. string 117. scratch 118. squirtgun 119. clock 120. yellow 121. drum 122. dentist 123. washcloth 124. hanger 125. teacher
151
Appendix D: Phonological Awareness Tasks
Blending Adapted from Larivee & Catts (1999)
Onset-rhyme
Item #
Stimulus Pictured Choices Score
(C-VC) BT1 t—æg top tag bag (training)
BT2 ȓ—it shoes meat sheet (training)
BT3 w—Ǻg wig weed pig (training)
B01 f—Ǻ ȓ dish fan fish 0 1
B02 tȓ—iz cheese knees chain 0 1
B03 ȓ—Ǻp chip ship shell 0 1
B04 ȅ—Ȝm thumb sun thief 0 1
B05 m—aȚs mouse house mouth 0 1
B06 f—eǺs feet vase face 0 1
3 phonemes
(C-V-C) BT4 s—i—d spoon seed knees (training)
BT5 r—oȚ—p soup rope rose (training)
B07 v—æ—n van fan vase 0 1
B08 s—Ȝ—n sub one sun 0 1
B09 ȴ—ǫ—t jet net juice 0 1
B10 n—aǺ—t light night knife 0 1
B11 k—aǺ—t kite bike cup 0 1
B12 k—oȚ-t cap coat boat 0 1
Total # correct: _________/12
152
Rhyme Matching Adapted from Bird, Bishop & Freeman (1995)
Training Items
Sue T1 bee Pat T3 witch fish shoe cow face cat T2 two T4 chair hat mud rose coat neck T5 bat ham jet leg
Experimental Items
Dan 1 mouse cap 0 1 Pete 9 feet cheese 0 1 spoon pan hat ham 2 cat fan 0 1 10 doll bean 0 1 run bike sheet nut 3 bone tap 0 1 11 keys bat 0 1 can night soap seat 4 van back 0 1 12 knees meat 0 1 pin house light knife Doug 5 nut rug 0 1 Ned 13 top seed 0 1 wig soup neck bed 6 mug pig 0 1 14 red mug 0 1 sub chin leg chair 7 pot bag 0 1 15 pen thief 0 1 jug cup weed head
8 tag run 0 1 16 food sled 0 1
door bug hen chain
Total correct _____/16
153
Onset Matching Adapted from Bird, Bishop & Freeman (1995)
Training Items Pictured Choices
/r/ T1 red dog T2 house rug one T3 chin run wig /m/ T4 mouth tape sheep rose T5 doll rope weed mouse
Experimental Items
/p/ 01 deer kite 0 1 /ttttȓȓȓȓ/ 06 bike chair 0 1
bug pin ship head 02 sock nut 0 1 07 coat pan 0 1 pan boat chain sheet 03 pig hen 0 1 08 keys fish 0 1 light bone chip shell 04 can pen 0 1 09 shoes rope 0 1 red back net cheese 05 tie pot 0 1 10 chin cat 0 1 bean seat witch sheep Total correct: __________/10
154
Onset Segmentation & Matching Adapted from Bird et al. (1995) Ben OST1 cow bone OST2 saw mud boat OST3 bed kite seed door OST4 van net bug tea OST5 sled sock fan back Tom Sam OS01 pin juice 0 1 OS06 two bat 0 1 tie door rug sun OS02 jug ham 0 1 OS07 tape cow 0 1 deer top bee saw OS03 toes dish 0 1 OS08 toes pen 0 1 food hen sock meat OS04 tap ship 0 1 OS09 bag knife 0 1 dog leg soup thumb OS05 doll tea 0 1 OS10 bed tag 0 1 mug hat soap jet
Total correct: __________/10
155
Appendix E: Syllable Repetition Task (from Shriberg et al., 2006)
2 Syllable (16 consonants)
/bada/
/dama/
/bama/
/mada/
/naba/
/daba/
/nada/
/maba/
3 Syllable (18 consonants)
/bamana/
/dabama/
/madaba/
/nabada/
/banada/
/manaba/
4 Syllable (16 consonants)
/bamadana/
/danabama/
/manabada/
/nadamaba/
Total: 50 Consonants
156
Appendix F: Rapid Naming Task
Monosyllable:
Disyllable:
Appendix G: Complete Correlation Matrix
Note: * Correlation is significant at 0.05 . ** Correlation is significant at 0.01 . r = Pearson’s correlation coefficient
GF
TA
-2 S
td
Sco
re
PP
VT
-4 S
td
Sco
re
Sen
t Str
uctu
re
CE
LF:P
-2
Con
cept
s &
D
irect
ions
C
ELF
:P-2
DA
S P
atte
rn
Con
stru
ctio
n T
Sco
re
Rhy
me
Ons
et
Mat
chin
g
Ons
et S
eg &
M
atch
ing
Ble
ndin
g
PA
Prin
cipa
l C
omp
Syl
labl
e R
epet
ition
P
CC
RN
Mon
osyl
l
RN
Dis
ylla
ble
RN
Z S
core
(A
vg)
PC
C
Dis
tort
ions
P
er C
ons
Typ
ical
Sou
nd
Cha
nges
Per
C
ons
Ayp
ical
Sou
nd
Cha
nges
Per
C
ons
r 1 .301* .121 .304* .100 .046 .050 .264 .119 .152 .391** .239 .081 .191 .818** .266 -.775** -.476 GFTA-2 Std Score Sig. .033 .401 .032 .495 .768 .749 .087 .448 .332 .010 .128 .605 .225 .000 .084 .000 .001
N 50 50 50 50 49 43 43 43 43 43 43 42 43 42 43 43 43 43
r .301* 1 .585** .575** .442** .400** .427** .443** .367* .517** .312* -.227 -.281 -.267 .402** .181 -.391** -.218 PPVT-4 Std Score Sig. .033 .000 .000 .001 .008 .004 .003 .015 .000 .042 .148 .068 .087 .008 .246 .009 .160
N 50 51 51 51 50 43 43 43 43 43 43 42 43 42 43 43 43 43 r .121 .585** 1 .581** .453** .295 .200 .059 .267 .252 .129 -.131 .002 -.065 .007 -.045 -.072 -.226
Sig. .401 .000 .000 .001 .055 .198 .707 .084 .103 .408 .407 .987 .681 .964 .774 .645 .146
Sent Structure CELF:P-2 N 50 51 51 51 50 43 43 43 43 43 43 42 43 42 43 43 43 43
r .304* .575** .581** 1 .499** .333* .203 .185 .361* .332* .411** -.165 -.017 -.100 .172 .077 -.168 -.055
Sig. .032 .000 .000 .000 .029 .191 .234 .018 .030 .006 .297 .915 .529 .271 .623 .282 .724
Concepts & Directions CELF:P-2 N 50 51 51 51 50 43 43 43 43 43 43 42 43 42 43 43 43 43
r .100 .442** .453** .499** 1 .074 .252 .159 .070 .180 .194 -.212 .004 -.085 -.163 -.099 .177 .044
Sig. .495 .001 .001 .000 .639 .103 .307 .656 .247 .212 .177 .979 .592 .297 .528 .257 .781
DAS Pattern Construct. T Score
N 49 50 50 50 50 43 43 43 43 43 43 42 43 42 43 43 43 43
Rhyme r .046 .400** .295 .333* .074 1 .621** .508** .356* .789** .296 -.073 .037 -.047 .159 .079 -.064 -.349*
Sig. .768 .008 .055 .029 .639 .000 .001 .019 .000 .054 .645 .815 .769 .307 .616 .684 .022
N 43 43 43 43 43 43 43 43 43 43 43 42 43 42 43 43 43 43
Onset Matching
r .050 .427** .200 .203 .252 .621** 1 .637** .401** .854** .313* -.287 -.081 -.199 .146 .040 -.117 -.285
Sig. .749 .004 .198 .191 .103 .000 .000 .008 .000 .041 .066 .607 .207 .351 .798 .454 .064
N 43 43 43 43 43 43 43 43 43 43 43 42 43 42 43 43 43 43 r .264 .443** .059 .185 .159 .508** .637** 1 .490** .841** .281 -.143 -.225 -.199 .285 .212 -.285 -.317* Onset Seg
& Matching
Sig. .087 .003 .707 .234 .307 .001 .000 .001 .000 .068 .367 .146 .207 .064 .172 .064 .038
N 43 43 43 43 43 43 43 43 43 43 43 42 43 42 43 43 43 43
Blending r .119 .367* .267 .361* .070 .356* .401** .490** 1 .681** .218 -.150 -.106 -.148 .103 .075 -.060 -.188
Sig. .448 .015 .084 .018 .656 .019 .008 .001 .000 .161 .342 .497 .351 .513 .634 .702 .228
N 43 43 43 43 43 43 43 43 43 43 43 42 43 42 43 43 43 43 r .152 .517** .252 .332* .180 .789** .854** .841** .681** 1 .351* -.207 -.120 -.187 .222 .129 -.171 -.362* PA Principal
Comp Sig. .332 .000 .103 .030 .247 .000 .000 .000 .000 .021 .188 .444 .235 .152 .409 .273 .017
N 43 43 43 43 43 43 43 43 43 43 43 42 43 42 43 43 43 43
GF
TA
-2 S
td
Sco
re
PP
VT
-4 S
td
Sco
re
Sen
t Str
uctu
re
CE
LF:P
-2
Con
cept
s &
D
irect
ions
C
ELF
:P-2
DA
S P
atte
rn
Con
stru
ctio
n T
Sco
re
Rhy
me
Ons
et
Mat
chin
g
Ons
et S
eg &
M
atch
ing
Ble
ndin
g
PA
Prin
cipa
l C
omp
Syl
labl
e R
epet
ition
P
CC
RN
Mon
osyl
l
RN
Dis
ylla
ble
RN
Z S
core
(A
vg)
PC
C
Dis
tort
ions
P
er C
ons
Typ
ical
Sou
nd
Cha
nges
Per
C
ons
Ayp
ical
Sou
nd
Cha
nges
Per
C
ons
r .391** .312* .129 .411** .194 .296 .313* .281 .218 .351* 1 .131 .302* .239 .475** .031 -.340* -.611** Syllable Repetition PCC-R
Sig. .010 .042 .408 .006 .212 .054 .041 .068 .161 .021 .408 .049 .128 .001 .842 .026 .000
N 43 43 43 43 43 43 43 43 43 43 43 42 43 42 43 43 43 43
RN Monosyllab.
r .239 -.227 -.131 -.165 -.212 -.073 -.287 -.143 -.150 -.207 .131 1 .548** .882** .141 .075 -.113 -.217
Sig. .128 .148 .407 .297 .177 .645 .066 .367 .342 .188 .408 .000 .000 .374 .636 .478 .167
N 42 42 42 42 42 42 42 42 42 42 42 42 42 42 42 42 42 42
RN Disyllable
r .081 -.281 .002 -.017 .004 .037 -.081 -.225 -.106 -.120 .302* .548** 1 .878** -.018 -.074 .081 -.257
Sig. .605 .068 .987 .915 .979 .815 .607 .146 .497 .444 .049 .000 .000 .910 .639 .607 .096
N 43 43 43 43 43 43 43 43 43 43 43 42 43 42 43 43 43 43 r .191 -.267 -.065 -.100 -.085 -.047 -.199 -.199 -.148 -.187 .239 .882** .878** 1 .087 .022 -.050 -.258 RN Z score
(Avg) Sig. .225 .087 .681 .529 .592 .769 .207 .207 .351 .235 .128 .000 .000 .583 .890 .755 .098
N 42 42 42 42 42 42 42 42 42 42 42 42 42 42 42 42 42 42
PCC r .818** .402** .007 .172 -.163 .159 .146 .285 .103 .222 .475** .141 -.018 .087 1 .301* -.924** -.600**
Sig. .000 .008 .964 .271 .297 .307 .351 .064 .513 .152 .001 .374 .910 .583 .050 .000 .000
N 43 43 43 43 43 43 43 43 43 43 43 42 43 42 43 43 43 43 r .266 .181 -.045 .077 -.099 .079 .040 .212 .075 .129 .031 .075 -.074 .022 .301* 1 -.440** -.180
Sig. .084 .246 .774 .623 .528 .616 .798 .172 .634 .409 .842 .636 .639 .890 .050 .003 .248
Distortions Per Cons
N 43 43 43 43 43 43 43 43 43 43 43 42 43 42 43 43 43 43 r -.775** -.391** -.072 -.168 .177 -.064 -.117 -.285 -.060 -.171 -.340* -.113 .081 -.050 -.924** -.440** 1 .344* Sig. .000 .009 .645 .282 .257 .684 .454 .064 .702 .273 .026 .478 .607 .755 .000 .003 .024
Typical Sound Changes Per Cons
N 43 43 43 43 43 43 43 43 43 43 43 42 43 42 43 43 43 43
r -.476** -.218 .226 -.055 .044 -.349* -.285 -.317* -.188 -.362* -.611** -.217 -.257 -.258 -.600** -.180 .344* 1
Sig. .001 .160 .146 .724 .781 .022 .064 .038 .228 .017 .000 .167 .096 .098 .000 .248 .024
Atypical Sound Changes Per Cons N 43 43 43 43 43 43 43 43 43 43 43 42 43 42 43 43 43 43
r -.386* -.091 -.123 -.257 -.212 .327* .268 .195 .023 .264 -.089 -.077 -.188 -.129 -.023 .191 -.030 -.030
Sig. .011 .561 .432 .096 .173 .032 .082 .210 .885 .087 .569 .627 .228 .417 .868 .219 .846 .846
Age
N 43 43 43 43 43 43 43 43 43 43 43 42 43 42 43 43 43 43
159
Appendix H: Measurement Issues
Validity and Reliability. One of the assumptions of any statistical analysis is that
the variables of interest accurately reflect what they are intended to reflect (validity), and
that they are accurately measured (reliability). However, with behavioral data, this is
often a significant challenge (hence, we rely on converging evidence of a particular
effect). The reliability statistics of the variables reported for this study, although roughly
similar to other studies, reflect the notion that there is always some error in measurement
(e.g., there was not 100% agreement between the original measure and the reliability
judges’ measures on any of the tasks or transcriptions). Similarly, phonetic transcription
is perceptually-based, and therefore it can vary depending on several listener and
environmental factors (Oller & Ramsdell, 2006; Shriberg et al., 1984). Complete
agreement on narrow (very detailed) phonetic transcription for children with SSD is not
achievable.
Although no systematic variation in quantification of errors was identified that
would have influenced the data, a potential criticism is that the individual who completed
the phonetic transcription (the author) was also the individual who obtained the
phonological processing data. The author made every attempt to be objective in the
transcriptions but this is a potential source of bias. However, the inter-judge reliability
agreements support the notion that this had relatively little influence on the author’s
transcriptions.
Sampling. Sampling issues are always a concern in studies involving clinical
populations (Kazdin, 2003). As reported in the Methods section, the participants, as a
group, scored significantly above the expected means on the PPVT-4 and the Pattern
160
Construction subtest of the DAS. Although the study was well advertised in Upstate
New York, it is possible that a particular ‘type’ of child was referred, perhaps based on
socioeconomic status, cooperativeness of the child, the clinician’s relationship with the
parents, or a host of other factors. Thus, it is not possible to generalize these results to all
children with idiopathic SSD.
Task demands. Measurement issues also come into play with the assessment of
phonological processing. Four phonological awareness tasks were combined to
approximate a global measure of PA, and 63% of the variance of the tasks was retained in
the Principal Component. However, it must be recognized that each of these tasks makes
different demands, even though all require nonverbal responses. For example, the tasks
tap different aspects of phonological awareness, such as rhyme, initial phoneme
identification, and phoneme synthesis (blending). If other tasks had been chosen (such as
identifying the number of syllables), the PA Principal Component scores might have been
different. Additionally, some of the PA tasks require different non-phonological
demands, such as attention, memory, and matching of visual and auditory information.
For example, the blending task requires retention and synthesis of up to three serially
presented individual phonemes (e.g., /k – o – t/ to form “coat”), whereas the other three
PA tasks required comparison of a target phonological feature (initial phoneme or rhyme)
to a group of four target words (e.g., Which one rhymes with Dan? cat, fan, run, bike.).
The blending task paradigm was slightly different than the other three tasks, and this is
likely one reason why the blending task was less strongly correlated with the PA
Principal Component than the other three tasks.
Chance performance. The PA tasks required nonverbal responses from a closed
161
set. This was done to avoid the complicating factor of ambiguous or unclear verbal
responses (e.g., by requiring children with SSD to produce a word that begins with a
specific phoneme or that rhymes with a particular word). However, using a closed set of
responses impacts the reliability of measurement when a relatively small number of trials
is used. For example, chance performance on the rhyme task would be 25% correct
(because there are four pictures to choose from). That is, a child could get 4 out of 16
items correct by guessing, so it is only when a child scores outside of the 95% confidence
interval for random guessing (1-7 items correct) that we can be relatively certain the child
is not guessing. Essentially, this means we are less confident of relatively low PA scores
than relatively high PA scores, because low PA scores may reflect chance performance.
Ideally, a very large number of trials on each task would be used to obtain a reliable
estimate of the child’s performance. However, constraints such as time and attention
make this impractical, leaving us with less reliable estimates of performance.
Power. Statistical power is the ability to correctly reject a false null hypothesis
(Keith, 2006, Baguley, 2004). It depends on sample size (here, n = 43), alpha level (here,
0.05), and the effect size (here, ∆R2 = 0.07 for the primary analysis with three types of
errors and 0.00 for the PCC analysis). Although one of the relatively unique features of
this study is its sample size of children with SSD, the effect size of atypical sound
changes in the prediction of PA (∆R2 = 0.07) is somewhat smaller than anticipated. The
study was designed to have adequate power to detect a change in R2 of 0.10 (i.e., an
increase in 10% variance explained above and beyond age and vocabulary). However,
because the effect size (∆R2 = 0.07) was found to be significant, this power limitation is
162
not a significant concern. 9
The lack of ∆R2 (0.00) when PCC is used to predict PA provides tentative support
for the use of the three-category system over PCC. A larger sample is required to have
adequate power to reject the possibility that PCC can predict unique variance in PA.
Note, however, that because there was NO change in R2 (0.00) when PCC is added to the
model, we do not have an actual effect size to estimate. Thus, observed power cannot be
determined for ∆R2 in the model that uses PCC to predict PA. It can only be stated with
relative certainty (i.e., power of 0.80) that ∆R2 is less than 0.10 when PCC is used to
predict variance in PA.
9 Observed power for the effect size found in this study is only 0.63, meaning that if the ∆R2 had not been statistically significant when Atypical Sound Changes per Consonant was added to the equation to predict PA, there would have been inadequate power to be certain that the effect size of 0.07 was not real. That is, there would have been adequate power (0.80) to suggest that the ∆R2 is not 0.10, but not enough power (0.63) to be relatively certain that the effect was not 0.07. However, Baguley has suggested that the calculation of observed (retrospective) power is “fundamentally flawed” and should be avoided (Baguley, 2004).
163
Appendix I: Regression Diagnostics
The dependent variable (PA) and all independent variables (age, receptive
vocabulary, and types of speech sound errors) were normally distributed based on
Kolmogorov-Smirnov tests for normality. A normal p-p plot of the regression
standardized residuals is shown below. This generally conforms to a straight line,
indicating that the residuals are normally distributed.
Normal p-p Plot of the Regression Standardized Residuals
1.00.80.60.40.20.0
Observed Cumulative Probability
1.0
0.8
0.6
0.4
0.2
0.0
Exp
ecte
d C
um
ula
tive
Pro
bab
ility
PA Principle Component
Residuals. Plots of the residuals against the predictor variables appear below.
These can be used to identify trends for nonlinearity, heteroscedacity, and outliers (Keith,
2006). The primary regression equation,
(1) PA = Vocab + Age
(2) PA = Vocab + Age + Atypical Sound Changes per Consonant
was re-run several times, eliminating some of the potential outliers (P04, P23, P34, P38,
P46) identified by residual plots or those with standardized residuals above 1.8. When
PA Principal Component
164
adding atypical errors to the equation, the R2 change remains significant with each of
these participants removed, with the exception of P04, in which ∆R2 is 0.059 (adjusted
∆R2 = 0.046, p = 0.055) . These values are judged to be relatively close to the overall
model that includes all participants. Note that participant P04 met all inclusionary
criteria, but he received a high score on Atypical Sound Changes because he frequently
deleted initial and medial consonants. When Weighted Least Squares regression is used
to reduce the influence of outliers, Atypical Sound Changes per Consonant is a
significant predictor of PA.
1501401301201101009080
PPVT4SS
2.0
0.0
-2.0Un
stan
dar
diz
ed R
esid
ual
46
38
34
23
706560555045
Age_Part_II
2.0
0.0
-2.0Un
stan
dar
diz
ed R
esid
ual
46
34
23
22
Multicollinearity. Multicollinearity (an excessively high correlation of
independent variables) would affect standard error measures, and therefore statistical
165
significance testing of the variables in the model (Keith, 2006). Collinearity diagnostics
are reported below for the regression equation. They indicate relatively little overlap
among the variables (tolerance levels are close to 1).
Correlations and Collinearity Statistics in two regression models predicting PA Principal Component
Model Correlations Collinearity Statistics
Zero-order
Partial Part Tolerance VIF
1 PPVT4SS .517 .563 .543 .992 1.008 Age .264 .365 .313 .992 1.008
2 PPVT4SS .517 .533 .474 .946 1.057 Age .264 .393 .321 .991 1.009 Atypical
Sound Changes Per Cons
-.361 -.332 -.265 .951 1.051
Interactions. Using the final equation, R2 = 0.435, F (3, 39) =10.0, p <0.001,
PA = Receptive Vocabulary + Age + Atypical Sound changes
the interaction terms were tested by adding them to the model. None of the interaction
terms contributed significant variance to the model:
Interaction R2 change p-value
Receptive Vocabulary * Age 0.001 0.818
Receptive Vocabulary * Atypical Errors 0.006 0.536
Age * Atypical Errors 0.003 0.674
Receptive Vocabulary * Age * Atypical Errors 0.015 0.322
166
Appendix J: Caveats and Limitations: The Role of Children’s Experiences
There is reason to believe that a child’s experiences may influence performance
on speech and phonological processing tasks. The home environment (e.g., parental
focus on speech and language, exposure to books, etc.), could have an impact on a child’s
speech sound development (Hauner et al., 2005; Law et al., 2004) and preliteracy skills
(Hall & Moats, 1999; Justice & Pullen, 2003; Nittrouer & Burton, 2005). Because there
is also reason to believe that genetic factors interact with environmental factors (McGrath
et al., 2007), the degree to which either of these factors independently predicts
phonological development is unclear. Furthermore, speech-language therapy programs
vary widely in terms of the amount of emphasis they place on phonological awareness
(Gillon, 2000, 2005; Hesketh et al., 2000). It is also unknown how much attention has
been given to the remediation of atypical sound changes for the participants in this study.
Thus, all of these factors could contribute to the findings, but they are difficult to measure
retrospectively.
Children obviously vary in the amount of exposure to and/or explicit teaching of
the alphabetic principle. However, one plausible influence that was not directly
considered in this study is that alphabet knowledge could also relate to PA development.
A bidirectional relationship has been reported between letter knowledge and PA (e.g.,
Burgess & Lonigan, 1998), and it could be argued that orthographic knowledge might aid
in the refinement of phonological representations. This issue was explored
retrospectively in the following way. The case history form that parents completed for
this study asked whether children knew any letters of the alphabet and, if so, how many.
167
Thirty-seven parents provided a descriptive or quantifiable response to this question, and
responses ranged from no alphabet knowledge to knowing all of the letters. Not all
parents listed an exact number of letters, so responses were categorized along a five-level
scale: 0= “No” letters (0-1 letters); 1= “Some” letters (2-10 letters); 2 = “Many” letters
(11-18 letters); 3 = “Most” or “almost all” letters (19-23); 4 = 24+ letters or “all” letters.
For these 37 children, this variable was found to be correlated with the PA composite
score, and also with age. When forced to enter into the regression with vocabulary and
age, alphabet knowledge was not a significant predictor of PA. This suggests that
alphabet knowledge, as reported by the parent, did not significantly predict PA. With
alphabet knowledge forced into the regression with vocabulary, and age in Step 1, and
atypical speech errors added in Step 2, only vocabulary and age are statistically
significant predictors of PA. However, there is limited power to detect the influence of
atypical errors (0.55), because there are more variables in the equation (4) and also fewer
participants (37). The overall R2 is comparable to before. Therefore, it could be argued
that one rival hypothesis that cannot be discounted in this study is that alphabet
knowledge is related to PA skills in such a way that speech sound errors no longer
contribute to the prediction of PA. The theoretical rationale behind this has not been put
forth, although it is possible that knowledge of letters and letter-sound associations helps
to sharpen or specify a child’s phonological representations (Gillon, 2005).
168
Appendix K: Speech Perception
It should be noted that a relationship between phonological awareness (PA) and
speech perception/discrimination has been discussed (e.g., Rvachew & Grawburg, 2006),
although it is not yet well understood. In fact, the relationship between speech
production and speech perception is not well understood. Some studies have found that
children with speech sound disorders, as a group, may have difficulty with speech
perception and/or discrimination compared to typically developing peers (Bird & Bishop,
1992; Ohde & Sharf, 1988; Rvachew et al., 2003; Shuster, 1998). However, other studies
have not found such differences (Sutherland & Gillon, 2005). Also, assessment of speech
perception may require stimuli that are individualized in order to be sensitive to
individual differences in perceptual abilities; hence, broad measures of speech perception
may not be useful. Although the potential relationship between speech perception,
speech production and PA is in need of further investigation, it is beyond the scope of
this study. The current investigation focused on how speech production relates to PA. It
is acknowledged, however, that there may be some potentially overlapping variance
between speech perception and speech production. Therefore the next step would be to
further explore the relationships between PA, speech perception, and speech production.
169
BIOGRAPHICAL DATA NAME OF AUTHOR: Jonathan L. Preston PLACE OF BIRTH: Oneonta, New York DATE OF BIRTH: March 1, 1978 UNDERGRADUATE AND GRADUATE SCHOOLS ATTENDED: Elmira College, Elmira, New York Syracuse University, Syracuse, New York DEGREES AWARDED: Bachelor of Science in Speech and Hearing, 2000, Elmira College Master of Science in Speech-Language Pathology, 2002, Syracuse University AWARDS AND HONORS: Syracuse University Fellowship American Speech-Language-Hearing Foundation Graduate Student Scholarship American Speech-Language-Hearing Foundation Grant in Early Childhood
Language PROFESSIONAL EXPERIENCE: Research Associate, Pediatric Audiology Laboratory, Syracuse University Clinical Supervisor, Gebbie Speech-Language-Hearing Clinic
Speech-Language Pathologist, Rochester City School District Speech-Language Pathologist, Rochester Hearing & Speech Center
top related