177
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

Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

  • Upload
    others

  • View
    6

  • Download
    0

Embed Size (px)

Citation preview

Page 1: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 2: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 3: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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: ____________________

Page 4: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

Copyright 2008 Jonathan Preston

All rights reserved

Page 5: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 6: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 7: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 8: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 9: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 10: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 11: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 12: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 13: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 14: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 15: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 16: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 17: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 18: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 19: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 20: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 21: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 22: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 23: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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,

Page 24: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 25: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 26: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 27: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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;

Page 28: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 29: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 30: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 31: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 32: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 33: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 34: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 35: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 36: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 37: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 38: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 39: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 40: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 41: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 42: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 43: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 44: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 45: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 46: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 47: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 48: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 49: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 50: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 51: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 52: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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).

Page 53: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 54: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 55: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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)

Page 56: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 57: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 58: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 59: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 60: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 61: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 62: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 63: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 64: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 65: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 66: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 67: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 68: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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?

Page 69: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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/

Page 70: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 71: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 72: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 73: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 74: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 75: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 76: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 77: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 78: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 79: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 80: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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,

Page 81: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 82: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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).

Page 83: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 84: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 85: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 86: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 87: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 88: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 89: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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).

Page 90: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 91: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 92: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 93: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 94: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 95: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 96: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 97: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 98: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 99: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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,

Page 100: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 101: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 102: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 103: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 104: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 105: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 106: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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).

Page 107: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 108: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 109: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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).

Page 110: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 111: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 112: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

104

understanding of which children are at particular risk for preliteracy and literacy

problems so that early intervention can be implemented.

Page 113: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 114: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 115: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 116: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 117: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 118: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 119: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 120: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 121: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 122: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 123: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 124: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 125: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 126: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 127: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 128: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 129: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 130: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 131: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 132: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 133: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 134: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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)

Page 135: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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).

Page 136: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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]

Page 137: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 138: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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]

Page 139: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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]

Page 140: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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]

Page 141: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 142: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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]

Page 143: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 144: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 145: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 146: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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]

Page 147: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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]

Page 148: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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]

Page 149: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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]

Page 150: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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]

Page 151: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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]

Page 152: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 153: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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]

Page 154: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 155: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 156: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 157: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 158: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 159: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 160: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 161: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 162: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 163: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 164: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

156

Appendix F: Rapid Naming Task

Monosyllable:

Disyllable:

Page 165: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 166: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 167: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 168: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 169: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 170: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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).

Page 171: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 172: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 173: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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

Page 174: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 175: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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).

Page 176: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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.

Page 177: Jonathan Preston Dissertationspeechproductionlab.syr.edu/_PDFs/Jonathan Preston Dissertation.pdf · Jonathan Preston B.S. Elmira College M.S. Syracuse University DISSERTATION Submitted

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