Dyslexia and Perception of Indexical Features in Speech Dissertation

Dyslexia and Perception of Indexical Features in Speech

Dissertation

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy

in the Graduate School of The Ohio State University

Gaylene Patrice Beam

Graduate Program in Speech and Hearing Science

The Ohio State University

2019

Dissertation Committee

Robert Fox, Advisor

Rebecca McCauley

Ewa Jacewicz

Copyrighted by

Gaylene Patrice Beam

2019

Abstract

Dyslexia is a specific learning disability that is neurobiological in origin. It is characterized by difficulties with accurate and/or fluent written word recognition and by poor spelling and decoding abilities. These difficulties are typically the result of a deficit in the phonological component of language. A variety of studies have pointed to an association of this impaired phonological processing ability with the perception of speech. This dissertation consists of three separate yet interrelated experiments designed to examine the roles that dyslexia and indexical features play in speech perception.

The purpose of Experiment 1 was to determine whether the underlying phonological impairment seen in adults and children with dyslexia is associated with a deficit in the ability to categorize regional dialects. Our results confirmed our hypothesis that individuals with dyslexia would perform more poorly than average reading controls in regional dialect categorization tasks. In addition, we found that listeners’ phonological processing ability (in specific, phonological short-term memory) was associated with listeners’ sensitivity to dialect. Children performed more poorly than did adults. Children with dyslexia performed more poorly than did the child control group.

Building on Experiment 1, Experiment 2 further inquired into sensitivity to indexical information (talker dialect and talker sex) in adults and children with dyslexia using stimuli that varied the nature and the redundancy of acoustic cues (namely, low-pass

iii filtered speech and noise-vocoded speech). Our results supported our previous findings.

Overall, listeners with dyslexia performed more poorly on categorization tasks than did controls. Children performed more poorly than adults in all conditions. We also found that for talker dialect identification, all listeners were most sensitive to dialect cues in clear speech, followed by vocoded speech. Listeners were least sensitive to dialect in low-pass filtered speech. For talker sex identification, listeners were again most sensitive to talker sex cues in the clear speech condition, yet for the degraded speech conditions the pattern was reversed. Listeners were more sensitive to talker sex cues in low-pass filtered speech than in the vocoded speech condition.

Experiment 3 addressed the question of how adults with dyslexia differ from average- reading adults in their ability to categorize indexical information (talker dialect and talker sex) when speech samples are systematically degraded by noise-vocoding. Talker speech was presented in five stimulus conditions: unprocessed speech and four levels of noise- vocoding (16-channel, 12-channel, 8 channel, and 4-channel). We also examined the intelligibility of this systematically degraded speech for adults with dyslexia and average- reading adults. The results seen in the indexical cue sensitivity portion of this experiment did not support the findings of our first two experiments. Individuals with dyslexia did not demonstrate a decreased sensitivity to indexical features compared to controls.

However, regarding speech intelligibility, our results did indicate that adults with dyslexia performed more poorly than controls in all stimuli conditions. In addition, all listeners demonstrated the native-dialect advantage, in that speech was more intelligible when talker and listener shared the same regional dialect.

Dedication

But if the while I think on thee, dear friend,

All losses are restored, and sorrows end.

William Shakespeare

Dedicated to the memory of Christine Beattie, Ph.D. Her friendship, vitality, and courage continue to inspire me.

Acknowledgments

I would like to express my deepest appreciation to my committee. Dr. Robert Fox, my advisor, provided indispensible support and rigorous mentorship through every stage of my doctoral program. I will fondly remember our conversations on linguistics,

Shakespeare, all things Appalachian, and the joy of a perfectly sharpened pencil. I am happy now to say that the many-headed Hydra lies slain at my feet. I would like to thank

Dr. Rebecca McCauley for providing such interesting, well thought-out, and thought- provoking courses. Her wise counsel, the depth and breadth of her knowledge in the fields of speech-language pathology and linguistics, and her unwavering confidence in my ability to attain this degree have been invaluable. I appreciate Dr. Ewa Jacewicz for modeling true dedication to scientific research and an enthusiasm for learning and engaging in new areas of content in order to expand scientific knowledge. Despite her own busy research schedule, she always found the time to help when needed. I thank Dr.

Dorothy Morrison for sharing her vast knowledge regarding the assessment and instruction of reading, her wisdom, and her determination to ensure that all children with reading disabilities, including those with dyslexia, have access to appropriate instruction.

vii

Vita

1984………………………………………….B.A. Humanities/Classics, Ohio Wesleyan

University

1990…………………………………………M.A. Speech Language Pathology, The

Ohio State University

Publications

Long, G. B., Fox, R. A., & Jacewicz, E. (2016). Dyslexia limits the ability to categorize

talker dialect. Journal of Speech, Language, and Hearing Research, 59, 900-914.

Fields of Study

Major Field: Speech and Hearing Science

viii

Table of Contents

Abstract ...... iii Dedication ...... v Acknowledgments...... vii Vita ...... viii List of Tables ...... x List of Figures ...... xi Chapter 1. Introduction ...... 1 Chapter 2. Background ...... 6 2.1 Phonological processing and dyslexia ...... 6 2.1.1 Phonological awareness ...... 7 2.1.2 Rapid automatic naming ...... 8 2.1.3 Phonological memory ...... 10 2.1.3.1 Phonological memory and sentence repetition ...... 12 2.1.3.2 Phonological memory and talker normalization processes in dyslexia ...... 13 2.2 Speech perception and dyslexia ...... 15 2.3 Indexical information ...... 20 2.4. Research using degraded speech ...... 21 2.4.1. Band-pass filtering ...... 22 2.4.2. Noise-vocoding ...... 23 2.5 Speech Intelligibility ...... 26 2.5.1 Speech intelligibility and dialect ...... 26 2.5.1.1 Speech intelligibility, dialect, and degraded speech ...... 27 2.5.2 Speech intelligibility and dyslexia ...... 28 2.5.2.1 Speech intelligibility, dyslexia, and degraded speech ...... 30 2.6 Current research questions ...... 32 2.6.1 Experiment 1 ...... 33 2.6.2 Experiment 2 ...... 35 2.6.3 Experiment 3 ...... 36 Chapter 3. Experiment 1 ...... 37 3.1 Methods...... 37 3.1.1 Participants ...... 37 3.1.2 Stimulus materials ...... 40 3.1.3 Procedure ...... 41

viii

3.2 Results ...... 41 3.2.1 Dialect sensitivity (A' measure) ...... 43 3.2.1.1 Main effects ...... 43 3.2.1.2 Two-way interactions...... 45 3.2.2 Response bias (B''D measure) ...... 47 3.2.3 Correlations between dialect categorization (A', B''D) and measures of phonological processing (CTOPP-2) ...... 50 Chapter 4. Experiment 2 ...... 54 4.1 Methods...... 54 4.1.1 Participants ...... 54 4.1.2 Stimulus materials ...... 55 4.1.3 Procedure ...... 59 4.2 Results ...... 61 4.2.1 Dialect Identification ...... 61 4.2.1.1 Main effects ...... 61 4.2.1.2 Two-Way Interaction ...... 62 4.2.1.3 Three-Way Interaction ...... 63 4.2.2 Sex Identification ...... 64 4.2.2.1 Main Effects ...... 65 4.2.2.2 Two-way interactions ...... 65 Chapter 5. Experiment 3 ...... 68 5.1 Methods...... 68 5.1.1 Participants ...... 68 5.1.2 Stimulus material ...... 70 5.1.3 Procedure ...... 76 5.2 Results ...... 78 5.2.1 Dialect identification ...... 78 5.2.1.1 Main effects ...... 78 5.2.2 Sex identification ...... 79 5.2.2.1 Main effects ...... 79 5.2.3 Correlations between dialect categorization (A') and measures of phonological processing (CTOPP-2) ...... 80 5.2.4 Intelligibility ...... 80 5.2.4.1 Main Effects ...... 81 5.2.4.2 Two-way interactions ...... 81 5.2.4.3 Three-way interactions ...... 83 5.2.4 Correlations between intelligibility scores (RAU) and measures of phonological processing (CTOPP-2) ...... 88 Chapter 6. Conclusion ...... 92 References ...... 109 Appendix A. Experiment 1 - Sentence Sets ...... 124 Appendix B. Experiment 2 – Sentence Sets ...... 127 Appendix C: Experiment 3 - Sentence Sets and Target Words ...... 142

List of Tables

Table 1. Mean (SD) scores for assessment measures of phonological ability on the Comprehensive Test of Phonological Processing – Second Edition (CTOPP-2) for participants with dyslexia (DYS) and participants with average reading (AR)…………………………………………………………………………...…………39

Table 2. Correlations between measures of phonological ability (CTOPP-2) and measures of dialect categorization (dialect sensitivity (A') and response bias (B''D)) for adults with dyslexia (DYS adult), children with dyslexia (DYS children), adults with average reading ability (AR) and children with average reading ability (AR)...... 51

Table 3. Mean (SD) for assessment measures of phonological memory on the Comprehensive Test of Phonological Processing- 2nd Edition (CTOPP-2) for participants with dyslexia (DYS) and participants with average reading ability (AR)….69

Table 4. Division frequencies (Hz) used for noise band vocoding processing……….….71

Table 5. Correlations between measures of phonological memory (CTOPP-2) and measures of intelligibility of speech (RAU) for participants (adult listeners with with dyslexia and controls) in all conditions…….....……………………………………87

List of Figures

Figure 1Mean dialect sensitivity scores (A’) for adult and child participants with average reading ability (AR) and adult and child participants with dyslexia (DYS), The error bars represent 1 SE...... 43

Figure 2. Mean dialect sensitivity scores (A’) for participants with average reading ability (AR) and participants with dyslexia (DYS) as a function of talker sex. The error bars represent 1 SE…………………………………………………….……………………...44 Figure 3. Mean dialect sensitivity scores (A’) for adult and child participants as a function of talker sex. The error bars represent 1 SE………………………………….....45

Figure 4. Mean response bias scores (B”D) for adult and child participants with average reading ability (AR) and adult and child participants with dyslexia (DYS). Negative values represent liberal bias (i.e., a bias toward Ohio). The error bars represent 1 SE………………………………………………………………………………………...47

Figure 5. Mean Response bias scores (B”D) for adult and child participants as a function of talker sex. The error bar represents 1 SE…………………………………………..….48

Figure 6. A spectrogram of a clear speech stimulus utterance. It is acoustically unprocessed except for amplitude equalization…………………………………………57

Figure 7. A spectrogram of a low-pass filtered utterance in which all spectral information above 400 Hz was removed (and the altered token amplitude equalized)………………57

Figure 8. A spectrogram of a vocoded stimulus processed through an 8-channel noise vocoder (and amplitude normalized)…………………………………………………… 58

Figure 9. A screen shot of response boxes used by the participant during the experiment to identify talker regional dialect and talker sex…………………………………………60

Figure 10. Mean dialect sensitivity scores (A’) representing interaction between talker sex and stimulus type. The error bars represent 1 SE……………………………………62

Figure 11. Mean dialect sensitivity scores (A') representing a three-way interaction between talker sex, stimulus type, listener age. The error bars represent 1 SE…….……63

Figure 12. Sensitivity to talker sex (A') as a function of listener group and stimulus type. The error bars represent 1 SE……………………………………………………….……65

Figure 13. Sensitivity to talker sex scores (A') as a function of stimulus type and talker dialect. The error bars represent a 95% confidence interval………………………..……66

Figure 14. Unprocessed sentence “We are in the record business” band-pass filtered from 200 to 8000 Hz………………………………………………………………..……72

Figure 15. 16-band vocoding of “We are in the record business” band-pass filtered from 200 to 8000 Hz…………………………………………………………………..……….72

Figure 16. 12-band vocoding of “We are in the record business” band-pass filtered from 200 to 8000 Hz……………………………………………………………..…………….73

Figure 17. 8-band vocoding of “We are in the record business” band-pass filtered from 200 to 8000 Hz……………………………………………………………….….……….73

Figure 18. 4-band vocoding of “We are in the record business” band-pass filtered from 200 to 8000 Hz………………………………………………………………….…….….74

Figure 19. A screen shot of response boxes used by the participant during the experiment to identify talker regional dialect and talker sex…………………………………………75

Figure 20. A Screen shot of the response prompt displayed on the monitor in the Intelligibility Task…………………………………………………………………….…76

Figure 21. Mean (SE) arcsine transformed intelligibility scores for individuals with dyslexia and average reading individuals at each stimulus condition. RAU, rationalized arcsine unit. Error bars represent 2 SE…………………………………………………...81

Figure 22. Mean (SE) arcsine transformed intelligibility scores for all individuals as a function of talker dialect and talker sex. RAU, rationalized arcsine unit. Error bars represent 2 SE………………………………………………………………………....…82

Figure 23. Mean (SE) arcsine transformed intelligibility scores for all individuals when listening to male talkers and female talkers as a function of stimulus condition and talker dialect. RAU, rationalized arcsine unit. Error bars 2 SE………………………………...83

Figure 24. Mean (SE) arcsine transformed intelligibility scores for individuals with dyslexia and average reading individuals as a function of talker sex in each stimulus condition. RAU, rationalized arcsine unit. Error bars 2 SE…………………….……….85

Figure 25. Scatterplot of correlation between intelligibility scores (RAU) in the unprocessed speech condition and CTOPP-2 Phonological Memory Composite Scores (expressed in standard scores) for all participants……………………………………….88

xii

Figure 26. Scatterplot of correlation between intelligibility scores (RAU) in the 16- channel vocoded condition and CTOPP-2 Phonological Memory Composite Scores (expressed in standard scores) for all participants…………………………………….…89

Figure 27. Scatterplot of correlation between intelligibility scores (RAU) in the 12- channel vocoded condition and CTOPP-2 Phonological Memory Composite Scores (expressed in standard scores) for all participants…………………………………….…89

Figure 28. Scatterplot of correlation between intelligibility scores (RAU) in the 8-channel vocoded condition and CTOPP-2 Phonological Memory Composite Scores (expressed in standard scores) for all participants…………………………………….………………..90

Figure 29. Scatterplot of correlation between intelligibility scores (RAU) in the 12- channel vocoded condition and CTOPP-2 Phonological Memory Composite Scores (expressed in standard scores) for all participants…………………………………….…90

xiii

Chapter 1. Introduction

Dyslexia is a specific learning disability that is neurobiological in origin. It is characterized by difficulties with accurate and/or fluent word recognition and by poor spelling and decoding abilities. These difficulties typically result from a deficit in the phonological component of language that is often unexpected in relation to other cognitive abilities and the provision of effective classroom instruction. Secondary consequences may include problems in reading comprehension and reduced reading experience that can impede growth of vocabulary and background knowledge (Lyon,

Shaywitz, & Shaywitz, 2003). Dyslexia is a lifelong condition that continues to affect individuals into adulthood.

A convergence of evidence from the fields of neuroscience, psychology, education, linguistics, and speech and hearing science points to a deficit in the underlying phonological component of language as a primary source of these reading and spelling difficulties (Fletcher, Lyon, Fuchs, & Barnes, 2007; Goswami, 2010; Kamhi & Catts,

2012; Pennington, 2009; Rayner, Pollatsek, Ashby, & Clifton, 2011; Robertson, Joanisse,

Desroches, & Ng, 2009; Shankweiler & Liberman, 1989; Shaywitz & Shaywitz, 2005).

This “phonological deficit,” in which accessing lexical phonological representations of words appears to be compromised, has been commonly found in individuals with dyslexia (Fraser, Goswami, & Conti-Ramsden, 2010; Ramus & Szenkovits, 2008). In particular, individuals with dyslexia have persistent difficulties processing and

manipulating phonological units, such as segmenting and blending phonemes in words, rhyming, and discriminating between speech sounds (Hulme & Snowling, 2009; Uhry,

2005). These difficulties, in turn, decrease their abilities to construct the robust phoneme- to-grapheme mapping required to develop accurate and fluent reading and spelling abilities (Lyon, Shaywitz, & Shaywitz, 2003). Importantly, the written language disorder of dyslexia appears to be part of a central phonological impairment rather than the result of a deficit in the cognitive components dedicated to decoding and encoding written language (Ferrer, Shaywitz, Holahan, Marchione, & Shaywitz, 2010; Tunmer & Greaney,

2010; Voller, 2004).

Performance on clinical measures of general phonological ability is strongly predictive of both reading and spelling abilities (Catts & Adolph, 2011; Torgesen, 2007). Research exploring the phonological deficit in dyslexia has revealed three correlated yet distinct aspects of phonological ability impairment, which are most relevant to the phonological processing of spoken language and the development of written language skills: phonological awareness, lexical retrieval, and phonological memory.

Phonological awareness refers to an individual’s awareness of and access to the phonological structure of their language, including their ability to manipulate sublexical units such as syllables and phonemes (Goswami, 2010; Mattingly, 1972; Miller, Sanchez,

& Hynd, 2003). There is an abundance of evidence that individuals with dyslexia have poor phonological awareness (Catts, Adolph, Hogan, & Ellis-Weismer, 2005; Robertson et al., 2009; Uhry, 2005), although there is ongoing debate whether these difficulties stem

from poorly specified phonological representations (Velluntino & Fletcher, 2007) or impaired access to intact phonological representations (Boets et al., 2013).

Lexical retrieval is the ability to quickly retrieve the phonological forms of words from long-term memory. The speed and accuracy of processing the combined visual and phonological forms of words is typically tested in rapid symbolic naming tasks (Lervag

& Hulme, 2009; Manis, Doi, & Badha, 2000; Neuhaus & Swank, 2002; Wolf, Bowers, &

Biddle, 2000). Again, individuals with dyslexia perform more poorly on these tasks relative to individuals with average reading ability, showing comparatively slower automatized naming, which may indicate slower lexical retrieval (Manis, Seidenberg, &

Doi, 1999; Parilla, Kirby, & McQuarrie, 2004; Powell, Stainthorp, Stuart, Garwood, &

Quinlan, 2007; Savage, Pillay, & Melidona, 2007).

Finally, phonological memory involves perceiving, phonetically analyzing and encoding speech sounds for temporary storage in working or short-term memory (Baddeley, 2003;

Gathercole & Baddeley, 1990; Savage, Lavers, & Pillay, 2007). Any one of these processes, which are believed to occur within a specific component of working memory called the phonological loop (Baddeley, 1986; 2007), might be the source of the phonological memory deficit found in individuals with dyslexia (Brady, 1986). The deficit might arise from problems with perception or problems in the initial phonetic encoding of correctly perceived information. Alternatively, the deficit might stem from a limitation in rehearsing and retrieving information in a phonetic code. More recently,

Ramus and Szenkovits (2008) proposed that the phonological processing deficit in

dyslexia emerges as a function of increased task demands when working memory is involved. In particular, the deficit is manifested in tasks that load heavily on working memory and require metalinguistic awareness or rapid processing. This implies that aspects of phonological processing including phonetic encoding, rehearsal and retrieval involve more resource allocation in individuals with dyslexia. An increase in processing effort may interfere with storing information in working memory within the central- storage capacity limits (Cowan, 2005; 2010) or, relatedly, in the episodic buffer component of the Baddeley’s working memory model (Baddeley, 2000). Less efficient processing is a plausible source of the phonological memory deficit in dyslexia which, in turn, inhibits the mapping of speech sounds to print in the decoding (reading) and encoding (spelling) of written words (Hulme & Snowling, 2009; Pennington & Bishop,

2009).

In this dissertation the view is adopted that the phonological deficit in dyslexia arises from impaired access to intact phonological representations rather than from poorly specified representations (Boets, et al., 2013; Dickie, Ota, & Clark, 2013; Hazan,

Messaoud-Galusi, & Rosen, 2013; Ramus & Szenkovits, 2008). In other words, the intrinsic properties of online phonological processing, not phonological representations per se, that may be impaired in dyslexia (Bloomert, Mitterer, &

Paffen, 2004).The difficulties with accessing phonological representations may arise, in part, from deficiencies in working memory processes that impede phonetic coding and constrain the access to information stored in long-term memory. The aim of this dissertation is to ascertain whether the presence of extensive indexical variation in speech

is more detrimental to performance of individuals with dyslexia relative to participants with average reading ability. Indexical information pertains to those aspects of variability in speech that are co-present with the linguistic information and cue both biological and social attributes about a talker such as age, sex, regional dialect or foreign accent, emotional state or socio-economic background. More recent models of the mental lexicon have included these indexical (or episodic) features in the underlying representations, which contain feature slots both for linguistic information and for indexical information

(Goldinger, 1996, 1998; Johnson, 2005, 2006; Pierrehumbert, 2001). The indexical variation is considered an integral part of lexical representations (Nygaard & Pisoni,

1998). It has also been proposed that the perception of spoken words is socially weighted and that listeners encode the same speech signal simultaneously to both linguistic representations and social (indexical) representations (Sumner, Kim, King, &

McGowan, 2014). However, it may be the case that the spoken word recognition system treats the linguistic and indexical variability differently. In particular, linguistic phonological variation may be represented more abstractly in memory and indexical variation may exert its influence later in the course of perception, being represented in a more specific format (Luce, McLennan, & Charles-Luce, 2003).

Chapter 2. Background

2.1 Phonological processing and dyslexia

According to Crystal (2001) and other linguists, the term phonology refers to the sound system of a language. The most important component of phonology, called phonetics, involves the study of significant speech sounds. The sounds (phonemes) are the building blocks used to construct words and sentences while speaking. Individuals who have difficulties processing phonological information may mispronounce words when speaking and misperceive the speech of others. These individuals may also experience difficulty with other speech abilities, such as rhyming, segmenting sounds in words, blending phonemes, and discriminating between speech sounds.

Traditionally, interest in phonology has been held mostly by linguists, speech pathologists, and speech scientists. Yet over the past 40 years many psychologists and educators have discovered that some phonological processing abilities play an important role in learning to read and spell. Evidence that some phonological processing abilities are involved in learning to read and write come from both correlational and training studies. Three kinds of phonological processing in particular appear to be especially relevant to the development of written language skills: phonological awareness, phonological memory, and rapid naming.

Phonological awareness, rapid naming, and phonological memory represent three correlated yet distinct kinds of phonological processing abilities. These abilities are thought to be correlated rather than independent based on confirmatory factor analysis studies. In general, phonological awareness and phonological memory tend to be more highly correlated with one another than with rapid naming. In addition, these three kinds of phonological processing tend to become less correlated with development. And for very young children, phonological awareness and phonological memory can be nearly perfectly correlated (Wagner, Torgesen, Laughon, Simmons, & Raschotte, 1993).

2.1.1 Phonological awareness

Phonological awareness refers to an individual’s awareness of and access to the sound structure of their oral language (Mattingly, 1972). The spoken words of a language represent strings of phonemes that signal differences of meaning. Phonological awareness is the ability to detect, manipulate, or analyze the auditory aspects of spoken language, including the ability to distinguish or segment words, syllables, or phonemes), independent of meaning. Phonological awareness is typically measured in phonemic awareness tasks, involving the identification or manipulation of single speech sounds.

However, phonological awareness also encompasses the manipulation of larger units of speech sounds.

Isabelle Y. Liebermann (1973) first suggested that learning to read and write depends on an individual’s phonological domain of skills and especially on the degree to which an individual is aware of the underlying phonological structure of words, i.e., phonological

awareness. This metalinguistic skill of phonological awareness of the phonological structure of words was shown to predict reading ability in children in a variety of languages (Bradley & Bryant, 1983; Lundberg, Olofsson, & Wall, 1980; Torneus, 1984).

The difficulties seen in individuals with dyslexia involving analyzing, memorizing, and accessing the sound structure of the language are well documented. Children with dyslexia show poor performance on tasks such as phoneme elision, blending, and segmentation, as well as decreased sensitivity to rhyming and alliteration (Ball &

Blachman, 1991; Bradley & Bryant, 1983; Muter, Holme, Snowling, & Taylor, 1998).

Additionally, adults with dyslexia were shown to have difficulties with tasks requiring explicit understanding of the phonological nature of words (Bradley & Bryant, 1983).

Adults with dyslexia have demonstrated poor phonological awareness in tasks involving segmentation, identification, discrimination, and blending of sublexical units (Goswami,

2003; Windfuhr & Snowling, 2001). Adults with dyslexia are less phonologically aware even than younger average-readers of similar reading ability (Bruck, 1992). Indeed reduced activation of the mid- to posterior temporal cortex was observed in adults with dyslexia compared to adult controls during a phonological awareness task (Rumsey,

Nace, Donohue, Wise, Maisog, & Andreason, 1997; Desroches, Newman, Robertson, &

Joanisse, 2012).

2.1.2 Rapid automatic naming

Rapid naming of digits, letters, objects, and colors requires efficient retrieval of phonological information for long-term, or permanent, memory. Unlike phonological awareness and phonological memory, which are entirely auditory-oral in mode, rapid

naming tasks involve visual components, most of which are graphemes or glyphs.

Because of this, rapid naming is best thought of as being a hybrid ability, in that successful performance depends on how quickly an individual can recognize the array of visual symbols and encode a phonological response. Obviously, the mixed modality nature of this ability is the same type that underlies decoding when reading aloud.

When reading, young readers presumably retrieve (a) phonemes associated with letters or letter pairs, (b) pronunciations of common word segments, and (c) pronunciations of whole words. The efficiency with which individuals are able to retrieve phonological codes associated with individual phonemes, word segments, or entire words should influence the degree to which phonological information is useful in decoding words

(Wolf, Bowers, & Biddle, 2000). The efficiency with which phonological codes are associated with items to be names will be affected by how well the items are known and how strong the mapping is between the item and its pronunciation.

Measures of rapid naming require speech and processing of visual as well as phonological information. Some researchers who study rapid naming suggest that rapid naming tasks assess the operation of a precise timing mechanism that is important for the developing knowledge of common letter patterns in printed words (Bowers & Wolf,

1993). Others have focused on the visual aspects of the rapid naming task as the reason it is predictive of reading and problematic for individuals with dyslexia (Geiger, Lettvin, &

Zeggara-Morgan, 2009), or have hypothesized that poor performance on rapid naming tasks is due to domain-general problems in speech of processing (Kail, Hall, & Caskey,

1999). Subsequent studies have not supported these alternate hypotheses, however

(Bonifacci & Snowling, 2008; Hawelka & Wimmer, 2005; Lervag & Hulme, 2009).

It is clear that rapid naming tasks predict poor performance in reading, and they do so independently of measures of phonological awareness (Manis, Doi, & Badha, 2000;

Parrila, Kirby, & McQarrie, 2004; Powell, Stainthorp, Stuart, Garwood, & Quinlan,

2007). Individuals who have double deficits – that is, deficits in both rapid naming and phonological awareness – appear to have greater difficulty learning to read words accurately and fluently than do individuals with either deficits in either rapid naming or phonological awareness alone (Bowers & Wolf, 1993). What is less clear is exactly what rapid naming tasks measure that is not captured by measures of phonological awareness or phonological memory. The most promising candidate appears to be something related to the ability to connect visual stimuli to phonological codes, as opposed to the adequacy of phonological codes themselves (Jones, Branigan, Hatzidaki, & Obregon, 2010).

2.1.3 Phonological memory

Phonological memory refers to coding information phonologically for temporary storage in working or short-term memory. The part of memory most involved in storing phonological information is called the phonological loop. The phonological loop provides a brief, verbatim storage of auditory information (Baddley, 1986, 1992). The phonological loop consists of two parts working together. The first is a phonological store, which can be thought of as a tape-recording loop that retains the most recent 2 seconds’ worth of auditory information that has been recorded. The second is an

articulatory control process that provides input to the phonological loop initially and can also refresh information already in the loop so that it can be stored longer that 2 seconds.

Phonological coding in working memory may not be all that important for reading common known words, although some controversy exists about this assertion (Crowder

& Wagner, 1991). It is clear, however, that phonological coding in working memory is potentially more useful when attempting to decode new unfamiliar words, particularly words that are long enough to decode bit by bit, as a means of storing intermediate sounds. Phonological processing also appears to by important to reading by supporting the role of working memory in comprehension (Rayner, Pollatsek, Ashby, & Clifton,

2011) because the meaning of a sentence depends not only on the words it contains but also on their order in the sentence.

Some studies have shown that a deficient phonological memory does not appear to impair either word-level processing or listening to a noticeable extent, providing the words involved are already in the individual’s vocabulary. However, phonological memory impairments can constrain the ability to learn new written and spoken vocabulary

(Gathercole, Willis, & Baddeley, 1991). Perhaps the most comprehensive investigation of impaired phonological memory is a series of more than 20 experiments that investigated the consequences of poor phonological memory for individuals with dyslexia (Torgesen,

1988, 1996). Early studies demonstrated that the origin of the memory deficit was a specific deficit in phonological coding of familiar verbal materials, such as digits and words. Alternate explanations, such as inattention, anxiety, poor use of mnemonic

strategies, and poor motivation, were ruled out. The individuals were not impaired for short-term memory for nonverbal material, long-term memory, or listening comprehension; however, they showed severe impairments in decoding visually presented nonsense syllables. When followed up with approximately a decade later, more than half of the individuals still had severe memory impairments, and those who did showed almost no improvement in reading skills. Finally, consistent with the studies of

Gathercole and Baddeley (1990), individuals with impaired phonological memories as children scored poorly on vocabulary measures a decade later, yet performed adequately on nonverbal measures.

2.1.3.1 Phonological memory and sentence repetition

The most obvious skill involved in sentence repetition is verbal working memory. Some researchers have proposed that problems with verbal memory processes (i.e., phonological short-term memory as measured by nonword repetition) compromise sentence processing (Adams & Gathercole, 2000). Individuals with phonological memory impairments, including individuals with dyslexia, might be expected to have problems with sentence processing, as put forth in the processing limitation hypothesis

(Shankweiler & Crain, 1986). According to the processing limitation hypothesis, constraints on phonological memory produce problems with oral and written language processing. However, the status of memory processes as predictors of language skill has been the subject of debate (MacDonald & Christiansen, 2002; Waters & Caplan, 1996).

Thus, rather than viewing poor working memory as a cause of poor language ability,

Hulme (2006) has argued that maintaining verbal information in working memory

depends critically on the adequacy of linguistic representations and, in particular, output phonological representations, though such linguistic representations are derived from item-based experiences (Lieven, 2008).

2.1.3.2 Phonological memory and talker normalization processes in dyslexia

Talker and linguistic information overlap in spoken language. Research suggests that phonological representations are affected by detailed talker-specific information (Creel &

Bregman, 2011) and that talker variability decreases the efficiency of speech processing.

For example, typical listeners are found to be slower and less accurate in identifying speech sounds when listening to several talkers in a speech perception task as opposed to a single talker (Green, Tomiak, & Kuhl, 1997; Mullennix & Pisoni, 1990; Nusbaum &

Morin, 1992). Memory for word-lists is also negatively affected by extensive talker variability and listeners recall fewer words from multiple-talker lists than from single- talker lists (Martin, Mullennix, Pisoni, & Summers, 1989). When the talker changes unpredictably, listeners usually experience increased difficulty encoding words in memory because rapid perceptual adjustment is required to reduce talker interference.

This perceptual adjustment, operationally termed talker normalization (Pisoni, 1997), is crucial in building associations between the nuances of an individual talker’s speech and the listener’s phonological representations of speech sounds encoded in long-term memory.

Recently, it has been shown that listeners with dyslexia appear to be impaired in their abilities to both recognize voices of multiple talkers (Perrachione, Del Tufo, & Gabrieli,

2011a) and to utilize talker normalization processes when listening to utterances in their native English language (Perrachione, Del Tufo, Ghosh, & Gabrieli, 2011b). In particular, listeners with dyslexia were deficient in recognizing different voices of unfamiliar talkers in English, which was in contrast to control participants whose accuracy was significantly greater. However, when both listener groups were trained on recognizing different voices in an unfamiliar language (Mandarin), average reading listeners were equally as poor as listeners with dyslexia in their talker identification abilities (Perrachione et al., 2011a). These results indicated that when both listener groups had access to stored phonological representations in their native language, the discrepancies between the listeners with dyslexia and average-reading listeners were due to their differential abilities in utilizing talker normalization processes. That is, average- reading adults were able to perceptually adjust to individual talker characteristics and retain the nuances of the talkers’ voices in working memory in order to associate these peculiarities with stored phonological representations of speech sounds. Listeners with dyslexia were limited in this ability, being unable to detect consistent deviations between the talker-specific phonetic variation and their stored abstract representations. The differences between the listeners with dyslexia and average-reading listeners were eliminated when neither group had access to stored phonological representations in

Mandarin, and thus were unable to utilize the correspondence between the phonetic variation in talker voices and the underlying phonological representation.

In a second study, Perrachione et al. (2011b) utilized fMRI data to demonstrate that, in

DYS listeners, brain regions typically engaged in talker normalization (superior temporal

gyrus, including primary auditory cortex and Wernicke’s area) displayed reduced neurophysiologic adaptation to a single talker during a speech perception task. Speech of a single talker is produced with the same idiosyncratic and consistent features, as opposed to the more variable speech characteristics of multiple talkers. The relative constancy of phonetic variability in the productions of a single talker results in strong adaptive neural processes related to talker normalization. Indeed, average-reading controls displayed extensive adaptation to the same single talker (in a one-talker condition) as opposed to four different talkers (in a multiple-talker condition) indicating strong talker- normalization effects. However, participants with dyslexia exhibited only limited adaptation, being unable to utilize the consistent and predictable indexical features and low phonetic variability in a single talker to establish the phonetic-phonemic correspondences during a speech perception task. Taken together, these results indicate that phonological processing in listeners with dyslexia is less efficient and more physiologically expensive relative to average-reading controls who share a common native language. The inability to fully utilize talker normalization in speech processing seems to impair the access to abstract phonological representations in dyslexia, interfering with the development of sound-to-letter mapping in reading.

2.2 Speech perception and dyslexia

Speech perception involves distinguishing speech sounds and categorizing acoustic signals into individual phonemes such as vowels, consonants, and lexical tones to make phonetic identifications and determine meaningful differences between phonetic contrasts

(Liu & Tsao, 2017). Many studies indicate that individuals with dyslexia demonstrate

poorer speech perception abilities than do age-matched or reading-level matched average- reading controls (Bogliotti, Serniclaes, Messaoud-Galusi, & Sprenger-Charolles, 2008;

Goswami, 2003). In addition, speech perception and phonological processing ability appear to be linked (Manis, et al., 1997; Mayo, Scobbie, Hewlett, & Waters, 2003). There is some evidence that weak phonological representations may be the cause of speech perception deficits in individuals with dyslexia (Liberman, 1983), but more recent studies indicate that the reverse might be true. Problems with speech processing may be the cause of phonological deficits. Speech perception deficits may arise from abnormal phonetic categorization as early as infancy (Bogliotti, et al., 2008; Serniclaes, Van

Heghe, Mousty, Carre, & Sprenger-Charolles, 2004). According to this view, the deficit in speech perception that affects individuals with dyslexia would be subtle and may not be noticeable in everyday speech that provides multiple contextual and redundant cues.

Yet, when individuals with dyslexia are presented with speech that is ambiguous (such as when background noise is introduced to categorical perception tests), these individuals may perform less well than average-reading individuals.

Some early research also pointed to impaired categorical speech perception in children with dyslexia than in age-matched average-reading children (Godfrey, Syrdal-Lasky,

Millay, & Cox, 1981). Discrimination for phoneme boundaries was poorer in individuals with dyslexia than in average-reading individuals, suggesting that phonemic categories are more confusable. In addition, stimulus identification was shown to be less consistent in individuals with dyslexia. These findings could indicate that individuals with dyslexia

could have less precise or overlapping phoneme categories (Bloomert, et al., 2004; Hazan

& Barrett, 2000).

Other studies have also found similar patterns of poor categorical speech perception in individuals with dyslexia (Boada & Pennington, 2006; Bogliotti, et al., 2008; Chiappe,

Chiappe, & Siegel, 2001; Mody, Studdert-Kennedy, & Brady, 1997; Nittrouer, 1999;

Rosen & Manganari, 2001). A study by Liu and Tsao (2017) revealed that Mandarin- speaking children with dyslexia exhibited less-categorized speech perception, suggesting that imprecise speech perception contributes to reading ability. Access to abstract phonological representations could be affected by poor phoneme categorization, because extracting invariant phonological representations from the speech signal would be more difficult. This would impact word identification when presented with limited acoustic information and would negatively affect the meta-linguistic access to phonemes and, therefore, the acquisition of reading. This view is supported by a study conducted by

Boets, Ghesquiere, van Wieringen, and Wouters, (2007). They tested categorical perception and speech-in-noise perception in a group of children with dyslexia, an age- matched control group, and an adult control group. The group of children with dyslexia performed significantly more poorly than either control group in the speech-in-noise perception tasks. In addition, a significant deficit was also seen in the categorization task involving discrimination. There was a strong positive correlation between performance on the speech perception tasks and measures of phonological awareness.

Hazan, Messaoud-Galusi, & Rosen (2013) conducted a study in which the authors sought to determine whether children with dyslexia were more affected than children with average reading ability by talker and intonation variability when perceiving speech in noise. Children were tested on their perception of consonants in naturally produced CV tokens in multi-talker babble. Twelve CVs were presented for identification in four conditions varying in the degree of talker and intonation variability. The authors investigated consonant place (/bi/-/di/) and voicing (/bi/-/pi/) discrimination with the same conditions. Children with dyslexia made more identification errors than the average-reading children but only for conditions with variable intonation. For discrimination tasks, which had a greater memory and cognitive load, children with dyslexia scored lower than the average-reading children across all conditions.

Chiappe, Chiappe, & Siegel (2001) examined the interaction between speech perception and lexical information in children with dyslexia. All participants’ performance was assessed on tasks targeting reading skill, phonological awareness (including phoneme deletion), and pseudoword repetition. They explored lexical influences in speech perception using the paired stimulus continua of /bis/-/pis/ and /bif/-/pif/ in the phoneme identification paradigm. Both sets had a real word (/pis/ and /bif/) at one end of the continuum and a pseudoword (/bis/ and /pif/) at the other end. The classic lexicality effect involves a boundary shift in which the category boundary for /bif/-/pif/ continuum is at longer voice onset time (VOT) values than that of the /bis/-/pis/ continuum. While the controls demonstrated clearly defined categorical perception in a phoneme identification task for both sets of continua, the category boundary for /bif/-/pif/ was at longer VOTs

than the boundary for /bis/-pis/, which characterizes the classic lexicality effect. The children with dyslexia showed less sharply defined categorical perception on both continua. Although children with dyslexia did not show the classic lexicality effect, lexicality did affect the overall rate with which phonemes were identified as /b/ or /p/ at each VOT. The authors state that these findings suggest that the lexicon may operate as a compensatory mechanism for resolving ambiguity in speech perception. Performing a statistical correction for group differences in phoneme identification made group differences in the phoneme deletion task disappear. The authors suggest that deficits in speech perception may play a causal role in the phonological core deficit associated with dyslexia. McBride-Chang (1996) also found that speech perception deficits may affect access to lexical information, especially when the stimuli is presented in noise. This research utilized structural equation modeling on a large sample of children with dyslexia. The results suggested that the effect of speech perception on reading ability was mediated by its relation to phonological processing abilities.

Another study has shown that the speech perception deficits of individuals with dyslexia may arise from their perception of within-category variants as distinct units. Serniclaes et al. (2004) demonstrated that while individuals with dyslexia show a poorer discrimination across phonemic categories, they also demonstrated an enhanced within-category discrimination when compared to average-readers. As a result they concluded that the phoneme inventories of children and adults with dyslexia may be “overcrowded” with more categories than is necessary for the perception of their native language. They called this phenomenon “allophonic mode of speech perception” and stated that this enlarged

number of possible grapheme-to phoneme correspondences could negatively affect reading acquisition.

2.3 Indexical information

Spoken language encodes two different forms of information that is used by the listener: linguistic information related to the message of the signal (e.g., phonemes, words, semantics, syntax) and indexical information about the speaker (Levi & Pisoni, 2007;

Clopper & Bradlow, 2009). Abercrombie (1967) made a distinction between language and medium. Language is form, whereas medium is substance. Spoken and written

English are the same language, but they are embodied in different mediums. A medium can carry language, yet it has its own independent properties as well. The medium is a human artifact that can indirectly reveal aspects such as mood, social status, and geographical status. This medium carries “extra-linguistic” properties that carry a separate complex system of non-linguistic information. Information of this type can be called an index and the features of the medium that carry these indices are called indexical features, as opposed to linguistic features. Abercrombie goes on to suggest a division of the indexical features into three classes that depend on the kinds of information they provide: (1) those that indicate membership of a group (e.g., regional, dialectal, and social aspects, (2) those that characterize an individual (e.g., age, sex, and size), and (3) those that reveal changing states of a speaker (e.g., anger, happiness, suspicion, or fatigue).

Recent research has revealed that listeners are sensitive to indexical features including the regional dialect spoken in their speech community (Clopper, Levi, & Pisoni, 2006;

Jacewicz & Fox, 2012). Dialect features are primarily conveyed by acoustic spectral cues such as vowel formant pattern, stop consonant releases, consonant cluster reduction or the occurrence of r-colored vowels (Clopper & Pisoni, 2004). Talker sex differences are cued primarily by voice characteristics (Skuk & Scheinberger, 2014). The perceived pitch of male voices is lower than the pitch of female voices. The average fundamental frequency

(F0) is about 100-120 Hz for adult males and 200-220 Hz for adult females. Listeners are able to use this pitch (voice) information to help determine the sex of a talker.

2.4. Research using degraded speech

Differing types of signal distortion have been utilized in conducting speech perception research in order to examine how the human auditory system processes speech when only partial information is available in the speech spectrum. As opposed to adding noise or background babble to speech signals, researchers have begun to use varying types of filters to remove portions of the speech spectrum in order to degrade speech. These filters often produce sounds that are similar to the output of a poorly tuned radio or a bad connection on a telephone call. The primary research interest here is to determine how listeners react to and compensate for these degraded signals. In addition, this research helps to identify neurocognitive functions that may underlie the processing of degraded speech, such as working or phonological memory, controlled selective attention, and inhibitory control (Roman, Pisoni, Kroenberger, & Faulkner, 2016).

2.4.1. Band-pass filtering

One type of speech signal filtering is band-pass filtering, which can be completed as either low-band pass or high-band pass filtering. In low-pass filtering, the lower frequency band of acoustic energy is retained and the high frequencies are cut off.

Conversely, in high-pass filtering, the higher band of acoustic energy is retained and the lower band is cut off. These two different band-pass filters allow for the examination of the contribution of various types of spectral cues to dialect, sex, and even speaker identification. Low-pass filtering, while removing spectral cues, has been shown to preserve prosodic features such as pitch, range, intonation contour, rhythm, rate, and pause information (French & Steinberg, 1947; Pollack, 1948).

A few studies have utilized low-pass filtering to investigate the contribution of prosodic cues to dialect identification. In 1999, van Bezooijen & Gooskens conducted a study that found that speech low-pass filtered as 350 Hz provided very little information to the identification of regional Dutch and British English. In another study, listeners were able to perceive the rhythmic differences between European Portuguese and Brazilian

Portuguese in female speech that was low-pass filtered at 400 Hz, but only with experimentally unaltered intonation contours (Frota, Vigaro, & Martins, 2002). Another study reported that male speech that was low-pass filtered at 300 Hz retained a sufficient number of prosodic cues to allow listeners to differentiate between two Scottish dialects

(van Leyden & van Heuven, 2006). Yet, again, listeners were only able to distinguish between the two dialects when the intonation was preserved. These listeners were unable to distinguish between the dialects when the intonation information was experimentally

removed in a monotonized (f0-flattened) condition. More recently, Fox, Jacewicz, &

Smith (2016) examined the effects of various filter frequency cut-offs on dialect and talker sex identification using low-pass filter bands ranging from 400 Hz to 1,000 Hz.

This study revealed that the lowest 400 Hz frequency band retained very little dialect information and improvement was observed with each higher filter up to 900 Hz. There was no further improvement at 1,000 Hz, but the unprocessed speech still provided the strongest dialect cues. Identification of talker sex was comparatively higher for all filter cut-off points and further improved with unprocessed speech. Finally, listeners benefitted more from cues to talker sex when they shared the same dialect with the talkers.

2.4.2. Noise-vocoding

Another type of degraded speech used in speech perception research is created by noise vocoding. Noise-vocoded speech refers to speech signals that have been processed to preserve gross temporal and amplitude information, but have degraded the fine spectral information. This is accomplished by applying a series of bandpass filters across the entire speech spectrum. This type of spectral degradation was created by Shannon, Zeng,

Kamath, Wygonski, & Ekelid (1995) to model the way a cochlear implant processes speech. Vocoded speech preserves temporal cues and amplitude of the original speech signal through introducing frequency-matched white noise to the signal. Yet it eliminates the tonal (prosodic) information because the signal is excited by noise and not by voice

(which provides the harmonic information). The frequency region that is modified by the noise filter can be as narrow or wide as is allowed by the acoustics of the original signal.

This results in differing frequency-width bands. The ability to identify speech is seen to

improve with a greater number of noise bands. This is due to the fact that the noise bands are narrower (Davis & Johnsrude, 2003). The speech signal must be divided into a set of separate frequency bands in order to create the noise-vocoded analog. When a greater quantity of frequency bands are replaced by noise the narrower the bands become and more spectral details of the original speech signal remain. Noise-vocoded speech having

12 or 16 channels is often easily intelligible. Yet, when fewer channels are used significant degradation to the original signal occurs. In this case, the intelligibility of the speech signal declines (Davis, Johnsrude, Hervais-Adelman, Taylor, & McGettigan,

2005).

Noise vocoding has been used to model the speech processing of cochlear implants as well as to examine the perception of degraded speech in listeners with normal hearing.

Newman, Chatterjee, Morini, & Remaz (2015) studied the speech perception of toddlers at 27 months of age. These researchers presented speech in noise-vocoded and clear speech conditions. They found that the toddlers were able to identify the eight-band noise-vocoded speech quite well in relation to the clear speech presented. Lowenstein,

Nittrouer, & Tarr (2012) examined how vocoding with four noise bands affected the intelligibility of speech in normal hearing adults and children. Participants were asked to repeat clear speech and noise-vocoded sentences. This study found that while the adults could repeat the noise-vocoded sentences that contained little spectral information, the children required intact spectral information in order to repeat a sentence. Yet, as children aged, they relied less and less on available spectral cues, but their performance was still significantly worse than the performance of the adults. The children in the study (4-year

olds) rely heavily on the intelligibility of whole words. This may be due to the fact that in young children early phonological organization is constituted of whole words and not the transition of sounds within words. Nittrouer and Lowenstein (2013) hypothesize that due to the fact that the dynamic co-articulatory nature of vowels in words is eliminated in vocoded speech, adults are better able to guess correctly based on stationary vowels.

Another study that examined the development of perceptual skills in children and compared them to adults used vocoding to look at perceptual abilities of adults and children in two different age ranges. The older children (10-12 years old) performed as well as adults given 4 to 8 noise bands. Yet the performance of the younger children was significantly worse with the same bands. The younger children needed much more spectral resolution (16 noise bands) in order to perform as well as the older children and adults in the study. The authors suggest that these results may indicate that a learning period of about 10 years may be necessary in order for children to be able to fully utilize the spectral information in the speech signal (Eisenberg, Shannon, Schaefer Martinez,

Wygonski, & Boothroyd, 2000). In Dorman, Loizou, Kemp, & Kirk (2000) and

Nittrouer, Lowenstein, & Packer (2009), the performance of children was significantly worse than that of adults for 8-band vocoded speech. These findings support claims that the auditory cortex is still developing into adolescence. Therefore it is expected for age- related improvements in speech perception to be seen in typically developing children

(Moore, Guan, & Wu, 1997). The above studies support the hypothesis that by decreasing the number of channels in noise-vocoded speech, an increase of age-related differences in speech perception are seen.

2.5 Speech Intelligibility

Speech intelligibility is a measure of how comprehensible speech is in given conditions.

Intelligibility is affected by characteristics of the speech signal, such as the volume level, the quality of the speech signal, the type and level of background noise, and reverberation. Intelligibility of speech can be measured as the proportion of speech that a listener can understand. In experimental settings, to determine the proportion of speech that is intelligible, a listener may be asked to transcribe or to repeat what they have heard.

For some experiments, the speech is presented in the form of imperatives and the listener must perform a specific task to indicate they have understood the speech.

2.5.1 Speech intelligibility and dialect

There is a convergence of scientific evidence that indicates that listeners are sensitive to pronunciation differences across regional dialects (Clopper & Pisoni, 2004, 2007). Cross- dialectal communication results in poorer performance than within-dialect communication in a variety of listening tasks. Cross-dialectal communication has been likened to listening to degraded speech signals (Mattys et al. 2012; Van Engen andPeelle

2014): effective communication is possible, but more difficult and more cognitively taxing than within-dialect communication. It has been established that familiarity with a dialect improves its intelligibility (Labov & Ash, 1997; Mason, 1946). Listeners also perform worse on a variety of language-processing tasks when stimuli are drawn from an unfamiliar dialect than when stimuli are in a familiar dialect, and worse yet when the stimuli are in a non-native accent (Adank, Evans, Stuart-Smith, & Scott, 2009). Yet listeners do show adaptation to unfamiliar dialects or accents through exposure or

training (Bent & Holt, 2013). This suggests that intelligibility differences between dialects are at least in part, if not predominantly, a reflection of listener experience.

2.5.1.1 Speech intelligibility, dialect, and degraded speech

While comprehension of any speech type is negatively impacted by noise, researchers have found a larger effect of noise on the intelligibility of unfamiliar, and/or non-native dialects (Munro, 1998; Nygaard et al., 1994; Walker, 2014), and further, that increased experience with a dialect improves performance within an experimental session (Bradlow and Bent, 2008), and across a life-time (Walker, 2014) Several studies have examined the effects of talker dialect and degraded speech on speech intelligibility. Clopper and

Bradlow (2008) conducted a study using four American English regional dialects mixed with noise. Listeners were instructed to transcribe the sentences they had heard. The researchers found that speech intelligibility in noise was generally poorer for more phonologically marked dialects (North, Mid-Atlantic) than less-marked dialects (General

American). This study demonstrated that dialect differences among talkers have a differential effect on intelligibility of degraded speech, yet these effects tend to be attenuated in more favorable listening conditions that provide listeners with a greater redundancy of cues.

In a second study, Adank et al. (2009) examined processing cost (measured in terms of response times) associated with comprehension of sentences produced by talkers by native speakers of these two regional varieties. Standard English listeners showed less efficient speech processing for an unfamiliar regional accent in moderate listening

conditions, most likely due to their lack of experience with the Scottish English. The

Scottish English listeners were equally fast when responding to both varieties, apparently due to their familiarity with both accents. It needs to be underscored that the difference between the familiar and unfamiliar accents was not significant in quiet, suggesting that both groups of listeners benefited from the redundancy of the speech signal in the favorable listening condition.

Finally, Jacewicz & Fox (2015) investigated whether dialect variations of the same language differentially affect intelligibility of speech in the presence of a multi-talker masker. Similar to Van Engen & Bradlow (2007), they found that the intelligibility of speech decreases as the level of the masking noise increases. The effects resulting from the manipulation of the signal-to-noise-ratio level are also applicable to listening environments that involve dialect variation within the same language. They concluded that dialect variation does influence listeners’ performance in a multi-talker environment.

2.5.2 Speech intelligibility and dyslexia

There is some evidence that individuals with dyslexia exhibit subtle spoken language comprehension deficits. McArthur, Hogben, Edwards, Heath, and Mengler (2002) found that in a sample of 110 children with dyslexia, over half of the children scored at least one standard deviation below the mean across tests of spoken sentence comprehension and production. Although not always observable in favorable listening conditions, there is some evidence that individuals with dyslexia exhibit a decreased ability to comprehend sentences when increased processing demands are present. These processing demands

include increased working memory loads, such as when listening to non-native dialect or accented speech, and the processing of speech in challenging conditions, such as speech- in-noise and speech that has degraded signals.

Sentence comprehension performance generally decreases as the working memory load increases. Robertson and Joanisse (2010) examined spoken sentence comprehension in children with dyslexia by comparing them to children with specific language impairment

(SLI), and to same-age and younger control children. They found that syntactic processing problems (marked by poorer performance on non-canonical compared to canonical sentences) were only observed when verbal working memory demands were high for the children with dyslexia and the age-matched control group; however, this effect was more pronounced in the children with dyslexia. Children with specific language impairment also have characteristic syntax deficits, yet often in the absence of apparently high working memory demands. They also investigated whether a significant relationship existed between children’s phonological short-term memory, measured by the CTOPP-2 non-word repetition subtest, and sentence comprehension accuracy and whether this relationship was affected by levels of working memory loads. They found both the children with specific language impairment and the children with dyslexia exhibited poor phonological short-tem memory compared to the age-matched control group. A significant correlation was observed between phonological short-term memory and sentence comprehension accuracy under working memory loads that involved more demanding processing and storage conditions. These results suggest that better phonological memory abilities facilitate spoken sentence comprehension. Therefore,

these results indicate that subtle sentence processing difficulties found in children with dyslexia might be explained as a result from these children’s phonological short-term memory limitations.

2.5.2.1 Speech intelligibility, dyslexia, and degraded speech

Research has demonstrated that individuals with dyslexia exhibit impaired speech-in- noise perception. Brady, Shankweiler, and Mann (1983) studied speech comprehension in individuals with dyslexia and demonstrated the importance of adding background noise to speech targets in order to examine comprehension difficulties found in that group.

Their results showed a clear speech-in-noise comprehension deficit in dyslexic children, with a decrease of intelligibility scores of about 10% compared to aged-matched controls.

They found that a speech-in-noise comprehension deficit could be reproduced in 5-year- old preschool children with a familial history of dyslexia.

Ziegler, Pech-Georgel, George, and Lorenzi (2009) demonstrated that speech-perception deficits in individuals with dyslexia do not only appear when extrinsic noise is added to target-speech but also when the signal is degraded (e.g. by eliminating temporal fine- structure cues). The authors interpreted their findings as evidence that the addition of background noise is not necessary for the speech-perception deficit to occur in individuals with dyslexia but that this deficit may stem from a general lack of speech robustness in the presence of any kind of degradation or interference affecting clarity of speech signals.

More recently, Dole, Hoen, and Meunier (2012) explored processing strategies used by adults with dyslexia during speech-in-noise perception. They looked at the influence of background type by presenting single target-words against backgrounds made of cocktail party sounds, modulated speech derived noise, and stationary noise. Their results confirmed the presence of a speech-in-noise perception deficit in adults with dyslexia, in particular when the competing signal is also speech. The authors state that the fact that speech constitutes the worst background for individuals with dyslexia could be interpreted in the context of a phonological deficit, as it would suggest that the deficit might be specific to speech-sounds used as maskers (Brady et al., 1983; Breier, Gray,

Fletcher, Foorman, & Klaas, 2002). In addition, it may be that noisy conditions just impose increased processing demands on individuals with dyslexia, preventing them from successfully using phonological information present in the target speech signals.

Nittrouer and Lowenstien (2013) examined the hypothesis that developmental dyslexia may be due to faulty perceptual organization of linguistically relevant sensory input.

Sentence-length speech signals were processed to create either sine-wave or noise- vocoded analogs. Seventy children between 8 and 11 years of age, with and without dyslexia participated. Both types of processed sentences were presented for recognition.

Results showed that children with dyslexia had poorer recognition scores than children without dyslexia for both kinds of degraded sentences. Older children with dyslexia recognized the sine-wave sentences better than younger children with dyslexia, but no such effect of age was found for the vocoded materials. Finally, matching young, typical readers with older children with dyslexia on reading abilities did not mitigate the group

difference in recognition of vocoded sentences. Conclusions were that children with dyslexia have difficulty organizing linguistically relevant sensory input, but learn to do so for the structure preserved by sine-wave signals before they do so for other sorts of signal structure. These perceptual organization deficits could account for difficulties acquiring refined linguistic representations, including those of a phonological nature.

2.6 Current research questions

The goal in this dissertation is to determine whether dyslexia limits in some way access to those indexical representations that cue regional dialects. The findings of these studies will contribute to a better understanding of the sources of phonological ability impairment in dyslexia and to the nature of phonological memory deficit, which may also include deficient encoding of indexical cues in speech.

This dissertation examines a specific type of perceptual talker normalization that also relies on implicit long-term memory associated with previous exposure to and experiences with phonetic variation in spoken dialects (Pisoni, 1993). Recent research has brought to light that regional accents cause disruptions in the auditory processing of speech (Floccia, Goslin, Girard, & Konopczynski, 2006) and that typical adult listeners are sensitive to the dynamic dialectal features of a talker’s speech (e.g., Clopper, Levi, &

Pisoni, 2006). Moreover, lifetime experience with a given dialect through prolonged exposure to the regionally accented variant leads to a native-dialect advantage in phonological processing. The own-dialect advantage becomes apparent when listeners are presented with only brief intervals of speech such as monosyllabic words (Fox &

Jacewicz, 2011; Jacewicz & Fox, 2012). Once an individual’s perceptual system becomes attuned to the systematic phonetic variation within a dialect, efficient talker normalization involves integration of both talker idiosyncrasies and the dynamic features of the spoken dialect.

2.6.1 Experiment 1

The purpose of this experiment was to determine whether the underlying phonological impairment seen in dyslexia is associated with a deficit in categorizing regional dialects.

The hypothesis is that, compared to average-reading controls, adults with dyslexia will be less sensitive to dialect features in the speech of multiple talkers. If the native-dialect advantage is the result of asymmetric cultural experiences with the speech of different talker populations (Perrachione, Chiao, & Wong, 2010), listeners with dyslexia are expected to have developed implicit long-term memory for features of their own dialect.

However, the retrieval of these features when listening to voices of unfamiliar talkers is expected to be less effective relative to average-reading controls due to both the less efficient processing under increased cognitive load (Ramus & Szenkovits, 2008) and the deficient talker normalization processes in dyslexia (Perrachione et al., 2011b). That is, the idiosyncratic characteristics of unfamiliar talkers will interfere with dialectal features and the processing of multiple sources of talker variability may further impede access to stored indexical representations (Nygaard & Pisoni, 1998; Sumner et al., 2014).

Consequently, the increased processing effort in listeners with dyslexia relative to average-reading controls will result in their comparatively lower ability to categorize talker dialect.

The second goal of Experiment 1 was to determine the extent to which school-age children with dyslexia, relative to average-reading children, have developed sensitivity to dialect features when presented with extensive between-talker variability. There is some evidence that children with dyslexia score lower than average-reading children in consonant discrimination tasks involving multiple talkers (Hazan, Messaoud-Galusi,

Rosen, Nouwens, & Shakespeare, 2009). However, research on children’s perception of talker variability in relation to regional dialects is scant. It is known that older 9-12 year- old average-reading children perform more poorly than adults when presented with indexical variation in voices of multiple talkers (Jacewicz & Fox, 2014). Their ability to cope with talker variability seems to be still maturing at this age. In particular, their perceptual decisions are comparatively less consistent most likely due to their less efficient encoding of acoustic-phonetic information in the speech of multiple talkers and relative inexperience with regional variation in speech. However, despite their lower performance scores, average-reading children in the Jacewicz and Fox (2014) study identified vowels in their native dialects more accurately than vowels in the non-native dialects, manifesting native-dialect advantage. In the current study, children with dyslexia are expected to perform more poorly than average-reading children and both child groups are predicted to underperform the adults due to their maturing abilities to cope with between-talker variability and relative inexperience with regional variation in speech.

2.6.2 Experiment 2

The previous experiment revealed that the underlying phonological impairment in dyslexia is associated with a deficit in recognizing indexical features in voices of multiple talkers, including talker dialect (Long, Fox, & Jacewicz, 2016). Individuals with dyslexia appear to be deficient in utilizing indexical features. In this experiment we were interested in examining how the amount of the acoustic cues available in stimulus speech affected participants’ ability to identify two differing dialects, central Ohio and western

North Carolina, and whether talkers were male or female. We used stimuli that varied the nature and the redundancy of acoustic cues: clear, unfiltered speech, low-pass filtered speech, and noise-vocoded speech. The low-pass filtered (400 Hz) speech retained voice information but little content. The noise-vocoded speech eliminated all harmonic information.

Building on Experiment 1, this experiment further inquired into listener sensitivity to indexical information (talker dialect and sex) in dyslexia using stimuli that varied the nature and the redundancy of acoustic cues (namely, low-pass filtered speech and noise- vocoded speech). Of interest was the extent to which dyslexia limits the ability to detect indexical features in difficult listening situations, including degraded speech. This study modifies the acoustic information that is available to the listener to examine how voice information may give the listener cues about the dialect and the sex of a talker. In modifying the acoustic information, much of the voice information was removed, yet the spectral information was retained in the low pass-filtered speech condition. In the noise- vocoded speech condition, the temporal information was retained, while the spectral

information was removed. We hypothesized that the performance of listeners with dyslexia will be significantly lower than that of the control group and that they will perform disproportionately lower in the low-pass filtered and noise-vocoded conditions.

The results of this experiment will give insight into how both dyslexic and average- reading individuals utilize both spectral and temporal information when categorizing talker dialect and sex.

2.6.3 Experiment 3

This experiment addresses the question of how adults with dyslexia differ from average- reading adults in their ability to categorize indexical information (talker dialect and sex) when the speech samples are systematically degraded (noise-vocoded). It also examines how intelligible this vocoded speech was to adults with dyslexia and average-reading adults. Since noise-vocoded speech preserves gross temporal and amplitude information but degrades fine spectral information, we hypothesize that cues available to categorize talker dialect and sex, and to understand the content, will be lost as the speech signal is increasingly degraded. We also hypothesize that since the indexical cues are primarily found in the fine spectral structure of speech, that the loss of spectral information will have more of an effect on listeners with dyslexia due to their decreased access to phonological representations.

Chapter 3. Experiment 1

3.1 Methods

3.1.1 Participants

There were a total of 80 participants in the study. Two groups consisted of individuals with dyslexia: 20 adults (10 male, 10 female) ranging in age from 18;2 to 72;8 years (M =

36.7, SD = 17.8) and 20 children (10 male, 10 female) ranging in age from 10;2 to

13;11years (M = 12.4, SD = 1.1). The two corresponding average-reading control groups consisted of 20 adults (10 male, 10 female) ranging in age from 18;0 to 65;1 years (M =

39.7, SD = 12.5) and 20 children (10 male, 10 female) ranging in age from 10;7 to

13;11years (M =12.7, SD = 0.8). An independent-samples t-test revealed no significant differences in chronological age between the two adult groups, t(38) = 0.6, p = .530, or between the two child groups, t(38) = 0.8, p =.470. There were 10 male and 10 female participants in each group. All of the children attended school and ranged from fifth- grade to eighth-grade. All participants were lifelong residents of central Ohio and native speakers of the dialect of American English spoken in that region. They were paid volunteers naïve to the purpose and methods of the study. The study protocol, including the consent, assent, and parental consent forms was approved by the Institutional Review

Board at The Ohio State University.

The children with dyslexia (a) had received a formal diagnosis of dyslexia from a qualified professional, (b) demonstrated a history of difficulty decoding (reading) and encoding (spelling) written language, and (c) were currently being served under an

Individualized Education Plan (IEP) with reading/spelling goals or utilized a Section 504

Plan under the Americans with Disabilities Act (ADA, 1990) with accommodations for reading and spelling. DYS adults reported (a) a childhood history of difficulty learning to read, (b) current difficulty or anxiety regarding reading aloud, (c) poor spelling ability,

(d) slower silent reading with good comprehension skills, and (e) received a scaled score of 7 (“below average”) or less on at least one subtest of the Comprehensive Test of

Phonological Processing – Second Edition (CTOPP-2, Wagner, Torgesen, Raschotte, &

Pearson, 2013). The average-reading adults and children had no history of difficulty learning to read or spell and no current difficulties reading material appropriate for their age and education level. Regarding cognitive function, the children in both groups had scored within the average to above-average range on Cognitive Abilities Test – Form 6

(CogAT, Lohman & Hagan, 2005), which was administered at school. All adults were, at minimum, high school graduates. All participants passed a bilateral pure tone hearing screening, having pure-tone thresholds of ≤ 20-dB HL at octave frequencies from 250 through 8000 Hz (ANSI, 2004).

The CTOPP-2 (a test of phonological processing ability) was administered for all study participants prior to their participation in the study. The CTOPP-2 scores, displayed in

Table 1, are of interest because the test measures three phonological processing abilities

that are related to reading ability: phonological awareness, rapid symbolic naming, and phonological memory (Wagner, et al., 1997). Participants with dyslexia obtained significantly lower scores on all components of CTOPP-2 than did average-reading participants as determined by independent-sample t-tests, p < .001 for each component.

Adults Children CTOPP-2 Dyslexic Non-dyslexic Dyslexic Non-dyslexic Phonological awareness composite 84 (13) 119 (17) 84 (12) 119 (11) Elision 8 (2) 11 (2) 8 (1) 12 (2) Blending 8 (2) 12 (2) 9 (2) 12 (2) Phoneme isolation 5 (2) 15 (4) 5 (2) 14 (3) Phonological memory composite 79 (17) 113 (13) 83 (15) 116 (12) Memory for digits 6 (3) 13 (3) 7 (3) 13 (3) Nonword repetition 7 (3) 11 (2) 7 (2) 12 (2) Rapid symbolic naming composite 81 (15) 109 (11) 85 (15) 109 (11) Rapid digit naming 7 (3) 11 (2) 7 (3) 11 (2) Rapid letter naming 7 (3) 11 (2) 7 (2) 11 (2)

Note. Composite scores are reported as standard scores. Subtest scores are reported as scaled scores.

Table 1. Mean (SD) scores for assessment measures of phonological ability on the Comprehensive Test of Phonological Processing – Second Edition (CTOPP-2) for participants with dyslexia (Dyslexic) and participants without dyslexia (Non-Dyslexic).

3.1.2 Stimulus materials

Meaningful short sentences selected from spontaneous conversations of 40 talkers, 20 from Ohio (OH) and 20 from North Carolina (NC), were the stimuli in the perception test. There were 10 male talkers and 10 female talkers in each group. The Ohio talkers were born and raised in Central Ohio and spoke the Midland variety of American

English, which was also the dialect of the listeners in the current study. The North

Carolina talkers came from Western North Carolina and spoke a variety of Southern

American English known for its strong Southern features (Labov, Ash, & Boberg, 2006).

These talkers were a subset of the talkers included in a larger recorded corpus of regional variation (e.g., Jacewicz, Fox, & Wei, 2010; Jacewicz, Fox, & Salmons, 2011). The talkers ranged in age between 52 and 68 years. The mean ages of the talkers from Ohio were M = 57.7 (SD = 3.4) for males and M = 60.8 (SD = 5.9) for females, and the mean ages of the talkers from NC were M = 58.8 (SD = 5.9) for males and M = 59.4 (SD = 2.8) for females. Each talker produced two sentences for a total number of 80 sentences from all 40 talkers. A complete set of sentences used in the study is included in the Appendix.

The sentences did not contain semantic cues suggestive of the geographic region or background of the talkers. The complete stimulus set was auditorily checked by the experimenters to ensure there were no dysfluencies present and that the speech was appropriate for each dialect. All sound files were equated for root mean square amplitude.

3.1.3 Procedure

The experimental task was a forced-choice identification task. The randomized speech samples were delivered diotically over Sennheizer HD 600 headphones in one block at a comfortable listening level. Upon hearing each sentence, the listener responded by clicking with the mouse on one of two boxes on the computer screen that displayed the two responses: “Ohio” and “North Carolina.” The participants were instructed that they would hear one sentence at a time and that they would need to decide whether the sentence was produced by an Ohio or North Carolina talker. No repetitions were allowed and the listeners were asked to guess if they were uncertain which response to choose.

Each listener was tested individually and the experiment was self-paced. The stimulus presentation and response collection were under the control of a custom MATLAB program. For familiarization with the task, an eight-token practice run was administered to each listener prior to the experiment. Both the sentences and talkers in the practice were different than those in the experiment. No feedback about the accuracy of listener’s responses was provided on either the practice block or the actual experiment.

3.2 Results

Overall, all listeners were able to categorize talker dialect relatively well (M = 78.6% correct), although both dyslexic (DYS) groups scored lower (M = 73.7%) than the corresponding average-reading (AR) groups (M = 83.5%) and adults (M = 82.3%) performed better than children (M = 74.9%). However, as is well known, percent correct

scores do not represent performance well as sensitivity to stimulus differences is not separable from the response bias used in making identification judgments. In addition, percent correct scores cannot easily account for a listener’s decisions made under different degrees of stimulus uncertainty (Lynn & Barrett, 2014). To better understand the above categorization decisions, participants’ responses were analyzed using signal detection theory (STD), which allows for separating listeners’ sensitivity to dialect from their response bias (Green & Swets, 1966; Macmillan & Creelman, 2005). The responses were initially measured in terms of hits (H) and false alarms (FA). The correct categorization of an OH talker was a hit and the categorization of a NC talker as an OH talker was a false alarm. This choice was made because OH was the most common dialect heard by the listeners who grew up in central Ohio. (Please note selection of the correct categorization of the North Carolina talker as a hit would produce equivalent results.) Each participant’s categorization performance was then transformed into A' values, a nonparametric analogue of d' sensitivity measure in STD (Grier, 1971;

Snodgrass & Corwin, 1988; Stanislaw & Todorov, 1999). A' is a distribution-free measure and is less dependent than d' on assumptions regarding normal distribution of scores. We calculated A' following Snodgrass and Corwin (1988) using the following two equations:

(퐻−퐹퐴)(1+퐻−퐹퐴) A'= 0.5 + when H ≥ FA (1) 4퐻(1−퐹퐴)

(퐹퐴−퐻)(1+퐹퐴−퐻) A' = 0.5 - when H < FA (2) 4퐹퐴(1−퐻)

Equation 2 was used in calculations of the A' score for only a single participant (in the

DYS children’s group).

To further investigate potential differences between the strategies of listeners with dyslexia and typical-reading listeners in terms of their inclination to respond “Ohio” rather than “North Carolina,” the responses were analyzed for the effects of listener bias.

We used B''D, a nonparametric measure of response bias proposed by Donaldson (1992), which is similar to B'' measure by Snodgrass and Corwin (1988) (see also Stanislaw &

Todorov, 1999) except that it demonstrates better performance at low levels of discrimination. B''D was computed using Equation 3:

[(1−퐻)(1−퐹퐴)−퐻퐹퐴] B''D = [(1−퐻)(1−퐹퐴)+퐻퐹퐴]

3.2.1 Dialect sensitivity (A' measure)

Using IBM SPSS Statistics v. 21 (2012), a repeated-measures analysis of variance

(ANOVA) was carried out on mean A' scores to investigate the within-subject effect talker sex (male, female) and the between-subject effects listener group (DYS, AR) and listener age (adults, children).

3.2.1.1 Main effects

A preliminary analysis indicated no significant effect of listener sex, F(1, 72) =.317, p =

2 .575, 휂푝 = .020, so this variable was subsequently excluded from further consideration.

Two main effects were significant. Average reading listeners were significantly more

2 sensitive to talker dialect than were the DYS listeners, F(1, 76) = 27.58, p < .001, 휂푝 =

.266 (AR: M = .909; DYS: M = .807) and adults were significantly more sensitive than

2 children, F(1, 76) = 15.24, p < .001, 휂푝 = .167 (adults: M = .877; children: M = .812). As shown in Figure 1, DYS listeners were less sensitive to talker dialect when compared to

AR listeners; of note is a comparatively greater variability in the responses of participants with dyslexia, particularly for children with dyslexia. The main effect of talker sex was

2 not significant, F(1, 76) = 0.17, p = .898, 휂푝 = .000.

Figure 1. Mean dialect sensitivity scores (A') for adult and children participants with average reading ability (AR) and adult and children participants with dyslexia (DYS). Error bars represent 1 SE.

3.2.1.2 Two-way interactions

The two-way interactions involving talker sex were significant, which were subsequently explored with post hoc t-tests. The significant Listener Group × Talker Sex interaction,

2 F(1, 76) = 9.14, p = .003, 휂푝 = .107, displayed in Figure 2, resulted from the fact that average reading listeners showed greater sensitivity to male talkers (M = .901) than to female talkers (M = .876), whereas this pattern was reversed for listeners with dyslexia

(male talkers: M = .789; female talkers: M = .812). However, paired t-tests showed that the male-female differences were significant only for average reading listeners, t(39) =

4.01, p <.001, and not significant for DYS listeners, t(39) = –1.5, p =.143.

Figure 2. Mean dialect sensitivity scores (A’) for participants with average reading ability (AR) and participants with dyslexia (DYS) as a function of talker sex. The error bars represent 1 SE.

There was also a significant Listener Age × Talker Sex interaction, F(1, 76) = 5.31, p =

2 .024, 휂푝 = .065. As illustrated in Figure 3, adult listeners were more sensitive to male talkers (M = .886) than to female talkers (M = .867), whereas the pattern was reversed for children, who were more sensitive to female talkers (M = .821) than to male talkers (M =

.803). Paired t-tests showed that the male-female differences were significant only for adults, t(39) = 2.39, p =.022, but were not significant for children, t(39) = –1.16, p =.253.

The remaining interactions were not significant, including Listener Group × Listener

2 Age, F(1, 76) = .007, p = .936, 휂푝 = .000, and Listener Group × Listener Age × Talker

2 Sex, F(1, 76) = 3.48, p = .066, 휂푝 = .044.

Figure 3. Mean dialect sensitivity scores (A’) for adult and child participants as a function of talker sex. The error bars represent 1 SE.

3.2.2 Response bias (B''D measure)

The contribution of bias to making decisions about talker dialect reflects how liberal or conservative listeners are under uncertainty (Lynn & Barrett, 2014) and is a function of where each listener places his/her criterion for responding “target.” That is, in case of doubt, a liberal listener tends to respond that the ambiguous talker was from Ohio (the target) rather than from North Carolina (the foil). In STD terms (Macmillan & Creelman,

2005), by choosing a target rather than a foil, a liberal listener will show a negative bias.

On the other hand, a conservative listener tends to respond that the talker was from NC

(i.e., will choose a foil rather than a target), showing a positive bias. For the B''D measure,

values lie between –1 and +1. A zero value indicates no bias. Negative values are associated with a liberal bias (since the listener’s criterion is shifted to the left of the two possible response distributions, target and foil), while positive values are associated with a conservative bias (as the response criterion is shifted towards the target distribution to the right). To reiterate, negative values indicate a bias towards responding “Ohio” whereas positive values indicate a bias towards responding “North Carolina.”

Shown in Figure 4 are the results for response bias broken down by listener group and listener age. A repeated measures ANOVA with the between-subject factors listener group and listener age and the within-subject factor talker sex was used to assess possible differences in the response bias. The main effect of group was significant, F(1, 76) =

2 11.71, p < .001, 휂푝 = .133. There were important differences between the groups.

Subsequent one-sample t-tests showed that average reading listeners had a liberal bias, i.e., were significantly more likely to respond “Ohio” (M = –.286), t(39) = –5.078, p

<.001, whereas DYS listeners were unbiased in their responses (M = –.001), t(39) = –

2 .018, p =.985. A significant main effect of talker sex, F(1, 76) = 19.93, p < .001, 휂푝 =

.208, indicated that talker sex influenced listeners’ response bias. In particular, all listeners were significantly more likely to choose “Ohio” in response to a female talker

(M = –.245), t(79) = –5.382, p <.001, but were unbiased in response to a male talker (M =

–.021), t(79) = –.444, p =.658. The main effect of listener age was not significant, F(1,7=

2 .044, p = .835, 휂푝 = .001.

There was a significant Listener Age × Talker Sex interaction, F(1, 76) = 7.07, p = .010,

2 휂푝 = .085, which is illustrated in Figure 5. In terms of bias, adults were significantly more likely to choose “Ohio” in response to a female talker (M = –.171), t(39) = –2.595, p

=.013, but were unbiased in response to a male talker (M = –.080), t(39) = –.999, p =.324.

Children’s responses had the same overall pattern: a liberal bias in response to a female talker (M = –.319), t(39) = –5.191, p <.001, and no bias in response to a male talker (M

=.037), t(39) = .709, p =.483. The significant interaction arose because the difference in bias in response to male versus female talkers was significant for the children, t(78) =

3.68, p <.001, but not for the adults, t(78) = 1.91, p =.060. No other interactions were

2 significant, including Listener Group × Listener Age, F(1, 76) = 1.89, p = .173, 휂푝 = .024,

2 Listener Group × Talker Sex F(1, 76) = 2.53, p = .116, 휂푝 = .032, and Listener Group ×

2 Listener Age × Talker Sex F(1, 76) = 0.36, p = .850, 휂푝 = .000.

Figure 5. Mean Response bias scores (B”D) for adult and child participants as a function of talker sex. The error bar represents 1 SE.

3.2.3 Correlations between dialect categorization (A', B''D) and measures of phonological processing (CTOPP-2)

Listeners’ performance was further examined in relation to clinical assessment measures of phonological processing ability. Pearson’s correlations were carried out separately on

each of the four groups to determine the strength of the associations between dialect categorization measures and CTOPP-2 tests. Here, a positive correlation indicates that a better ability to categorize dialects is associated with a better ability to process phonological information.

For all groups, only phonological memory tests scores (i.e., phonological memory composite, memory for digits and non-word repetition) were strongly associated with dialect sensitivity (A') and no significant correlations among other CTOPP-2 measures with A' scores were obtained (see Table 2). The strongest positive significant correlations among all three phonological memory tests with A' were for adults with dyslexia

(Pearson’s r ranging from .554 to .710). The correlations were also strong for average reading children (from .430 to .670) and moderately strong for average reading adults

(from .249 to .550) but the two average reading groups differed on phonological memory subtests. In particular, while the correlations between phonological memory composite and A' scores were significant for both average reading groups, memory for digits scores significantly correlated with A' scores only for average reading adults and non-word repetition scores significantly correlated with A' scores only for average reading children.

The correlations for children with dyslexia, although positive, were weak (from .145 to

.394) and not significant. This outcome clearly differentiates children with dyslexia from the remaining groups, showing that their weaker sensitivity to dialect features is not significantly associated with their performance on phonological memory tests.

Phonological Elision Blending Phoneme Phonological Memory Nonword Rapid Rapid Rapid awareness isolation memory for digits repetition symbolic digit letter composite naming naming naming composite A' DYS .171 .432 -.099 .353 .710*** .554* .596** -.327 -.334 -.307 adult A' AR -.124 -.208 -.073 -.026 .550* .446* .249 .106 .078 .097 adult A' DYS .176 .101 .096 .192 .394 .358 .145 -.054 -.017 -.070 children A' AR .250 .293 .242 .235 .624** .430 .670** .143 .411 -.172 children

B''D -.130 -.324 -.229 -.034 -.128 -.151 -.032 -.267 -.129 -.372 DYS adult

B''D AR -.204 -.184 .099 -.242 .457* .544* .182 -.091 -.174 .036 adult

B''D DYS .047 -.185 .227 .049 -.089 -.013 .322 -.216 -.005 -.427 children

B''D AR .351 .258 -.037 .412 .538** .356 .524** .123 .269 -.049 children *p < .05, **p < .01, ***p < .001 Table 2. Correlations between measures of phonological ability (CTOPP-2) and measures of dialect categorization (dialect sensitivity (A') and response bias (B''D)) for adults with dyslexia (DYS adult), children with dyslexia (DYS children), adults with average reading ability (AR) and children with average reading ability (AR).

A similar analysis was carried out for the response bias (B''D). For both groups of listeners with dyslexia, no significant correlations among CTOPP-2 scores and B''D scores were obtained. For average-reading listeners, the overall pattern was similar to that for A'.

That is, for adults with average reading ability B''D scores were significantly correlated with the phonological memory composite and with the memory for digits scores; for children with average reading ability, B''D scores were significantly correlated with phonological memory composite and with the non-word repetition scores (see Table 2).

No significant correlations among any other CTOPP-2 measures with B''D scores were obtained for any average reading group.

In summary, the analyses for the average reading groups revealed that both dialect sensitivity and response bias, i.e., listeners’ greater likelihood to respond “Ohio” when uncertain, were significantly correlated with selected measures of phonological memory ability. However, for the groups with dyslexia, significant correlations between dialect sensitivity (A') and phonological memory tests were obtained only for adults with dyslexia. No similar correlations were found for the children with dyslexia. Moreover, no significant correlations between response bias (B''D) and phonological memory tests were obtained for either group with dyslexia, consistent with the finding that listeners with dyslexia were not biased when making decisions about talker dialect.

Chapter 4. Experiment 2

4.1 Methods

4.1.1 Participants

Eighty participants were subjects in this study. All participants were lifelong residents of central Ohio (here defined as the geographic area within a one hour drive from

Columbus) and spoke the dialect of American English common to this region (Midland).

The participants were recruited by word of mouth and were compensated for their time.

All participants reported normal hearing and were naïve to the purpose of the study. The study protocol, including consent, assent, and parental consent forms, was approved by the institutional review board at The Ohio State University.

There were four groups of participants in this study. The two DYS groups consisted of 20 adults ranging in age from 18 to 39 years (M=23.8, SD=5.7) and 20 children ranging in age from 11 to 14 years (M=13.2, SD=1.1). The two corresponding control groups consisted of 20 adults with average reading ability ranging in age from 21 to 45 years

(M=25.5, SD=5.8) and 20 children ranging with average reading ability ranging in age from 11 to 14 years (M=13.1, SD=3.2). There were 10 male and 10 female participants in each group. All children were school aged and attended sixth to eighth grade.

All participants with dyslexia reported difficulty learning to read, difficulty or anxiety associated with reading aloud, poor spelling, and slower than average silent reading with good comprehension. All children with dyslexia had received a formal diagnosis of dyslexia from a qualified professional and were currently being provided services in reading and spelling at school through an individualized education plan (IEP) or utilized a Section 504 plan under the Americans with Disabilities Act (ADA, 1990) with accommodations and/or modifications for reading and spelling. The groups with dyslexia had significantly lower scores than did the control groups on the Comprehensive Test of

Phonological Processing (CTOPP-2; Wagner, Torgesen, Roschotte, & Pearson, 2013) as determined by independent-samples t tests (p < .01). The test measures three phonological processing abilities that are associated with reading ability: phonological awareness, phonological memory, and rapid symbolic naming (Torgesen, 2007). None of the control subjects reported any difficulty learning to read or spell. Also, their general reading skills were appropriate to their age and educational level. Regarding cognitive function, all children in the study scored within the average to above-average range on the school administered Cognitive Abilities Test – Form 6 (Lohman & Hagan, 2005). All adult participants were at least high school graduates.

4.1.2 Stimulus materials

Stimulus tokens were extracted from previously recorded spontaneous conversation of 40 speakers, 20 from central Ohio (OH) and 20 from western North Carolina (NC). Each dialect group was comprised of 10 male and 10 female talkers. The talkers ranged in age

from 52-68 years. For the central Ohio group, the mean age was 57.7 (SD = 3.4) for males and 60.8 (SD = 5.9) for females; for the North Carolina group, the mean age was

58.8 (SD = 5.9) for males and 59.4 (SD = 2.80) for females. The Ohio talkers were all raised and currently lived in central Ohio and spoke the Midland variety of American

English, which was the same dialect spoken by the listeners in this study. The North

Carolina talkers all came from western North Carolina and spoke a variety of Southern

American English. Three utterances from each speaker were selected to use as stimuli for a total of 120 tokens.

In selecting the utterances to be used in this experiment, the conversational speech sample from each speaker was reviewed by an experimenter using audio editing software

(Adobe Audition). The experimenter extracted nine utterances from each speaker based on its overall intelligibility. The experimenter also carefully selected utterances that fell within the conversational pauses of the speaker to insure that the natural speech pattern of the talker was not interrupted. In addition, the selected samples did not contain any lexical information that might link a speaker to any specific region (e.g., “The mountains aren’t a very good place to make a living.”). Also, male-female specific information was not included (e.g., “My wife went to the store.”).

All stimuli were placed into one of three conditions by random assignment. The three conditions of speech presented to the listeners were clear speech, low-pass filtered, and vocoded speech. In the clear speech condition, the stimuli consisted of the original,

unprocessed waveforms (except for amplitude equalization) (see Figure 6). In the low- pass filtered speech condition the stimulus utterances were filtered so that all spectral information above 400 Hz was removed (see Figure 7). Low-pass filtering retains prosodic information, while eliminating spectral detail (and semantic/syntactic content).

In the vocoded speech condition, the stimulus sentences were processed through an 8- channel noise vocoder (see Figure 8). Noise vocoded speech eliminates voice (harmonic) information, while maintaining much of the spectral envelope, which provides detail regarding the semantic/syntactic content of the speech. Speech processed through an 8- channel noise vocoder is considered to contain enough information for sufficient good speech intelligibility (Louizou, 1999). There were three utterances from each speaker in each of the three conditions, for a total of 120 stimuli in each condition. No individual utterance appeared in more than one experimental condition.

Figure 6. A spectrogram of a clear speech stimulus utterance. It is acoustically unprocessed except for amplitude equalization.

Figure 7. A spectrogram of a low-pass filtered utterance in which all spectral information above 400 Hz was removed (and the altered token amplitude equalized).

Figure 8. A spectrogram of a vocoded stimulus processed through an 8-channel noise vocoder (and amplitude normalized).

4.1.3 Procedure

The experiment involved a forced-choice identification task. Participants were asked to identify the gender and dialect of 40 different speakers (20 from central Ohio and 20 from western North Carolina). Each participant was seated in a sound-attenuating booth and heard speech tokens from these 40 different talkers over Sennheiser 640 headphones at a comfortable listening level. After hearing each utterance the participant responded by clicking with a mouse on one of four boxes that appeared on the computer screen: Ohio male, Ohio female, North Carolina male, and North Carolina female (see Figure 9). The entire task was completed in one 45-minute session and participants were compensated

$15 for their time.

Figure 9. A screen shot of response boxes used by the participant during the experiment to identify talker regional dialect and talker sex.

Each participant received verbal instructions regarding the task. The experimenter explained that the participant would hear one sentence at a time and that they had to decide whether the talker was from Ohio (OH) or North Carolina (NC), male or female.

No repetitions were allowed and listeners were asked to guess if they were unsure of which response to choose. Each participant heard stimuli under each of the three stimulus conditions: clear speech (CS), low-pass filtered speech (LP), and vocoded speech (VC).

To insure that the participants understood the task, ten practice items preceded each stimulus set condition. The practice items were followed by 120 test utterances. The order of presentation was counterbalanced across the study. Half of the participants heard

the utterance condition in the order: CS, LP, VC and the other half heard condition order:

CS, VC, LP.

4.2 Results

All data were analyzed using Signal Detection Theory (Snodgrass & Corwin, 1988). A' values are reported as a measure of sensitivity.

4.2.1 Dialect Identification

A repeated-measures ANOVA was carried out on the A' scores with the within-subject factors being stimulus type (CL, LP, VC) and talker dialect (OH, NC). The between- subject factors were group (DYS, AR) and listener age (adults, children). A repeated- measures ANOVA was carried out on the A' scores. The within-subject factors were stimulus type (clear speech, low-pass filtered speech, and vocoded speech) and speaker sex. The between-subjects factors were group (average-readers, individuals with dyslexia) and listener age (adults, children). The results of the analysis revealed that all four of the main effects were statistically significant.

4.2.1.1 Main effects

There was a significant main effect for stimulus type, F(1.684, 128) = 303.3, p < .001, 2

= .800 (Greenhouse-Geisser). Listeners were most sensitive to dialect cues in clear speech (M = 0.885, SE = .007), followed by vocoded speech (M = 0.784, SE = .012).

Listeners were least sensitive to low-pass filtered speech (M = 0.607, SE = 011).

Contrasts revealed a significant difference between clear speech and low-pass filtered speech (p < .001), clear speech and vocoded speech (p < .001), and low-pass filtered speech and vocoded speech (p > .001). There was also a significant main effect of talker sex, F(1, 76) = 39.58, p < .001, 2 = .342 (Greenhouse-Geisser). Listeners were more sensitive to dialect cues in male talkers (M = 0.780, SE = .009) than in female speakers

(M = 0.737, SE = .008). A significant main effect of listener age was also seen, F(1, 76) =

6.564, p = .012, 2 = .080. Adults were more sensitive to dialect in all talkers in all conditions (M = 0.778, SE = .011) than were children (M = 0.739, SE = .011).

A statistically significant main effect of group was seen as well, F(1, 76) = 8.838, p =

.004, 2 = .104. The listeners with dyslexia were less sensitive dialect cues in talkers in all conditions (M = 0.736, SE = .011) than average readers (M = 0.781, SE = .011).

4.2.1.2 Two-Way Interaction

Figure 10 shows a statistically significant two-way interaction between talker sex and stimulus type, F(1.364, 103.7) = 23.66, p < .001, 2 = .237 (Greenhouse-Geisser).

Listeners were more sensitive to dialect cues in male talkers than in female talkers in the clear speech condition (Male: M = 0.918, SE = .007; Female: M = .853, SE = .008) and the low-pass filtered condition (Male: M = 0.651, SE = .015; Female: M = .562, SE =

.014) . In the vocoded speech condition, the pattern was reversed (male talkers: M =

0.772, SE = .012; female talkers: M = 0.797, SE = .013). Contrasts revealed that for each

stimulus condition, sensitivity to dialect was significantly different for male talkers versus female talkers (p < .001).

Figure 10. Mean dialect sensitivity scores (A’) representing interaction between talker sex and stimulus type. The error bars represent 1 SE.

4.2.1.3 Three-Way Interaction

Figure 11 illustrates a statistically significant three-way interaction between talker sex, stimulus type, and age, F(1.364, 103.7) = 4.064, p = .034, 2 = .051 (Greenhouse-

Greisser). In the vocoded speech condition, for adults female speech (M = .838, SE =

.018) provided more dialect information than did male speech (M = .782, SE = .017), but children did not show the same preference (male: M = .761, SE = .017; female: M = .755,

SE = .018). Contrasts revealed that for adults there was a significant difference between

sensitivity to dialect for male versus female talkers in the vocoded condition (p < .01), but for children there was not a significant difference in sensitivity to dialect for male versus female talkers in the vocoded condition (p = .64).

Figure 11. Mean dialect sensitivity scores (A') representing a three-way interaction between talker sex, stimulus type, listener age. The error bars represent 1 SE.

4.2.2 Sex Identification

A repeated-measures ANOVA was carried out on A-prime scores. The within-subject factors were stimulus type (clear speech, low-pass filtered speech, and vocoded speech) and state (Ohio, North Carolina). The between-subjects factors were group (average- readers, individuals with dyslexia) and listener age (adults, children).

4.2.2.1 Main Effects

Again, all four main effects were significant. There was a significant main effect for stimulus type, F(1.957, 76) = 175.8, p < .001, 2 = .698 (Greenhouse-Geisser). Listeners were more sensitive to talker sex in the clear speech condition (M = 0.986, SE = .002), followed by the low-pass filtered speech condition (M = .0960, SE = .004). Listeners were least sensitive to cues for talker sex in the vocoded speech condition (M =0.898, SE

= .066). Contrasts revealed a significant difference between clear speech and low-pass filtered speech (p < .001), clear speech and vocoded speech (p < .001), and low-pass filtered speech and vocoded speech (p > .001). There was a significant main effect for listener group, F(1,76) = 7.112, p = .009, 2 = .086. The average-reading group was more sensitive to talker sex (M = 0.956, SE = .004) than the group with dyslexia (M = 0.939,

SE = .004). There was also a significant main effect for listener age, F(1,76) = 6.008, p =

.016, 2 = .074. The adults were more sensitive to talker sex (M = 0.956, SE = .004) than were the children in this study (M = 0.940, SE = .004). We also found a significant main effect for dialect, F(1, 76) = 13.29, p < .001, 2 = .149. North Carolina dialect provided more cues to talker sex (M = .952, SE = .003) than the Central Ohio dialect (M = .944, SE

= .004).

4.2.2.2 Two-way interactions

There was a significant two-way interaction between listener group and stimulus type,

F(1.975, 76) = 5.505, p = .005, 2 = .068 (Greenhouse-Geisser) (see Figure 12).

Listeners with dyslexia were disproportionately less sensitive to talker sex in the vocoded speech condition (M = 0.880, SE = .008) than the average-reading group (M = 0.995, SE

= .008). Contrasts revealed that there was no difference in sensitivity to talker sex between listeners with dyslexia and controls in the low-pass filtered condition (p = .48).

However, there were significant differences in listener sensitivity to sex between dyslexics and controls in both the clear speech (p = .04) and the vocoded conditions (p =

.004).

Figure 12. Sensitivity to talker sex (A') as a function of listener group and stimulus type. The error bars represent 1 SE.

There was also a significantly two-was interaction between talker dialect and stimulus type, F(1.342, 1) = 34.36, p < .001, 2 = .311 (Greenhouse-Geisser) (see Figure 13).

Listeners were equally sensitive to talker sex cues in the clear speech condition (M =

.986, SE = .002). In the low-pass filtered condition sensitivity to talker sex was lower for

North Carolina dialect (M = .955, SE = .004), whereas for the vocoded condition sensitivity to talker sex was lower for the Ohio dialect (M = .881, SE = .068). Contrasts revealed there was no difference in listener sensitivity to talker sex between Ohio and

North Carolina dialects in the clear speech condition (p = .75). However, there was a significant difference between listener sensitivity to sex between Ohio and North

Carolina dialects in the low-pass filtered condition (p < .001) and in the vocoded speech condition (p < .001).

Figure 13. Sensitivity to talker sex scores (A') as a function of stimulus type and talker dialect. The error bars represent a 95% confidence interval.

Chapter 5. Experiment 3

5.1 Methods

5.1.1 Participants

There were 40 adult participants in this study: 20 in the group with dyslexia and 20 in the average-reading group. Each group was comprised of 10 males and 10 females. The age range in the group with dyslexia was 18 to 57 years of age (M = 25.7, SD = 10.73). The age range for the average-reading group was 22 to 29 years of age (M = 24.0, SD = 1.97).

In the DYS group all participants self-identified as having dyslexia and all reported having had difficulty in learning to read during their childhood. In addition, as adults, all participants in the group with dyslexia reported having experienced difficulty with oral reading fluency, slower silent reading (with average to above-average comprehension), and/or spelling skills. In the average-reading group, no participant reported previous experience involving speech therapy or reading intervention. All AR participants reported average to above-average reading ability. No participants in either group reported having hearing difficulties. All participants were natives of central Ohio and were recruited by word of mouth. All participants either spoke or recognized the Midland dialect of

American English.

Assessment using the Comprehensive Test of Phonological Awareness-2nd edition

(CTOPP-2; Wagner, Torgesen, Raschotte, & Pearson, 2013) was conducted with all

participants. We administered the two subtests in the area of phonological memory

(Memory for Digits and Non-word Repetition). These two subtests were scored individually to attain two scaled scores. These scaled scores were then combined to attain a composite Phonological Memory standard score. We chose to administer only these subtests because in Experiment 1, and in Perichione (2011), only the Phonological

Memory Composite Scores and the Nonword Repitition subtest scores were correlated to performance, dialect sensitivity in the former and voice recognition in the latter. These scores are used to measure the construct of Phonological Memory, that is, one’s ability to encode and store phonological information in short-term or working memory (see table

3). These subtests were administered to each participant prior to participating in the experimental listening tasks. Obtaining these scores allowed researchers to ascertain whether the participants with dyslexia exhibited phonological memory deficits (which we found were correlated to sensitivity measures in Experiment 1) and to ensure that the AR participants did not exhibit phonological memory deficits. A comparison of the two sets of scores (using an independent-samples t-test) allowed us to demonstrate that the individuals in each group did indeed constitute two separate groups. Participants with dyslexia obtained significantly lower scores on the component and composite scores of phonological memory than did the participants with average reading ability as determined by the independent-samples t-tests (p < .001). The means and standard deviations are presented in Table 3 below.

CTOPP-2 scores DYS AR

Phonological Memory 88.1 (11.44) 119.2 (11.93) Composote Score

Memory for Digits 9.5 (3.01) 14.3 (2.27) Subtest Score

Nonword Repitition 6.5 (1.62) 11.7 (2.69) Subtest Score

Note. Composite score is reported as a standard score. Subtest scores are reported as scaled scores.

5.1.2 Stimulus material

For this experiment the stimuli were short phrases and sentences spoken by 40 different speakers: 20 from central Ohio (OH) and 20 from western North Carolina (NC). The talkers from Ohio represented the Midland dialect of American English, while the North

Carolina talkers represented a subset of the Southern dialect of American English. Within each of these regional groups, 10 speakers were male and 10 were female. The talkers’ ages ranged from 51 to 65 years. Each talker contributed 10 different and unique utterances for a total of 400 utterances used in this experiment. These utterances were

taken from previously recorded interviews and conversations. The 400 utterances used in this experiment were recorded during informal talks. Noise-band vocoding was performed using a MATLAB program. This procedure was comparable with the method described in Shannon et al. (1995). The stimulus sentences were first band-pass filtered from 200-8000 Hz, following this each sentence was band-pass filtered into 4, 8, 12, or

16 channels (see Table 4). The original spectral fine structure of each filtered channel was then replaced with band-passed noise and the temporal envelope was used to modulate the noise bands. The frequency bands were then summed to create the final vocoded version of the stimulus sentences. All sentences in all conditions were then amplitude equalized.

Division Number of Channels No. 16 12 8 4 1 278 308 373 629 2 373 448 629 1562 3 489 629 1006 3590 4 629 864 1562 5 799 1168 2382 6 1006 1562 3590 7 1257 2072 5372 8 1562 2734 9 1932 3590 10 2382 4700 11 2927 6738 12 3590 13 4395 14 5372 15 6559

Table 4. Division frequencies (Hz) used for noiseband vocoding processing.

The 400 utterances were then divided into ten stimulus sets. Each talker contributed one sentence to each set. Each set contained 40 utterances, representing an even number of male and female talkers from both dialectal regions (OH and NC). Listeners heard 20 of these ten sets in each of the five conditions: unprocessed clear speech and 4-channel, 8- channel, 12-channel, and 16 channel noise vocoded speech (see Figures 14-18). The ten

sets were then randomly assigned to one of these five conditions for each listener and for each listening task (identification and intelligibility). The individual sentences within each of the ten sets remained the same and appeared in the same order.

Figure 14. Unprocessed sentence “We are in the record business” band-pass filtered from 200 to 8000 Hz.

Figure 15. 16-band vocoding of “We are in the record business” band-pass filtered from 200 to 8000 Hz.

Figure 16. 12-band vocoding of “We are in the record business” band-pass filtered from 200 to 8000 Hz.

Figure 17. 8-band vocoding of “We are in the record business” band-pass filtered from 200 to 8000 Hz.

Figure 18. 4-band vocoding of “We are in the record business” band-pass filtered from 200 to 8000 Hz.

5.1.3 Procedure

There were two listening tasks involved in this experiment. The first task was a forced- choice identification task. Participants were asked to identify the regional dialect (OH or

NC) and the sex of each talker in each of the five conditions (one clear speech condition and four different vocoded speech conditions). The listeners heard one utterance at a time over Sennheiser 640 headphones in a sound-attenuating booth. After hearing each utterance, the participants chose whether they thought the talker was male or female and whether the talker was from central Ohio or North Carolina. These selections were made by clicking on one of the four response boxes on a computer screen in front of them (see

Figure 19). This task took approximately one hour to complete and participants were compensated $15 for their time.

Figure 19. A screen shot of response boxes used by the participant during the experiment to identify talker regional dialect and talker sex.

The second task was completed at least three days after the participant completed the first task (the identification task). In the second task the participants again listened to the same stimuli presented in exactly the same order. After hearing each sentence, the computer screen in front of them prompted them to repeat the as much of the content as they understood. The participants listened to each utterance though Sennheiser 640 headphones in the sound-attenuating booth. They repeated what they heard into a table- mounted SHURE SM58 microphone. Once finished with repeating the utterance as best they could, the listeners clicked “OK” to hear the next utterance (see Figure 20). The intelligibility task lasted approximately 70 minutes for each participant and participants were compensated $20 for their time. Each response was recorded and labeled with the appropriate sentence set and condition by the experimenter. Finally, the participants’ responses were transcribed and scored for completeness by the experimenters.

Figure 20. A Screen shot of the response prompt displayed on the monitor in the Intelligibility Task.

Each task began with ten practice sentences to familiarize each participant with the tasks.

These practice sentences were different than the sentences presented in the experiment.

The participants were told to listen carefully as the utterances could not be repeated. They were instructed that if they were unsure of their response that they should provide their best guess. For the intelligibility task, listeners were instructed to respond “no words” if they did not understand any of the words in the presented utterance. The order of presentation in each task was counterbalanced across all participants. Specifically, the identification task began with four female and six male voices, while the intelligibility task began with four male and six female voices.

5.2 Results

All data were analyzed using Signal Detection Theory. A-prime values are reported as a measure of sensitivity.

5.2.1 Dialect identification

A repeated-measures ANOVA was carried out on A-prime scores. The within-subject factors were stimulus type (unprocessed speech, 4-channel, 8-channel, 12-channel, and

16-channel vocoded speech) and talker sex. The between-subject factor was group

(average-readers, individuals with dyslexia).

5.2.1.1 Main effects

There was a significant main effect for stimulus type, F(2.374, 38) = 58.618, p < .001, 2

= .004 (Greenhouse-Geisser). Listeners were most sensitive to dialect cues in

clear speech (M = 0.915, SD = .050), followed by 16-channel vocoded speech (M =

0.868, SD = .072), 12-channeled vocoded speech (M = 0.865, SD = .077), then 8-channel vocoded speech (M = .844, SD = .081). Finally, listeners were least sensitive to 4-channel vocoded speech4-channel vocoding (M = 0.739, SD = .110) Contrasts revealed significant differences in listener sensitivity to dialect cues between all levels of stimulus type (p < .02) except for between the 16-channel and 12-channel vocoding (p = .71).

5.2.2 Sex identification

A repeated-measures ANOVA was carried out on A-prime scores. The within-subject factors were stimulus type (unprocessed speech, 4-channel, 8-channel, 12-channel, and

16-channel vocoded speech) and talker state (Ohio, North Carolina). The between-subject factor was group (average-readers, individuals with dyslexia).

5.2.2.1 Main effects

There was a significant main effect for stimulus type, F(1.357, 38) = 138.371, p < .001,

2 = .785 (Greenhouse-Geisser). Listeners were most sensitive to speaker sex in clear speech (M = 0.987, SD = .027), followed by 16-channel vocoded speech (M = 0.969, SD

= .022), 12-channeled vocoded speech (M = 0.960, SD = .023), and 8-channel vocoded speech (M = .943, SD = .035). Finally, listeners were least sensitive to 4-channel vocoded speech (M = 0.7786, SD = .097). Contrasts revealed a significant difference between all stimulus conditions (p < .001).

5.2.3 Correlations between dialect categorization (A') and measures of phonological processing (CTOPP-2)

All listeners’ sensitivity to dialect cues in each of the 5 conditions was further examined in relation to their performance on the CTOPP-2 assessment of phonological memory. In

Experiment 1 analysis for the groups (adults and children) with average reading ability revealed that dialect sensitivity was significantly correlated with measures of phonological memory. In addition, for adult listeners with dyslexia, but not for the children, the sensitivity to dialect was significantly correlated to phonological memory scores. Pearson’s correlations were run for each group to determine whether there were any correlations between sensitivity to dialect or sex and phonological memory scores.

The analyses revealed that for both groups (individuals with dyslexia and average reading individuals) there was no significant relationship between the Memory for Digits subtest,

Non-word Repetition subtest, or the Phonological Memory Composite scores and any of the A’ scores for state or for sex identification, p ≥ .13.

5.2.4 Intelligibility

Intelligibility scores expressed in proportional data were transformed to Rationalized

Arcsine Units (RAU), because arcsine transformation stabilizes variance and the normalizes proportional data. A repeated measures ANOVA was carried out on the RAU scores. Within subjects factors were: speaker state, speaker sex, and stimulus condition.

The between subjects factor was listener group. Mauchly’s Test of Sphericity revealed that the assumption of sphericity was met for all main effects except stimulus condition.

The degrees of freedom were corrected using Greenhouse-Geisser estimates of sphericity.

5.2.4.1 Main Effects

There was a significant main effect of talker state on the listeners’ ability to understand the speech tokens. Listeners were better able to understand the degraded speech of the

Ohio speakers than the degraded speech of the North Carolina speakers, F (1, 38) =

282.6, p < .001, η2 = .881 (Ohio: M = 80.59; North Caroline: M = 70.54). There was also a significant main effect of stimulus condition on listeners’ ability to understand the talkers. Listeners were better able to understand speech in the unprocessed speech condition. As the level of degradation of the utterance increased, the ability of listeners to understand the speech decreased, F(3, 16) = 506.313, p < .001, η2 = .930 (Greenhouse-

Geisser). Analysis using t-tests indicated that each of the conditions were significantly different from one another, p < .003. There was a significant main effect of talker sex on the ability of listeners to understand the speech tokens. Listeners were better able to understand the female talkers than the male talkers, F(1, 38) = 26.68, p < .001, η2 = .413

(female: M = .76.86; male: M = 74.27). In addition, there was also a main effect of listener group on the intelligibility of talkers’ speech. Listeners without dyslexia were better able to understand talkers than listeners with dyslexia in all conditions, F(1, 38) =

24.01, p < .001, η2 = .389 (listeners without dyslexia: M = 80.97; listeners with dyslexia:

M = 70.16

5.2.4.2 Two-way interactions

There was a significant two-way interaction effect between stimulus condition and listener group (see Figure 21). Listeners with dyslexia were less able to understand all

talkers than listeners without dyslexia in all conditions of degraded speech. However, in the unprocessed speech condition, there was not a significant difference in the two groups’ performance, F(4,312) = 3.206, p = .015, η2 = .078. Contrasts revealed a significant difference in intelligibility scores between individuals with dyslexia and individuals without dyslexia in all conditions of degraded speech (p < .01). Additionally, contrasts did indicate that there was a significant difference in intelligibility scores between the two groups in unprocessed speech as well (p < .05).

There was also a two-way interaction effect between speaker sex and speaker state (see

Figure 22). Listeners were better able to understand female talkers from North Carolina than male talkers from North Carolina. However, there was not a significant difference in

listeners’ ability to understand male and female talkers from Ohio, F(1,78) = 23.56, p <

.001, η2 = .381. Contrasts revealed a significant difference in intelligibility scores for all listeners when listening to male versus female talkers in the North Carolina dialect (p <

.01). However, there was not a significant difference between intelligibility scores when listening to male and female talkers in the Ohio dialect (p < .63).

Figure 22. Mean (SE) arcsine transformed intelligibility scores for all individuals as a function of talker dialect and talker sex. RAU, rationalized arcsine unit. Error bars represent 2 SE.

5.2.4.3 Three-way interactions

There was a three-way interaction effect between speaker state, speaker sex, and stimulus condition (see Figure 23). Listeners were better able to understand male talkers in their own dialect (Ohio) in all conditions. However, listeners were better able to understand

Ohio female talkers in most conditions, yet for the 4-channel and 8-channel conditions, there was no difference in performance, F(1, 152) = 12.45, p < .001, η2 = .245. Contrasts revealed that for all listeners there was a significant difference between intelligibility scores between dialects when listening to male talkers in all conditions (p > .01). When listeners heard female talkers, there was a significant difference in intelligibility scores between dialects for all listeners in all conditions except for the 4-channel vocoded condition where there was no difference (p = .87).

There was also a three-way interaction effect between stimulus condition, group, and speaker sex (see Figure 24). Listeners without dyslexia were better able than listeners with dyslexia to understand female talkers in all conditions. Listeners without dyslexia were better able to understand male talkers in all conditions except in the unprocessed speech condition. In the unprocessed speech condition there was no significant difference in performance between the listeners without dyslexia and the listeners with dyslexia,

F(4,152) = 3.418, p = .010, η2 = .063. Contrasts revealed a significant difference in intelligibility scores between individuals with dyslexia and average reading individuals for male talkers in all conditions (p < .01). When listening to male talkers there was also a significant difference in intelligibility scores between individuals with dyslexia and average reading individuals in all conditions except the unprocessed speech condition (p

< .38).

5.2.4 Correlations between measures of phonological memory scores (CTOPP-2) and intelligibility scores (RAU)

Overall, both phonological memory subtest scores (i.e., Memory for Digits and Non-

Word Repetition) and the Phonological memory Composite Score were either moderately to strongly associated with intelligibility of speech across all conditions (see Table 5).

Memory for Nonword Phonological Digits Repetition Memory Unprocessed Speech .395* .444** .480**

16-Channel Vocoded .326* .494** .465**

12-channel Vocoded .503*** .605*** .631***

8-Channel Vocoded .457** .498** .542***

4-Channel Vocoded .591*** .527*** .626*** *p < .05, **p < .01, ***p < .001 Table 5. Correlations between measures of phonological memory (CTOPP-2) and measures of intelligibility (RAU) for participants (adult listeners with with dyslexia and controls) in all conditions.

The strongest positive significant correlations were for the Phonological Memory

Composite scores with measures of intelligibility (RAU) across all conditions (Pearson’s r ranging from .542 to .626) (see figures 25-29). The correlations were also strong for the

for the Nonword Repetition Subtest scores (from .444 to .527) and moderately strong correlations for the Memory for Digits Subtest scores (from .326 to .591). Scatterplots representing correlations between intelligibility scores (RAU) and CTOPP-2

Phonological Memory Composite Scores for all participants in all stimulus conditions are shown below (fFgures 25-29).

Figure 26. Scatterplot of correlation between intelligibility scores (RAU) in the 16- channel vocoded speech condition and CTOPP-2 Phonological Memory Composite Scores (expressed in standard scores) for all participants.

Figure 27. Scatterplot of correlation between intelligibility scores (RAU) in the 12- channel vocoded speech condition and CTOPP-2 Phonological Memory Composite Scores (expressed in standard scores) for all participants.

Figure 28. Scatterplot of correlation between intelligibility scores (RAU) in the 8-channel vocoded speech condition and CTOPP-2 Phonological Memory Composite Scores (expressed in standard scores) for all participants.

Figure 29. Scatterplot of correlation between intelligibility scores (RAU) in the 4-channel vocoded speech condition and CTOPP-2 Phonological Memory Composite Scores (expressed in standard scores) for all participants.

Chapter 6. Conclusion

The experiments in this dissertation build on previous reports found in the literature that listeners with dyslexia are deficient in the ability to recognize different voices of unfamiliar talkers in their native language (Perrachione et al., 2011a). We hypothesized that compared with average-reading controls individuals with dyslexia would be less sensitive to indexical features (talker dialect and talker sex) in the speech of multiple talkers because perception of regional accents crucially depends on the ability to detect consistent fine-grained phonetic variations. Furthermore, the details of indexical information in ought to be retained in working memory for association with underlying indexical (or episodic) representations (Goldinger, 1996, 1998; Nygaard & Pisoni, 1998).

We expected individuals with dyslexia to be less efficient in utilizing indexical cues in accommodating talker variability due to both increased demands on working memory caused by the processing of speech produced by 40 different talkers, as well as differing types and levels of speech degradation. In addition, we speculated that the individuals with dyslexia would exhibit lower indexical sensitivity due to impaired access to stored phonological representations (Boets et al., 2013; Ramus & Szenkovits, 2008).

Confirming our predictions, the results of Experiment 1 and 2 found that adults with dyslexia were indeed limited in their ability to categorize talker dialect. Compared with average-reading adults, adults with dyslexia were significantly less sensitive to dialect features and the distribution of their responses showed greater variability. The results of

second experiment supported these findings, as well. Participants with dyslexia in this experiment performed significantly worse than did the individuals with average reading abilities when categorizing dialect in all three stimulus conditions (unprocessed, low-pass filtered, and vocoded speech). The fact that listeners with dyslexia are less sensitive to dialect cues than are average reading individuals may be due to their limited access to implicit long-term memory representations. This impeded access may occur because less efficient operations in working memory including deficiencies in utilizing talker normalization processes. Presumably, the access is hampered by increased processing costs associated with resolving phonetic and dialect-indexical differences presented, not only in clear speech, but also in degraded speech, among multiple talkers in working memory, including those components that are involved in talker category learning (Levi,

2014).

As predicted, overall, children were less sensitive to indexical cues than adults.

In experiment 2, children preformed significantly worse than did adults in identifying talker dialect and talker sex in all three stimulus conditions (unprocessed speech, low- pass filtered speech, and noise vocoded speech). In addition, children with dyslexia were less sensitive to indexical cues than average-reading children. The results for average- reading children are consistent with other reports in the literature demonstrating that processing talker information is more cognitively demanding for children than for adults

(Levi & Schwartz, 2013) and that even older children are still less efficient than adults in accommodating talker variability (Jacewicz & Fox, 2014). The current findings add to the

mounting evidence that phonological processing skills are still maturing in late childhood and early adolescence (Hazan & Barrett, 2000).

The comparatively poorer performance of children with dyslexia in this study is highly suggestive of an underlying phonological deficit, which impedes their abilities to utilize talker normalization processes in the ways the average children do. A related finding was reported for preschool children with specific language impairment, who demonstrated a weaker ability to perceive individual talker variations compared to their age-matched peers (Dailey, Plante & Vance, 2013). In the context of the current study, the phonological deficit was manifested in children with dyslexia’s less effective processing of indexical information. Some aspects of this deficit may be developmental in that they are related to children with dyslexia’s slower development of phonological grammar, that is, the implicit rules or constraints that govern how sounds are assembled into words

(Marshall & Van Der Lely, 2009). Limited experience with regional variation in speech relative to adults is another contributing factor. Of relevance here is a study with younger average reading children by Wagner, Clopper and Pate (2014), who found that 5-6 year- olds were still in the process of mastering regional dialect as an indexical category, not being able to successfully separate their home dialect from another regional dialect. With regard to the current study, it may be the case that a slower development of phonological grammar in children with dyslexia also contributed to their limited ability to categorize indexical properties of speech although this possibility remains to be tested in a focused design. Presumably, the locus of the poorer performance of the children with dyslexia was in the combined effects of deficient phonological ability, maturation of phonological

processing skills, relative inexperience with regional accents including delayed formation of indexical categories, and less effective talker normalization strategies.

A talker-sex-related asymmetry was observed in the first two experiments, as talker sex was seen to interact with the ability to make dialect categorization decisions in both

Experiment 1 and 2. Overall, listeners were more sensitive to dialect cues in male talkers than in female talkers. When comparing average reading-listeners to listeners with dyslexia, the average-reading listeners were more sensitive to dialect in response to male talkers, whereas listeners with dyslexia did not benefit from differences in talker sex.

Similarly, talker sex interacted with listener age so that adults were more sensitive to dialect in response to males whereas children’s decisions were not significantly influenced by the sex.

We note that talker sex has been previously found to interact with dialect features in speech perception in average-reading listeners. In a study examining perceptual sensitivity to different dialects, perceptual ratings on same-sex pairs were higher than ratings on different-sex pairs (Clopper et al., 2006). Importantly, McCloy, Wright, and

Souza (2015) found that when listeners responded to target utterances in two different dialects under noisy conditions, male talkers were more intelligible than female talkers.

This finding suggests that male speech may provide a more distinct set of acoustic features associated with regional variation that become more salient in challenging listening conditions. This may be due to the fact that men typically use more of the local dialect markers in their speech than women (e.g., Barbu, Martin & Chevrot, 2014;

Jacewicz, et al., 2011; Labov, 1990). Our results provide further support for this view: those listeners who showed significantly higher dialect sensitivity (i.e., average-reading listeners versus listeners with dyslexia and adults versus children) were also more sensitive to dialect differences when sentences were spoken by men.

The talker-sex-related asymmetry found in the current studies suggests that male productions may supply greater phonetic constancy with regard to socioacoustic features.

The redundancy of dialect markers in male speech would place a smaller load on working-memory components, facilitating dialect recognition in speech of multiple talkers. Mature phonological processing skills in adults seem to take advantage of this redundancy and their listening strategies may include a form of pattern-recognition implementation, which would impose lower loads on working memory. However, a signal that is highly redundant for an adult may be only minimally redundant for a child

(Mills, 1975), and older children’s perceptual strategies may still employ a different type of acoustic cue weighting which impedes their flexibility in attending to indexical features in speech. Curiously, the less efficient strategies of average-reading children resulted in performance levels comparable to those in adults with dyslexia, whose inability to benefit from the indexical redundancy in male speech can be attributed to their deficient processing skills associated with their impairment.

Regarding listeners’ sensitivity to talker sex, in Experiment 2, all listeners found more cues to talker sex in the unprocessed speech condition, followed by the low-pass filtered condition. Noise-vocoded speech supplied the least cues to talker sex. These results

indicate that gender information is conveyed better by the speaker’s voice (the low-pass condition), and less so when voice cues are unavailable (noise-vocoded speech). In addition, all listeners found more cues to talker sex when listening to talkers from Ohio

(native dialect) in the low-pass filtered stimulus condition. The reverse was true in the noise-vocoded condition: all speakers were more sensitive to talker sex when listening to

North Carolina (non-native dialect) speakers when the speech was vocoded. For the clear speech condition, listeners showed no preference for dialect when identifying talker sex.

Adults were more sensitive to talker sex cues than were children. Average-reading adults were better able to identify talker sex than individuals with dyslexia in both the clear speech and low-pass filtered speech, but there was no difference in their performance in the vocoded speech condition. For Experiment 3, for all listeners, only stimulus condition

(clear speech and 4 levels of speech vocoding) affected listeners’ ability to categorized talker sex.

Next we consider the findings regarding talker dialect and talker sex identification from

Experiment 3. Experiment 3 examined participants’ sensitivity to talker dialect and talker sex in 5 conditions: clear speech and 4 differing levels of vocoding (16-channel, 12- channel, 8-channel, and 4-channel). As opposed to the results for Experiment 1 and

Experiment 2, for both talker dialect and talker sex identification, in this experiment, the only main effect was that of stimulus condition. There was no difference in the performance of both groups of participants: the adults with dyslexia and average-reading adults. Both groups preformed best in the clear speech condition. As the speech signal became more degraded, the performance of all listeners decreased. There was no

significant statistical difference in sensitivity scores between individuals with dyslexia and individuals with average reading skills in categorizing talker dialect and talker sex differences. The reason for this finding is unclear. Perhaps, it is due to the fact that the tasks in Experiment 3 differed in some important way from the first two experiments. In the first experiment, the listener heard unprocessed speech only, heard a total of 80 speech token, and the entire task took approximately 15 minutes to complete. In the second experiment, the listeners heard speech in three differing stimulus conditions: unprocessed, low-pass filtered, and vocoded (8-channel). They responded to a total of

360 speech tokens (120 tokens for each stimulus condition) and the length of the experimental session was 45 minutes. Yet, for the third experiment, the participants heard

400 tokens in five different conditions (unprocessed and four levels of vocoding). The experimental task was divided into two sessions, each held on a separate day. In one session, the participants were asked to categorize dialect and sex of the talker. This took approximately one hour. In the other session, the participants were asked to repeat what they understood the talker to saying. This task required one hour and 15 minutes. Perhaps the varied complexity and length of the sessions impacted the listeners’ sensitivity to talker dialect and talker sex in some way. When the sensitivity scores (A’) among all three experiments were compared using an ANOVA, the results revealed that there was a statistically significant difference between all three sets of scores (Experiment 1: M =

.889, SD = .04; Experiment 2: M = .935, SD = .02; Experiment 3: M = .924, SD = .05).

Another possible explanation for the differing results may be related to the listeners themselves. The groups that participated in the three experiments could have been different in some way. First, we compared the CTOPP-2 scores for each group using an

ANOVA. We found for the variable of phonological processing ability, which is measured by the test, there was no statistically significant difference between the groups of participants for each experiment. Another variable that may have influenced the sensitivity scores is the participants’ ages. The ages of the three groups of participants were compared using an ANOVA. While there was not a statistically significant difference between the ages of the participants of Experiment 2 and Experiment 3, we did find a significant difference between the ages of the participants of Experiment 1 and

Experiment 2 (Experiment 1: M = 39.74 years; Experiment 2: M = 25.5 years;

Experiment 3: M = 24 years). It is unclear how this difference may have impacted the results of these experiments.

For experiment 1, we also examined how the listeners with dyslexia and average-reading listeners differed in terms of response bias. In particular, the average-reading listeners were significantly more likely to respond “Ohio” when uncertain about talker dialect whereas listeners with dyslexia were unbiased in their responses. Clearly, the lack of bias in the listeners with dyslexia, both in the children and adults, corresponds to their reduced dialect sensitivity. When the speaker sounded ambiguous, listeners with dyslexia were not inclined to choose one dialect over the other as average-reading listeners did. This outcome may to some extent reflect inherent differences between individuals with dyslexia and average-reading individuals associated with the presence or absence of dyslexia-related deficiencies. On the other hand, both listeners with dyslexia and average- reading listeners showed similar bias as a function of talker sex so that both groups were significantly more likely to choose “Ohio” in response to a female talker and both were

unbiased in response to a male talker. This result suggests that certain indexical features in female voices triggered a similar liberal response bias, eliminating the differences between the groups. We have no explanation for this curious result at present. Similarly, it is unclear why children (and not the adults) were significantly more likely to respond

“Ohio” to female talkers as indicated by a significant Listener Age × Talker Sex interaction in the analysis of B''D.

In Experiment 2 and 3, we examined the extent to which dyslexia affects the ability to detect indexical features in difficult listening situations, such as degraded speech. For both experiments, we modified the acoustic information that was available to the listeners in order to examine how voice information may give the listener cues about the dialect and sex of a talker. In Experiment 2, we presented clear speech, low-pass filtered speech, and noise-vocoded speech conditions. In Experiment 3, we presented clear speech and four different noise-vocoded conditions (16-channel, 12-channel, 8-channel, and 4- channel). We hypothesized that the performance of listeners with dyslexia would be significantly lower than that of the control group in all conditions. For Experiment 3, we did not find evidence of that discrepancy. Yet, we did find that, for Experiment 2, when compared to controls, listeners with dyslexia had significantly lower sensitivity to talker dialect in all conditions. From these findings, there is evidence that individuals with dyslexia may be deficient in utilizing indexical features in speech, and even more so when coping with impoverished acoustic information.

100

For Experiment 2, we also hypothesized that listeners would perform disproportionately lower in the low-pass filtered and noise-vocoded conditions. Additionally, since noise- vocoded speech preserves gross temporal and amplitude information but degrades fine spectral information, we hypothesized that cues available to categorize talker dialect and sex are lost in vocoded speech because the cues are primarily found in the fine spectral structure of speech and that this loss of spectral information would have more of an effect on listeners with dyslexia due to their decreased access to phonological representations.

For all listeners, the overall pattern of sensitivity to indexical features in all conditions of degraded speech was consistent and the sensitivity to indexical cues was lower in listeners with dyslexia, as predicted. For all listeners, clear speech provided the most cues to both talker dialect and talker sex. Low-pass filtered speech, in which much of the voice information is removed, while the spectral information is retained, provided more cues to talker sex than talker dialect. While vocoded speech, in which the temporal information is retained, while the spectral information was removed, provided more cues to talker dialect than to talker sex. Due to the fact that listener sensitivity to talker dialect was higher in the vocoded speech condition than the low-pass filtered condition, it appears that spectral cues give the listener more information regarding dialect features. The reverse is true for talker sex identification. It appears that voice information contributes more information to talker sex features. Overall, children preformed significantly worse than did adults regarding sensitivity to talker dialect and talker sex in all stimulus conditions.

101

Next we examined any possible correlations between the CTOPP-2 phonological processing scores and A’ scores. It is noteworthy that the correlations between CTOPP-2 scores and A' scores were significant only for the phonological memory components, indicating that dialect sensitivity was most strongly associated with phonological memory abilities. In addition, in Experiment 3, phonological memory scores were moderately to strongly correlated to participants’ intelligibility of speech scores (RAU). These result are consistent with the findings of Robertson & Joanisse (2010) and Perrachione et al.

(2011a). In Robertson and Joanisse, a significant correlation was observed between phonological memory and sentence comprehension in children. In Perrachione et al.

(2011a), adults with dyslexia scored particularly low on phonological memory tests (i.e., on the non-word repetition subtest), which significantly correlated with their poorer voice recognition performance. However, unlike in Perrachione et al., the current participants must have also relied on implicit long-term memory, which underlies acquired knowledge of regional variation in speech. In the task at hand, the listeners were to associate the indexical nuances of the unknown talkers’ voices with both phonological representations of speech sounds and indexical dialect-related information encoded in long-term memory through prior experience with regional variation.

The perceptual operations used to recognize the indexical features in unfamiliar voices of

40 talkers most certainly imposed a processing cost due to heavier working memory load and the current task was also demanding for average-reading adults. The significant correlations between CTOPP-2 scores and A' scores for average-reading adults and average-reading children are therefore hardly surprising. However, it needs to be

102

underscored that adults with dyslexia must have also overcome their reduced ability to retain talker information in working memory in order to associate the indexical features with implicit long-term memory representation for dialect-related variations. Their low scores on phonological memory tests are thus predictive of their reduced dialect sensitivity as evident in strong and significant correlations with A' scores. The lack of similar correlations for children with dyslexia may be explained by a range of developmental factors such as maturation of phonological processing skills, relative inexperience with regional accents or delayed formation of indexical categories as discussed above. These additional variables make the direct relation between phonological memory scores and dialect sensitivity less straightforward. Again, the analysis of the data in Experiment 3 did not reveal any positive correlation between phonological memory scores and listeners’ sensitivity to talker dialect and talker sex

(A’). These results are in line with another study by Hazan, Messaoud-Galusi, & Rosen

(2013). These authors note that the lack of significant correlation between scores reflecting phonological short-term memory and discrimination test results is surprising and supports an argument against a simple link with short-term memory.

Finally, in Experiment 3, we examined how individuals with and without dyslexia performed in the area of speech intelligibility. The speech was presented in multiple stimulus conditions and the talkers in this experiment included both male and female talkers and had differing dialects (Central Ohio and Eastern North Carolina). Our findings support the previous literature in which typical listeners found both non-native dialects (Labov & Ash, 1997) and degraded speech (Clopper & Barlow, 2008) more

103

difficult to understand. As predicted, all listeners were better able to understand the degraded speech of the Ohio speakers (the native dialect of all listeners) than the degraded speech of the North Carolina speakers, which supports the “native dialect advantage”. This advantage comes from listeners’ far greater experience with their own dialect. Listeners were also better able to understand speech in the unprocessed speech condition. As the level of degradation of the utterance increased, the ability of listeners to understand the speech decreased. This points to the relative processing cost of an unfamiliar dialect under adverse listening conditions, as seen in Adank, Evans, Stuart-

Smith, & Scott (2009).

In conclusion, the majority of our results support two findings in previous literature: that individuals with dyslexia have more difficulty than listeners without dyslexia in both understanding speech in a non-native dialect (McArthur, et al., 2000) and in conditions of degraded speech (Nittrouer, 2010). Listeners with dyslexia were less able to understand talkers than listeners without dyslexia in all conditions of degraded speech. The reason for this may be that processing of a non-native dialect, while simultaneously processing degraded speech, over tax an already compromised phonological processing system.

Female talkers provided more cues to intelligibility for listeners without dyslexia. The reverse was true for listeners with dyslexia were better able to understand male talkers, except in the unprocessed speech condition. In the unprocessed speech condition there was no significant difference in performance between the listeners without dyslexia and the listeners with dyslexia, perhaps due to the fact that speech becomes more intelligible

104

in more favorable conditions that provide listeners with greater redundancy in the speech signal.

In general, we found that adults and children with dyslexia had more difficulty categorizing the indexical features talkers’ speech (talker dialect and talker sex) than did the adults and children without dyslexia. We speculate this is due to the increased processing load that these tasks place on the already impoverished phonological processing system of individuals with dyslexia. When talker speech was degraded by low-pass filtering or noise-vocoding, all listeners identification patterns were similar: low-pass filtering yielded better talker sex identification and noise-vocoding yielded better talker dialect identification. In Experiment 2, individuals with dyslexia performed disproportionately poorer than average reading individuals. It is unclear why we did not find the same discrepancy in Experiment 3 as was found in Experiment 1 and 2.

Children performed worse than adults on these categorization tasks, as well. We attribute this result to the fact that children have less experience with non-native dialect than adults.

A correlation was observed between the CTOPP-2 (The Comprehensive Test of

Phonological Processing – 2nd Edition) phonological memory scores and the sensitivity to talker dialect and talkers sex scores (A’) for all participants in Experiment 1. This echoes the findings in Perrochione (2011) who also found a positive correlation between CTOPP phonological memory scores and speaker voice recognition ability. Again, however, the results of Experiment 3 do not support this finding

105

Yet, Experiment 3 did reveal a discrepancy in intelligibility scores between listeners with dyslexia and listeners without dyslexia. Listeners with dyslexia scored more poorly on measures of intelligibility across all conditions. We also found that intelligibility scores were higher for all listeners when the talker was from central Ohio (the native dialect advantage). The fact that listeners performed worse with non-native speech replicates previous findings (Munro 1998).

The current results carry some important implications for individuals with dyslexia. First, the current results point to a need for additional assessment tools to evaluate the phonological processing abilities of individuals with dyslexia, specifically measures to assess an individual’s sensitivity to indexical cues in speech. The results from such an assessment could aid in identifying those individuals who have difficulty processing indexical information, especially as it pertains to understanding speech. Second, our results suggest that the influence of regional dialect variation on phonological category formation in children with dyslexia may underlie some of the difficulties with accurate and fluent word recognition and poor spelling abilities. For example, talker-listener dialect mismatches or an extensive accentedness of input speech may lead to an rise in processing effort in children with dyslexia and interfere with their sound-to-letter mapping.

For individuals with dyslexia (especially children with dyslexia), listening to the speech of various talkers who speak differing dialects may overtax an already impoverished phonological processing system and inhibit comprehension of speech. The adverse

106

effects of this increased workload on the ability to access phonological representations would likely be heightened in difficult listening environments. Additionally, individuals with dyslexia may have more difficulty understanding speech when listening to the speech of a talker who does not speak the native dialect of the listener. It is important to consider how dialect mismatch may affect students with dyslexia when their daily instruction is delivered in a non-native regional dialect of English. Speech comprehension may be compromised. This effect would be intensified if the speech occurs in less than optimal listening conditions, such as in a busy airport, a public restaurant, or a loud classroom. Trying to comprehend non-native speech in these types of listening environments could have a negative impact on work or academic performance of individuals with dyslexia.

Importantly, this difficulty should be considered regarding a wide variety of educational settings – large crowded lecture halls with poor acoustical properties, physical education classes held in gymnasiums, and classrooms full of very vociferous students. Being exposed to these types of listening environments and a dialect mismatch on a daily basis may have a negative impact on the academic growth of students with dyslexia. Of special consideration is the possible negative impact of dialect mismatch in noisy conditions when teaching skills which require students to attend to the finely-grained acoustic information, such as when teaching sound-to -symbol and symbol-to sound mapping. In a noisy classroom environment it may be helpful to provide multiple opportunities for students to listen to instruction, especially when a student with dyslexia is learning to read and spell. For students with dyslexia who struggle daily to process non-native dialect

107

in less than optimal listening environments, the negative impact on their literacy skills would have long-lasting and far-reaching consequences.

108

References

Abercrombie, D. (1967). Elements of General Phonetics. Chicago: Aldine.

Adams, A. M., & Gathercole, S. E. (2000). Limitations in working memory: Implications For language development. International Journal of Language and Communication Disorders, 35, 95-116.

Adank, P., Evans, B. G., Stuart-Smith, J., & Scott, S. K. (2009). Comprehension of familiar and unfamiliar native accents under adverse listening conditions. Journal of Experimental Psychology: Human Perception and Performance, 35(2), 520- 509.

Americans With Disabilities Act of 1990, Pub. L. No. 101-336, 104 Stat. 328 (1990).

ANSI. (2004). Specification for audiometers (ANSI S3.6-2004). American National Standards Institute, New York.

Baddeley, A. (1986). Working memory. Oxford: Clarendon.

Baddley, A. (1993). Working memory. Science, 255, 556-559.

Baddeley, A. (2000). The episodic buffer: A new component of working memory? Trends in Cognitive Science, 4, 417–423.

Baddeley, A. (2003). Working memory and language: An overview. Journal of Communication Disorders, 36, 189–208.

Baddeley, A. (2007). Working memory, thought, and action. Oxford: Oxford University Press.

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.

Barbu, S., Martin, M., & Chevrot, J. P. (2014). The maintenance of regional dialects: A matter of gender? Boys, but not girls, use local varieties in relation to their friends’ nativeness and local identity. Frontiers in Psychology, 5, Article 1251.

109

Bent, T., & Holt, R. F. (2013). The influence of talker and foreign-accent variability on spoken word identification. The Journal of the Acoustical Society of America, 133(3), 1677-86.

Blomert, L., Mitterer, H., & Paffen, C. (2004). In search of the auditory, phonetic, and/or phonological problems in dyslexia: Context effects in speech perception. Journal of Speech, Language, and Hearing Research, 47, 1030-1047.

Boada, R., & Pennington, B. F. (2006). Deficient implicit phonological representations in children with dyslexia. Journal of Experimental Child Psychology, 95, 153-193.

Boets, B.,Wouters, J.,van Wieringen, A., & Ghesquière, P.(2007).Auditory processing, speech perception and phonological ability in preschool children at high-risk for dyslexia: A longitudinal study of the auditory temporal processing theory. Neuropsychologia, 45, 1608–1620.

Boets, R., Ghesquiere, P., van Wieringen, A., & Wouters, J. (2007). Speech perception in preschoolers at family risk for dyslexia: Relations with low-level auditory processing and phonological ability. Brain and Language, 101, 19-30.

Boets, B., Op de Beeck, H. P., Vandermosten, M., Scott, S. K., Gillebert, C. R., Mantini, D., Bulthé, J., Sunaert, S., Wouters, J., & Ghesquière, P. (2013). Intact but less accessible phonetic representations in adults with dyslexia. Science, 342(6163), 1251-1254.

Bogliotti, C., Serniclaes, W., Messoud-Galusi, S., & Sprenger-Charolles, L. (2008). Discrimination of speech sounds by children with dyslexia: Comparisons with chronological age and reading level controls. Journal of Experimental Child Psychology, 101, 137-155.

Bonifacci, P., & Snowling, M. J. (2008). Speed of processing and reading disability: A cross-linguistic investigation of dyslexia and borderline intellectual functioning. Cognition, 107, 999-1017.

Bowers, P. G., & Wolf, M. (1993). Theoretical links between naming speed, precise timing mechanisms, and orthographic skill in dyslexia. Reading and Writing: An Interdisciplinary Journal, 5, 69-85.

Bradley, L., & Bryant, P. E. (1983). Categorizing sounds and learning to read – a causal connection. Nature, 301(3), 419-421.

Bradlow, A. R., & Bent, T. (2007). Perceptual adaptation to non-native speech. Cognition, 106(2), 707-729.

Brady, S. (1986). Short-term memory, phonological processing and reading ability. Annals of Dyslexia, 36, 138-153.

110

Brady, S., Shankweiler, D., & Mann, V. (1983). Speech perception and memory coding in relation to reading ability. Journal of Experimental Child Psychology, 35, 345- 367.

Breier, J. I., Gray, L. C., Fletcher, J. M., Foorman, B., & Klaas, P. (2002). Perception of speech and non-speech stimuli by children with and without reading disability and attention deficit hyperactivity disorder. Journal of Experimental Child Psychology, 82, 226–250.

Bruck, M. (1992). Persistence of dyslexic’s phonological awareness deficits. Developmental Psychology, 28, 874-886.

Catts, H. W., & Adolph, S. M. (2011). Phonological and other language deficits associated with dyslexia. In S. A. Brady, D. Braze, & C. A. Fowler (Eds.), Individual differences in reading: Theory and evidence. (pp. 137-151). New York, NY: Psychology Press.

Catts, H. W., Adolph, S. M., Hogan, T. P., & Ellis-Weismer, S. E. (2005). Are specific language impairments and dyslexia distinct disorders? Journal of Speech, Language, and Hearing Research, 48, 1378-1396.

Chiappe, P., Chiappe, D. L., & Siegel, L. S. (2001). Speech perception, lexicality, and reading skill. Journal of Experimental Child Psychology, 80, 58-74.

Clopper, C. G., & Bradlow, A. R. (2008). Perception of dialect variation in noise: Intelligibility and classification. Language and Speech, 51, 175–198.

Clopper, C. G., & Bradlow, A. R. (2009). Free classification of American English dialects by native and non-native listeners. Journal of Phonetics, 37, 436-451.

Clopper, C. G., Levi, S. V., & Pisoni, D. B. (2006). Perceptual similarity of regional dialects of American English. Journal of the Acoustical Society of America, 119, 566-574.

Clopper, C. G., & Pisoni, D. B. (2004). Some acoustic cues for the perceptual categorization of American English regional dialects. Journal of Phonetics, 32(1), 111-140.

Clopper, C. G., & Pisoni, D. B. (2007). Free classification of regional dialects of American English. Journal of phonetics, 35(3), 421-438.

Cowan, N. (2005). Working memory capacity. Hove, East Sussex, UK: Psychology Press.

Cowan, N. (2010). The magical mystery four: How is working memory capacity limited, and why? Current Descriptions in Psychological Science, 19, 51-57.

111

Creel, S. C., & Bregman, M. R. (2011). How talker identity relates to language processing. Language and Linguistic Compass, 5, 190-204.

Crowder, R. G., & Wagner, R. K. (1991). The psychology of reading. New York, NY: Oxford University Press.

Crystal, A. E. (2001). A dictionary of language (2nd ed.). Chicago, IL: University of Chicago Press.

Dailey, N. S., Plante, E., & Vance, R. (2013). Talker discrimination in preschool children with and without specific language impairment. Journal of Communication Disorders, 46, 330-337.

Davis, M. H., Johnsrude, I. S., Hervais-Adelman, A., Taylor, K., & McGettigan, C. (2005). Lexical information drives perceptual learning of distorted speech: Evidence for the comprehension of noise-vocoded sentences. Journal of Experiemental Psychology: General, 134(2), 222.

Davis, M. H., & Johnsrude, I. S. (2003) Hierarchical processing in spoken language comprehension. Journal of Neurosicence, 23(8), 3423-3431.

Desroches, A. S., Newman, R. L., Robertson, E. K., & Joanisse, M. F. (2012). Journal of Speech, Language, and Hearing Research, 56, 250-264).

Dickie, C., Ota, M., & Clark, A. (2013). Revisiting the phonological deficit in dyslexia: Are implicit non-orthographic representations impaired? Applied Psycholinguistics, 34, 649-672.

Dole, M., Hoen, M., & Meunier, F. (2012). Speech-in-noise perception deficit in adults with dyslexia: Effects of background type and listening configuration. Neuropsychologia, 50, 1543–1552.

Donaldson, W. (1992). Measuring recognition memory. Journal of Experimental Psychology: General, 121, 275-277.

Dorman, M. F., Loizou, P. C., Kemp, L. L., & Kirk, K. I. (2000). Word recognition by children listening to speech processed into a smaller number of channels: Data from normal-hearing children and children with cochlear implants. Ear and Hearing, 21(6), 590-596.

Eisenberg, L. S., Shannon, R. V., Schaefer Martinez, A., Wygonski, J., & Boothroyd, A. (2000). Speech recognition with reduced spectral cues as a function of age. Journal of the Acoustical Society of America, 107(5), 2701-2710.

112

Ferrer, E., Shaywitz, B. A., Holahan, J. M., Marchione, K., & Shaywitz, S. E. (2010). Uncoupling of reading and IQ over time: Empirical evidence for a definition of dyslexia. Psychological Science, 21(1), 93-101.

Fletcher, J. M., Lyon, G. R., Fuchs, L. S., & Barnes, M. A. (2007). Learning disabilities: From identification to intervention. New York, NY: Guilford Press.

Floccia, C., Goslin, J., Girard, F., & Konopczynski, G. (2006). Does a regional accent perturb speech processing? Journal of Experimental Psychology: Human Perception and Performance, 32, 1276-1293.

Fox, R. A., & Jacewicz, E. (2011). Perceptual sensitivity to dialectal and generational variation in vowels. Interspeech 2011: Proceedings of the International Speech Communication Association (pp. 2921-2924). Florence, Italy: International Speech Communication Association.

Fox, R. A., Jacewicz, E., & Smith, J. T. (2016). Effects of low-pass filtering on dialect and gender perception. Journal of the Acoustical Society of America, 139(4), 2122.

Fraser, J., Goswami, U., & Conti-Ramsden, G. (2010). Dyslexia and specific language impairment: The role of phonology and auditory processing. Scientific Studies of Reading, 14, 8-29.

French, N. R., & Steinberg, J. C. (1947). Factors governing the intelligibility of speech Sounds. Journal of the Acoustical Society of America, 19, 90-119.

Frota, N. R., Vigario, M., & Martins, F. (2002). Language discrimination and rhythm classes: Evidence from Portuguese. In Speech Prosody 2002, International Conference.

Gathercole, S. E., & Baddeley, A. D. (1990). Phonological memory deficits in language of disordered children: Is there a causal connection? Journal of Memory and Language, 29, 336-360.

Gathercole, S. E., Willis, C. S., & Baddeley, A. D. (1991). Differentiating phonological memory and awareness of rhyme: Reading and vocabulary development in children. British Journal of Psychology, 82, 387-406.

Geiger, G., Lettvin, J. Y., Zegarra-Morgan, O. (2009). Task determined strategies of visual process. Cognitive Brain Research, 1, 39-52.

Godfrey, J. J., Syrdal-Lasky, A. K., Millay, K. K., Knox, C. M. (1981). Performance of dyslexic children on speech perception tests. Journal of Experimental Child Psychology, 32, 401-424.

113

Goldinger, S. D. (1996). Words and voices: Episodic traces in spoken word identification and recognition memory. Journal of Experimental Psychology: Learning, Memory, and Cognition, 22, 1166–1183.

Goldinger, S. D. (1998). Echoes of echoes? An episodic theory of lexical access. Psychological Review, 105, 251–279.

Goswami, U. (2003). Why theories about developmental dyslexia require developmental designs. Trends in Cognitive Sciences, 7, 534-540.

Goswami, U. (2010). Phonology, reading, and reading difficulty. In K. Hall, U. Goswami, C. Harrison, S. Ellis, & J. Soler (Eds.), Interdisciplinary perspectives on learning to read: Culture, cognition, and pedagogy (pp. 103-116). London: Routledge.

Green, D. M., & Swets. J. A. (1966). Signal detection theory and psychophysics. New York: Wiley.

Green, K. P., Tomiak, G. R., & Kuhl, P. K. (1997). The encoding of rate and talker information during phonetic perception. Perception & Psychophysics, 59, 675- 692.

Greenwood, D. D. (1990). A cochlear frequency‐ position function for several species— 29 years later. The Journal of the Acoustical Society of America, 87(6), 2592- 2605.

Grier, J. B. (1971). Nonparametric indexes for sensitivity and bias: Computing formulas. Psychological Bulletin, 75, 424-429.

Hawelka, S., & Wimmer, H. (2005). Impaired visual processing of multi-element arrays is associated with increased number of eye movements in dyslexic reading. Vision Research, 45, 855-863.

Hazan, V., & Barrett, S. (2000). The development of phonemic categorization in children aged 6-12. Journal of Phonetics, 28, 377-396.

Hazan, V., Messaoud-Galusi, S., & Rosen, S. (2013). The effect of talker and intonation variability on speech perception in noise in children with dyslexia. Journal of Speech, Language, and Hearing Research, 56, 44-62.

Hazan, V., Messaoud-Galusi, S., & Rosen, S., Nouwens, S., & Shakespeare, B. (2009). Speech perception abilities of adults with dyslexia: Is there any evidence for a true deficit? Journal of Speech, Language, and Hearing Research, 52, 1510-1529.

114

Hulme, C., Neath, I., Stuart, G., Shostak, L., Surprenant, A. M., & Brown, G. D. (2006). The distinctiveness of the word-length effect. Journal of Experimental Psychology: Learning, Memory, and Cognition, 32(3), 586.

Hulme, C. & Snowling, M. J. (2009). Developmental disorders of language learning and cognition. West Sussex, United Kingdom: Wiley-Blackwell.

Jacewicz, E., & Fox, R. A. (2012). The effects of cross-generational and cross-dialectal variation on vowel identification and classification. Journal of the Acoustical Society of America, 131, 1413-1433.

Jacewicz, E., & Fox, R. A. (2014). The effects of indexical and phonetic variation on vowel perception in typically developing 9- to 12- year-old children. Journal of Speech, Language, and Hearing Research, 57, 389-405.

Jacewicz, E., & Fox, R. A. (2015). The effects of dialect variation on speech intelligibility in a multi-talker background. Applied Psycholinguistics, 36(3), 729- 746.

Jacewicz, E., Fox, R. A., & Salmons, J. (2011). Cross-generational vowel change in American English. Language Variation and Change, 23, 45-86.

Jacewicz, E., Fox, R. A., & Wei, L. (2010). Between-speaker and within-speaker variation in speech tempo of American English. Journal of the Acoustical Society of America 128, 839-850.

Johnson, K. (2005). Speaker normalization in speech perception. In D. B. Pisoni & R. E. Remez (Eds.), The handbook of speech perception (pp. 363–389). Malden, MA: Blackwell Publishing Ltd.

Johnson, K. (2006). Resonance in an exemplar-based lexicon: The emergence of social identity and phonology. Journal of Phonetics, 34, 485–499.

Jones, M. W., Branigan, H. P., Hatzidaki, A., & Obregon, M. (2010). Is the “naming” deficit in dyslexia a misnomer? Cognition, 116, 56-70.

Kail, R., Hall, L. K., & Caskey, B. J. (1999) Processing speed, exposure to print, and naming speed. Applied Psycholinguistics, 20, 303-314.

Kamhi, A. G., & Catts, H. W. (2012). Language and reading disabilities (3rd ed.). Boston, MA: Pearson.

Labov, W. (1990). The intersection of sex and social class in the course of linguistic change. Language Variation and Change, 2, 205-254.

115

Labov ,W., & Ash, S. (1997). Understanding Birmingham. In C. Bernstein C, T. Nunnally, & R. Sabino, (Eds), Language variety in the south revisited (pp. 508- 573). Tuscaloosa, AL: University of Alabama Press.

Labov, W., Ash, S., & Boberg, C. (2006). Atlas of North American English: Phonetics, phonology, and sound change. Berlin, Germany: Mouton de Gruyter.

Lervag, A., & Hulme, C. (2009). Rapid automatized naming (RAN) taps a mechanism that places constraints on the development of early reading fluency. Psychological Science, 20, 1040-1048.

Levi, S. V. (2014). Individual differences in learning talker categories: The role of working memory. Phonetica, 71, 201-226.

Levi, S. V., & Pisoni, D. B. (2007). Indexical and linguistic channels in speech perception: Some effects of voiceovers on advertising outcomes. Psycholinguistic Phenomena in Marketing Communications (T. M. Lowrey, ed.). Boca Raton, FL: Lawrence Erlbaum Associates.

Levi, S. V. & Schwartz, R. G. (2013). The development of language-specific and language-independent talker processing. Journal of Speech, Language, and Hearing Research, 56, 913–920.

Liberman, I. Y. (1973). Segmentation of the spoken word and reading acquisition. Bulletin of the Orton Society, 23, 665-77.

Liberman, I. (1983). A language-oriented view of reading and its disabilities. In H. Myklebust (Ed.), Progress in learning disabilities (Vol. V), New York, NY: Grune & Stratton.

Lieven, E. (2008). Building language competence in first language acquisition. European Review, 16, 445-456.

Liu, H. M., & Tsao, F. M. (2017). Speech Perception Deficits in Mandarin-Speaking School-Aged Children with Poor Reading Comprehension. Frontiers in psychology, 8, 2144.

Lohman, D. F., & Hagan, E. P. (2005). Cognitive Abilities Test – Form 6. Rolling Meadows, IL: Riverside Publishing.

Long, G. B., Fox, R. A., & Jacewicz, E. (2016). Dyslexia limits the ability to categorize talker dialect. Journal of Speech, Language, and Hearing Research, 59, 900-914.

Lowenstein, J. H., Nittrouer, S., & Tarr, E. (2012). Children weight dynamic spectral structure more than adults: Evidence from equivalent signals. Journal of the Acoustical Society of America, 132(6), EL443-EL449.

116

Luce, P. A., McLennan, C. T., & Charles-Luce, J. (2003). Abstractness and specificity in spoken word recognition: Indexical and H variability in long-term repetition priming. In Bowers, J. & Marsolek, C. (Eds.), Rethinking implicit memory (pp. 197-214). Oxford: Oxford University Press.

Lynn, S. K., & Barrett, L. F. (2014). “Utilizing” signal detection theory. Psychological Science, 25, 1663-1673.

Lyon, G. R., Shaywitz, S. E., & Shaywitz, B. A. (2003). A definition of dyslexia. Annals of Dyslexia, 53, 1-14.

Manis, F. R., Doi, F. M., & Badha, B. (2000). Naming speed, phonological awareness, and orthographic knowledge in second graders. Journal of Learning Disabilities, 33, 325-333.

McDonald, M. C., & Christiansen, M. H. (2002). Reassessing working memory. Psychological Review, 109, 35-54.

Macmillan, N. A., & Creelman, C. D. (2005). Detection theory: A user’s guide (2nd ed). Mahwah, NJ: Erlbaum.

Manis, F. R., Doi, L. M., & Badha, B. (2000). Naming speed, phonological awareness, and orthographic knowledge in second graders. Journal of Learning Disabilities, 33, 325-333.

Manis, F. R., McBride-Chang, C., Seidenberg, M. S., Keating, P., Doi, L. M., Munson, B., & Petersen, A. (1997). Are speech perception deficits associated with developmental dyslexia? Journal of Experimental Child Psychology, 66, 211-235.

Manis, F. R., Seidenberg, M. S., & Doi, L. M. (1999). “See Dick RAN.” Rapid naming and the longitudinal prediction of reading subskills in first and second graders. Scientific Studies of Reading, 3, 129-157.

Marshall, C. R., & Van Der Lely, H. K. J. (2009). Effects of word position and stress on onset cluster production: Evidence from typical development, specific language impairment, and dyslexia. Language, 85, 39-57.

Martin, C. S., Mullennix, J. W., Pisoni, D. B., & Summers, W. V. (1989). Effects of talker variability on recall of spoken word lists. Journal of Experimental Psychology: Learning, Memory, and Cognition, 15, 676-684.

Mason, H. M. (1946) Understandability of speech in noise as affected by region of origin of speaker and listener. Speech Monographs, 13(2), 54-58.

117

Mattingly, I. G. (1972). Reading, the linguistic process and linguistic awareness. In J. Kavanagh & I. Mattingly (Eds.), Language by ear and by eye (pp. 133-147). Cambridge, MA: MIT Press.

Mattys, S., Davis, M., Bradlow, A., & Scott, S. (2012). Speech recognition in adverse conditions: A review. Language and Cognitive Processes, 27, 953–978.

Mayo, C., Scobbie, J. M., Hewlett, N., & Waters, D. (2003). The influence of phonemic awareness development on acoustic cue weighting strategies in children’s speech perception. Journal of Speech, Language, and Hearing Research, 46, 1184-1196.

McArthur, G. M., Hogben, J. H., Edwards, S. M., Heath, S. M., & Mengler, E. D. (2000). On the “specifics” of specific reading disability and specific language impairment. .Journal of Child Psychology and Psychiatry, 41, 869–874.

McBride-Chang, C. (1996). Models of speech perception and phonological processing in reading. Child Development, 67, 1836-1856.

McCloy, D. R., Wright, R. A., & Souza, P. E. (2015). Talker versus dialect effects on speech intelligibility: A symmetrical study. Language and Speech, 58, 371-386.

Miller, C. J., Sanchez, J., & Hynd, G. W. (2003) Neurological correlates of reading disability. In H. L. Swanson, K. R. Harris, & G. Graham (Eds.), Handbook of Learning Disabilities (pp. 242-255). New York, NY: Guilford Press.

Mills, J. H. (1975). Noise and children: A review of literature. Journal of the Acoustical Society of America, 58, 767-779.

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.

Moore, J. K., Guan, Y. L., & Wu, B. J. C. (1997). Maturation of the human auditory cortex: Laminar cytoarchitecture and axonal growth. Association for Research in Otolaryngology Abstracts, 20, 8.

Mullennix, J. W., & Pisoni, D. B. (1990). Stimulus variability and processing dependencies in speech perception. Perception & Psychophysics, 47, 379-390.

Munro, M. J. (1998). The effects of noise on the intelligibility of foreign accented speech. Studies in Second Language Acquisition, 20, 139–153.

Muter, V., Hulme, C., Snowling, M., & Taylor, S. (1998). Segmentation, not rhyming, predicts early progress in learning to read. Journal of Experimental Child Psychology, 71, 3-27.

118

National Institutes of Health, National Institute of Child Health and Human Development (NICHD). (2002). Definition of dyslexia. Retrieved from https://www.nichd.nih.gov/publications/pubs/council_db_2004/ documents/

Neuhaus, G. F., & Swank, P. R. (2002). Understanding the relations between RAN letter subtest components and word reading in first-grade students. Journal of Learning Disabilities, 35, 158-174.

Newman, R. S., Chatterjee, M., Morini, G., & Remez, R. E. (2015). Toddlers’ comprehension of degraded signals: Noise-vocoded versus sine-wave analogs. Journal of the Acoustical Society of America, 138(3), EL311-EL317.

Nittrouer, S. (1999). Do temporal processing deficits cause phonological processing problems? Journal of Speech, Language, and Hearing Research, 42, 925-942.

Nittrouer, S., & Lowenstein, J. H. (2010). Learning to perceptually organize speech signals in native fashion. The Journal of the Acoustical Society of America, 127(3), 1624-1635.

Nittrouer, S., & Lowenstein, J. H. (2013). Perceptual organization of speech signals by children with and without dyslexia. Research in Developmental Disabilities, 34(8), 2304-2325.

Nittrouer, S., Lowenstein, J. H., & Packer, R. R. (2009). Children discover spectral skeletons in their native language before the amplitude envelopes. Journal of Experimental Psychology: Human Perception and Performance, 35(4), 1245.

Nusbaum, H. C., & Morin, T. M. (1992). Paying attention to differences among talkers. In Y. Tohkura, E. Vatikiotis-Bateson, & Y. Sagisaka (Eds.), Speech perception, production and linguistic structure (pp. 113-154). Washington: IOS Press.

Nygaard, L. C., & Pisoni, D. B. (1998). Talker-specific learning in speech perception. Perception & Psychophysics, 60, 355–376.

Nygaard, L. C, Sommers, M., & Pisoni, D. B. (1994). Speech perception as a talker- contingent process. Psychological Science, 51, 42–46.

Olofsson, A., & Lundberg, I. (1983). Can phonemic awareness be trained in kindergarten? Scandinavian Journal of Psychology, 24, 35-44.

Parilla, R., Kirby, J. R., & McQuarrie, L. (2004). Articulation rate, naming speed, verbal short-term memory, and phonological awareness: Longitudinal predictors of early reading development? Scientific Studies of Reading, 8, 3-26.

119

Pennington, B. F. (2009). Diagnosing learning disorders: A psychological framework (2nd ed.). New York, NY: Guilford Press.

Pennington, B. F., & Bishop, D. V. M. (2009). Relations among speech, language, and reading disorders. Annual Review of Psychology, 60, 283-306.

Perrachione, T. K., Chiao, J. Y., & Wong, P. C. M. (2010). Asymmetric cultural effects on perceptual expertise underlie an own-race bias for voices. Cognition, 114, 42- 55.

Perrachione, T. K., Del Tufo, S. N., & Gabrieli, J. D. (2011a). Human voice recognition depends on language ability. Science, 333, 595.

Perrachione, T. K., Del Tufo, S. N., Ghosh, S. S., & Gabrieli, J. D. (2011b). Phonetic variability in speech perception and the phonological deficit in dyslexia. In W.-S. Lee & E. Zee (Eds.), Proceedings of the 17th International Congress of Phonetic Sciences (pp. 1578-1581). Hong Kong, China.

Pierrehumbert, J. B. (2001). Exemplar dynamics: Word frequency, lenition and contrast. In J. Bybee & P. Hopper (Eds.), Frequency effects and the emergence of linguistic structure (pp. 137–158). Amsterdam: John Benjamins.

Pisoni, D. B. (1993). Long-term memory in speech perception: Some new findings on talker variability, speaking rate and perceptual learning. Speech Communication, 13, 109-125.

Pisoni, D. B. (1997). Some thoughts on “normalization” in speech perception. In K. Johnson, & J. W. Mullennix (Eds.), Talker variability in speech processing (pp. 9- 32). San Diego, CA: Academic Press.

Pollack, I. (1948). Effects of high pass and low pass filtering on the intelligibility of speech in noise. Journal of the Acoustical Society of America, 20, 259-266.

Powell, D., Stainthorp, R., Stuart, M., Garwood, H., & Quinlan, P. (2007). An experimental comparison between rival theories of rapid automatized naming performance and its relationship to reading. Journal of Experimental Child Psychology, 98, 46-68.

Ramus, F., & Szenkovits, G. (2008). What phonological deficit? Quarterly Journal of Experimental Psychology, 61, 129-141.

Rayner, K., Pollatsek, A., Ashby, J., & Clifton, C. (2011). The psychology of reading (2nd ed.). New York, NY: Psychology Press.

120

Robertson, E., & Joanisse, M. (2010). Spoken sentence comprehension in children with dyslexia and language impairment: The roles of syntax and working memory. Applied Psycholinguistics, 31(1), 141-165.

Robertson, E. K., Joanisse, M. F., Deroaches, A. S., and Ng, S. (2009). Categorical speech perception deficits distinguish language and reading impairments in children. Developmental Science, 12(5), 753-757.

Roman, A. S., Pisoni, D. B., Kronenberger, W. G., Faulkner, K. F. (2016). Some neurocognitive correlates of noise-vocoded speech perception in children with normal hearing: A replication and extension of Eisenberg et al. (2002). Ear and Hearing, 38(3), 344-356.

Rosen, S., & Manganari, E. (2001). Is there a relationship between speech and nonspeech auditory processing in children with dyslexia? Journal of Speech, Language, and Hearing Research, 44, 720-736.

Rumsey, J. M., Nace, K., Donohue, B., Wise, D., Maisog, J. M., & Andeason, P. (1997). A positron emission tomographic study of impaired word recognition and phonological processing in dyslexic men. Archives of Neurology, 54, 562-573.

Savage, R., Lavers, N., and Pillay, V. (2007). Working memory and reading difficulties: What we know and what we don’t know about the relationship. Educational Psychology Review, 19, 185-221.

Savage, R., Pillay, V., & Melidona, S. (2007). Deconstructing rapid automatic naming: Component processes and the prediction of reading difficulties. Learning and Individual Differences, 17, 129-146.

Serniclaes, W., van Heghe, S., Mousty, P., Carre, R., & Sprenger-Charolles, L. (2004). Allophonic mode of speech perception in dyslexia. Journal of Experimental Child Psychology, 87, 336-361.

Shankweiler, D., & Crain, S. (1986). Language mechanisms and reading disorder: A modular approach. Cognition, 24, 139-168.

Shankweiler, D., & Liberman, I. (1989). Phonology and reading disability [IARLDMonograph No. 6]. Ann Arbor, MI: University of Michigan Press.

Shannon, R. V., Zeng, F. G., Kamath, V., Wygonski, J., & Ekelid, M. (1995). Speech recognition with primarily temporal cues. Science, 270, 303-304.

Shaywitz, S. E., & Shaywitz, B. A. (2005). Dyslexia (Specific Reading Disability). Biological Psychiatry, 57, 1301-1309.

121

Skuk, V. G., & Schweinberger, S. R. (2014). Influences of fundamental frequency, formant frequencies, aperiodicity, and spectrum level on the perception of voice gender. Journal of Speech, Language, and Hearing Research, 57, 285-296.

Snodgrass, J. G., & Corwin, J. (1988). Pragmatics of measuring recognition memory: Applications to dementia and amnesia. Journal of Experimental Psychology: General, 117, 34-50.

Stanislaw, H., & Todorov, N. (1999). Calculation of signal detection theory measures. Behavior Research Methods, Instruments, & Computers, 31, 137-149.

Sumner, M., Kim, S. K., King, E., & McGowan, K. B. (2014). The socially weighted encoding of spoken words: A dual-route approach to speech perception. Frontiers in Psychology, Article 1015. doi: 10.3389/fpsyg.2013.01015

Torgeson, J. K. (1988). Studies of children with learning disabilities who perform poorly on memory span tasks. Journal of Learning Disabilities, 21, 605-612.

Torgesen, J. K. (1996). A model of memory from an information processing perspective: The special case of phonological memory. In G. R. Lyon & N. A. Krasnegor (Eds.). Attention, memory, and executive function (pp. 157-184). Baltimore, MD: Brookes.

Torgesen, J. K. (2007). Recent discoveries from research on remedial intervention forchildren with dyslexia. In M. Snowling & C. Hulme (Eds.), The science of reading: A handbook (pp. 521-537). Malden, MA: Blackwell.

Torneus, M., Hedstrom, G. B., & Lundberg, I. (1991). Phonological awareness and reading: A chicken and egg problem? Journal of Educational Psychology, 76, 1346-1358.

Tunmer, W., & Greaney, K. (2010). Defining Dyslexia. Journal of Learning Disabilities, 43, 229-243.

Uhry, J. K. (2005). Phonemic awareness and reading: Research, activities, and instructional materials. In J. R. Birsh (Ed.), Multisensory teaching of basic language skills (pp. 83-111). Baltimore, MD: Paul H. Brookes. Van Bezooijen, R., & Gooskens, A. G. (1999). Identification of language varieties: The contribution of different linguistic levels. Journal of Language and Social Psychology, 18(1), 31-48.

Van Engen, K. J., & Bradlow, A. R. (2007). Sentence recognition in native- and foreign- language multi-talker background noise. The Journal of the Acoustical Society of America, 121(1), 519-26.

122

Van Engen, K., & Peelle, J. (2014). Listening effort and accented speech. Frontiers in Human Neuroscience 8, 577.

Van Leyden, K., & van Heuven, V. J. (2006). On the prosody of Orkney and Shetland dialects. Phonetica, 63(2-3), 149-174.

Velluntino, F. R., & Fletcher, J. M. (2007). Developmental dyslexia. In M. J. Smowling & C. Hulme (Eds.), The science of reading: A handbook (pp. 362-378). Maldon, MA: Blackwell.

Voller, K. K. S. (2004). Dyslexia. Journal of Child Neurology, 19, 740-744.

Wagner, L., Clopper, C. G., & Pate, J. K. (2014). Children’s perception of dialect variation. Journal of Child Language, 41, 1062-1084.

Wagner, R. K., Torgesen, J. K., Laughon, P., Simmons, K., & Raschotte, C. A. (1993). Development of young readers’ phonological processing abilities. Journal of Educational Psychology, 85, 83-103.

Wagner, R. K., Torgesen, J. K., Raschotte, C. A., & Pearson, N. A. (2013). Comprehensive Test of Phonological Processing – Second Edition. Austin, TX: Pro-Ed.

Wagner, R. K., Torgesen, J. K., Raschotte, C. A., Hecht, S. A., Barker, T. A., Burgess, S. R., & Garon, T. (1997). Changing causal relations between phonological processing abilities and word-level reading as children develop from beginning to fluent readers: A five-year longitudinal study. Developmental Psychology, 33, 468-479.

Walker, A. (2014). Crossing Oceans with Voices and Ears: Second Dialect Acquisition and Topic-Based Shifting in Production and Perception. Doctoral dissertation, The Ohio State University.

Waters, G. S., & Caplan, D. (1996). The capacity theory of sentence comprehension: Critique of Just and Carpenter (1992). Psychological Review, 103, 761-772.

Wolf, M. A., Bowers, P. G., & Biddle, K. (2000). Naming-speed processes, timing, and reading: A conceptual review. Journal of Learning Disabilities, 33, 387-407.

Ziegler, J. C., Pach-Geogel, C., George, F., & Lorenti, C. (2009). Speech perception in noise deficits in dyslexia. Developmental Science, 12, 732-745.

123

Appendix A. Experiment 1 - Sentence Sets

The 80 stimulus sentences used in the dialect categorization task listed by talker dialect and talker sex.

Central Ohio – male talkers

1. Several years ago it was a very small town. 2. And now it’s becoming a much larger city. 3. My son doesn’t have a major yet. 4. My daughter played volleyball and softball. 5. I’m going to talk to you about my son Bradley. 6. We are involved in our college search right now. 7. We had a great time going up in Jim’s van. 8. The drive home was very long it was tiring. 9. And about a two-hour wait in one spot. 10. The sun come out which was nice. 11. It was Linguistics 201 I think. 12. Particularly in the region of the country where we are. 13. I’m told that I was born in nineteen thirty-seven. 14. The next event was the move itself. 15. Met at church at a young adults’ group. 16. Her aunt and uncle used to be the advisers of the group. 17. I guess she’s a freshman in college next. 18. My dad was a sort of hobby farmer. 19. A side yard that was big enough to play baseball. 20. That he was going to move back to Italy.

Central Ohio – female talkers

21. My parents always had a business there. 22. Because my dad had had a nervous breakdown. 23. He ended up not going into the service. 24. The neighborhood was always a safe place to be. 25. I don’t know why I don’t like bees. 26. And I should tell you about my new kitten. 27. One shy one is medium and the other one is very outgoing.

124

28. And of course there is the barking beagle. 29. More ethnically diverse than it was back then too. 30. My mother’s house was sold last year. 31. “C” is my middle initial. 32. And so now I try not to use any middle initial. 33. And so I became a librarian. 34. It only lasted for three years. 35. We were married in October of nineteen seventy-seven. 36. Our children are all on the tall side also. 37. Our oldest son lives in Salt Lake City. 38. The youngest works for the Washington Speakers Bureau. 39. Some people do not appreciate their family. 40. He comes to work and he cannot see.

Western North Carolina – male talkers

41. My mother was a schoolteacher for forty-two years. 42. I had one brother his name was John. 43. A lifetime of listening to my parents and grandparents. 44. And you know people still do that today. 45. I have a twenty year-old daughter named Paige. 46. My father came over here after the Second World War. 47. We did a program called vector analysis. 48. The boss superintendent sent me to Providence, Rhode Island. 49. Donna had the most beautiful wedding dress. 50. Exactly, precisely like the great Ellingberg wedding. 51. I was standing in a cafeteria line. 52. It was a fun place to go. 53. He was born in nineteen thirty-eight. 54. She had to have me in a hospital. 55. Cassie’s supposed to come home today in fact. 56. Bryce is going to have surgery on his shoulder 57. And I’ve done some other interesting things along the way. 58. Meant a lot of good things to me over the years. 59. I have no idea of how my mom and dad met. 60. We grew up and went to school together.

Western North Carolina – female talkers

61. My granddaughter was born two months after my mother died. 62. Four generations that I did not get. 63. Janet the words are there in your head if you would just write them down. 64. Well, no, but I figured I could save it up. 65. I’m going to tell you about my middle daughter. 66. Eight baby copperheads under the blanket.

125

67. I got a master’s degree in history. 68. I have seven grandchildren and they are all very healthy. 69. My mother came from a family of eight. 70. Fifteen cents to get into a ball game. 71. I’m a retired banker of twenty-five and a half years. 72. And we do horse and buggy weddings. 73. I was the middle of three children. 74. I guess you could say Indian-type herbs. 75. People here had never seen anything like it. 76. I can’t think of a particular example. 77. I remember both my grandmothers very well. 78. My sister got me interested in genealogy. 79. I have three children, two boys and one girl. 80. And we each have a portion of this land.

126

Appendix B. Experiment 2 – Sentence Sets

# of Dur State Sex Syll (ms) Ord Sentence OH fem 5 1203 1 I have two children. NC fem 12 2660 2 And I found many of their death certificates. NC m 13 4767 3 My paternal grandfather was a Baptist preacher. OH fem 10 1821 4 And I should tell you about my new kitten. NC m 12 2153 5 So that was an interesting experience. NC fem 10 1881 6 And I could remember her very well. NC m 6 1392 7 They like to hear me talk. NC m 6 1576 8 He calls me everyday. NC m 6 1426 9 That's all I've got to say. My granddaughter was born 2 months after my NC fem 14 3254 10 mother died OH fem 13 2358 11 I enjoy doing it and they enjoy getting them. NC fem 6 1839 12 I guess her heart was big. OH m 12 2306 13 I've been practicing law since 1982. NC m 5 2071 14 We had a string band. NC fem 7 2129 15 And I have three grandchildren. OH fem 5 1288 16 It's my married name. OH fem 11 2148 17 We moved back here in 1987. NC fem 9 2823 18 My mother is eighty three years old. OH m 12 2070 19 I finished my master's here in seventy one. OH fem 5 1347 20 I learned to tap dance. OH fem 12 1903 21 He had to go to the veteran's hospital. OH m 9 1548 22 That was the best part of the whole thing. OH m 13 2240 23 You could make a radio or a burglar alarm.

127

OH fem 5 1590 24 He is very tall. NC fem 5 1503 25 He's a fisherman. NC m 12 2906 26 The superintendent sent me to Providence OH fem 6 2009 27 There's the barking beagle. OH fem 8 1370 28 It only lasted for three years. NC fem 8 1502 29 I kept scrubbing the shower stall. And it's the most wonderful thing- being a NC fem 14 2592 30 grandparent. NC m 6 1069 31 You have to deal with her. OH m 6 1083 32 They would live to the south OH m 5 1545 33 The seats were great seats. NC m 5 1210 34 Let me hear you talk. OH m 10 1796 35 There's currently a job opening there. OH m 10 2204 36 My dad was sort of a hobby farmer. OH m 5 1327 37 The high school's the same. OH m 7 1811 38 My oldest is twenty three. NC fem 6 1569 39 I'm not ashamed of it. NC m 14 2095 40 They decided to open a laundry and dry cleaners.

State Sex # of Syll Dur (ms) Ord Sentence Traffic was obviously getting worse as we came OH m 14 2437 1 home. NC fem 9 1703 2 It was a bustling community. NC fem 14 2488 3 She was not dependent on a lot of other people. OH fem 7 1269 4 Yeah so they come over here. NC fem 7 2169 5 So I ran down to the shed. NC m 10 2131 6 You've got to go see this woman. NC m 6 1032 7 That was a bad mistake. OH fem 11 1938 8 The love you receive is unconditional. NC fem 5 964 9 I love the mountains. OH fem 11 2744 10 And it has been in my family every since.

128

He would walk to the church and then back home on NC m 13 3887 11 Sundays. NC fem 7 1646 12 They used to play full court then. NC fem 12 3539 13 And I've made a book for each of our three children. OH fem 9 2124 14 I went to their college of nursing. NC m 7 1346 15 There was a little fellow. OH fem 5 879 16 But it was worth it. OH fem 5 1380 17 Mom, you've burned me out. NC m 9 1824 18 Kathy came out by the wall one day. OH m 7 1282 19 We got married in August. NC m 6 1689 20 And he's lived here for years. NC m 7 1508 21 I hadn't checked up on her. OH fem 8 1053 22 I became a librarian. NC fem 11 2315 23 In their married years they were around Cleveland OH m 12 2709 24 Well it seemed like it was really high at the time. OH fem 12 1718 25 He has an international studies degree. OH m 7 1420 26 She needed a break from that. NC fem 7 1766 27 We've sort of retired here. OH m 14 3129 28 And I arrived in my driveway at 2:30 AM. More words were coming but they were a different NC fem 12 2908 29 verse. OH fem 6 1355 30 That's always a good thing. NC m 10 1401 31 That's how politicians get elected. NC m 9 1650 32 And I ended up staying twelve years. OH fem 6 1627 33 I like to sew and quilt. OH m 8 1365 34 He would grow things in the summer. OH m 5 1311 35 So he moved back here. NC m 10 2476 36 She was fifteen days over a hundred. NC fem 8 1987 37 She's pretty special in our lives. OH m 10 2242 38 It's a very competitive program. OH m 7 1510 39 A survey of linguistics. OH m 6 1065 40 That was a lot of fun.

129

State Sex # of Syll Dur (ms) Ord Sentence OH m 8 1372 1 We've been married thirty-nine years. NC m 9 1997 2 It takes something to get me started. NC m 8 1999 3 She's been real good for stuff like that. OH m 8 1860 4 She had no worries about that. NC fem 7 1371 5 It was a beautiful day. OH m 10 1855 6 He's trying to get into nursing school. OH fem 10 2049 7 So I had two teaching certificates. NC fem 8 3000 8 You had to find a white ash tree. NC fem 7 1524 9 My life is pretty simple. OH m 8 1286 10 My mother was a teacher too. OH fem 5 983 11 Try to adopt one. NC fem 8 1510 12 Now I live on Riverwood Hill. NC m 8 1435 13 My wife had just left on a trip. OH fem 8 1742 14 He has dreams of saving the world. NC m 8 1267 15 Somebody has to write those games. NC m 7 1837 16 I was born during the war. NC fem 9 2600 17 Where in the world have I heard this song? NC fem 11 3205 18 We actually didn't eat on the table OH fem 8 1766 19 My husband is an attorney. NC fem 8 1850 20 She takes it right after him. OH fem 12 2326 21 I do like bumblebees 'cause they help my garden. NC m 8 2055 22 They all were very successful. OH fem 8 1339 23 But she lives in Albuquerque. OH fem 9 1742 24 My passion is Disney trivia. OH fem 6 859 25 I could teach anything. OH m 10 2126 26 But I'm a huge college basketball fan. OH m 11 2096 27 I've been married for almost 25 years. NC m 8 3888 28 We had him saying "Fried Chicken". NC fem 7 1855 29 For about fifty-five years

130

OH m 9 1822 30 It's an interest of mine, a hobby. NC m 10 3134 31 We would sail the bucket lids through the air. NC fem 10 2258 32 I never recognized any difference. NC m 10 1850 33 I finished in 1979. OH m 10 1151 34 Our trip back was kind of uneventful OH fem 8 1917 35 My family still likes coming back. NC fem 10 2416 36 She graduated from high school this year. OH m 7 1163 37 Yeah it was a lot of fun. OH fem 6 1396 38 I'm making wall hangings. OH m 7 1425 39 I work for the school system. NC m 10 2072 40 It was in the rose garden over there.

# of Dur State Sex Syll (ms) Ord Sentence OH m 9 1653 1 And he wants to be a pharmacist If you looked hard enough you could find the good in NC fem 14 2703 2 people NC m 7 1969 3 So what else can I tell you? OH fem 8 1838 4 She was kind of a reverse snob. NC fem 9 1875 5 I don’t think it’s putting on airs. NC m 8 2111 6 My dad was one of eight children NC m 7 1547 7 People still do that today. NC fem 7 1424 8 She can play the piano. NC fem 5 1393 9 And I heard her scream. NC fem 8 1455 10 What’re you doing with shoes on? OH m 8 1877 11 We just came back from Savannah. NC m 9 1645 12 This has been an exciting school year. NC m 9 1452 13 We had to go to the library. NC m 9 2037 14 Last year Kathy had back surgery. NC m 9 1388 15 We just had a lot of fun with that. NC fem 12 2326 16 I remember both my grandmothers very well.

131

OH fem 6 1699 17 Her name is Junie Mae. NC fem 8 1244 18 I got in the shower later. OH fem 7 1975 19 I worked at Westland High School. OH fem 7 1867 20 So I almost came back home. NC m 8 1955 21 There’s a cemetery up there. NC m 12 2063 22 I was standing in the cafeteria line. OH fem 6 1577 23 Tom Tyrone starred in it. NC m 9 1833 24 I don’t know if you need to go back. OH m 8 1828 25 I now coach baseball in college OH fem 12 3314 26 Make sure they don’t eat any foreign objects. OH m 8 2153 27 My dad worked with electronics. OH fem 10 1729 28 I could see changes in the area. OH m 9 1496 29 She’s a very talented young girl. NC fem 9 1935 30 I was the middle of three children. OH m 11 2570 31 We had a great time going up in Jim’s van. OH fem 9 2590 32 I have great joy in witnessing that. NC fem 7 1859 33 But I don’t like boiled okra. OH fem 7 1156 34 I had a double major. NC fem 12 1736 35 And he was the superintendent of schools there. OH m 10 1296 36 Try to get on the other side of it. OH m 10 1732 37 I just sort of listen to speech patterns OH m 8 1604 38 Well we met right before Easter. OH m 8 2001 39 So I had a lot of earaches. OH fem 8 1293 40 I have a little dog, Scooter

State Sex # of Syll Dur (ms) Ord Sentence NC fem 6 1365 1 I think that's wonderful. OH fem 13 2237 2 Now I try not to use any middle initial. OH fem 6 1336 3 I had hoped he would be. OH m 5 781 4 It was a good time. NC m 6 1666 5 My son is twenty five.

132

There was about approximately twenty three acres NC m 14 2037 6 here. OH fem 6 903 7 Her body was worn out. OH fem 11 2594 8 And it was always a very friendly place. NC fem 12 2008 9 We're now thoroughly enjoying our grandchildren. NC fem 5 1595 10 I love the greenery. OH fem 10 2278 11 We hope to get another dog someday. NC fem 10 2071 12 And I started drying off with a towel. NC fem 10 2286 13 He was a very interesting person. NC m 6 1570 14 He's my brother Joe's age. OH m 5 874 15 We're glad to be home. NC m 6 1519 16 Y'all sure do sound funny. NC fem 11 2073 17 And my husband only works one day a week. OH m 7 2147 18 So it was a little strange. OH fem 7 1587 19 Our life has changed quite a bit. OH m 10 3375 20 It was linguistics 201 I think. NC fem 7 2321 21 Mom, I'm standing on a snake NC m 7 1166 22 I heard I had a good time. And I don't want to miss his senior year of high OH m 13 2146 23 school. NC fem 12 2477 24 They all had the different accent from what we did. NC m 12 2175 25 My grandmother never drove an automobile. OH fem 5 1719 26 Now I don't like bees. NC m 7 1173 27 I think that's kind of unique. OH fem 11 2895 28 The older I got the more involved I got. NC m 8 1399 29 Everybody didn't do that. NC m 10 2189 30 They don't have to go through these corrections NC m 10 2101 31 I know she's worked with several of those kids. OH fem 5 892 32 That's all they could teach. OH m 5 855 33 We have two daughters. NC fem 11 1959 34 That was the most horrible experience. NC fem 8 2597 35 She could look after herself without any trouble.

133

OH fem 7 1390 36 But you do have to have patience. I was able to ride up there with five very good OH m 14 3367 37 friends. OH m 11 1921 38 There's a lots of shopping opportunities. OH m 7 2283 39 He would sell things to groceries. OH m 12 2966 40 My daughter is majoring in psychology.

State Sex # of Syll Dur (ms) Ord Sentence OH m 10 1706 1 We lived on one corner of the acreage. NC m 6 2213 2 They grew a large garden. It was called Christian Outreach School of OH m 12 2266 3 Ministries. OH fem 8 1930 4 We've been married forty two years. NC fem 12 2809 5 I have never tried to change my vernacular. NC m 5 1148 6 I met my wife there. NC fem 7 1898 7 I really enjoy my job. OH fem 6 1301 8 Well I have thirty years. NC m 7 1375 9 It's kind of special to me. NC m 13 1720 10 I've been an equal opportunity employee. NC fem 13 2625 11 They're pitiful looking in their little uniforms. NC fem 8 2084 12 I like to cut the scraps for her. OH fem 6 1847 13 We moved into the house. NC fem 12 2159 14 I finally decided to close the pewter shop. We had to go around the room and introduce NC m 14 2784 15 ourselves. NC m 11 2177 16 Some of the places are pretty nice places. NC m 6 1145 17 I'm not sure how she is. NC fem 7 1466 18 He got acquainted with us. NC m 6 2577 19 Daddy's people came here. OH fem 9 2296 20 We have been married thirty plus years. OH fem 11 2542 21 I had to do six years to get both of them.

134

NC fem 8 1605 22 What are you trying to tell me? NC fem 7 2926 23 My mother was born in France. They don't have that country sound, that Appalachian NC m 13 4842 24 sound OH m 8 1322 25 It's far more than pop and soda. OH fem 10 1766 26 My father had died at seventy-five. NC fem 11 3134 27 She's a very strong-willed and strong-headed child. OH m 7 1701 28 The sun came out which was great. OH fem 8 1335 29 That's always a priority. OH m 9 1062 30 She's applied to do that for two years. OH m 10 1719 31 And it sounded like interesting stuff. OH fem 6 1599 32 I'm lucky I've got three. OH fem 6 1145 33 My daughter had problems. OH m 7 1860 34 He can call triple A dad. OH fem 10 1840 35 Poor little Mickey was the apprentice. OH m 7 1502 36 And so he had to choose one. OH m 11 2019 37 On Friday I had a great experience. OH m 7 1540 38 She works for the school system. NC fem 6 1515 39 Well I like fried okra. NC m 6 2116 40 I ride horses and hunt.

State Sex # of Syll Dur (ms) Ord Sentence NC m 10 1777 1 She had to have me in a hospital. OH fem 5 1579 2 They're a lot of fun. NC fem 7 1567 3 I was fifteen when she died. NC fem 9 2107 4 He had a gold Elgin pocket watch. OH fem 8 2124 5 My mother's house was sold last year. OH fem 7 1613 6 I also make baby quilts. NC fem 8 1962 7 He taught and coached there for five years. OH m 10 1916 8 It's becoming a much larger city. NC m 7 1248 9 You just can't go anymore.

135

OH m 8 1283 10 She has some friends that have done it. OH m 9 2135 11 My son doesn't have a major yet. OH fem 10 2017 12 You know at three I was tapping along. OH fem 8 2041 13 We have a daughter and two sons. OH fem 8 1521 14 We don't go back very often. OH m 13 2830 15 We were able to find a fairly close parking space. NC fem 8 1599 16 We do horse and buggy weddings. OH m 8 1522 17 My dad was a high school teacher. NC m 7 1286 18 Her eyes are getting bigger. OH m 10 1584 19 It was basically very blue collar. NC fem 8 1664 20 We used to have a lot of snakes. OH fem 6 1057 21 They just get politics. NC m 11 1609 22 Brady's gonna have surgery on his shoulder. NC fem 11 3332 23 And that's one of my proudest accomplishments. NC m 7 1841 24 Well I'll think about it Tom. OH fem 12 4033 25 It gave me something to focus on besides grief. NC m 8 1961 26 That's a whole lot different today. NC m 9 2506 27 Both the kids have been involved in sports. NC m 9 2475 28 My parents were uneducated. NC fem 10 2064 29 But I didn't like the sewing part much. OH m 8 1489 30 He wants to be independent. OH fem 6 1679 31 Sometimes I don't like birds. OH m 8 1536 32 So that's why there's such a difference. OH fem 10 2519 33 My daughter moved into the other half. NC m 12 1666 34 I wanted to be a radio announcer. OH m 8 2743 35 They also had two small children. NC fem 11 5026 36 We never got beyond three generations NC fem 6 1292 37 My husband has a boat

The words are there in your head, if you would just write them NC fem 14 3025 38 down. NC m 7 2107 39 I grew up on Pressey Creek.

136

OH m 7 1318 40 We'd gone to our favorite spot.

State Sex # of Syll Dur (ms) Ord Sentence OH fem 7 1805 1 My daughter now owns the home. NC fem 8 1324 2 And they don't have any children. OH m 10 1705 3 I do a lot of political work. NC fem 11 2892 4 And I went back on all the old microfilm. NC fem 10 2183 5 My mom was the only one that moved South. OH fem 9 1740 6 You're looking for the tomato soup. OH m 11 1747 7 The temperature wasn't gonna change at all. NC fem 8 1623 8 But to him it was not fiction. OH m 7 1759 9 The drive home was very long. NC m 7 1121 10 I had a boy named Tommy. OH m 6 1554 11 Who has the best sports team? NC m 8 2124 12 I can't think of anything else. OH fem 6 1570 13 We used to have two dogs. OH m 8 1113 14 They just don't want to hire me. OH fem 8 1645 15 They need to be spayed or neutered. NC fem 9 1624 16 Still couldn't put my finger on it. OH fem 7 1515 17 My bank account was like that. NC m 6 1580 18 My wife's from Michigan. NC m 9 1901 19 We've been divorced for lots of years now NC m 8 1411 20 He was eighty seven years old. NC fem 9 3227 21 Homes, livestock, people were washed away. NC m 10 2380 22 I was just blown away by rock and roll. NC m 11 2154 23 I have a twenty year old daughter named Paige. NC m 9 2582 24 It showed how much energy you had OH fem 9 1480 25 She didn't like the hypocrisy. NC fem 6 2272 26 Oh, good, mine's tomatoes. NC m 7 1582 27 He got hurt playing football. OH m 7 1488 28 That's a lot of fun as well.

137

OH m 11 3083 29 My interest is regional variations OH fem 12 2527 30 She died about six years ago at ninety-six. OH m 10 2047 31 The middle son didn't like sports at all. OH m 8 1795 32 They will contact her by email. NC fem 11 2383 33 Rachel I'm sorry but it's a copperhead. OH fem 7 1172 34 I was born in fifty-four. OH m 10 1389 35 We lived in a pretty good neighborhood. NC m 11 1983 36 The wedding proceeded according to plan. OH fem 7 1214 37 Take care of them when needed. NC fem 9 2351 38 There's a lot of things you can't replace. NC fem 11 3441 39 I'm married to a man from this area. OH fem 8 1696 40 There's been people here for ten years.

State Sex # of Syll Dur (ms) Ord Sentence OH m 8 1889 1 I'm the second of four children. NC fem 9 1926 2 Are you trying to tell me something? NC m 9 1955 3 I golf and fish when I have the time. OH m 7 1222 4 My father was a lawyer. OH m 10 1732 5 A lot of that is migration patterns. NC fem 7 1805 6 We call her the snake lady. OH fem 9 2760 7 Walt actually wrote under that name. OH fem 7 1589 8 I call her hotdog hound dog. NC m 7 3291 9 Now, I don't wanna be cruel The mountains are not the best place to make a NC fem 13 2098 10 living. OH m 12 1747 11 The weather was not gonna get any better. NC m 9 2279 12 Donna had the most beautiful dress NC m 10 1720 13 It was a different time to say the least. NC m 7 1481 14 I can't think of anything. NC fem 8 1687 15 But that didn't happen for us. OH m 10 1652 16 Because everybody around us stood.

138

NC fem 7 2124 17 They neither one ever drove. OH m 7 1610 18 I taught for twenty-three years. My mother taught my daughter and my NC fem 12 3672 19 granddaughter. NC m 9 1653 20 I've been a musician all my life. NC m 8 1938 21 My phone was that way for a while. NC m 9 2474 22 I enjoyed that tenure very much. OH m 8 1736 23 I took care of all the sports fields. NC fem 8 1490 24 So she was very free-hearted NC fem 14 2423 25 It's not really a vegetable, as I read, it's a fruit. OH m 10 2120 26 My daughter played volleyball and softball. OH fem 7 1578 27 That's about all my pet peeves. OH fem 6 1233 28 Find me the skinny one. OH m 7 1291 29 And we can take care of it. OH fem 11 2302 30 I have a master's degree in special ed. OH m 10 2570 31 Well I had one brother and three sisters. NC fem 7 1774 32 I had a hard time in school. OH fem 8 1501 33 We live in one of the suburbs. OH fem 7 2399 34 Now they have nine grandchildren OH fem 9 2669 35 We lived in that home for fifty years. NC fem 6 1802 36 But you can't buy pictures. OH fem 8 1814 37 You know that's somewhere down the line. OH fem 8 2060 38 And we've lived there forty four years. NC m 7 1973 39 We're in the wrecker business. NC m 10 1908 40 His name was Ed, I'll never forget him.

State Sex # of Syll Dur (ms) Ord Sentence OH m 10 1732 1 So I know what really good sweet corn is. While I was in the shower stall, more words would NC fem 13 2581 2 come in. NC fem 10 1871 3 They didn't have any way to make ice.

139

NC m 10 2632 4 I have no ideal what to talk about OH fem 5 1276 5 I have three children. NC fem 7 1310 6 We stood around the table. OH fem 6 1503 7 He had problems with peers. OH m 11 2097 8 Then I also do all the logo painting. NC m 12 2254 9 They become disoriented very quickly. OH m 6 1878 10 South is down on Ann Street. NC m 8 1983 11 I know all those Foxes up there. NC m 7 1624 12 Oh yeah, he got it down pat. OH fem 5 1483 13 You start from square one. NC fem 8 3047 14 I can cook but I don't like to. OH m 12 2498 15 I know that Route 40 is a linguistic line. OH fem 5 1742 16 She lives in one half. OH m 5 1196 17 We had some chickens. OH fem 11 2094 18 I have a new granddaughter who's five months old. NC fem 8 1946 19 She is one lucky little dog. NC fem 6 1411 20 I have her school letter. NC m 5 977 21 She was born down there. OH m 6 1514 22 Traffic was pretty good. NC m 10 2044 23 That's the immediate plans anyway. NC fem 10 2330 24 I'll tell you one good story on myself. OH m 12 1973 25 The two older ones aren't doing anything now. OH fem 5 1624 26 My patience was less. NC m 6 3447 27 We played with the bucket lids. OH fem 14 3179 28 I have credit cards with "C" as my middle initial. NC fem 9 2328 29 She never talked like my other aunts. OH m 11 1860 30 We're getting close to the end now I take it. NC m 8 1248 31 The local people didn't care. NC m 10 1187 32 Now I'm not yelling at anybody. NC fem 11 5061 33 Tomatoes , green beans, onions, pepper, cabbage. OH fem 12 2059 34 Now the two older ones can go into a bar.

140

OH m 8 1686 35 The sun was shining where he was. NC fem 10 2219 36 I have three children- two boys and one girl. OH fem 6 1092 37 You're looking up and down. OH fem 12 2260 38 Our daughter is finishing up medical school. NC m 10 1435 39 I remember being at a concert. OH m 8 1637 40 We'll see what happens on that front.

141

Appendix C: Experiment 3 - Sentence Sets and Target Words

Bolded words = target words Set 1

# of Duration State Sex Order Targets Sentence Syllables (ms)

OH female 5 1203 1 2 I have two children.

And I found many of their death NC female 12 2660 2 2 certificates.

My paternal grandfather was a Baptist NC male 13 4767 3 2 preacher.

And I should tell you about my new OH female 10 1821 4 2 kitten.

NC male 12 2153 5 2 So that was an interesting experience.

NC female 10 1881 6 2 And I could remember her very well.

NC male 6 1392 7 2 They like to hear me talk.

NC male 6 1576 8 2 He calls me everyday.

NC male 6 1426 9 2 That's all I've got to say.

My granddaughter was born 2 months NC female 14 3254 10 3 after my mother died

I enjoy doing it and they enjoy getting OH female 13 2358 11 2 them.

NC female 6 1839 12 2 I guess her heart was big.

OH male 12 2306 13 2 I've been practicing law since 1982.

NC male 5 2071 14 2 We had a string band.

NC female 7 2129 15 2 And I have three grandchildren.

OH female 5 1288 16 2 It's my married name.

OH female 11 2148 17 2 We moved back here in 1987.

NC female 9 2823 18 2 My mother is eighty three years old.

142

I finished my master's here in seventy OH male 12 2070 19 2 one.

OH female 5 1347 20 2 I learned to tap dance.

OH female 12 1903 21 2 He had to go to the veteran's hospital.

That was the best part of the whole OH male 9 1548 22 2 thing.

You could make a radio or a burglar OH male 13 2240 23 3 alarm.

OH female 5 1590 24 2 He is very tall.

NC female 5 1503 25 1 He's a fisherman.

The superintendent sent me to NC male 12 2906 26 2 Providence

OH female 6 2009 27 2 There's the barking beagle.

OH female 8 1370 28 2 It only lasted for three years.

NC female 8 1502 29 2 I kept scrubbing the shower stall.

And it's the most wonderful thing- being NC female 14 2592 30 2 a grandparent.

NC male 6 1069 31 2 You have to deal with her.

OH male 6 1083 32 2 They would live to the south

OH male 5 1545 33 2 The seats were great seats.

NC male 5 1210 34 2 Let me hear you talk.

OH male 10 1796 35 2 There's currently a job opening there.

OH male 10 2204 36 3 My dad was sort of a hobby farmer.

OH male 5 1327 37 1 The high school's the same.

OH male 7 1811 38 1 My oldest is twenty-three.

NC female 6 1569 39 2 I'm not ashamed of it.

They decided to open a laundry and dry NC male 14 2095 40 2 cleaners.

Set 2 state sex #syll dur order targets Sentence

Traffic was obviously getting worse as OH male 14 2437 1 2 we came home.

NC female 9 1703 2 2 It was a bustling community.

She was not dependent on a lot of other NC female 14 2488 3 2 people.

OH female 7 1269 4 2 Yeah so they come over here.

143

NC female 7 2169 5 2 So I ran down to the shed.

NC male 10 2131 6 2 You've got to go see this woman.

NC male 6 1032 7 2 That was a bad mistake.

OH female 11 1938 8 3 The love you receive is unconditional.

NC female 5 964 9 2 I love the mountains.

And it has been in my family every OH female 11 2744 10 2 since.

He would walk to the church and then NC male 13 3887 11 2 back home on Sundays.

NC female 7 1646 12 2 They used to play full court then.

And I've made a book for each of our NC female 12 3539 13 3 three children.

OH female 9 2124 14 2 I went to their college of nursing.

NC male 7 1346 15 2 There was a little fellow.

OH female 5 879 16 1 But it was worth it.

OH female 5 1380 17 2 Mom, you've burned me out.

NC male 9 1824 18 2 Kathy came out by the wall one day.

OH male 7 1282 19 2 We got married in August.

NC male 6 1689 20 2 And he's lived here for years.

NC male 7 1508 21 2 I hadn't checked up on her.

OH female 8 1053 22 2 I became a librarian.

In their married years they were around NC female 11 2315 23 2 Cleveland

Well it seemed like it was really high at OH male 12 2709 24 2 the time.

OH female 12 1718 25 2 He has an international studies degree.

OH male 7 1420 26 2 She needed a break from that.

NC female 7 1766 27 1 We've sort of retired here.

And I arrived in my driveway at 2:30 OH male 14 3129 28 2 AM.

More words were coming but they were NC female 12 2908 29 2 a different verse.

OH female 6 1355 30 2 That's always a good thing.

NC male 10 1401 31 2 That's how politicians get elected.

NC male 9 1650 32 2 And I ended up staying twelve years.

144

OH female 6 1627 33 2 I like to sew and quilt.

OH male 8 1365 34 2 He would grow things in the summer.

OH male 5 1311 35 2 So he moved back here.

NC male 10 2476 36 2 She was fifteen days over a hundred.

NC female 8 1987 37 2 She's pretty special in our lives.

OH male 10 2242 38 2 It's a very competitive program.

OH male 7 1510 39 2 A survey of linguistics.

OH male 6 1065 40 2 That was a lot of fun.

Set 3 state sex #syll dur order targets Sentence

OH male 8 1372 1 2 We've been married thirty-nine years.

NC male 9 1997 2 2 It takes something to get me started.

NC male 8 1999 3 2 She's been real good for stuff like that.

OH male 8 1860 4 1 She had no worries about that.

NC female 7 1371 5 2 It was a beautiful day.

OH male 10 1855 6 2 He's trying to get into nursing school.

OH female 10 2049 7 2 So I had two teaching certificates.

NC female 8 3000 8 2 You had to find a white ash tree.

NC female 7 1524 9 2 My life is pretty simple.

OH male 8 1286 10 2 My mother was a teacher too.

OH female 5 983 11 2 Try to adopt one.

NC female 8 1510 12 2 Now I live on Riverwood Hill.

NC male 8 1435 13 2 My wife had just left on a trip.

OH female 8 1742 14 2 He has dreams of saving the world.

NC male 8 1267 15 2 Somebody has to write those games.

NC male 7 1837 16 2 I was born during the war.

Where in the world have I heard this NC female 9 2600 17 2 song?

NC female 11 3205 18 2 We actually didn't eat on the table

OH female 8 1766 19 2 My husband is an attorney.

NC female 8 1850 20 2 She takes it right after him.

I do like bumblebees 'cause they help OH female 12 2326 21 2 my garden.

145

NC male 8 2055 22 2 They all were very successful.

OH female 8 1339 23 2 But she lives in Albuquerque.

OH female 9 1742 24 2 My passion is Disney trivia.

OH female 6 859 25 2 I could teach anything.

OH male 10 2126 26 2 But I'm a huge college basketball fan.

OH male 11 2096 27 2 I've been married for almost 25 years.

NC male 8 3888 28 2 We had him saying "Fried Chicken".

NC female 7 1855 29 2 For about fiftyfive years

OH male 9 1822 30 2 It's an interest of mine, a hobby.

We would sail the bucket lids through NC male 10 3134 31 3 the air.

NC female 10 2258 32 2 I never recognized any difference.

NC male 10 1850 33 1 I finished in 1979.

OH male 10 1151 34 2 Our trip back was kind of uneventful

OH female 8 1917 35 2 My family still likes coming back.

She graduated from high school this NC female 10 2416 36 2 year.

OH male 7 1163 37 2 Yeah it was a lot of fun.

OH female 6 1396 38 2 I'm making wall hangings.

OH male 7 1425 39 3 I work for the school system.

NC male 10 2072 40 2 It was in the rose garden over there.

Set 4 state sex #syll dur order Targets Sentences

OH male 9 1653 1 2 And he wants to be a pharmacist

If you looked hard enough you could find NC female 14 2703 2 2 the good in people

NC male 7 1969 3 2 So what else can I tell you?

OH female 8 1838 4 2 She was kind of a reverse snob.

NC female 9 1875 5 2 I don’t think it’s putting on airs.

NC male 8 2111 6 3 My dad was one of eight children

NC male 7 1547 7 2 People still do that today.

NC female 7 1424 8 2 She can play the piano.

NC female 5 1393 9 2 And I heard her scream.

146

NC female 8 1455 10 2 What’re you doing with shoes on?

OH male 8 1877 11 2 We just came back from Savannah.

NC male 9 1645 12 2 This has been an exciting school year.

NC male 9 1452 13 1 We had to go to the library.

NC male 9 2037 14 3 Last year Kathy had back surgery.

NC male 9 1388 15 2 We just had a lot of fun with that.

I remember both my grandmothers very NC female 12 2326 16 2 well.

OH female 6 1699 17 2 Her name is Junie Mae.

NC female 8 1244 18 2 I got in the shower later.

OH female 7 1975 19 2 I worked at Westland High School.

OH female 7 1867 20 2 So I almost came back home.

NC male 8 1955 21 1 There’s a cemetery up there.

NC male 12 2063 22 2 I was standing in the cafeteria line.

OH female 6 1577 23 2 Tom Tyrone starred in it.

NC male 9 1833 24 2 I don’t know if you need to go back.

OH male 8 1828 25 2 I now coach baseball in college

Make sure they don’t eat any foreign OH female 12 3314 26 2 objects.

OH male 8 2153 27 2 My dad worked with electronics.

OH female 10 1729 28 2 I could see changes in the area.

OH male 9 1496 29 2 She’s a very talented young girl.

NC female 9 1935 30 2 I was the middle of three children.

We had a great time going up in Jim’s OH male 11 2570 31 2 van.

OH female 9 2590 32 2 I have great joy in witnessing that.

NC female 7 1859 33 2 But I don’t like boiled okra.

OH female 7 1156 34 2 I had a double major.

And he was the superintendent of NC female 12 1736 35 2 schools there.

OH male 10 1296 36 2 Try to get on the other side of it.

OH male 10 1732 37 2 I just sort of listen to speech patterns

OH male 8 1604 38 2 Well we met right before Easter.

OH male 8 2001 39 2 So I had a lot of ear aches.

147

OH female 8 1293 40 2 I have a little dog, Scooter

Set 5 state sex #syll dur order Targets Sentence

NC female 6 1365 1 1 I think that's wonderful.

OH female 13 2237 2 2 Now I try not to use any middle initial.

OH female 6 1336 3 1 I had hoped he would be.

OH male 5 781 4 2 It was a good time.

NC male 6 1666 5 2 My son is twenty five.

There was about approximately twenty NC male 14 2037 6 2 three acres here.

OH female 6 903 7 2 Her body was worn out.

OH female 11 2594 8 2 And it was always a very friendly place.

We're now thoroughly enjoying our NC female 12 2008 9 2 grandchildren.

NC female 5 1595 10 2 I love the greenery.

OH female 10 2278 11 2 We hope to get another dog someday.

NC female 10 2071 12 2 And I started drying off with a towel.

NC female 10 2286 13 2 He was a very interesting person.

NC male 6 1570 14 2 He's my brother Joe's age.

OH male 5 874 15 2 We're glad to be home.

NC male 6 1519 16 2 Y'all sure do sound funny.

And my husband only works one day a NC female 11 2073 17 2 week.

OH male 7 2147 18 1 So it was a little strange.

OH female 7 1587 19 3 Our life has changed quite a bit.

OH male 10 3375 20 1 It was linguistics 201 I think.

NC female 7 2321 21 3 Mom, I'm standing on a snake

NC male 7 1166 22 2 I heard I had a good time.

And I don't want to miss his senior year OH male 13 2146 23 3 of high school.

They all had the different accent from NC female 12 2477 24 2 what we did.

My grandmother never drove an NC male 12 2175 25 2 automobile.

148

OH female 5 1719 26 2 Now I don't like bees.

NC male 7 1173 27 2 I think that's kind of unique.

OH female 11 2895 28 2 The older I got the more involved I got.

NC male 8 1399 29 2 Everybody didn't do that.

They don't have to go through these NC male 10 2189 30 2 corrections

I know she's worked with several of NC male 10 2101 31 2 those kids.

OH female 5 892 32 2 That's all they could teach.

OH male 5 855 33 2 We have two daughters.

NC female 11 1959 34 2 That was the most horrible experience.

She could look after herself without any NC female 8 2597 35 2 trouble.

OH female 7 1390 36 2 But you do have to have patience.

I was able to ride up there with five very OH male 14 3367 37 3 good friends

There's a lots of shopping OH male 11 1921 38 2 opportunities.

OH male 7 2283 39 2 He would sell things to groceries.

OH male 12 2966 40 2 My daughter is majoring in psychology.

Set 6 state sex #syll dur order Targets Sentence

OH male 10 1706 1 2 We lived on one corner of the acreage.

NC male 6 2213 2 2 They grew a large garden.

It was called Christian Outreach School OH male 12 2266 3 2 of Ministries.

OH female 8 1930 4 2 We've been married forty two years.

I have never tried to change my NC female 12 2809 5 3 vernacular.

NC male 5 1148 6 2 I met my wife there.

NC female 7 1898 7 2 I really enjoy my job.

OH female 6 1301 8 2 Well I have thirty years.

NC male 7 1375 9 2 It's kind of special to me.

149

I've been an equal opportunity NC male 13 1720 10 2 employee.

They're pitiful looking in their little NC female 13 2625 11 2 uniforms.

NC female 8 2084 12 2 I like to cut the scraps for her.

OH female 6 1847 13 2 We moved into the house.

I finally decided to close the pewter NC female 12 2159 14 2 shop.

We had to go around the room and NC male 14 2784 15 2 introduce ourselves.

Some of the places are pretty nice NC male 11 2177 16 2 places.

NC male 6 1145 17 2 I'm not sure how she is.

NC female 7 1466 18 1 He got acquainted with us.

NC male 6 2577 19 2 Daddy's people came here.

OH female 9 2296 20 2 We have been married thirty plus years.

OH female 11 2542 21 2 I had to do six years to get both of them.

NC female 8 1605 22 2 What are you trying to tell me?

NC female 7 2926 23 2 My mother was born in France.

They don't have that country sound, that NC male 13 4842 24 2 appalachian sound

OH male 8 1322 25 2 It's far more than pop and soda.

OH female 10 1766 26 2 My father had died at seventy five.

She's a very strong-willed and strong- NC female 11 3134 27 2 headed child.

OH male 7 1701 28 2 The sun came out which was great.

OH female 8 1335 29 1 That's always a priority.

OH male 9 1062 30 2 She's applied to do that for two years.

OH male 10 1719 31 2 And it sounded like interesting stuff.

OH female 6 1599 32 2 I'm lucky I've got three.

OH female 6 1145 33 2 My daughter had problems.

OH male 7 1860 34 2 He can call triple A dad.

OH female 10 1840 35 3 Poor little Mickey was the apprentice.

OH male 7 1502 36 2 And so he had to choose one.

OH male 11 2019 37 2 On Friday I had a great experience.

150

OH male 7 1540 38 2 She works for the school system.

NC female 6 1515 39 2 Well I like fried okra.

NC male 6 2116 40 2 I ride horses and hunt.

Set 7 state sex #syll dur order Targets Sentence

NC male 10 1777 1 1 She had to have me in a hospital.

OH female 5 1579 2 2 They're a lot of fun.

NC female 7 1567 3 2 I was fifteen when she died.

NC female 9 2107 4 2 He had a gold Elgin pocket watch.

OH female 8 2124 5 3 My mother's house was sold last year.

OH female 7 1613 6 2 I also make baby quilts.

He taught and coached there for five NC female 8 1962 7 2 years.

OH male 10 1916 8 2 It's becoming a much larger city.

NC male 7 1248 9 2 You just can't go anymore.

OH male 8 1283 10 2 She has some friends that have done it.

OH male 9 2135 11 2 My son doesn't have a major yet.

OH female 10 2017 12 2 You know at three I was tapping along.

OH female 8 2041 13 2 We have a daughter and two sons.

OH female 8 1521 14 2 We don't go back very often.

We were able to find a fairly close OH male 13 2830 15 2 parking space.

NC female 8 1599 16 2 We do horse and buggie weddings.

OH male 8 1522 17 3 My dad was a high school teacher.

NC male 7 1286 18 2 Her eyes are getting bigger.

OH male 10 1584 19 2 It was basically very blue collar.

NC female 8 1664 20 2 We used to have a lot of snakes.

OH female 6 1057 21 1 They just get politics.

Brady's gonna have surgery on his NC male 11 1609 22 2 shoulder.

And that's one of my proudest NC female 11 3332 23 2 accomplishments.

NC male 7 1841 24 2 Well I'll think about it Tom.

151

It gave me something to focus on OH female 12 4033 25 2 besides grief.

NC male 8 1961 26 2 That's a whole lot different today.

Both the kids have been involved in NC male 9 2506 27 3 sports.

NC male 9 2475 28 2 My parents were uneducated.

NC female 10 2064 29 2 But I didn't like the sewing part much.

OH male 8 1489 30 1 He wants to be independent.

OH female 6 1679 31 2 Sometimes I don't like birds.

OH male 8 1536 32 2 So that's why there's such a difference.

OH female 10 2519 33 2 My daughter moved into the other half.

NC male 12 1666 34 2 I wanted to be a radio announcer.

OH male 8 2743 35 2 They also had two small children.

NC female 11 5026 36 2 We never got boyond three generations

NC female 6 1292 37 2 My husband has a boat

The words are there in your head, if NC female 14 3025 38 2 you'd write them down.

NC male 7 2107 39 2 I grew up on Pressey Creek.

OH male 7 1318 40 2 We'd gone to our favorite spot.

Set 8 state sex #syll dur order Targets Sentence

OH female 7 1805 1 2 My daughter now owns the home.

NC female 8 1324 2 1 And they don't have any children.

OH male 10 1705 3 2 I do a lot of political work.

NC female 11 2892 4 2 And I went back on all the old microfilm.

My mom was the only one that moved NC female 10 2183 5 3 South.

OH female 9 1740 6 2 You're looking for the tomato soup.

The temperature wasn't gonna change OH male 11 1747 7 2 at all.

NC female 8 1623 8 1 But to him it was not fiction.

OH male 7 1759 9 2 The drive home was very long.

NC male 7 1121 10 2 I had a boy named Tommy.

152

OH male 6 1554 11 2 Who has the best sports team?

NC male 8 2124 12 2 I can't think of anything else.

OH female 6 1570 13 2 We used to have two dogs.

OH male 8 1113 14 2 They just don't want to hire me.

OH female 8 1645 15 2 They need to be spayde or neutered.

NC female 9 1624 16 1 Still couldn't put my finger on it.

OH female 7 1515 17 2 My bank account was like that.

NC male 6 1580 18 2 My wife's from Michigan.

We've been divorced for lots of years NC male 9 1901 19 2 now

NC male 8 1411 20 2 He was eighty seven years old.

Homes, livestock, people were washed NC female 9 3227 21 3 away.

NC male 10 2380 22 2 I was just blown away by rock and roll.

I have a twenty year old daughter NC male 11 2154 23 2 named Paige.

NC male 9 2582 24 2 It showed how much energy you had

OH female 9 1480 25 2 She didn't like the hypocrisy.

NC female 6 2272 26 2 Oh, good, mine's tomatoes.

NC male 7 1582 27 2 He got hurt playing football.

OH male 7 1488 28 2 That's a lot of fun as well.

OH male 11 3083 29 2 My interest is regional variations

She died about six years ago at ninety OH female 12 2527 30 2 six.

OH male 10 2047 31 2 The middle son didn't like sports at all.

OH male 8 1795 32 2 They will contact her by email.

NC female 11 2383 33 2 Rachel I'm sorry but it's a copperhead.

OH female 7 1172 34 2 I was born in fifty four.

We lived in a pretty good OH male 10 1389 35 2 neighborhood.

The wedding proceeded according to NC male 11 1983 36 2 plan.

OH female 7 1214 37 2 Take care of them when needed.

NC female 9 2351 38 3 There's a lot of things you can't replace.

NC female 11 3441 39 2 I'm married to a man from this area.

153

OH female 8 1696 40 2 There's been people here for ten years.

Set 9 state sex #syll dur order Targets Sentence

OH male 8 1889 1 2 I'm the second of four children.

NC female 9 1926 2 2 Are you trying to tell me something?

NC male 9 1955 3 2 I golf and fish when I have the time.

OH male 7 1222 4 2 My father was a lawyer.

OH male 10 1732 5 2 A lot of that is migration patterns.

NC female 7 1805 6 2 We call her the snake lady.

OH female 9 2760 7 2 Walt actually wrote under that name.

OH female 7 1589 8 2 I call her hotdog hound dog.

NC male 7 3291 9 2 Now, I don't wanna be cruel

The mountains are not the best place to NC female 13 2098 10 2 make a living.

The weather was not gonna get any OH male 12 1747 11 2 better.

NC male 9 2279 12 2 Donna had the most beautiful dress

NC male 10 1720 13 2 It was a different time to say the least.

NC male 7 1481 14 2 I can't think of anything.

NC female 8 1687 15 2 But that didn't happen for us.

OH male 10 1652 16 2 Because everybody around us stood.

NC female 7 2124 17 2 They neither one ever drove.

OH male 7 1610 18 2 I taught for twenty-three years.

My mother taught my daughter and my NC female 12 3672 19 2 granddaughter.

NC male 9 1653 20 2 I've been a musician all my life.

NC male 8 1938 21 2 My phone was that way for a while.

NC male 9 2474 22 2 I enjoyed that tenure very much.

OH male 8 1736 23 2 I took care of all the sports fields.

NC female 8 1490 24 2 So she was very free-hearted

It's not really a vegetable, as I read, it's a NC female 14 2423 25 2 fruit.

My daughter played volleyball and OH male 10 2120 26 2 softball.

154

OH female 7 1578 27 2 That's about all my pet peeves.

OH female 6 1233 28 2 Find me the skinny one.

OH male 7 1291 29 2 And we can take care of it.

OH female 11 2302 30 2 I have a master's degree in special ed.

Well I had one brother and three OH male 10 2570 31 2 sisters.

NC female 7 1774 32 2 I had a hard time in school.

OH female 8 1501 33 2 We live in one of the suburbs.

OH female 7 2399 34 2 Now they have nine grandchildren

OH female 9 2669 35 2 We lived in that home for fifty years.

NC female 6 1802 36 2 But you can't buy pictures.

You know that's somewhere down the OH female 8 1814 37 2 line.

OH female 8 2060 38 2 And we've lived there forty four years.

NC male 7 1973 39 2 We're in the wrecker business.

NC male 10 1908 40 2 His name was Ed, I'll never forget him.

Set 10 state sex #syll dur order Targets Sentence

So I know what really good sweet corn OH male 10 1732 1 3 is.

While I was in the shower stall, more NC female 13 2581 2 2 words would come in.

NC female 10 1871 3 2 They didn't have any way to make ice.

NC male 10 2632 4 2 I have no ideal what to talk about

OH female 5 1276 5 2 I have three children.

NC female 7 1310 6 2 We stood around the table.

OH female 6 1503 7 2 He had problems with peers.

OH male 11 2097 8 2 Then I also do all the logo painting.

NC male 12 2254 9 2 They become disoriented very quickly.

OH male 6 1878 10 2 South is down on Ann Street.

NC male 8 1983 11 2 I know all those Foxes up there.

NC male 7 1624 12 2 Oh yeah, he got it down pat.

OH female 5 1483 13 2 You start from square one.

155

NC female 8 3047 14 2 I can cook but I don't like to.

OH male 12 2498 15 2 I know that Route 40 is a linguistic line.

OH female 5 1742 16 2 She lives in one half.

OH male 5 1196 17 1 We had some chickens.

I have a new granddaughter who's five OH female 11 2094 18 2 months old.

NC female 8 1946 19 2 She is one lucky little dog.

NC female 6 1411 20 2 I have her school letter.

NC male 5 977 21 2 She was born down there.

OH male 6 1514 22 2 Traffic was pretty good.

NC male 10 2044 23 2 That's the immediate plans anyway.

NC female 10 2330 24 2 I'll tell you one good story on myself.

The two older ones aren't doing OH male 12 1973 25 2 anything now.

OH female 5 1624 26 2 My patience was less.

NC male 6 3447 27 2 We played with the bucket lids.

I have credit cards with "C" as my OH female 14 3179 28 2 middle initial.

NC female 9 2328 29 2 She never talked like my other aunts.

We're getting close to the end now I take OH male 11 1860 30 2 it.

NC male 8 1248 31 2 The local people didn't care.

NC male 10 1187 32 2 Now I'm not yelling at anybody.

Tomatoes , green beens, onions, NC female 11 5061 33 2 pepper, cabbage.

Now the two older ones can go into a OH female 12 2059 34 2 bar.

OH male 8 1686 35 2 The sun was shining where he was.

I have three children- two boys and one NC female 10 2219 36 2 girl.

OH female 6 1092 37 2 You're looking up and down.

Our daughter is finishing up medical OH female 12 2260 38 2 school.

NC male 10 1435 39 2 I remember being at a concert.

OH male 8 1637 40 2 We'll see what happens on that front.

156