MEASURING PHONOLOGICAL PROCE SSING AND PHONOLOGICAL WORKING MEMORY IN ADULTS WITH DEVELOPMENTAL DYSLEXIA: A FUNCTIONAL MAGNETIC RE SONANCE IMAGING STUDY By TIMOTHY W. CONWAY A DISSERTATION PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY UNIVERSITY OF FLORIDA 2003
Copyright 2003 by Timothy Wayne Conway
This dissertation is dedicated to my parents, Walter and Julie Conway, who have always provided me with unconditional support and love , as well as modeled an insatiable desire to learn, and a compassion for others.
ACKNOWLEDGMENTS I would like to acknowledge my wife (Amy) and children (Sheridan and Chloe), immediate family, and friends for their tireless support. I would especially like to acknowledge Dr. Bruce Crosson for showing remarkable patience with me during this unexpectedly lengthy process, as well as Dr. Ken Heilman for the encouragement to pursue a doctoral degree by seeing â€œa racehorse pulling a cart.â€ I extend sincere gratitude to my committee members for their support and guidance. I extend special thanks to Kaundinya Gopinath, Kyung Peck, Harris Slepian, Leslie Gonzalez-Rothi, Debbie Moncrieff, Tom Chesnes, Ann Alexander, Leeza Maron, and Stacey Hoffman for assistance with many aspects of this scholarly endeavor. The faculty and staff of the University of Florida's Department of Clinical and Health Psychology provided the remarkable education and training that were necessary for conceptualizing and completing this research project. Outstanding technical assistance was provided by the staff of the CIRCA Electronic Thesis and Dissertation Laboratory. Financial assistance for this research was graciously provided by the Donald D. Hammill Foundation, American Psychological Foundation, Patricia C. Lindamood, the Fred J. Wellington Foundation, and Grant HD 30988 awarded to Joseph K. Torgesen by the National Institute of Child Health and Human Development. iv
TABLE OF CONTENTS Page ACKNOWLEDGMENTS ................................................................................................. iv LIST OF TABLES............................................................................................................ vii LIST OF FIGURES ........................................................................................................... ix LIST OF OBJECTS.............................................................................................................x ABSTRACT....................................................................................................................... xi CHAPTER 1 INTRODUCTION........................................................................................................1 2 PHONOLOGICAL PROCESSING..............................................................................7 3 MODELS OF READING...........................................................................................12 4 FUNCTIONAL NEUROIMAGING ..........................................................................26 Methodological Considerations of Functional Neuroimaging ...................................28 PET Neuroimaging of Pure Tone and Phonological Stimuli......................................29 Pure Tone Stimulus .............................................................................................29 Phonological Stimulus.........................................................................................32 FMRI of Pure Tone and Phonological Stimuli...........................................................34 Pure Tone Stimulus .............................................................................................35 Phonological Stimulus ........................................................................................36 5 RATIONALE AND HYPOTHESES .........................................................................38 Statement of the Problem............................................................................................38 Hypotheses for Experiment One-Counting and Segmenting......................................40 Hypothesis 1â€“Tone Counting..............................................................................40 Hypothesis 2â€“Phoneme Counting .......................................................................41 Hypothesis 3â€“Pseudoword Segmenting ..............................................................42 Hypotheses for Experiment Twoâ€“Working Memory Comparisons...........................43 Hypothesis 1â€“Tone Comparison .........................................................................43 Hypothesis 2â€“Pseudoword Comparison..............................................................44 v
6 METHODS.................................................................................................................45 Participants and Selection Criterion ...........................................................................45 Apparatus and Scanning Procedures ..........................................................................49 Stimulus Parameters ...................................................................................................53 Experiment One-Counting and Segmenting........................................................55 Tone counting...............................................................................................57 Phoneme counting........................................................................................57 Pseudoword segmenting...............................................................................58 Experiment Twoâ€“Working Memory Comparisons.............................................58 Tone comparison..........................................................................................60 Pseudoword comparison ..............................................................................60 Functional Neuroimage Analysis ...............................................................................61 7 RESULTS...................................................................................................................64 Behavioral Results......................................................................................................64 FMRI Results..............................................................................................................69 Experiment Oneâ€“Tone and Phoneme Counting, and Pseudoword Segmenting .69 Temporal region ...........................................................................................71 Frontal region ...............................................................................................73 Parietal region. .............................................................................................74 Experiment Two-Tone Comparison and Pseudoword Comparison....................74 Tone comparison..........................................................................................76 Pseudoword comparison versus baseline white Noise.................................81 8 DISCUSSION.............................................................................................................87 APPENDIX A. NEUROPSYCHOLOGICAL ASSESSMENT.........................................................106 B. EXPERIMENT ONEâ€“EXPLORATORY ANALYSIS............................................108 LIST OF REFERENCES.................................................................................................113 BIOGRAPHICAL SKETCH ...........................................................................................124 vi
LIST OF TABLES Table page 5-1 Experiment Oneâ€“Tasks and Hypothesized Regions of Activity per Group.............43 5-2 Experiment Twoâ€“Tasks and Hypothesized Regions of Activity per Group............44 6-1 Matching Variables for Control and Dyslexic Groups.............................................46 6-2 Demographic Data per Group..................................................................................46 6-3 Variables for Defining Group Membership.............................................................49 6-4 Counting and Segmenting Task Descriptions..........................................................56 6-5 Counting and Segmenting Cognitive Components per Tasks..................................57 6-6 Working Memory Comparison Description of Tasks..............................................59 6-7 Working Memory Comparison Cognitive Components per Task...........................59 7-1 State and Trait Anxiety Levels Prior to FMRI.........................................................64 7-2 Reading skills per Group..........................................................................................65 7-3 Part 1 Phonological Processing Abilities per Group................................................66 7-4 Part 2 Phonological Processing Abilities per Group................................................66 7-5 Auditory Working Memory and Attention Abilities per Group..............................66 7-6 Rapid Naming Abilities per Group..........................................................................67 7-7 Experiment One Response Accuracy per Group......................................................67 7-8 Experiment Two Response Accuracy per Group.....................................................68 7-9 Active Temporal Regions from Within-groups Analyses........................................72 7-10 Temporal Region Activity per Group for Tone Comparison versus Baseline White Noise..............................................................................................................77 vii
7-11 Temporal Region Activity for Dyslexic Group Greater than Control Group during Tone Comparison..........................................................................................77 7-12 Frontal Region Activity per Group for Tone Comparison versus baseline white noise.........................................................................................................................78 7-13 Parietal Region Activity per Group for Tone Comparison Versus Baseline White Noise.........................................................................................................................79 7-14 Temporal Region Activity per Group for Pseudoword Comparison Versus Baseline White Noise...............................................................................................81 7-15 Frontal Region Activity per Group for Pseudoword Comparison Versus Baseline White Noise...............................................................................................84 7-16 Parietal Region Activity per Group for Pseudoword Comparison Versus Baseline White Noise...............................................................................................85 viii
LIST OF FIGURES Figure page 5-1 Articulatory Awareness Model of Speech Perception (AAMSP)............................40 6-1 Experiment One Stimulus Blocks Alternated in Each Run......................................53 6-2 Experiment Two Stimulus Blocks Alternated in Each Run.....................................54 7-1 Left Temporal Region Activity for Dyslexic Group Greater than Control Group during Tone Counting..............................................................................................73 7-2 Left Parietal Region Activity for Control Group greater than Dyslexic Group during Pseudoword Segmenting...............................................................................74 7-3 Frontal Region (dorsal premotor) Activity for the Control Group During Tone Comparison versus Baseline White Noise...............................................................78 7-4 Frontal Region Activity for the Dyslexic Group During Tone Comparison versus Baseline White Noise...............................................................................................79 7-5 Significantly Different Parietal Region Activity (BA 40) for Dyslexic Group greater than Control Group During Tone Comparison............................................80 7-6 Movie of Whole Brain Activity per Group During Tone Comparison....................81 7-7 Significant Temporal Activity (red) for Dyslexic Group (yellow) greater than Control Group (blue) During Tone Comparison......................................................82 7-8 Temporal Region HRF for Dyslexic greater than Control During Pseudoword Comparison..............................................................................................................83 7-9 Frontal Region Activity for the Dyslexic Group (yellow) and Control Group (blue) During Pseudoword Comparison...................................................................84 7-10 Significant Parietal Region Activity (red) for Dyslexic Group (yellow) greater than Control Group (blue) During Pseudoword Comparison..................................86 7-11 Movie of Whole Brain Activity per Group During Pseudoword Comparison .......86 ix
LIST OF OBJECTS Object 1. WAV sound file for Experiment One Stimulus Blocks Alternated in Each Run 2. WAV sound file for Experiment Two Stimulus Blocks Alternated in Each Run 3. MPEG Movie of Whole Brain Activity per Group During Tone Comparison 4. MPEG Movie of Whole Brain Activity per Group During Pseudoword Comparison x
Abstract of Dissertation Presented to the Graduate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Doctor of Philosophy MEASURING PHONOLOGICAL AWARENESS AND PHONOLOGICAL WORKING MEMORY IN ADULTS WITH DEVELOPMENTAL DYSLEXIA: A FUNCTIONAL MAGNETIC RESONANCE IMAGING STUDY By Timothy W. Conway May 2003 Chairman: Bruce Crosson, Ph.D. Major Department: Clinical And Health Psychology Developmental phonological dyslexia (DPD) is characterized by an impaired ability to convert graphemes to phonemes, with common errors in addition, omission, substitution or repetition of phonemes. A cognitive deficit that can cause DPD is impaired perception and manipulation of speech sounds in words (phonological awareness). A measure of phonological awareness that is highly predictive of reading development involves analytic phonological processing, orally segmenting words into their phonemes. Comparator function requires analytic phonological processing and working memory, and is correlated with reading ability in adults with DPD. The Articulatory Awareness Model of Speech Perception (AAMSP) hypothesizes deficient cortical networks to explain impaired phonological processing and working memory in DPD. Hence, it was hypothesized that dyslexics should show decreased activity in Brocaâ€™s area, somatosensory cortex, supplementary motor area, and dorsal lateral prefrontal cortex during analytic phonological and phonological working memory tasks. xi
Two auditory FMRI experiments compared adults with DPD to age, gender, IQ, education, handedness, and SES matched controls. Experiment oneâ€™s three tasks (counting pure tones, counting phonemes, and segmenting pseudowords) indicated that the dyslexic group exhibited greater left superior temporal gyrus (BA 42) activity during pure tone counting. In contrast, the control group showed greater left parietal lobe (BA 40, 7, 39) activity during pseudoword segmenting. Experiment twoâ€™s tasks (identifying two series of tones that contain the same number of tones or identifying pairs of pseudowords with an identical number of phonemes), indicated greater left superior temporal gyrus (BA 42, 22) activity both during the dyslexicâ€™s pure tone comparison and during the phoneme comparison task, and greater supramarginal gyrus (BA 40) activity during their pseudoword segmenting. Both groups reported a primary strategy of auditory rehearsal and segmentation for both experimentsâ€™ tasks. Overall, the AAMSPâ€™s predictions were not supported. However, the dyslexicâ€™s over-activity in primary auditory cortex for acoustic and phonological stimuli may indicate a principal deficit or inefficiency in primary auditory cortex. Inefficient processing in primary auditory cortex, regardless of the stimuliâ€™s temporal characteristics, may contribute to impaired analytic phonological awareness and phonological working memory abilities in DPD. The limitations of this study and the implications of these findings for empirically based treatments are discussed. xii
CHAPTER 1 INTRODUCTION Developmental dyslexia (Morgan, 1897) is a disorder characterized by an impaired ability to read words. The hallmark of this impairment is difficulty converting letters (graphemes) to sounds (phonemes). This difficulty results in reading errors that include adding, omitting, substituting, reversing, or repeating sounds (phonemes) in words (e.g. reading steam as stream, gril as girl, or frustrated as flustrated), as well as whole or partial substitutions of words (e.g. reading construction as constitution, pacific as specific, or exploring as expressive). The prevalence of developmental dyslexia has been estimated to range from 3-6% (Hynd & Cohen, 1983), 3-10% (Pennington, 1990) or as high as 20% (Liberman, 1987) of the general population. The negative impacts of dyslexia can be occupational, social, interpersonal, financial, and psychological (Feldman et al., 1993; Maughan, 1995). Therefore, developmental dyslexia is a well-defined and prevalent disorder with significant negative effects. Investigations into the causes of developmental dyslexia identified a deficit in phonological awareness that directly hinders the development of accurate reading skills (Liberman, 1973; Wagner & Torgesen, 1987). Phonological awareness is the ability to determine the number, identity, and order of speech sounds (phonemes) in a word (Lindamood, Bell, & Lindamood, 1992). The primary effect of impaired phonological awareness is an inability to accurately sound out words (Liberman, Shankweiler, Liberman, 1989; Bradley, & Bryant, 1985; Bradley, & Thompson, 1984). Thus, one type of developmental dyslexia that has been identified from a consensus of years of research 1
2 is developmental phonological dyslexia. The phonological awareness deficits that are the hallmark of this disorder are usually evident in childhood and persist into adulthood, despite remedial reading instruction (Maughan, 1995; Bruck, 1993; Lefly, & Pennington, 1991). Therefore, individuals with developmental phonological dyslexia are forced to memorize as many words as possible through a â€œwhole wordâ€ approach to reading and are unable to reliably use phonological skills for reading, spelling, and speech. To illustrate how phonological skills are related to reading, two types of phonological skills are reviewed. Analytic and synthetic phonological skills are two different aspects of phonological processing with differential predictive abilities relative to reading achievement (Torgesen, Wagner, & Rashotte, 1994). Analytic phonological skills are measured by tasks that require one or more of the following: segmenting a word into its phonemes, identifying the number of phonemes in a word, or omitting a phoneme from a word. In contrast, synthetic phonological skills include blending individual phonemes or syllables together to form a word. Both of these phonological skills are beneficial for reading; however, analytic phonological skills have been shown to be more predictive of reading development than synthetic. Longitudinal studies of children with developmental dyslexia reported that analytic phonological skills (identifying sounds within words) are the best predictors of reading achievement (Torgesen, Wagner, & Rashotte, 1994). Also, cross-sectional studies of adults with and without developmental dyslexia reported that phoneme segmentation (an analytic phonological skill) is a reliable discriminator of the two groups (Bruck, 1993). Similarly, a longitudinal study of adults with developmental dyslexia indicated that a test of phonological awareness (Lindamood & Lindamood, 1979) that includes phonological
3 working memory and analytic phonological skills (phoneme manipulation and phoneme segmentation) is a reliable discriminator of adults with and without dyslexia (Felton, Naylor, & Wood, 1990). This same test has been found to be highly predictive of reading achievement in children and adults (Calfee, Lindamood, & Lindamood, 1973). This unique combination of analytic phonological skills and phonological working memory has been called comparator function (Lindamood, Bell, & Lindamood, 1992), because it involves an internal auditory comparison of the phonemes in two phonological structures (either syllables or words). Lindamood, Bell, and Lindamood (1997) maintain that comparator function is essential for independent phonological decoding, including self-correcting reading errors. Comparator function is believed to comprise two skills: 1. analytic phonological skills for determining the specific identity and order of each phoneme in a word, and 2. phonological working memory skills which are essential for maintaining two words in auditory working memory. Overall, comparator function allows an individual to compare the sameness or difference of the phonemes in two words. For example, during a phonological approach to self-correction, the reader determines if the word he/she said matches the phonemes that are represented by the wordâ€™s graphemes. For words that have a one-to-one mapping between graphemes and phonemes, the reader can use analytic phonological skills and phonological working memory to make this distinction. Analytic phonological skills allow the reader to separate both the word he/she said aloud and the actual printed word into their constituent phonemes. The phonological working memory skills allow the reader to compare the phonemes in the spoken word, which he/she heard himself/herself read aloud, with the phonemes represented by each grapheme in the printed word. This
4 self-correction process involves comparing two readings of the word to be sure their phonemes are identical. This comparison process allows the reader to determine if she/he has added, omitted, shifted, repeated, or substituted one or more phonemes in the word. These same analytic phonological and phonological working memory skills facilitate the ability to identify the phonemes in a word, as well as synthesize or blend the phonemes when reading a word. Thus, additional research on comparator function is very relevant to illiteracy, as comparator function is believed to comprise both analytic phonological skills and phonological working memory skills. Hence, intact comparator function facilitates a phonological approach to reading, and equally important, a phonological approach to self-correcting reading errors. The construct of comparator function does not posit that good readers must literally read every word twice to ensure their decoding accuracy. Rather, comparator function proposes that when a reader is presented with a word that is difficult to read, then she/he has the ability to use comparator function to ensure her/his decoding accuracy. Also, comparator function is proposed as an automatic cognitive process for individuals with good decoding skills. It is not a fully conscious process. However, the reader can bring comparator function under conscious control when necessary to help him/her decode a difficult word. The relevance of aural tasks, e.g. phonological processing, to higher cognitive functions, such as reading, is the primary focus of this study. However, previous functional neuroimaging studies have focused on identifying the neuroanatomical correlates of primary or simplistic phonological processes, like word rhyming.
5 Unfortunately, these primary phonological processing tasks have only a limited relation to reading achievement (Hickok & Poeppel, 2000). Studies of primary phonological processing using an aurally presented stimulus (e.g. word rhyming) have reported several cortical regions of activity (Millen, Haughton, & Yetkin, 1995; Rumsey et al., 1992; Zatorre, Meyer, Gjedde, & Evans, 1992, 1996; Demont et al., 1992; Demont, Price, Wise, & Frackowiak, 1994; Fiez et al., 1995). While these studies have identified cortical regions involved in basic phonological processing, the extent of involvement of these regions in complex phonological processing, e.g. comparator function, is unknown. Hickok and Poeppel (2000) contend that many laboratory-based speech perception tasks (e.g. syllable identification or discrimination) only include a small portion of the cognitive load that is present in functional speech perception. Since comparator function is highly predictive of reading achievement in children and adults and comparator function is believed to represent a cognitive load similar to functional speech perception (Lindamood, Bell, & Lindamood, 1992), a functional neuroimaging study with tasks that approximate comparator function could contribute to a more detailed knowledge of the cortical systems involved in phonological tasks which are highly predictive of reading achievement. The purpose of the current investigation is to acquire functional magnetic resonance imaging (FMRI) and behavioral performance data from two experiments with tasks that represent a hierarchy of phonological processing demands. This hierarchy attempts to approximate the cognitive demands of comparator function and to elucidate
6 the cortical networks involved 1 . Experiment one focuses on a hierarchy of tasks that progress from basic auditory processing of acoustic sounds (counting pure tones) to tasks with increasing demands for analytic phonological skills (subvocally counting phonemes and subvocally segmenting meaningless words or pseudowords into their component phonemes). Experiment two involves a closer approximation of comparator function. This experiment measures phonological working memory (holding two pseudowords in auditory memory and comparing their number of phonemes) and contrasts its cortical demands with those necessary for pure tone working memory (holding in auditory memory and comparing pure tones). Both experiment one and two contrast the performance and cortical activity from two groups of adult men, those with normal reading performance versus those with deficient phonological processing (including comparator function) and impaired reading abilities. This investigation is based on findings from studies of acquired and developmental reading disorders, as well as investigations of differences in cerebral cytoarchitecture in individuals with developmental or acquired reading disorders. First, the nature and effects of phonological awareness and phonological working memory deficits are reviewed. Next, models of reading based on acquired reading disorders, due to injury or cerebrovascular accidents, are reviewed. As a contrast and complement to the neuroscientific and lesion based models of reading impairments, developmental models of reading and phonological processing are presented. Lastly, several aspects relative to the appropriate interpretation of functional neuroimaging data (e.g. general assumptions, methodological limitations, and their interplay in these studies) are presented. 1 Due to issues relative to FMRI experiments that will be discussed in detail later, the most difficult task in this study provided only a close approximation of comparator function.
CHAPTER 2 PHONOLOGICAL PROCESSING Nature and effects of impaired phonological awareness. The following chapter will present the background and rationale for this studyâ€™s operational definition of developmental dyslexia, which guided subject selection and group assignment (control or dyslexic), as well as highlight some of the complexities in this field of research that are problematic for deriving a consensus. Although the majority of the research on developmental dyslexia has focused on children, several studies report cross-sectional and longitudinal findings concerning adults with developmental dyslexia. The most important finding is that reading and phonological awareness impairments are resistant to maturation and persist into adulthood (Maughan, 1995; Bruck, 1993; Lefly & Pennington, 1991). As stated by Lindamood (1994), this strongly refutes early views of developmental dyslexia as being primarily due to a delay in maturation (Stanovich, 1986). Children with developmental dyslexia show the same phonological reading impairments in adulthood. In addition to the persistent reading impairments, negative outcomes have been reported to correlate strongly with developmental dyslexia. Adults with developmental dyslexia may have a lower SES, more episodes of unemployment, lower occupational attainment, lower educational attainment, and may experience symptoms of depression and/or anxiety more frequently than adults without developmental dyslexia (Feldman et al., 1993; Maughan, 1995). However, firm conclusions about the negative effects of developmental dyslexia are difficult to make, because multiple criteria were used to operationalize developmental dyslexia. 7
8 Forming a consensus from the research on developmental dyslexia and phonological processing is confounded by the variety of ways developmental dyslexia has been operationalized among studies with children and adults. For example, a discrepancy definition includes performance on a standardized measure of reading that is 1.5 to 2 standard deviations below an intellectual quotient. A different discrepancy definition involves performance on a standardized measure of reading that is 2 years below an individual's current grade level. Other definitions of developmental dyslexia focus on poor reading achievement and an impaired cognitive process, such as phonological awareness (Sattler, 1993). A completely different approach to defining developmental dyslexia involves a regression equation (based on education, intelligence quotient, and passage reading accuracy) that compares an impaired readerâ€™s standard error of prediction (SEP) with a normal readerâ€™s SEP (Finucci, Whitehouse, Isaacs, & Childs, 1984). An additional issue in operationalizing developmental dyslexia is whether reading achievement should be measured by tests of single word reading, tests that involve reading passages of text, or both. Also, two recent evaluations of the operationalization of developmental dyslexia concluded that children who met the IQ-achievement discrepancy criteria had cognitive profiles that were indistinguishable from children who did not meet this discrepancy, but did demonstrate poor reading performance (Fletcher et al., 1994; Stanovich & Siegel, 1994). Therefore, Fletcher et al. (1994) stated that one option is to allow the use of either IQ-achievement or low reading performance definitions to diagnose individuals with developmental dyslexia.
9 The variability in definitions of developmental dyslexia has also affected estimates of its prevalence in the general population. The prevalence of developmental dyslexia has been estimated to range from 3-6% (Hynd & Cohen, 1983), 3-10% (Pennington, 1990) or as high as 20% (Liberman, 1987) of the general population. Typically, the larger estimates of prevalence have defined developmental dyslexia based on a processing deficit, such as phonological awareness, whereas smaller estimates use discrepancy based definitions. Therefore, an accurate assessment of the prevalence of developmental dyslexia is dependent upon a consensus regarding how this disorder is operationalized. Likewise an accurate understanding of how speech perception is performed by the human brain will be facilitated by neuroimaging research that provides adequate documentation of the cognitive abilities and disabilities of the studyâ€™s participants. It is noteworthy that not only should the abilities of participants in the experimental group be well assessed and documented, but also the abilities of the controls need to be adequately documented. Although obvious, it is an issue that has received very scant attention from researchers using FMRI to investigate the human language system. Consistent findings. Despite inconsistent definitions and inconsistent approaches to operationalizing developmental dyslexia, a well-replicated finding indicates that phonological awareness is a primary facilitator of nonlexical reading. Nonlexical reading, or sounding words out, is performed by converting graphemes into phonemes and accurately blending the phonemes to form a word. Thus, an individual must learn to associate the abstract visual forms of the graphemes with their respective phoneme(s). As the phonemes are blended, the reader must monitor his/her output for accuracy in identity, order, and number of phonemes or syllables. This ability to judge the number,
10 order and identity of phonemes in words has been referred to as phonological awareness (Liberman, 1973; Lindamood, Bell, & Lindamood, 1992). Phonological awareness has been conceptualized as the primary skill necessary for nonlexical reading (Liberman, Shankweiler, & Liberman, 1989; Beauvois & Derouesne, 1979). Individual s with intact phonological awareness are able to self-correct their errors (omissions, additions, repetitions, or shifts of phonemes or syllables) during reading as well as sound out unfamiliar words using their knowledge of the English alphabetic code (Stanovich, 1986; Liberman, Shankweiler, & Liberman, 1989). A comprehensive study of adults with developmental dyslexia (Felton, Naylor, & Wood, 1990) evaluated 115 adults who had been diagnosed during childhood by June Lyday Orton, wife of the late Samuel T. Orton a neurologist who pioneered the assessment and treatment of dyslexia among children. A regression formula was used to calculate deviation scores that classified the group members as either reading disabled, borderline, or normal readers (Finucci, Whitehouse, Isaacs, & Childs, 1984). The reading disabled and normal readers performed significantly differently on three specific measures: pseudoword reading from the Word Attack subtest of the Woodcock Reading Mastery Test (Woodcock, 1987), rapid automatized naming (Denckla & Rudel, 1976), and the Lindamood Auditory Conceptualization Test (LAC) (Lindamood & Lindamood, 1979). The Word Attack subtest, rapid automatized naming, and the LAC test all reflect several aspects of phonological processing. The Word Attack subtest is a well-standardized and normed measure of pseudoword reading. The pseudowords in this subtest require the use of nonlexical reading. Rapid automatized naming involves
11 presenting a stimulus array (letters, digits, or colors) to be named as quickly and accurately as possible. The LAC is a unique measure that assesses a complex level of phonological awareness and requires auditory working memory, together these abilities are known as comparator function (Lindamood et al., 1992). The first section of the LAC measures the ability to judge the number, order, and identity of a string of 2-3 phonemes. The second section measures a complex level of phonological awareness termed comparator function. Comparator function is â€œthe ability to hold the phoneme segments of two phonological structures in mind [auditory working memory] and analyze variations in their number, identity, and orderâ€ (Lindamood, Bell, & Lindamood, 1992). Comparator function is believed to reflect the skills necessary for self-correcting phonological reading errors, such as additions, omissions, substitutions, repetitions or shifts of phonemes (Lindamood, 1994). The LAC assesses comparator function by an auditory presentation of two pseudoword speech syllables that differ by a single phoneme addition, omission, substitution, repetition, or shift. The participant must use colored wooden blocks to indicate the type of change that has occurred from one pseudoword to another (and thereby avoids errors that derive solely from grapheme (letter) to phoneme (sound) conversion). Thus, the goal of this study was to select adults with developmental dyslexia who were impaired on the LAC test and pseudoword reading, in an attempt to identify neuroanatomical correlates for combined cognitive processing, auditory working memory and phonological awareness together.
CHAPTER 3 MODELS OF READING Post-mortem examinations of acquired language disorders have contributed significantly to research on reading. These studies highlighted the cognitive complexity of reading, which is unique to human language. Human language includes a variety of means of communication, such as speech, reading, and writing. Historically, the neurobiology of language has been explored through investigations of acquired language disorders known as aphasia. Aphasia is characterized by a deficit or loss of language following some form of brain injury (Benson & Ardila, 1996). Aphasia can occur along with comorbid impairments in reading, speech, and/or writing. Pierre Paul Broca reported one such case of aphasia in 1861. Broca described an individual who exhibited a primary deficit in the fluency, vocabulary, and intelligibility of speech. A post-mortem examination of the individual revealed a lesion that subsumed portions of the left inferior frontal lobe, insula, and anterior temporal lobe. In 1874 Karl Wernicke wrote a monograph about an individual with impaired comprehension of speech. Upon post-mortem examination of this individual, Wernicke found damage in the posterior superior temporal gyrus of the left hemisphere. Additional examinations (Broca 1865 and 1868 in Benson & Ardila, 1996; Dax, 1836 in Goodglass, 1993; Wernicke, 1881 in Benson, 1993) of individuals with either Wernickeâ€™s or Brocaâ€™s aphasia confirmed that lesions in these two areas were associated with disordered speech fluency or impaired language comprehension, respectively. Also, reading impairments were found in cases of both Wernickeâ€™s and Brocaâ€™s aphasia. Therefore, the investigation of clinical/pathological 12
13 correlations of acquired language impairments, including reading, has an extensive historical basis (Benson, 1993). Some of the first anatomically based models of language function arose from the initial studies of the clinical/pathological correlations of brain lesions and acquired language deficits. Wernicke postulated an anatomically based model of language functioning, which Lichtheim (1884) modified as an anatomical-connectionist model of aphasia (stated in Goodglass, 1993). Dejerine modified the Wernicke-Lichtheim model of aphasia to incorporate the cognitive processes involved in reading (Dejerine, 1891 in Bub, Arguin, & Lecours, 1993). Thus, the beginning models of aphasia provided the basis for some of the initial and current models of reading. The dual route model of reading. One model that has developed from the Wernicke-Lichtheim anatomical-connectionist model is the dual-route model (Marshall & Newcombe, 1973). The dual-route model of reading proposes that reading is performed by either one of two possible routes: i) the memorized whole-word (lexical) route, or ii) the active conversion of letters-to-sounds (nonlexical) route (Marshall & Newcombe, 1973; Ellis & Young, 1988; Coltheart, Curtis, Atkins, & Haller, 1993). The lexical route has several components that facilitate the conversion of visual inputs (letters in a word) to phonological outputs (reading words). The lexical route is believed to involve a storage of whole words called the orthographic input lexicon. It can be used to read words without performing a grapheme (letter) to phoneme (sound) conversion. Evidence for this lexical reading route came from the examination of individuals who acquired reading deficits following some form of brain injury (Shallice & Warrington, 1980; Marshall & Newcombe, 1973; Patterson, Marshall, & Coltheart,
14 1985). These individuals showed a primary deficit in reading words that do not possess a one-to-one mapping between graphemes and phonemes. Words without this one-to-one mapping are termed orthographically irregular words, because their graphemes (the orthography of a word) do not coincide with the phonemes in the pronunciation of the word. For example, the word "though" is orthographically irregular because four of the graphemes (ough) do not produce their common phonemes (/ou/ as in out, /g/ as in get, and /h/ as in he). Instead, the graphemes ough represent only one phoneme /oe/, as in toe. Orthographically irregular words are learned by repeated exposure and reinforcement. The repeated exposure and reinforcement develops a memory for a wordâ€™s letters and its pronunciation. This memory or storage of orthographically irregular words and their pronunciations is known as an orthographic lexicon. Therefore, individuals who have significant difficulty reading orthographically irregular words exhibit a selective impairment in the lexical route to reading. They read real words or pseudowords that follow a one-to-one mapping of graphemes to phonemes better than orthographically irregular words. The nonlexical route to reading involves converting the letters of the word (graphemes) into their corresponding sounds (phonemes). This conversion process is commonly referred to as a phonological approach to reading, because it requires the awareness of the order and identity of the speech sounds that the graphemes represent. The evidence for the nonlexical route comes from individuals with an acquired selective deficit in reading pseudowords or novel real words (Beauvois & Derouesne, 1979; Funnell, 1983; Patterson, 1982; Shallice & Warrington, 1980). Pseudowords are meaningless words that possess a one-to-one mapping between graphemes and
15 phonemes, but follow the normal conventions of English. For example, â€œblepâ€ has a consonant blend, a vowel, and another consonant that are common to English, but the word is meaningless (Berndt, Dâ€™Autrechy, & Reggia, 1995). If the nonlexical route is intact, then even novel pseudowords are easily read by blending the sounds represented by the graphemes. Nonlexical reading is not dependent upon a memory of the wordâ€™s correct pronunciation, learned from prior experience with the word as a whole. Hence, the term nonlexical reading indicates a reading route that is not reliant on a store (lexicon) of previously learned whole words. One nonlexical task, namely pseudoword reading, is a consistently strong predictor of reading achievement and correlates with measures of phonological awareness (Felton, Naylor, & Wood, 1990). The single route model of reading. A single-route model of reading has also been postulated (Seidengberg & McClelland, 1989). This model contends that while lexical and nonlexical processing does occur during reading, these processes are supported by a single computational mechanism that differs fundamentally from a dual-route conceptualization. Thus, there are no rules that govern grapheme to phoneme transcoding (such as the letter c says the sound /s/ if immediately followed by a letter e, i, or y), and there is no storage of word forms in a lexicon. Instead, four processors facilitate reading: orthographic, phonologic, semantic, and contextual. The processors are joined through a series of units and these connections can be strengthened by exposure, facilitating faster word recognition without a lexicon of words. Similar to the central nervous system, this network is believed to have both excitatory and inhibitory connections. However, tests of this computational model have found it to have limited success with accurately reading pseudowords and some orthographically irregular words (Besner, Twilley, McCann, &
16 Seergobin, 1990). Also, words that have been accurately decoded by the single-route model are limited to monosyllable words that are formed from a selected group of English letters (Coltheart, Curtis, Atkins, & Haller, 1993). The balance model of reading. In contrast to the neuroscientific investigations and the lesion based dual-route model, developmental models propose a different account of the reading system. The balance model (Bakker, Moerland, & Goekoop-Hoetkens, 1981) proposes that developmental dyslexia is the direct result of a failure to establish cerebral dominance. Support for the balance model comes from Bakker, Smink, and Reitsmaâ€™s (1973) report that proficient young readers perform symmetrically on dichotic listening tasks. Conversely, proficient older readers show asymmetrical performance on these tasks. These data are interpreted to indicate that normal readers show increasing lateralization for processing language as they mature. Additionally, Bakker et al., (1973) found that children with developmental dyslexia showed a persistent deficit in cerebral dominance (a lack of lateralization for language), as measured by dichotic listening. Bakker contends that early reading development involves the perceptual analysis of text while later stages of proficient reading rely on the semantic analysis of text. Thus, Bakker believes his model accounts for two types of developmental dyslexia, P-type and L-type. The P-type dyslexic exhibits an over-reliance on a perceptual synthesis approach to reading (right-hemisphere dominant) and is described as reading slowly with many fragmentation errors. The L-type dyslexic is believed to rely on left-hemisphere processes (perceptual analysis) and is characterized as reading hastily and inaccurately (Bakker, 1992). Therefore, Bakkerâ€™s hypothesis is that deficient cerebral dominance impairs the language and reading skills of individuals with developmental dyslexia.
17 Several criticisms of the balance model challenge its validity. One of the most striking criticisms is that the experiments supporting Bakkerâ€™s theory have not been replicated by anyone other than himself (Hynd, 1992). This may be due in part to reports of lateralization of language processes in young children (Ogden, 1996), which directly contradict Bakkerâ€™s hypothesized stages of language development. Also, it has not been well established that the right hemisphere is involved in early reading (Hynd, 1992). Bakker has failed to demonstrate any convergent validity between his measure of cerebral dominance and processes known to occur during reading, such as phonological awareness. Lastly, the lack of cerebral dominance, as measured by his dichotic listening task, may reflect an underlying neuroanatomical anomaly in individuals with developmental dyslexia. This neuroanatomical anomaly may cause the symmetrical performance of individuals with developmental dyslexia and cause their reading impairments. The symmetry may be a symptom of a cortical abnormality, which has been found in a few post-mortem studies of individuals with documented developmental dyslexia (Galaburda, Sherman, Rosen, Aboitiz, & Geschwind, 1985). Therefore, while Bakker presents a unique perspective on the reading process, the lack of convergent data precludes consideration of the balance modelâ€™s tenets in the present investigation of the neuroanatomical basis of phonological processing. Developmental dyslexia and the motor-articulatory feedback hypothesis. A recent developmental model of reading contends that a deficit in motor-articulatory feedback contributes to the development of deficient phonological awareness and nonlexical reading in developmental dyslexia. Heilman, Voeller, and Alexander (1996) propose this motor-articulatory feedback hypothesis of developmental dyslexia
18 (MAFHDD). They contend that an impairment in the oral-kinesthetic feedback system may hinder the otherwise normal development of phonological awareness: the perception of spoken words is associated with the production of intended articulatory gestures. Therefore, for each phoneme in a word there is a new movement of the articulators. Awareness of the movements of the articulators would allow the child to parse a word into its component phonemes. If awareness of the position and movement of the articulatory apparatus is critical for phonemic parcellation of words and dyslexic children are unaware of the position and movements of the articulatory apparatus, they would be less able to perceive the low-contrast, rapid temporal presentation of the phonemic components of spoken words. (p. 409) The Motor-Articulatory Feedback Hypothesis of Developmental Dyslexia is based in part on the Motor-Theory of Speech Perception (Liberman & Mattingly, 1985). Liberman and Mattingly present empirical evidence that the most stable characteristics of phonemes are the oral motor-articulatory gestures which produce the phonemes. An awareness of these motor-articulatory gestures provides an in-depth perception of each phoneme and facilitates its association with a grapheme. Thus, the in-depth perception of phonemes (oral-motor feedback along with auditory and visual input) may be the basis for developing phonological awareness and nonlexical reading (Ehri, 1998). Support for the motor-articulatory feedback hypothesis of developmental dyslexia (MAFHDD) (Heilman et al, 1996) is found in a variety of disciplines using several technological methods. The support for the MAFHDD comes from behavioral assessment (Montgomery, 1981), electromyography (Sokolov, 1972), electrical cortical stimulation (Ojemann & Mateer, 1979), positron emission tomograpny (PET) (Rumsey, Nace, & Andreason, 1995; Zatorre et al., 1992, 1996), magnetoencephalographic recordings (Sams et al., 1991), and psychophysiology (McGurk & McDonald, 1976) findings. While none of these studies are a direct test of the MAFHDD, they do provide compelling evidence in support of the hypothesized relationships between deficient sensory
19 integration and an ensuing behavioral disorder. Therefore, this data will be reviewed in more detail to describe the cumulative support for the tenets of the MAFHDD from this wide range of studies. Behavioral assessment data comparing children with and without dyslexia provides some intriguing information about the possible links between sensory input or processing and reading abilities. Montgomery (1981) used an oral articulatory awareness task to reliably discriminate between children with normal and disabled reading skills. She assessed motor-articulatory awareness by having each child repeat a phoneme after the examiner and then choose the sagittal lineâ€“drawing of an articulating mouth that best matched the movements the childâ€™s mouth made to produce that phoneme. For each phoneme the child chose from a group of three different mid-sagittal drawings of the mouth cavity that showed the teeth, tongue, and lips in different positions of articulation. Montgomery found that children with a reading disability were not aware of the position of their articulators (tongue, teeth, and lips) during the production of various phonemes. Conversely, children with normal reading skills adeptly identified the position of their articulators when producing individual phonemes. Interestingly, infants as young as 4 months of age have demonstrated an accurate ability to visually discriminate between matched and mismatched auditory speech and oral motor movements of a speaker (Eimas, Siqueland, Jusczyk, & Vigorito, 1971). Studies with electromyography indicate that a dyslexicâ€™s poor awareness of the articulators may have a neurological basis. Electromyographic studies of individuals believed to be normal readers have reported that during silent reading (no mouth movements being made) more electrical activity is present in the tongue when reading
20 less familiar native words and during more difficult to pronounce foreign words than for familiar words (Sokolov, 1972). The increased electrical activity during reading less familiar words may reflect the increased use of motor-articulatory feedback to facilitate reading less familiar words. Even though the articulators were not moving during the reading, the increased electrical activity implies that motor programs for phoneme or word production may still be activated by covert speech mechanisms. Therefore, an interesting relationship may exist between a developmental dyslexicâ€™s sensory feedback from his/her oral articulators and the ability to use motor-articulatory feedback or stored oral motor programs to assist the parsing of phonemes when attempting to read words. While anecdotal evidence of the linkage between motor-articulatory feedback and phonological processing or reading is not as convincing as empirical support, it is important to consider some anecdotal evidence that could lead to additional empirical support. One piece of anecdotal support for the potential role that motor-articulatory feedback may play in parsing phonemes to facilitate reading comes from observations of children and adults during reading. It has been reported (Ehri, 1998) that children and adults overtly move their mouths and subvocalize or read aloud as they decode a new or less familiar word. Again, the child and adultâ€™s use of motor-articulatory feedback appears to facilitate their ability to segment and blend phonemes. Additionally, it is commonly observed that when a parent or teacher is trying to help a child correctly pronounce a word during reading or speaking, the adult does not say the word more loudly and clearly into the child's ear; if reading were purely an auditory process, then speaking more loudly and clearly into a learnerâ€™s ear would be a likely way to help develop reading or speaking skills. Rather, the adult orients his/her face and mouth in full
21 view of the child and the adult moves his/her mouth in a slower, more exaggerated fashion. The adult pronounces the word more slowly and more clearly. Similarly, many individuals innately recognize that seeing someone's mouth move can aid their own production and perception of language, especially in the presence of background noise. Thus, these anecdotal reports indicate that a percentage of individuals innately recognize that motor-articulatory feedback can aide the perception and production of language. PET and electrical cortical stimulation findings indicate that anterior perisylvian cortex seems to be involved in both the auditory perception or discrimination of phonemes and the oral-motor movements necessary to articulate a particular phoneme (Ojemann & Mateer, 1979; Zatorre et al., 1992, 1996). Also, this anterior perisylvian area has been shown to become more active during the reading of pseudowords than irregular real words (Rumsey et al., 1995). Magnetoencephalography (MEG) involves the recording of weak magnetic signals associated with neural currents. One such study found that visual feedback of the mouth of a speaker provided a measurable change in the superior temporal gyrus, believed to represent the integration of auditory and visual information. These authors contend that the MEG technique provides evidence that speech can be seen before it is heard, such that â€œvisual cues, e.g., from lip movements, exist in some cases hundred of milliseconds before the corresponding auditory stimulusâ€ (Sams et al., 1991, p. 144). Another manifestation of the integration of visual and auditory information is the â€œMcGurk effectâ€, where an individualâ€™s auditory perception is altered by visual information (McGurk & McDonald, 1976; MacDonald & McGurk, 1978). For example, individuals who hear an auditorily presented syllable /ba/ while watching a videotaped face saying /ga/ will report that they heard /da/ or /ga/ instead of
22 /ba/. Therefore, oral motor-articulatory feedback may serve as a bridge between learning visual symbols and the sounds that are associated with these symbols. The dual-route model of reading and the motor-articulatory feedback hypothesis of developmental dyslexia (MAFHDD) are complementary. The MAFHDD provides a reasonable explanation for how impaired phonological awareness may develop. While this theory has not been directly tested, there is a considerable amount of indirect support for the MAFHDD from empirical findings in a wide array of disciplines. In addition, both the increasing evidence that phonological processing has a causal link to reading skill (Stanovich & Siegel, 1994; Wagner & Torgesen, 1987) and the prevalent findings of impaired pseudoword reading among developmental dyslexics provide support for the dual-route modelâ€™s description of a nonlexical reading route. The single-route and the balance model provide interesting alternative conceptualizations of developmental dyslexia, but they do not provide an adequate account for the role of phonological awareness. Thus, portions of the dual route model, motor-articulatory feedback hypothesis, and Alexanderâ€™s (2002) developmental model of language were used as a foundation for generating the conceptual framework for the current study. This conceptual framework was then linked to functional neuroimaging findings to specify anatomic substrates for the current model of phonological processing. Also, exploring the neuroanatomical regions involved in phonological processing and comparator function (Broca's area providing motor-articulatory feedback, supramarginal gyrus functioning as a phonological storage center, superior temporal gyrus providing primary auditory perception, somatosensory cortex facilitating articulatory tactile feedback, and the lateral prefrontal cortex active during phonological working memory operations) was a primary
23 goal of this study. However, it is noteworthy that the issues raised in this chapter regarding variability in diagnosis and approaches to operationalizing dyslexia also pose difficulties for interpreting the increasing number of functional neuroimaging studies of reading and developmental dyslexia. Alternative theories of developmental dyslexia. Alternative theories of sensory-perceptual deficits (auditory, visual and motor) that may lead to reading impairments, either separate from or along with phonological deficits, have also been proposed. Although these models are not directly addressed by the methodology of this study, the tenets of these models may be relevant to this studyâ€™s findings. One theory is the magnocellular theory (Stein & Walsh, 1997), named after the subcortical cell types that are believed to cause sensory-perceptual deficits and, ultimately, reading impairment. This magnocellular theory includes deficits in two separate perceptual systems, namely impaired visual perception and impaired auditory perception as well as dysfunctions in the cerebellum. While auditory, visual and cerebellum functioning involves different sensory, motor, and perceptual systems, the etiology of their deficits is reported to be based on the same type of cells (magnocellular) (Stein, 2001). Therefore, specific descriptions of the visual and auditory perceptual deficits, and cerebellar dysfunction that may lead to reading problems are presented. The visual theory of reading proposes that reading impairments are secondary to visual processing deficits (Lovegrove, Bowling, Badcock, & Blackwood, 1980; Livingstone, Rosen, Drislane, & Galaburda, 1991). The cytoarchitecture of the visual system is comprised of both parvocellular and magnocellular subdivisions or pathways. The parvocellular pathway is responsible for visual perception of high spatial (high
24 contrast) and low temporal (slow) frequencies of visual stimuli. In contrast, the magnocellular pathway is specialized for the perception of low spatial (low contrast) and high temporal (fast) frequencies of visual stimuli. This visual model of dyslexia proposes that the magnocellular system is responsible for the perception of words (binocular control and visuospatial attention) during the reading process. This model is viewed as being complementary to the phonological deficit theory and does not propose to account for the etiology of phonological deficits. Rather, anatomical abnormalities in the cells of the magnocellular system and behavioral data (Livingstone et al., 1991) which indicate visual perception difficulties are cited as evidence that reading impairments may arise from aberrant visual processing by the magnocellular system. Therefore, the visual model proposes that visual processing deficits lead to reading difficulties by inhibiting accurate and efficient perception of letters and words on a page. The rapid temporal auditory processing theory of developmental dyslexia (Tallal, 1980; Tallal, Miller, & Fitch, 1993) involves subcortical structures (medial geniculate nucleus) and cortical pathways in the auditory system that have cellular structures and functions which parallel the magnocellular and parvocellular divisions of the visual system. Similar to the visual model, the rapid temporal auditory processing theory reports cytoarchitectural evidence of cellular anomalies in the magnocellular like division of the auditory system (Galaburda, Menard, & Rosen, 1994). Behaviorally, this theory proposes that the cellular anomalies result in an impaired ability to discriminate rapid acoustic cues that help distinguish phonemes. Tallal et al. (1993) contend that this sensory-perceptual deficit undermines the development of the phonological system by not allowing stable
25 phonological representations to be formed. The poor phonological representations are believed to lead to poor phonological processing and impaired nonlexical reading. The models of reading and the theories of sensory-perceptual deficits that may affect reading achievement cover a broad range of cognitive, sensory, perceptual, and motor abilities. Also, these systems highlight the broad range of behavioral manifestations of impairments in one or more of these systems that may occur in developmental dyslexia. However, based on research findings from remedial treatment of developmental phonological dyslexia, preventive treatment of developmental dyslexia, and rehabilitation of acquired phonological alexia, only when treatment has focused on the integration of sensory, motor, perceptual and cognitive abilities relevant to reading have robust changes in behavioral abilities (phonological awareness, reading, and pseudoword reading) been reported (Alexander et al., 1991; Conway, et al., 1998; Torgesen, Wagner, Rashotte, Rose, Lindamood, & Conway, 1999; Kendall et al., 2001).
CHAPTER 4 FUNCTIONAL NEUROIMAGING Prior to reviewing the current FMRI literature of developmental dyslexia, some essential background issues regarding neuroimaging are reviewed in this chapter. Functional neuroimaging captures local changes in metabolism or related brain substrates in a pictorial format, which are a direct result of engaging the local region in some activity (Nadeau & Crosson, 1995). However, proper use of FMRI technologies necessitates several points of clarification. Currently, Blood Oxygen Level Dependent (BOLD) FMRI (Ogawa et al., 1992) is the most widely used FMRI technique. This technique is based on the magnetic susceptibility effects of deoxyhemoglobin, which is paramagnetic. A reduction in deoxyhemoglobin occurs in areas of high activity, due to a greater increase in blood flow than metabolic needs demand. The ensuing decrease in amount of paramagnetic deoxyhemoglobin present is associated with an increase in magnetic resonance signal intensity. However, it is important to consider whether this signal change is originating from neuron cell bodies or synapses. Changes in local metabolism are the product of local synaptic activity. If a neuron has primarily distant projections, then increased or decreased metabolism will appear at the projection site of the neuron. This is due to the terminal buttons on the axonal arboration and the dendritic tree having the largest amount of metabolic change in activity. Their high surface-to-volume ratio for metabolism and frequent depolarization results in a larger expenditure of energy than other portions of the neuron (Nadeau & Crosson, 1995; Logothetis, 2000). Therefore, interpretation of FMRI requires a 26
27 knowledge of the cytoarchitecture and connections of neurons or known axonal pathways within the brain. The interpretation of activity increases, activity decreases, or a lack of activity in brain regions during functional neuroimaging is a complicated process. Brain regions that do not show marked changes in metabolism during functional neuroimaging may actually be involved in the behavioral task that is being measured. For example, if increases and decreases of synaptic activity occur simultaneously within a single region of cortex, then marked changes in metabolism or neuronal activity may not be apparent in that region of cortex. However, changes in the output from this cortex may be present in its synaptic connections with other cortical regions further downstream, as long as no other source affects the activity of this downstream region. Secondly, if the two behavioral tasks that are contrasted are too similar, then changes in metabolism in all the regions that are involved in a behavioral task may not be evident. The extensive similarity between the two tasks would result in very few brain regions reaching the necessary level of significance for a change in metabolism to be evident. Thus, functional neuroimages of increased activity, decreased activity, or a lack of activity may be misleading without the guidance of theoretical models of cortical and subcortical function and the connections or pathways between these structures. Because FMRI measures cerebral metabolism indirectly, it depends on a constant relationship between blood flow and metabolism. While this is likely to be true for many participants, there are occasions in which blood flow may not remain constant, e.g. stroke. Therefore, the relationship between blood flow and metabolism is an additional consideration for interpretation of functional images (Logothetis, 2000).
28 Methodological Considerations of Functional Neuroimaging The PET and FMRI methods each have unique technical limitations and advantages that need to be addressed when designing a study of the neural substrates of cognition. PET has several disadvantages compared to FMRI. These include low spatial resolution, moderate signal-to-noise ratio, slow temporal resolution, limited number of studies per participant, exposure to radioactive isotopes, variable image registration and ROI placement accuracy, and moderate interstudy interval. A few significant advantages of PET over FMRI include low susceptibility to movement in the scanner, and no susceptibility to image degradation by air spaces nearby cortical or subcortical structures. Other disadvantages of FMRI compared to PET include susceptibility to pulsatile blood and cerebral spinal fluid flow and significant limitations of study environment due to scanner noise (DeYoe et al., 1994; Binder & Rao, 1994; Nadeau & Crosson, 1995). Two other important issues include evidence that nonspecific learning and practice effects occur during FMRI responses that are associated with cognitive activation tasks, and that putative rest states are misleading baselines and a potential source of confusion since the brain never reaches a point of inactivity (Binder, 1995). Early assumptions that rest provided a cortically inactive task for a baseline condition have been shown to be problematic and far from accurate. Binder, Frost, Hammeke, Belgowan, Raod, and Cox (1999) reported an extensive semantic network of cortical activity during â€œrestâ€ conditions. This network was interpreted as reflecting the always-on nature of the human brain, which may involve ongoing retrieval, representation and manipulation of semantic information even during conscious rest conditions. Thus, the condition chosen for a baseline period may have a differential effect on the patterns and areas of cortical activity
29 detected (Newman, Twieg, & Carpenter, 2001). Consequently, the selection and design of a baseline comparison task is as important as the task chosen to elicit brain activity. One remaining FMRI consideration involves the use of an auditory stimulus in an environment with inherent background acoustic noise. Because the scanner in FMRI experiments produces a significant amount of noise during imaging, great care is required to ensure that as much of the scanner noise as possible is attenuated. Presently, expandable foam ear inserts which surround the stimulus delivery tubing, headphones, or foam padding on the outside the participantâ€™s ears are all commonly used to reduce the impact of background noise during FMRI. PET Neuroimaging of Pure Tone and Phonological Stimuli Auditory processing is a primary construct of the AAMSP model and the neural substrates of auditory processing have been studied with PET. However, while this study will test auditory processing with FMRI, it is also important to consider the findings from PET studies of individuals with normal auditory processing. Pure Tone Stimulus Normal participants. PET has been used to identify brain regions involved in several different types of tasks with aurally presented pure tones. Rumsey et al. (1992) administered a continuous performance tone detection task to a group of unimpaired readers whose reading skills were well documented. The comparison condition was a resting state with eyes closed. This group showed activity in left hemisphere regions near the angular gyrus and near Wernickeâ€™s area. Rumsey et al. (1994) used a tonal memory task that involved presentation of pairs of tone sequences, with no more than 4 tones per sequence. The comparison task was an eyes closed passive resting condition. Participants responded to the two sequences of
30 tones by pressing a key whenever the two sequences were identical. Stimulus presentation rate was very rapid (16 pairs per minute). Rumsey et al. (1994) reported that unimpaired readers exhibited right frontotemporal (superior and middle temporal gyri, middle and inferior frontal gyri) and left temporal activity (anterior and middle temporal regions) on this task. Fiez et al. (1995) investigated cortical activity during passive listening to tone triplets. Participants included a group of unimpaired readers, however, participantsâ€™ reading abilities were not formally assessed. Fiez et al. (1995) reported significant bilateral temporoparietal activity during this task. Dmonet et al. (1992) administered a pure tone task that involved detecting a rise in pitch in the third component of a triplet of pure tones among distractor triplets. The distractor triplets had no change or a lower pitch on the third tone. This task was reported to result in right superior temporal gyrus activity. Johnsrude, Zatorre, Milner, and Evans (1997) used a same or different judgment task with tones that included spectral changes (glides of long or short duration). Findings did not support predictions that left hemisphere auditory cortices are specialized for rapid acoustic processing. Rather, activity was reported in the left fusiform gyrus and was interpreted to indicate this region's potential involvement in judgment of short duration frequency changes. Dyslexic participants. Rumsey et al. (1992) administered a continuous performance test tone detection task to a group of developmental dyslexics. The reading skills of the developmental dyslexic group were thoroughly assessed with standardized measures and participants met discrepancy and regression definitions of developmental
31 dyslexia. The comparison condition was a passive rest state with eyes closed. Developmental dyslexic participants showed significant activity in the right parietal and the left anterior temporal region. Since the region in the left was also activated by a rhyme detection task, Rumsey et al. (1992) interpreted this region to be generally related to auditory processing. Rumsey et al. (1994) used a tonal memory task that involved presentation of pairs of tone sequences, with no more than 4 tones per sequence. The comparison task was an eyes closed passive rest condition. Participants responded to the two sequences of tones by pressing a key whenever the two sequences were identical. Stimulus presentation rate was very rapid (16 pairs per minute) and developmental dyslexics performed significantly worse on this task than controls without reading impairments. Developmental dyslexic participants showed activity in the left anterior and middle superior temporal lobe (registration and encoding of sounds), right anterior superior temporal lobe (right greater than left), and right posterior frontal lobe (manipulation and judgment of change in tones). The only significant difference between unimpaired controls and developmental dyslexics was a right hemisphere advantage for the controls in the middle temporal (tone registration) and two anterior inferior frontal lobe regions (tone registration). Findings from the PET studies of auditory processing with controls and dyslexics indicate some potential differences between the groups with pure tone stimuli. From the present studies, it is evident that differences across the temporal and frontal lobes are likely. However, only a limited number of studies are available and a larger number of studies are necessary to make more definitive determinations. Therefore, in the next
32 section (p. 31) additional anatomical regions will be reviewed from FMRI studies of pure tone processing. Phonological Stimulus Normal participants. Rumsey et al. (1992) administered a word rhyming task to a group of unimpaired readers whose reading skills were thoroughly assessed. The comparison condition was a passive rest state with eyes closed. Unimpaired readers showed activity in left hemisphere regions near the angular gyrus and near Wernickeâ€™s area. Hagman et al. (1992) studied a syllable identification task that involved listening for a target syllable presented among a series of phonemes and syllables. This study evaluated performance of unimpaired adults on a visual and a verbal speech syllable CPT task. The auditory CPT group evidenced higher metabolism bilaterally in the middle, superior, and inferior temporal lobe and lower frontal/parietal metabolism compared to the visual CPT group (left hemisphere metabolism was greater than the right for the individual tasks, auditory and visual CPT). Demont et al. (1994) administered four different phonological tasks and a comparison pure tone task. Two of the phonological tasks were described as involving perceptual ambiguity as the target phoneme was embedded within a speech syllable. These tasks involved detecting a phoneme in a target word (e.g. /b/) or detecting the phoneme /b/ if a /d/ precedes /b/ in the word. The two tasks also differed regarding the presence of nonsequential and sequential features respectively. Both of these tasks were reported to result in activity in left fusiform gyrus and Brocaâ€™s area. The remaining two unambiguous phonological tasks (one with sequential features and one without) were reported to activate Wernickeâ€™s area.
33 Fiez et al. (1995) investigated three different phonological tasks (steady state vowels, consonant-vowel syllables [CV], and consonant-vowel-consonant syllables [CVC]), and a visual fixation control task. Each task required responding whenever a corresponding target stimulus (vowel, CV, or CVC) was perceived among a presented class of stimulus. This study is difficult to interpret as the authors collapsed the data across the three experimental tasks. However, cortical activity was reported along the superior temporal gyrus, bilateral frontal opercular, and near the supplementary motor area. Fiez et al. (1996) had participants retain five pseudowords in memory and contrasted this task with visual fixation. Findings included bilateral activity in the dorsolateral prefrontal cortex. Supplementary motor area and cerebellum activity were also reported and their degree of activity appeared to be related to the amount of memory load the task entailed. Zatorre, Evans, Meyer, and Gjedde (1996) reported a reanalysis of findings from an experiment in 1992 and some additional experiments. One task, phonetic discrimination, was assessed by having participants attend to two real words and respond if the words ended in the same sound. Another task, phonetic monitoring, involved responding if the phoneme /b/ was perceived in either of a pair of words or providing an alternate response if it did not. These tasks were compared to a passive listening to words task. Both phonetic monitoring and discrimination resulted in significant activity (cerebral blood flow) in Brocaâ€™s area. The phonetic monitoring also resulted in occipital cortex and fusiform activity. A reanalysis of the 1992 data for phonetic discrimination and phonetic
34 monitoring tasks, compared to a control task of pairs of filtered noise bursts, revealed activity primarily in Brocaâ€™s area. Dyslexic participants. Rumsey et al. (1992) administered a word rhyming task to a group of developmental dyslexics who met both regression and discrepancy definitions of developmental dyslexia. The comparison condition was a passive rest state with eyes closed. Developmental dyslexics failed to show any significant activity in regions of the left hemisphere during this task. Hagman et al. (1992) studied a syllable identification task that involved listening for a target syllable presented among a series of phonemes and syllables. Developmental dyslexics performed more poorly than unimpaired controls on this task. No differences were found in lateral cortical metabolism, but developmental dyslexics did evidence higher metabolism bilaterally in medial temporal regions with an anterior to posterior gradient. FMRI of Pure Tone and Phonological Stimuli To add to the findings from the PET research on auditory processing of phonological and tonal stimuli, a brief review of the more recent findings from FMRI research is presented. The use of FMRI technology provides better temporal resolution and the opportunity for a wider combination of tasks within a study. These aspects may provide new findings or an elaboration of the findings from the PET research. Also, the present study will be using FMRI technology and is more directly comparable to prior studies of phonological and acoustic processing with FMRI. Therefore, a brief review of FMRI research relative to phonological and acoustic processing is presented.
35 Pure Tone Stimulus Binder et al. (1997) administered a pure tone monitoring task to 30 reportedly normal participants (no quantitative data was provided to document the subjects status as a normal). Participants heard sequences of 3 to 7 tones and were required to press a button if the sequence contained two high tones (750 Hz) among a group of low tones (500 Hz). Compared with a resting state, bilateral activity was found in the primary and association auditory cortex of superior temporal gyrus, supramarginal gyrus, premotor cortex, supplementary motor area, anterior cingulate, and anterior insula. Right hemisphere activity in the supramarginal gyrus (SMG) was greater than the left, and the medial temporal gyrus and prefrontal areas were only active in the right hemisphere. Compared to a language contrast task, the SMG appears to be involved in the short-term storage of auditory information, which is in agreement with findings by Paulesu, Frith, and Frackowiak (1993). Binder et al. (1994, 1995, 1996) presented the same tone task to the same group of subjects described above, but the baseline was changed to a white noise presentation. Activity was primarily found in the superior temporal gyrus bilaterally. Millen, Haughton, and Yetkin (1995) presented pulsed pure tones at 1000 Hz and 4000 Hz to 8 volunteers (no testing to verify cognitive functioning in the normal range and 4 participants were being scanned for intracranial disease). Activity was predominately in the superior temporal gyrus. However, no control task was used and some of the participants made a motor response to the stimulus while others did not. Therefore, the findings of this study are compromised by significant confounds and provide very preliminary findings at best.
36 Bilecen, Seifritz, Scheffler, Henning, and Schulte (2002) reported the presence of amplitopicity in the auditory cortex of 5 male and 4 female volunteers (no data reported to validate subjectsâ€™ reportedly normal range of cognitive and sensory abilities). By presenting a pulsed sine tone of 1000 Hz at three sound pressure levels (SPL) of 70, 82, and 90 decibels (dB), they identified a differential pattern of cortical activity for 70 versus 90 dB. The pattern of activity in the transverse temporal gyrus (TTG) of the superior temporal region for the 90 dB stimulus included a medial and ventral region of the TTG. Conversely, the 70 dB stimulus produced activity in a more lateral and ventral region of the TTG. These differential patterns of activity seem to indicate that the auditory cortex may have amplitopicity characteristics. The auditory cortex may make use of more than one pathway. The authors propose the possibly that a medial-ventral and lateral-ventral pathways that handle different aspects of speech perception and spatial location. Phonological Stimulus Dhankhar et al. (1997) presented normal participants with a series of single syllable, common English nouns at different rates of speed (10, 50, 90, 130 words per minute). Results indicated significant activity in the transverse left temporal gyrus, left transverse temporal sulcus, and bilateral superior temporal gyrus. An increase in signal was noted with increased rate of presentation of words, but not beyond 90 words per minute. Corina et al., (2001) compared the speech perception abilities of 10-13 year-old boys (all right-handed). The presence and absence of phonological and reading deficits were well document with standardized testing for both groups of subjects. Tasks were aurally presented words, pseudowords or series of tones. The word and pseudoword tasks required a same or different distinction regarding a pair of words rhyming or both being
37 real words, while the tone task required a determination of the presence or absence of a change in pitch in a series of five tones. Results indicated that for the phonological task the dyslexic boys had more activity in the right inferior temporal gyrus and in the left precentral gyrus. The pitch judgment task resulted in the controls producing more activity in the inferior parietal lobe. Overall, the authors concluded that dyslexics and controls differ in not only auditory language processing, but also in executive control and attention for switching between language codes (phonological, semantic and lexical); the phonological task in this study included the combination of runs with pseudowords and real words. Therefore, the differences in activation may reflect the activation of semantic networks during the phonological tasks. Also, the dyslexics had comorbid rapid naming deficits, which may indicate a more severe impairment in the language system than individuals with only phonologically based language deficits (Torgesen, Wagner, Rashotte, & Burgess, Hecht, 1997; Wolf & Bowers, 1999)
CHAPTER 5 RATIONALE AND HYPOTHESES Statement of the Problem Investigations of phonological processing with functional neuroimaging have identified specific brain regions involved in this perceptual/cognitive skill. The presence of some replication across these studies supports the reliability of functional neuroimaging for measuring such perceptual/cognitive events (Demont, Fiez, Paulesu, Petersen, and Zatorre, 1996). These studies have provided a basis for investigating the phonological processing and phonological awareness deficits of developmental dyslexics with FMRI. The Articulatory Awareness Model of Speech Perception (AAMSP) indicates how an aurally presented pseudoword may lead to a spread of activity that facilitates the ability to segment a pseudoword into its phonemes (Figure 1). The initial auditory input of a pseudoword creates or is stored as an auditory representation (superior temporal gyrus, STG). This input triggers activity of separate phonological and articulatory representations of speech sounds. The phonological representation includes some awareness of the individual phonemes within the pseudoword, a more fine-grained representation of the pseudoword (supramarginal gyrus, SMG) than the initial auditory representation (superior temporal gyrus), but not enough to quickly and accurately perform segmentation tasks. The articulatory representation includes the oral-motor programs for the movement of the articulators to produce the pseudowordâ€™s phonemes. Also, the articulatory representation consists of the somatosensory awareness of the 38
39 distinctive features within these motor programs (Brocaâ€™s area and somatosensory cortex, SSCx). Thus, through back propagation between the auditory representation, articulatory representation, and phonological representation, a multisensory system is established that facilitates the ability to segment a pseudoword into its component phonemes and make use of the alphabetic code of English. The addition of Baddeleyâ€™s (1986) central executive system to the AAMSP expands this model to include auditory working memory. The expanded model incorporates the importance of auditory working memory in phonological processing, accommodating Lindamood's (1992) concept of comparator function. The cortical activity associated with the working memory aspect of auditory or phonological working memory tasks is reportedly in the left supplementary motor area (SMA) and dorsolateral prefrontal cortex (DLPC) (Fiez et al., 1996; Paulesu et al., 1995; Dâ€™Esposito, Detre, Alsop, Shin, Atlas, & Grossman, 1995; Salmon et al., 1996). During reading, spelling or phonological activities the central executive engages the system in subvocal rehearsal to maintain the phonological and articulatory representations to compare the number, order and identity of the wordsâ€™ individual phonemes. Unlike Baddeleyâ€™s (1986) model of working memory, the AAMSP postulates that the articulatory representation is held online to maintain an articulatory representation parallel to a phonological representation, both facilitating subvocal rehearsal of complex phonological input. Two separate experiments were designed to evaluate the AAMSP. Experiment one focuses on identifying differential activity patterns between the control and dyslexic groups for cortical regions involved in tone and phoneme counting, and pseudoword segmenting. Experiment two attempts to identify differences in cortical activity across
40 the groups when phoneme or tone segmentation is combined with auditory working memory demands in a same/different comparison format. Auditory Representation (STG) Auditory Input Articulatory Representations (Brocaâ€™s & SSC) Phonological Representations (SMG) Central Executive(DLPC & SMA) Figure 5-1. Articulatory Awareness Model of Speech Perception (AAMSP) The tasks for each experiment were designed to provide a hierarchy of auditory discrimination, phoneme perception, phonological segmentation, and auditory working memory combined with phonological segmentation demands or auditory discrimination of tones. Hypotheses for Experiment One-Counting and Segmenting Hypothesis 1â€“Tone Counting During Tone Counting, the participant heard a variety of series of pure tones (two, three, four, or five tones per series), but only responded with a button press when the series contained three tones. Counting pure tones was chosen as a nonspeech control task that should exhibit bilateral superior temporal gyrus activity for basic auditory perception. Binder et al. (1994, 1995, 1996, 1997) reported tone stimulus activating superior temporal gyrus among a group of normal control participants. Rumsey et al. (1992, 1994) reported no difference in patterns of cortical activity on a tone task between
41 controls and individuals with dyslexia. Therefore, hypothesis one was that both groups of participants should show equivalent activity in the bilateral superior temporal gyrus when comparing tone counting to baseline white noise, and that no between-groups differences should be present (Table 5-1). Hypothesis 2â€“Phoneme Counting During Phoneme Counting, the participant heard a series of segmented phonemes (two, three, four, or five voiced consonant phonemes per series), but only responded with a button press when the series contained three phonemes. Breier, Gray, Fletcher, Foorman, and Klaas (2002) reported that speech and acoustic stimuli activate the superior temporal gyrus for individuals with and without phonological dyslexia. However, a long-standing controversy regarding whether individuals with reading disability have a speech-specific deficit for auditory discrimination or a general auditory discrimination deficit is still unresolved (Tallal, Miller, & Fitch, 1993; Mody, Studdert-Kennedy, & Brady, 1997). Thus, there is evidence that individuals with dyslexia may have difficulty judging rapid transitions in speech, when required to make comparative judgments between syllables or phonemes. Nonetheless, the phoneme counting task in this study did not require a direct comparison of phonological features between two stimuli. Rather, phoneme counting only required the perception of phonemes, along with the attention and memory demands for counting the number of phonemes in a series and identifying the target series (3 phonemes). Deficits in the gross discrimination of a phonological versus a pure tone stimulus have not been reported for individuals with or without dyslexia. Therefore hypothesis two was that the phoneme counting tasks should result in bilateral superior temporal gyrus activity for both groups, with no between-groups differences (Table 5-1).
42 Hypothesis 3â€“Pseudoword Segmenting During Pseudoword Segmenting, the participant heard one pseudoword and should subvocally segment it into individual phonemes. The pseudowords will have two, three, four, or five phonemes, but the participant should only respond with a button press when the pseudoword has exactly three phonemes. Because Pseudoword Segmenting requires analytic phonological skills, it should be the most difficult task for the dyslexic participants and should be relatively easier for the controls. Pseudoword segmenting was designed to stimulate cortical activity associated with the parcellation of a word into its component phonemes. Control group participants should show cortical activity in the superior temporal gyrus (auditory analysis), supramarginal gyrus (phonological analysis), Brocaâ€™s area (articulatory programs), and nearby sensorimotor cortex (oral somatosensory perception) (Demont, 1992; Zatorre et al., 1992, 1996). The dyslexic group should show little or no activity in Brocaâ€™s area and nearby sensorimotor cortex, but they should exhibit an equal amount of activity as the control group in the superior temporal gyrus and the supramarginal gyrus. Overall, the Pseudoword Segmenting task should test some of the AAMSP's predicted regions of activity, regarding tasks without significant working memory demands (STG, SMG, Broca's, and SSC). Therefore, hypothesis three states that the control group will evidence significantly more activity than the dyslexic group in Brocaâ€™s area and nearby somatosensory cortex during pseudoword segmenting (Table 5-1).
43 Table 5-1. Experiment One Tasks and Hypothesized Regions of Activity per Group. TASK Regions of Cortical Activity STG SMG Brocaâ€™s SSC SMA DLPC Tone Counting C=D Phoneme Counting C=D Pseudoword Segmenting C=D C=D C>D C>D C = control group, D = dyslexic group, STG = superior temporal gyrus, SMG = supramarginal gyrus, SSC = somatosensory cortex, SMA = supplementary motor area, DLPC = dorsal lateral prefrontal cortex. Hypotheses for Experiment Twoâ€“Working Memory Comparisons Hypothesis 1â€“Tone Comparison During the tone comparison task, the participant heard one series of tones and then a second series of tones. After counting the number of tones in each series, the participant provided a button press response when the two series contained exactly the same number of tones (e.g. both series had two, three, four, or five tones). The tone comparison task was selected as a nonspeech control task. Such a control task should identify cortical regions that may be specialized for auditory working memory of a speech stimulus as opposed to an acoustic nonspeech stimulus. Dissociable working memory systems have been identified in the visual modality (Courtney, Petit, Maisog, Ungerleider, Haxby, 1998). Similarly, such a dissociable system for auditory processing has recently been proposed by Kaas and Hackett (1999). However, since the stimulus duration and interstimulus intervals were moderately low, tone comparison was expected to cause equal activity for both groups in the superior temporal gyrus (auditory analysis), supplementary motor area (working memory), and dorsolateral prefrontal cortex
44 (working memory). This task was not expected to produce a differential amount of activity between the groups. Hypothesis 2â€“Pseudoword Comparison Because the pseudoword comparison task was expected to be the most challenging activity for the dyslexic group, there were several regions of differential activity that were predicted. Brocaâ€™s area, nearby sensorimotor cortex, dorsolateral prefrontal cortex, and supplementary motor area were expected to be active for the control group. However, only activity in the superior temporal gyrus, and supramarginal gyrus were expected for the dyslexic group. The direction of the difference was hypothesized to favor the control group with greater activity by control subjects in the above stated areas. The dyslexic group participants were expected to show significantly less activity in Brocaâ€™s area, dorsolateral prefrontal cortex, and the supplementary motor area, as it has been suggested that â€œdyslexic children use speech coding in an articulatory loop just as normal children do, though the system in dyslexic children appears to operate less efficientlyâ€ (Hulme & Roodenrys, 1995, p. 384). A summary of experiment twoâ€™s tasks and the hypothesized areas of cortical activity for each group are presented in Table 5-2. Table 5-2. Experiment Two Tasks and Hypothesized Regions of Activity per Group. TASK Cortical Regions STG SMG Brocaâ€™s SSC SMA DLPC Tone Comparison C=D C=D C=D Pseudoword Comparison C=D C=D C>D C>D C>D C>D C = control group, D = dyslexic group, STG = superior temporal gyrus, SMG = supramarginal gyrus, SSC = somatosensory cortex, SMA = supplementary motor area, DLPC = dorsal lateral prefrontal cortex.
CHAPTER 6 METHODS Presented below are the methods used for experiments one and two. The final experiments are based on two pilot studies that helped identify appropriate parameters for the tasks. Specifically, the pilot studies helped identify an appropriate duration for the response period following each stimulus. Also, from pilot studies it was determined that when target pseudowords contained three phonemes, then controls achieved nearly perfect response accuracy. With three phoneme pseudoword targets, the dyslexic group performed greater than chance, but measurably less than controls. Participants and Selection Criterion Twenty-two men (11 controls, 11 dyslexic) with English as their only language participated in the study and were reimbursed for their participation. One control participant was dropped from the study after the MRI revealed an unknown mass in his temporal lobe (follow-up care was provided by the participantâ€™s physician) and one dyslexic participant was dropped after a potential history of Aspergerâ€™s Syndrome was revealed by his parents. Only males were recruited for this study, as lateralization of phonological processing may be gender specific (Shaywitz et al., 1995; Frost et al., 1999; Pugh et al., 1996; Lambe, E. K., 1999). To increase homogeneity of language lateralization, all participants were strongly right handed as measured by the Edinburgh Handedness Inventory (Oldfield, 1971; dyslexic group M = 78.76, SD = 14.05; controls M = 88.82, SD = 10.23, p > 0.07). Also, the dyslexic and control groups were matched 45
46 on age, education and general intelligence (IQ) (age t = -.067, p > .95; education t = .441, p > .66; intelligence t = 1.636, p > .12; Table 6-1). Table 6-1. Matching Variables for Control and Dyslexic Groups Control Dyslexic M SD M SD Age a 34-10 9-03 35-02 10-10 Education b 15.18 2.41 14.73 2.43 IQ c 128.82 7.87 121.45 12.68 Note. a years-months. b years. c WAIS-III two-subtest estimated IQ (Block Design and Vocabulary). Demographic data on socioeconomic status, and relevant history (medical, educational and developmental) were collected via a semi-structured interview. No significant differences were found for socio-economic status (SES; Hollingshead, 1975), paternal years of education, and maternal years of education (Table 6-2). Table 6-2. Demographic Data per Group. Control Dyslexic M SD M SD SES 47.32 * 7.80 44.36 * 6.40 Paternal Education a 14.00 ** 4.36 13.50 ** 2.35 Maternal Education a 13.27 *** 2.60 13.58 *** 0.97 Note. a years. * p > 0.34. ** p > 0.78. *** p > 0.97. Half of the dyslexic group had a positive family history as well as a personal childhood history of a learning disability diagnosis. All control participants had a negative childhood history of learning disabilities and only one participant had a positive family history of a learning disability. Participants were excluded from the study if they had a history of a speech disorder, head injury, epilepsy, substance abuse, neurological or psychiatric disorders, visual or auditory acuity impairments, were musicians, or had more than two years of musical training. Potential risks from undergoing an MRI were explained and participants were screened for MRI contraindications. Each participantâ€™s
47 informed consent was obtained pursuant to the guidelines of the University of Floridaâ€™s Health Science Center Institutional Review Board. In addition to collecting FMRI data, all participants were assessed with a neuropsychological battery of tests (see Appendix A). Tests were selected to document the presence and absence of phonological processing and reading deficits common to developmental dyslexia (Stanovich & Siegel, 1994; Torgesen, Wagner, & Rashotte, 1994), as well as measures of rapid naming, auditory working memory, and articulatory awareness (Montgomery, 1981) to clearly define the language, memory, and sensorimotor abilities of the two groups. As mentioned previously, the findings from many investigations of developmental dyslexia were compromised by poorly documented reading and phonological abilities among the participants, both control and experimental subjects. The difficulty with proper documentation of reading and language skills was partly related to the existence of multiple operational definitions of dyslexia. The various operational definitions ranged from a specific discrepancy between IQ and achievement testing, placement in a learning disabilities classroom as evidence of a reading disability, low-achievement in reading skills, to a regression-equation-based definition of a reading disability. However, convincing evidence has indicated that no matter how a reading disability was defined, either with or without IQ and achievement discrepancy, the best indicators of differences between children with reading difficulties and children without reading difficulties were measures of phonological awareness (Fletcher et al., 1994). Children who met a discrepancy definition of reading disability as well as children who had reading difficulties, but did not meet discrepancy criteria, exhibited deficient phonological and
48 orthographic skills that subserve reading (Stanovich & Siegel, 1994). Deficits in these sub-skills have persisted into adulthood (Bruck, 1993). Therefore, participants in the dyslexic group met either an IQ and achievement discrepancy definition of reading disability, or they met a low-achievement definition of dyslexia. For the dyslexic group, deficient phonological processing was operationalized with measures of estimated IQ, pseudoword decoding, and phonological awareness. Their estimated IQ (Sattler, 2001) was greater than or equal to a standard score of 90 and ranged from average to superior (Table 6-1). Phonological processing deficits were identified by a measure of phonological awareness and auditory working memory (Lindamood Auditory Conceptualization Test (LAC) (Lindamood & Lindamood, 1979). Dyslexic participants performed below the LACâ€™s adult age-expectancy, minimum expectations for children in 5 th grade. Pseudoword decoding skills, as measured by the Word Attack subtest of the Woodcock Reading Mastery Test-Revised (WRMT-R) (Woodcock, 1987), were either greater than or equal to twenty-two standard score points below the mean standard score (100) or the participantâ€™s estimated IQ. The dyslexic groupâ€™s performance on the WRMT-R Word Attack subtest was in the low average range (Table 6-3). Participants in the control group performed in the average to superior ranges on measures of estimated IQ (WAIS-III) (Sattler, 2001) and pseudoword decoding (WRMT-R). They demonstrated intact phonological awareness by performing at the adult age-expectancy on the LAC or in the average to superior ranges on phonological processing measures (CTOPP) (Torgesen, Wagner, & Rashotte, 1998) (Table 6-3).
49 Table 6-3. Variables for Defining Group Membership Control Dyslexic M SD M SD IQ ab 128.82 7.87 121.45 12.68 Woodcock-Reading Mastery Test-III Word Attack b 108.55 * 6.06 89.00 * 9.07 Lindamood Auditory Conceptualization Test-Revised c 96.64 * 3.59 82.73 * 9.74 Note. a Wisc-III Block Design and Vocabulary 2-subtest estimated IQ. b standard scores. c weighted raw scores. * p < 0.001. Apparatus and Scanning Procedures Scanning was conducted on a 3 Tesla Signa MRI scanner (General Electric Medical Systems, Milwaukee) equipped with resonant gradients, using fast gradient echo imaging (spiral). A dome-shaped quadrature RF head coil provided improved signal-to-noise ratio (especially at the cortical surface) compared to the standard GE head coil. Three-plane localizer scout images were conducted prior to functional neuroimaging to localize the participantâ€™s head and act as a reference image to prescribe operative images. Alignment was judged by visual inspection and if necessary, the participantâ€™s head was realigned and three-plane localizer images were repeated. An MR angiogram image was acquired to allow exclusion of vessel â€œactivationâ€ artifacts (all 1-3 minutes; TE = min full (8ms); TR = 40ms; flip angle = 60; NEX [number of excitations] = 1; slice thickness = 5.5 mm; FOV = 180 cm; acquisition matrix size = 256 x 128; acquisition time = 0:55 seconds). Functional images were obtained using a gradient echo two-shot spiral sequence (Noll, Cohen, Meyer, & Schneider, 1995) (TE (echo time) = 18 ms; TR (repetition time) = 1250 ms; FA (flip angle) = 70 deg; FOV (field of view) = 20 x 20 cm; matrix size = 128
50 x 128, 24 contiguous sagittal 5.5 mm thick slices covering both hemispheres). A 124-slice high-resolution T1 weighted 3D spoiled GRASS (gradient-recalled acquisitions at steady-state) image set was collected as an anatomic reference (TE = min full (8ms); TR = 23 ms; flip angle = 25; NEX = 1; slice thickness = 1.3 mm; FOV = 24 cm; reconstruction matrix size = 256 x 256; acquisition time = 9:28). In summary, participants were inside the scanner approximately 55-65 minutes, including initial scout scans, magnetic resonance angiography (MRA), functional images, and the anatomic scan. Prior to scanning the first participant, a spiral sequence was run and an audiologist measured background scanner noise in the scanner room with a portable sound field measurement device (Bruel & Kjer 2237 Controller). Peak background noise was 99.4db. Sound output from the sound delivery system (described below) was evaluated with a measuring amplifier (Bruel & Kjer type 2609) and an artificial ear (Bruel & Kjer type 4152). Binaural output for all stimulus conditions was found to be approximately 95db. Background scanner noise was constant throughout all the task and rest conditions, but foam ear inserts attenuated scanner noise by approximately 20-30db. Based on these measures, the audiologistâ€™s professional opinion, and retrospective report by participants, the sound delivery in each ear was determined to be sufficient for appropriate stimulus clarity. Prior to the last participantâ€™s scanning session, a follow-up sound assessment of scanner background noise and sound delivery levels was performed. Since all levels were equivalent to the initial measurements, the end of the experiment measurement of the background noise and sound levels documented that no changes in the auditory environment of the scanner could account for changes in functional cortical activity.
51 Prior to entering the MRI scanner, all participants received instructions and practice trials on each experimentâ€™s tasks to ensure proper performance of the desired tasks in the scanner. Because phonemes are an arbitrary construct for most laypersons, the examiner first modeled proper segmentation of real words and then had the examiner practice phoneme segmentation of real words. Next the examiner modeled proper segmentation of pseudowords and then the participant again practiced this task. Participants were specifically informed to count the sounds they heard in a word and not itâ€™s letters (as there may not be a one-to-one correspondence) and were also informed that initial consonant blends (e.g. bl, pl) and final consonant blends (e.g. gr, nd, gd, bd) were considered to be two phonemes. Next a sample run from both experiments one and two were played on a CD player (Lenoxx Sound Model CD-52) and the participant practiced performing each of the tasks for both experiments. This last practice provided the participant with a sample of the interstimulus intervals and the duration of the response periods. Participants repeated the practice trials until they demonstrated mastery of each task and voiced confidence in their ability to perform the task in the scanner. Lastly, participants completed a measure of state and trait anxiety (Spielberger, 1983) to assess their current level of state anxiety prior to entering the scanner, as well as to determine if the groups differed in their typical level of trait anxiety (anxiety that is inherent within an individual). In the scanner, participants were fitted with a sound delivery system that included right-angle horn ear tips (Etymotic ER13-14). The ends of the ear tips were embedded in occlusive expandable foam to dampen external noise and to hold the ear tips in the ear canal. The head coil was placed over the participant's head and foam rubber pads were
52 placed over the ears and around the sides of the head to stabilize the head and minimize movement. These pads provided tactile feedback to assist the participants in remaining motionless and provided secondary support for keeping the ear tips motionless. Silicon air conduction tubes (20 feet long, 4mm diameter) connected the ear tips to transducers (Etymotic Research ER-30 Earphone). A cable with one-quarter-inch jacks (Etymotic Research) connected the transducers to a penetration panel cable. A serial cable with one quarter of an inch jacks interfaced the penetration panel and the audiometer (MAICO model MA52). The audiometer maintained binaural sound output at 80db per right and left channels. A portable compact disc player (Lenoxx Sound Model CD-52) was connected to the audiometer. The experimental stimulus for all runs was played from wave files stored on compact disc media. Participants responded to a target stimulus with a hand-held button device that connected to an LED light box in the control room. Participants were instructed to keep the button device in their right hand and to completely and firmly depress the button for each target. In the control room, sound output and response accuracy were monitored during each run. Sound output was monitored with headphones (Lenoxx TM101) connected to the compact disc player. Visual monitoring of an LED light box provided data for recording response accuracy on pre-printed scoring sheets. Each participantâ€™s response accuracy was recorded during every run in a scanning session. Prior to each task, participants were reminded of the directions. Using a pseudorandom mixed block design, a run consisted of three blocks of each of three experimental tasks (described in detail below) which alternated with blocks of white noise. The signal in voxels of tissue which were active during task performance were
53 expected to rise and fall with the alternation of task performance and white noise. The order of presentation of experiment one and experiment two were counterbalanced within and between participant groups. Stimulus Parameters In experiment one, blocks of pure tones, phonemes, and pseudowords were pseudorandomly alternated with 12.5, 15, and 17.5 second blocks of white noise. Three blocks of each stimulus type (tone, phoneme, and pseudoword) and nine blocks of white noise constituted a single run. Each run began with 17.5 seconds of white noise to allow adequate time for proper estimation of baseline FMRI signal. Each participant performed four runs for a sum total of 12 blocks per stimulus type. Each run required 5 minutes and 21 seconds. The total performance time was approximately 21 minutes and 21 seconds. white noise white noise white noise tone series p honeme series p seudowords Figure 6-1. Experiment One Stimulus Blocks Alternated in Each Run In experiment two, blocks of pseudoword pairs, pure tone pairs and white noise were alternated pseudorandomly. Three blocks of each stimulus (tone pairs and pseudoword pairs) and six blocks of white noise (12.5, 15, and 17.5 second durations) occurred in each run. Each run began with 17.5 seconds of white noise to allow adequate time for proper estimation of baseline FMRI signal. Each participant performed four runs for a sum total of 12 blocks of each stimulus type. Each run took 4 minutes and 36 seconds. The total performance time was approximately 17 minutes and 44 seconds.
54 white noise white noise tone pairs Pseudoword pairs Figure 6-2. Experiment Two Stimulus Blocks Alternated in Each Run The two sets of pseudowords created for experiments one and two, although no words were shared across the experiments, were specifically designed to decrease the potential effects of variability in the difficulty of word segmentation and variability in the novelty of the pseudowords. As mentioned previously, the work of Sokolov (1972) indicated that more novel or less familiar words create more oral-sensory activity, as measured by electromyography, even with subvocal speech. Therefore, an effort was made to equalize all the pseudowords along the dimension of frequency of occurrence. This would decrease the likelihood that some runs could cause more cortical activity due to the novelty of the stimuli. Also, unequal novelty or frequency of the stimuli could relate to discrepant levels of difficulty between the pseudowords in the runs and ultimately lead to variable difficulty levels between the runs. Less frequent pairings of phonemes would mean the participant has likely had less exposure to these particular phoneme combinations. Consequently, this limited exposure or lack of familiarity could make some stimuli significantly more difficult to segment than others. Therefore, to decrease the possibility that the pseudowords created for this study had variable frequencies of occurrence and variable levels of segmentation difficulty, Robertâ€™s (1965)
55 frequency of occurrence values for each phoneme and phoneme blend were utilized to create the pseudowords. Roberts (1965) calculated the frequency of occurrence of each initial, medial, and final phoneme or phoneme blend in a corpus of 8,000 highly frequent English words. The frequency of occurrence (phoneme frequency per word of the selected corpus) provided a representation of the novelty of the phoneme or phoneme blend. Thus, specific phonemes (g, b, d, m, n, v, j, l, r, and z), as well as short vowel sounds (e, i, o, u a) and phoneme blends (initial blends-bl, br, dr, gl, gr, and final blends-nd, nz, nj, md, mz, bd, gd, gz, dz, zd, ld, jd, vd, va,) included were used to create the pseudowords. Each individual phonemeâ€™s frequency of occurrence in English and/or each phoneme blendsâ€™ frequency of occurrence in English (Roberts, 1965) was summed for each pseudoword. For experiment one, the average frequency of occurrence for all four runs of pseudowords ranged from 21.33 to 22.31 (M = 21.75, SD = 0.63). For experiment twoâ€™s working memory comparisons the average frequency of occurrence for all four runs of pseudoword pairs ranged from 20.32 to 20.66 (M = 20.5, SD = 0.18). Runs were designed such that no significant difference existed between all four runsâ€™ average frequency of occurrence values, within experiment one and two. Experiment One-Counting and Segmenting In experiment one, three behavioral tasks (pure tones, consonant phonemes, and pseudowords) were alternated with white noise. During all behavioral tasks the target item contained only three units of the specific stimulus (tones, phonemes or phonemes in a pseudoword). Foils contained 2, 4 or 5 units of the specific stimulus (Table 6-4). Within each block, targets were pseudorandomly ordered in first, second, third, fourth or fifth ordinal position in the block. The pseudorandom ordering provided an equal distribution
56 of targets across the ordinal positions of each block (M = 3.6 targets per position per block across all four runs). For each behavioral task (tone, phoneme, and pseudowords) Table 6-4. Counting and Segmenting Task Descriptions Task Description Tone Counting Sequences of two, three, four or five pure tones in a series were presented aurally and the participant responded only when the sequence contained three tones. Phoneme Counting Sequences of two, three, four, or five segmented phonemes in a series (voiced consonants) were presented aurally and the participant responded only when a sequence contained three phonemes. Pseudoword Segmenting Monosyllable pseudowords with two, three, four, or five phonemes were presented aurally and the participant responded only when the pseudoword contained three phonemes. a sum total of eighteen targets per stimulus type occurred across all four runs. The total duration of each block across all three stimulus types was 20 seconds, in order to maintain this fixed block interval the response periods after a stimulus ranged from 2.5 to 3.5 seconds (minimally longer response periods after a long stimulus series and shorter response periods after a shorter stimulus series, but based on pilot study findings the response times allowed adequate time for participantsâ€™ responses). The experiments were counterbalanced within each group, such that half of the control and dyslexic group performed experiment one first and the other half performed experiment two first. Likewise, the order of the four runs in an experiment was counterbalanced, such that the order of the runs was equally distributed within and between each group. Participants were instructed to remain motionless and relaxed during the white noise and not to think about task stimulus. Participants were instructed to maintain a visual fixation during all tasks or to close their eyes. All stimulus types (tone, phoneme and pseudoword) were created with equal amplitudes to reduce the possibility of variations in cortical activity due to variations in the amplitude of the specific stimulus (Bilecen et al., 2002).
57 Tone counting Pure tone stimuli included a series of 500, 750 and 1000 Hz tones, digitally sampled at 44.1 kHz with a duration of 200 milliseconds. Tones were generated with Cool Edit 2000 software (Syntrillium Software Corporation, 2000). Each series of tones contained two, three, four or five tones that were separated by 250 msec of silence. Five series of tones constituted a block and each block was followed by a silent response period. The blocks in each run contained a pseudorandom number of target (3 tones) and non-target series (2, 4, or 5 tones). The average number of tones per block for runs one through four was 2.93, 3.07, 2.93, and 3.07, respectively. Across all four runs, the average number of tones in a block was 3, decreasing the possibility that changes in cortical activity would be due to variations in the number of stimulus across the runs. Also, the average number of 500, 750 and 1000 Hz tones was consistent across all blocks. To prevent masking by the white noise, each block of pure tone stimuli was preceded by 250 milliseconds of silence. The hypothesized cognitive components of tone counting are presented in Table 6-5. Table 6-5. Counting and Segmenting Cognitive Components per Tasks. Task Cognitive Components Tone Counting basic auditory discrimination Phoneme Counting basic auditory discrimination + phoneme perception Pseudoword Segmenting basic auditory discrimination + phoneme perception + segmentation Phoneme counting Stimuli for the phoneme counting task included random combinations of three digitally sampled male speech sounds (voiced consonant phonemes /g/, /b/, and /d/). All phonemes were produced by a male speaker of a Midwestern dialect of American English. Using Cool Edit 2000 (Syntrillium Software Corporation, 2000) the phonemes
58 were normalized and compressed to equalize their amplitudes, limiting the possibility that changes in cortical activity would occur due to variations in stimulus amplitude. Identical to the tones task, each series of phonemes included either 2, 3, 4, or 5 phonemes, such as /d/ /d/ /d/ /g/ /g/. Five different series of phonemes constituted a block (e.g. a 5 phoneme series, may be followed by a 2 phoneme, 3 phoneme, 4 phoneme, and another 2 phoneme series). Each block of phoneme stimuli was preceded by silence (250 msec) and followed by a period of white noise (12.5 seconds, 15 seconds, or 17.5 seconds in duration). Total duration for each block was consistent at 20 seconds. The hypothesized cognitive components of phoneme counting are presented in Table 6-5. Pseudoword segmenting Sixty monosyllable pseudowords were constructed from a subset of English phonemes. Only voiced consonants and vowels that had a one-to-one grapheme to phoneme relationship were included. All pseudowords were produced by a male speaker of a midwestern dialect of American English. Using Cool Edit 2000 (Syntrillium Software Corporation, 2000), pseudowords were digitally sampled at 44.1 kHz, normalized and compressed to equalize their amplitudes. The pseudowords contained 2, 3, 4, or 5 phonemes in one of the following consonant and vowel combinations CV/VC, CVC, CCV/VCC, CVCC/CCVC or CCVCC. The participants heard one pseudoword at a time and were asked to press a button when the pseudoword only contained 3 phonemes (CVC, CCV or VCC). The hypothesized cognitive components of pseudoword segmenting were presented in Table 6-5. Experiment Twoâ€“Working Memory Comparisons During working memory comparisons, two explicit behavioral tasks (auditory working memory comparisons of pure tones or pseudowords) were contrasted with
59 baseline white noise. Targets contained two sets of stimuli (either two series of segmented tones or two pseudowords) with an equal number of units (tones or phonemes) in each pair of stimuli. For example, during tone comparison a target contained two series of tones wherein both series contained an identical number of tones, either two, three, four, or five tones. Foils included mismatched pairings of the possible tone series, e.g. a series of two tones paired with a series of three tones. The total duration of each stimulus block was fixed at 30 seconds. To maintain this fixed block duration, the response time that followed a stimulus tone series pair or pseudoword pair ranged from 2.4 to 3.7 seconds, with longer response periods for pseudoword comparison and shorter response periods for tone comparison; based on pilot study findings. Participants were instructed to remain motionless and relaxed, and not to think about the experimental task stimulus during the white noise. They were instructed to maintain a visual fixation during all tasks or to close their eyes. A description of each task in experiment two is presented in Table 6-6 and the hypothesized cognitive components per task are displayed in Table 6-7. Table 6-6. Working Memory Comparison Description of Tasks. Task Description Tone Comparison Pairs of two, three, four or five pure tones in a series were presented aurally. Participants responded only when both series in a pair had the same number of pure tones (either two, three, four or five). Pseudoword Comparison Pairs of monosyllable pseudowords with two, three, four or five phonemes per pseudoword were presented aurally. Participants responded only when both pseudowords in a pair had the same number of phonemes (either two, three, four or five). Table 6-7. Working Memory Comparison Cognitive Components per Task. Task Cognitive Components Tone Comparison auditory discrimination + tone perception + auditory working memory Pseudoword Comparison auditory discrimination + phoneme perception + phoneme segmentation + auditory working memory
60 Tone comparison Two series of tones (either 2, 3, 4, or 5 tones per series) were separated by 1000 msec of silence. Each series of tones was made of combinations of 500, 750, and 1000 Hz tones, digitally sampled at 44.1 kHz with equal maximum amplitudes of 12 db. The pairs of tone series were followed by a silent response period that ranged from 2.4 to 2.7 seconds in duration. The participant pressed the response key only when the two series had the exact same number of tones. Pseudoword comparison Stimuli for the pseudoword comparison task were pairs of monosyllable pseudowords; no pseudowords were shared across the two experiments. One hundred and twenty pseudowords were digitally sampled at 44.1 kHz, normalized and compressed to equalize their amplitudes (Cool Edit 2000 Syntrillium Software, 2000). A male speaker with a midwestern dialect of American English produced all pseudowords. The pseudowords were comprised of two, three, four or five phonemes (consonants (C) and vowels (V) in the following combinations: CV, VC, CVC, CCV, VCC, CVCC, CCVC, or CCVCC). For each pseudoword pair, participants heard two unique pseudowords that contained similar or different numbers of phonemes. The pseudowords in a pairing were separated by 1000 milliseconds of silence. Participants were instructed to respond by a button press if the two pseudowords had the exact same number of phonemes, either two, three, four, or five. Therefore, after each pair of syllables a pause of 3300 to 3700 milliseconds occurred to provide the participants with enough time to decide whether the pair met target criteria and to make a button press if a target was detected.
61 Functional Neuroimage Analysis Identical procedures were used for analyzing data in experiment one and experiment two. Data were analyzed using the latest version of the Analysis of Functional Neuroimages (AFNI)(Cox, 1996) software package. Each image was aligned using a 3D rigid-body registration method and realigned with a base image (spiral image closest in time to the anatomic images) to reduce in-plane (sagittal) movement artifact. For each participant, the four FMRI runs were reordered to the same sequence, concatenated, and the linear trends of the time series were removed. Deconvolution analysis was used to estimate the impulse response function for each condition (Cox, 1996) with a maxlag of 16 images (40 seconds) to allow the hemodynamic response to relax completely. Deconvolution uses all trials of a task to estimate a single hemodynamic response function for each task in each voxel of an individual subjectâ€™s functional neuroimages. The area under the curve (AUC) of each estimated hemodynamic response function (HRF) was calculated on a voxel by voxel basis for each task stimulus. The AUC value quantified BOLD signal change for each block. Voxels in which the standard deviation of the signal change exceeded 5% of the mean signal were set to zero to remove â€œfalse activationâ€ caused by residual motion, other nuisance artifact, or random signal fluctuation. For each individual participant, the reconstructed FMRI anatomic data sets and structural images were interpolated to 1mm 3 voxels, co-registered, and converted into stereotactic coordinate space (Talairach & Tournoux, 1988). For each participant group, a merged anatomic dataset was created by blurring and combining the individual anatomic datasets. Individual subject functional images for each task were â€œblurredâ€ using a 5mm
62 full-width half-maximum isotropic Gaussian filter to compensate for interparticipant variability in structural and functional anatomy. For within-group analyses, voxel-wise repeated measures t-tests were conducted comparing each experimental task to baseline (white noise), as well as the appropriate comparison task with the normalized AUC dependent variable; AUC was divided by the mean intensity of the entire time series to produce a normalized AUC. For between-groups analyses (dyslexic versus controls), voxel-wise unpaired t-tests were conducted comparing the AUC variable between the two groups across each stimulus type. These t-tests generated a t-value statistical parametric maps (SPM) for functional overlay. These SPMâ€™s were thresholded at t-values corresponding to p-values of 0.005 and underwent cluster analyses. Voxels were selected for interpretation only if they were equal or greater than the minimum t-value limit for the specific comparison, maintained a cluster connection radius of 1.8mm and formed clusters with a minimum volume threshold of > 150 microliters. In experiment one, comparisons across a stimulus type were performed both within and between-groups. Separate comparisons of an individual stimulus type to baseline were performed within each group to verify that significant activity was present for each type of stimulus (tone counting, phoneme counting and pseudoword segmenting). Between-groups (dyslexic versus controls) independent contrasts in experiment one included: (1) tone counting, (2) phoneme counting and (3) pseudoword segmenting. In experiment two, comparisons across a stimulus type were performed both within and between-groups (dyslexic participants versus controls). Separate comparisons of an individual stimulus type with baseline white noise were performed to verify that
63 significant activity was present for each type of stimulus (tone comparison and pseudoword comparison). Between-groups independent contrasts included: (1) pseudoword comparison (PC) and (2) tone comparison (TC).
CHAPTER 7 RESULTS Behavioral Results As reported in the Methods section, no differences were evident between-groups on independent variables of age, education, handedness, SES, and estimated intelligence. Similarly, no differences existed between the groups in their level of anxiety prior to entering the scanner, as well as the participantsâ€™ general level of anxiety on any given day (Table 7-1.). To determine the presence of any other cognitive differences between the groups, neuropsychological assessment findings were collected for the dyslexic group and the control group. Studentâ€™s t-tests were conducted to identify significant differences on neuropsychological measures and performance accuracy on the various tasks used for functional neuroimaging. Overall, these measures provided quantitative documentation of the abilities that significantly differed between the two groups (phonological processing, auditory working memory, reading achievement, reading fluency). Table 7-1. State and Trait Anxiety Levels Prior to FMRI Control Dyslexic M SD M SD State Anxiety 41.6 * 6.58 45.5 * 8.54 Trait Anxiety 45.3 ** 8.37 48.6 ** 13.10 * p > 0.2. **p > 0.5. The control group performed significantly better than the dyslexic group on almost all neuropsychological assessments, despite no significant difference between the groups on IQ. On the dependent measures of pseudoword decoding and phonological awareness, the dyslexic group performed significantly worse than the controls, word attack (t = 64
65 6.330, p < 0.001) and phonological awareness (LAC t = 4.936, p < 0.001). Additional measures of phonological processing, auditory working memory, word identification, and passage reading also indicated significant differences between the groups, with the control group performing better than the dyslexic group on all significant differences. Overall, the dyslexic group evidenced significant impairments in their decoding skills (accuracy and fluency) (Table 7-2), additional measures of phonological processing (Tables 7-3 and 7-4), and auditory working memory (Table 7-5), when compared to the control group. However, no differences existed between the groups on two measures of rapid naming (Table 7-6), Therefore, the dyslexic group evidenced deficient phonological processing as their primary deficit, additional deficits in auditory working memory, but they did not meet criteria for a double-deficit type of reading difficulty (deficient reading skills and impaired rapid naming abilities; Wolf & Bowers, 1999). Table 7-2. Reading skills per Group Control Dyslexic M SD M SD Gray Oral Reading Test-III a Rate 13.63 ** 1.86 8.00 ** 5.02 Accuracy 14.91 * 1.14 7.18 * 4.99 Passage Score 14.73 * 1.56 7.36 * 4.78 WRMT-R b Word Attack 108.55 * 6.06 89 * 9.07 Word Identification 109.55 * 7.35 91.82 * 11.02 Note. WRMT-R = Woodcock Reading Mastery Test Revised. a scaled score M = 10 SD = 3. b standard score M = 100 SD = 15. *p < 0.001. **p < 0.002.
66 Table 7-3. Phonological Processing Abilities per Groupâ€“part 1 Control Dyslexic M SD M SD Comprehensive Test of Phonological Processing CTOPP a Nonword Repetition 11.64 * 1.4 7.64 * 1.6 Elision 10.73 ** 1.3 7.36 ** 3.6 Phoneme Reversal 13 * 2.4 7.82 * 2.3 Segmenting Nonwords 11 *** 1.8 8.18 *** 3 Note. a standard scores M = 10 SD = 3. *p < 0.001. **p < 0.009. ***p < 0.16. Table 7-4. Phonological Processing Abilities per Groupâ€“part 2 Control Dyslexic M SD M SD LAC-MAC ab Level 1â€“counting syllables 87.27 ** 14.21 70 ** 18.97 Level 2â€“syllable changes 92.73 7.86 89.09 9.44 Level 3â€“syllable and phoneme changes 90 * 11.27 62.91 * 16.54 Note. a Lindamood Auditory Conceptualization Multisyllable Auditory Conceptualization â€“Research Version. b weighted raw scores. *p < 0.001. **p < 0.025. Table 7-5. Auditory Working Memory and Attention Abilities per Group Control Dyslexic M SD M SD Auditory Consonant Triagrams a 0 delay 100 ** 0 92.73 ** 10.09 3 sec delay 85.45 12.93 65.45 18.09 9 sec delay 69.09 * 28.79 47.27 * 24.12 18 sec delay 52.73 *** 25.73 30.91 *** 18.68 Note. a percent correct. *p < 0.007. **p < 0.027. ***p < 0.034.
67 Table 7-6. Rapid Naming Abilities per Group. Control Dyslexic M SD M SD Comprehensive Test of Phonological Processing CTOPP a Rapid Letter Naming 10.36 2.84 9.82 3.40 Rapid Color Naming 11.00 2.50 9.64 2.30 Note. a standard scores M = 10 SD = 3. Behavioral measures recorded during scanning sessions included each participantâ€™s response accuracy on the experimental tasks. On experiment one (tone counting, phoneme counting, pseudoword segmenting) the control groupâ€™s percent of correct responses was 99, 98 and 88, respectively (Table 7-7). The dyslexic group demonstrated a very similar mean percentage correct for each task (93, 95, and 77) (Table 7-7). However, there was a significant difference between the groupâ€™s performance on tone counting and pseudoword segmenting. Table 7-7. Experiment One Response Accuracy per Group. Control Dyslexic M SD M SD Tone counting 99 * 4.21 93 * 14.15 Phoneme counting 98 4.77 95 11.59 Pseudoword segmenting 88.64 ** 19.32 77.88 ** 21.51 Note. * p < 0.008. **p < 0.016. On experiment two (tone comparison, pseudoword comparison) the control group achieved 99 and 64 percent accuracy, respectively (Table 7-8). Similarly, the dyslexic group showed 91 percent accuracy on tone comparison and 56 percent accuracy on pseudoword comparison. The greater than chance accuracy scores indicated that both groups of participants made valid efforts to perform the tasks used in scanning. Thus,
68 changes in cortical activity were attributed to participants performing the specified tasks. However, it is noteworthy that there was a reliable difference between the groups on accuracy of tone comparison, most likely attributable to the large variance in performance by the dyslexic group. Table 7-8. Experiment Two Response Accuracy per Group. Control Dyslexic M SD M SD Tone Comparison 99.09 * 4.21 91.36 * 19.95 Pseudoword Comparison 64.55 21.07 56.82 21.1 Note. *p < 0.014. While greater than chance accuracy scores on all tasks were important for both the control and dyslexic group, differential accuracy scores between the groups on two of the tasks were also targeted. The Pseudoword Segmenting and Pseudoword Comparing tasks were designed to ensure that the control group could perform them with nearly perfect accuracy. In comparison, the dyslexic group was expected to perform above chance and well below the control group on both pseudoword tasks, given that deficient phonological segmenting was the hallmark of their reading difficulty. On the remaining tasks (tone counting, phoneme counting and tone comparison), the control and dyslexic groups were expected to achieve similar levels of accuracy. Surprisingly, response accuracies were significantly different between the two groups on all stimuli, except experiment oneâ€™s Phoneme Counting and experiment twoâ€™s pseudoword comparison. However, the similar response accuracies for both groups on Phoneme Counting and pseudoword comparison provides interesting considerations when discussing the differential cortical activity between the groups on tasks; this issue is addressed further in the Discussion section.
69 FMRI Results All within and between-groups comparisons of FMRI data were performed with cluster analyses based on studentâ€™s t-tests. A statistical probability threshold of p < 0.005 was used to identify regions with significant activity changes, either between or within-groups. In addition, a contiguity threshold of a volume > 150L was applied to each of the clusters. The FMRI results were reported separately for each experiment (experiment oneâ€“counting and segmenting, experiment twoâ€“working memory comparison). For experiment one, the results were broadly categorized by cortical region (temporal, frontal, parietal) to relate the findings to the AAMSP. Within each cortical region the results were subdivided by within-groups and between-groups comparisons to identify general patterns of differential activity and to highlight significant differences between the control and dyslexic groups. In experiment one all within-groups comparisons were between active tasks (tone, phoneme, and pseudoword) and baseline white noise. Likewise, in experiment two all within-groups comparisons were between active tasks (tone comparison and pseudoword comparison) and baseline white noise. However, many more regions of cortical acstivity were identified across the groups and the stimulus types in experiment two, which indicated significantly better sensitivity and power than experiment one (tone comparison and pseudoword comparison). Therefore, the results for experiment two were grouped by stimulu type (tone comparison or pseudoword comparison) and then subgrouped according to the region of activity (temporal, frontal or parietal) and the type of analysis (within or between-groups). Experiment Oneâ€“Tone and Phoneme Counting, and Pseudoword Segmenting Because FMRI is dependent on cognitive functioning, it is important to consider the participantâ€™s strategy and accuracy on each task. Despite being trained on and
70 demonstrating mastery of each task immediately prior to scanning, it was possible that a participant did not follow the directed strategy or abandoned one strategy for another during the scanning. Such changes in strategy or the use of a strategy that does not match the experimentâ€™s design would directly affect the cortical activity of each participant. Also, if the participantâ€™s accuracy was below the level of chance, then the validity of their performance was compromised and the ensuing cortical activity was confounded by poor effort. Again, such compromises could directly affect the cortical activity in an experiment. Therefore, the accuracy of performance and reported strategies will be described prior to functional activity results, for each task. In experiment one, the three tasks required the participant to count the number of segmented pure tones, segmented phonemes (voiced consonants) or phonemes in a pseudoword. Each segmented series or pseudoword contained either 2, 3, 4, or 5 tones or phonemes, while only stimuli with three tones or phonemes (either segmented or in a pseudoword) were targets. After completing the scanning session, each participant completed a debriefing interview and identified the particular strategy or strategies used for each scanning task. In experiment one, participantsâ€™ strategies were categorized as either auditory (counting items â€œin my headâ€), visual (mentally imaging graphemes or hand gestures, or looking at bumps on magnetâ€™s bore), or combined (more than one approach used). Largely, the control group reported only using the auditory approach for all three tasks. The dyslexic group reported only using the auditory approach for the tone and phoneme segmenting tasks. For the pseudoword segmenting, 81% of the dyslexic group reported primarily using an auditory approach and the remaining 19% used a combined auditory analysis and visual imagery (mentally picturing an object to assist
71 with counting, like a number or hand) approach. Overall, both the control and dyslexic groups were primarily subvocally counting the tones or phonemes in all tasks without moving their mouths. Temporal region Control group. The tone counting and pseudoword segmenting tasks produced no clusters that met volume and probability value thresholds. However, phoneme counting resulted in a large region of activity in the right superior temporal gyrus (Brodmannâ€™s areas 22, 21, and 42) and a smaller region in the left superior temporal gyrus (Brodmannâ€™s areas 22, 42, and 41). To summarize, bilateral superior temporal lobe activity was noted on phoneme counting for the control group. Dyslexic group. Tone counting lead to no clusters that met volume and probability value thresholds. Phoneme counting yielded activity in the left superior temporal gyrus (Brodmannâ€™s areas 42 and 22). Unlike the control group, pseudoword segmenting generated active clusters in the left superior temporal gyrus (Brodmannâ€™s areas 42 and 22). Overall, left superior temporal gyrus activity was produced during phoneme counting and pseudoword segmenting for the dyslexic group (Table 7-9). In contrast, the control group evidenced bilateral superior temporal gyrus activity during phoneme counting.
72 Table 7-9. Active Temporal Regions from Within-groups Analyses Phoneme Counting Versus Baseline White Noise Group Region of Vol Maximum Interest L Anatomic Areas Local Max t test value Control R-STG 584 BA 22, 21, 42 (58, -26, 4) 5.44 L-STG 227 BA 22, 42, 41 (-51, -20, 5) 5.05 Dyslexic L-STG 254 BA 42, 22 (-54, -20, 6) 7.12 Pseudoword Segmenting Versus Baseline White Noise Dyslexic L-STG 253 BA 42, 22 (-55, -19, 6) 5.01 Note. BA = Brodmann's Area. STG = superior temporal gyrus. L = left. R = right. Volumes of activity change >150 L and p < 0.005. Between-groups analyses. Comparisons between control and dyslexic groups were performed for each task (tone counting, phoneme counting, or pseudoword segmenting). Only the tone counting produced one region with differential activity between the groups. The left superior temporal gyrus (BA 42) showed significantly greater activity for the dyslexic group than the control group (Figure 7-1). In the temporal region, no clusters met volume and probability value thresholds for the between-groups analyses on the phoneme counting and pseudoword segmenting. In summary, the two groups only differed reliably when counting segmented pure tones. This between-groups difference is interesting in light of no within-groups differences on the pure tones task. The between-groups difference implies that the control group may have exhibited decreased activity in comparison to white noise. In contrast, the dyslexic group may have evidenced an increase in activity compared to white noise. Therefore, when comparing the two groups together, the different directions of each groupsâ€™ activity (increased versus
73 decreased) to the pure tones produced a reliable difference between the groups in the superior temporal gyrus (BA 42, x = -50, y = -31, z = 10, volume = 158 L, maximum t value = 4.555) (see Figure 7-1). Figure 7-1. Left Temporal Region Activity for Dyslexic Group Greater than Control Group during Tone Counting (p < 0.005, cluster > 150L, radiologic images). Frontal region Control group. Separate comparisons of each task (tone counting, phoneme counting, pseudoword segmenting) versus baseline white noise were performed. No clusters met volume and probability value thresholds in the frontal region. Dyslexic group. Comparing pseudoword segmenting with baseline white noise found an active region in the left middle and inferior frontal gyri (Brodmannâ€™s areas 9 and 44). This cluster met the minimal volume threshold of exactly 150 microliters (x = -42, y = 13, z = 28, volume = 150 L, maximum intensity = 4.764). For the tone and phoneme counting tasks no clusters met volume and probability value thresholds. Between-groups analyses. Between-groups analyses were performed for the tone counting, phoneme counting, and pseudoword segmenting tasks to identify differential activity in the frontal region. However, no clusters with differential activity between the two groups in the frontal regions met minimum probability value and volume thresholds groups.
74 Parietal region. W ithin-groups analyses. For both the control and dyslexic groups, no activity in was evident on within-groups analyses. Despite the lack of activity from wity superior and inferior parietal regions (BA 40, 7 and 39) than the dysleFigure 7-2. Left Parietal Region Activity for Control Group greater than Dyslexic Group during Pseudoword Segmenting (p < 0.005, cluster > 150L, radiologic images). Similar to experiment one, the tasks in experiment two required the participant to count the number of pure tones in two separate series of tones or count the number of phonemes in each of two pseudowords. A target stimulus contained a pair of stimuli the parietal region ithin-groups analyses, with the prescribed volume and probability thresholds, the between-groups analyses demonstrated one region with a significant difference in activbetween the two groups. Between-groups analyses. On pseudoword segmenting, the controls exhibited greater activity in the left xic group (see Figure 7-2) (x = -44, y = -53, z = 54, volume = 176L, maximum intensity = 4.555). However, for the tone and phoneme counting tasks there were no activity differences in the parietal region between the control and dyslexic groups that met volume and probability value thresholds. Experiment Two-Tone Comparison and Pseudoword Comparison
75 (either two series of tones or two pseudowords) with an equal number o f units (tones or phonepant y ent arily mes) in both stimuli. Thus each series of tones or pseudowords in a pair could contain 2, 3, 4, or 5 tones or phonemes. However, to assess whether participants were following directions in the scanner, after completing the scanning session each particicompleted a debriefing interview and identified the particular strategy or strategies used for each scanning task. In experiment two, participantsâ€™ strategies were categorized aseither auditory (counting items â€œin my headâ€), visual (mentally imaging graphemes or hand gestures, or looking at bumps on magnetâ€™s bore), or combined (more than one approach used). One hundred percent of the controls reported using the auditory approachfor the tone comparison task and 82 percent used an auditory approach for the pseudoword comparison task; the remaining 18 percent used a visual approach (mentallpicturing an object to assist with counting, like a number or hand). One hundred percof the dyslexic group reported using only the auditory approach for the tone comparison task. For the pseudoword comparison, 91% of the dyslexic group reported primusing an auditory approach; the remaining 9% used a combined auditory and visual imagery approach. Overall, both the control and dyslexic groups reported using primarilya strategy of subvocalizing the stimuli without moving their mouths as they counted the tones or phonemes in the tasks on experiment two. Experiment two produced many more regions of cortical activity on within andbetween-groups analyses for each type of stimulus than did experiment one, even though the same p-value (p < 0.005) and volumetric thresholds (> 150 L) were utilized for the analyses in both experiments. The greater number of regions of activity and the larger volum of e of each cluster appeared to indicate greater power and sensitivity in the design
76 experiment two than experiment one. Therefore, for experiment two the cortical activity was grouped and reported by the task (tone comparison or pseudoword comparison) anthen subgrouped and reported by the region of activity (temporal, frontal, and parietal lobe) and the type of analysis (within-groups or between-groups). Tone comparison Temporal region within-groups analyses. For the control group, the tone comparison task yielded bilateral activity in the superior temporal g d yrus (BA 41, 21, 22, also extended into the temporal pole (BA 38) in the right hemisp t In and en rol group in the and 42). One cluster here. Similarly, the dyslexic group evidenced bilateral activity in the superiortemporal gyrus (BA 22, 41, and 20) on the tone comparison task. In contrast to the controlsâ€™ bilateral activity, the dyslexic group exhibited activity in only the righhemisphere for Brodmannâ€™s area 42 (superior temporal gyrus). Likewise, the dyslexicgroup exhibited activity in the left temporal pole (BA 38), opposite of the controls. summary, both groups evidenced predominately bilateral superior temporal gyrusunilateral temporal pole activity during the tone comparison task (Table 7-10). Temporal region between-groups analyses. Comparing activity patterns betwethe control and dyslexic groups identified differential activity in the superior temporal region. The dyslexic group evidenced significantly greater activity than the cont left superior temporal lobe (BA 22, 42). For all analyses of the tone comparison task that yielded clusters of activity in the temporal region, the Talairach coordinates ofthe voxel of maximum intensity in each cluster, the clustersâ€™ volume in microliters and the Broadmann region(s) of each cluster(s) are reported in Table 7-11.
77 Table 7-10. Temporal Region Activity per Group for Tone Comparison versus Baseline White Noise. Maximum GroupRegion of Vol t-value Interest L Anatomic Areas Local Max for cluster Control TT G, Ins, G, MTGA 41, 221 43, -18, 6) STG, TTG, MTG 1930BA 22, 41, 42, 21 (52, -20, 4) 7.12 G( TTG, STG, IPL 321 BA 41, 22, 42, 40 (-44, -34, 17) 5.51 STG 228 BA 22, 38 (49, 4, -5) 5.41 yslexic STG, TTG, IPL 5244BA 22, 41, 20, 38, 40(-56, -35, 15) 8.65 Note. BA = Brodmann's Area. Ins = insula. IPL = inferior parietal lobe. MTG = middle temporal gyrus. ST 2345 B , 2 (8.04 STG, MT 858 BA 42, 22, 21 56, -22, 14) 7.17 D STG 4814 BA 22, 42, 41, 20 (61, -23, 12) 8.22 S TG = supeporalyrus. TTG = tyrus. L = leftlumes ge > 150 L and p < 0.00ble 7-1poraegion AcGretrolp g To Comparis rior tem g ransve rse temporal g . R = right. Vo of Activity cha n 5. Ta 1. Temdurin l Rne tivity fon. or Dyslexic Group ater than Con Grou Vol Maximum -STG, L-IPL 389 BA 2 Region L Anatomic Areas Local Max t test value 2, 42, 40 (-57, -39, 20) 4.84 L L-STG 349 BA 42 (-57, -26, 10) 5.65 rodmann's Area. IPL = inferior pa Note. BA = Brietal lobe. L = left. STG = superior temporal gy rus. Volus of activity change >150 L d p < 0.005. oal regionin-For toup, thsotask prodacte bilateral dorsal premotor region (BA 6) (Table one comparison task yielded several areas vity in me an Fr nt with groups analyses. he control gr e tone compari n uced ivity in th 7-12 and Figure 7-3). For the dyslexic group, the t of activity in the frontal region. The dyslexic group exhibited bilateral actithe inferior frontal region (BA 44), left middle frontal gyrus (BA 6), left precentral gyrus(BA 4 and 8), and the right prefrontal cortex (BA 9) (Table 7-12 and Figure 7-4).
78 Frontal region between-groups analyses. Between-groups analyses were performed for the tone comparison task to identify differential activity in the front al regiond white noise. Vol Anatomic Maximum al regiond white noise. Vol Anatomic Maximum . However, no frontal regions of differential activity, between the control andyslexic groups, were evident that met the selected minimum probability value and volume thresholds. Table 7-12. Frontal Region Activity per Group for Tone Comparison versus baseline . However, no frontal regions of differential activity, between the control andyslexic groups, were evident that met the selected minimum probability value and volume thresholds. Table 7-12. Frontal Region Activity per Group for Tone Comparison versus baseline Group Region L Areas Local Max t-value Dyslexic L-PrCG, MFG B(-48, -3, 52) FG BA-IFG, MFG 542 A 47, 44 (46, 19, -3) 6.72 Control Note. BA =dFG al s. IFG = irus. 1233 A 6, 4 5.35 L-IFG, M 607 44, 8 (-45, 9, 29) 5.32 R B R-MFG, IFG 462 BA 9, 44 (48, 22, 28) 5.34 L-PrCG, MFG 180 BA 6, 4 (-38, 1, 33) 4.75 R/L-dFG, SFG 560 BA 6 (3, 7, 57) 5.36 Brodmann's Area. = dors frontal gyru nferior frontal gy L = left. MF gyruG l gyrus. Res of ae > 150 L and p < 0.00 Figure 7-3. Frontal Region (dorsal premotor) Activity for the Control Group During Tone Comparison versus Baseline White Noise (p < 0.005, cluster > 150L, radiologic image). G = middle frontal s. PrC = pre-centra = right. Volum activity ch ng 5.
79 Figure 7-4. Frontal Region Activity for the Dyslexic Group During Tone Comparison versus Baseline White Noise (p < 0.005, cluster > 150L, radiologic im age). Parietal region within-groups analyses. The control group evidenced inferior parietal lobe activity (BA 40) the tone comparison task. The dyslexic group showed activity in the right parietal region (BA 7, 39, and 40) and the left inferior parietal region (BTable 7-13. Parietal Region AcWhite Noise. Maximum in the left hemisphere during A 40) for the tone comparison task (Table 7-13). tivity per Group for Tone Comparison Versus Baseline Vol L Anatomic Areas Local Max 321 BA 41, 22, 42, 40 (-44, -34, 17) Group Region t value Control L-IPL 5.51 L R-SPL, IPL 153 BA 7, 40, 39 (36, -52, 50) 4.87 Note. BA Dyslexic -IPL 5244 BA 22, 41, 20, 38, 40(-56, -35, 15) 8.65 = Brodmann's Area. IPL = inferior parietal lobe. L = left. R = right. SPL = superior parietal lobe. Volumes of activity change >150 .L and p < 0.005. Parietal region between-groups analyses. Significantly greater activity was evident for the dyslexic group than the control group in the left superior parietal lobe (BA 40) during tone comparison. The voxel with maximum intensity was located at X = -57, Y = -39, and Z = 20, with a maximum t-value of 4.836 and a volume of 389 L (see Figure 7 5).
80 Figure 7-5. Significantly Different Parietal Region Activity (BA 40) for Dyslexic Grogre15 up ater than Control Group During Tone Comparison (p < 0.005, cluster > 0L, radiologic image). A composite mity from the within and between-groups analyses is presented in Figure shows the activity that was specific to each group (yellow for dyslexic acortical activity thdyslexic and control subjects are displayed in green. Lastly, based on the between-groups analyses, only the regions that were significantly more activk) were represented in red. ovie of the regions of active 7-6. This 3-dimensional movind blue for controls). Also, regions of at overlapped between the e (dyslexics were greater than controls for this tas
81 F igure7-6. Movie of Whole Brain Activity per Group During Tone Comparison (red = significant between-groups difference, yellow = dyslexic, blue = control, green = overlap, p < 0.005, cluster > 150L, left to right). seudoword comparison versus baseline white Noise Temporal region wworking memory comior temporal gyrus (BA 22, 41, 21, and 42) (see Tactivity in the superioisphere, the dyslexic groupâ€™s activity exe superior temporal gyrus (BA 41) (Table 7-14). Table 7-14. Temparison Versus P ithin-group analyses. For the control group, the pseudoword parison task revealed bilateral activity in the superable 7-14). Similarly, the dyslexic group exhibited r temporal gyrus bilaterally (BA 22 and 42). In the left hemtended into another area of thporal Region Activity per Group for Pseudoword Com Baseline White Noise. Vol Anatomic Maximum egion L Areas Local Ma Group Rx t test value Control R-STG, TTG, MTG 2605BA 22, 41, 42, 21 (49, -4, -4) 7.17 L-STG, TTG, MTG 2288BA 22, 41, 21, 42(-62, -18, 6) 7.56 Dysle L-STG, TTG, (IPL) 3929BA 22, 42, 41, 40(-57, -39, 21) 8.01 xic R-STG, (PoCG) 3092BA 42, 22, 40 (55, -14, 9) 7.12 Note. BA = Brodmann's area. IPL = inferior parietal lobe. L = left. MTG = middle temporal gyrus. PoCG = post-central cingulate gyrus. R = right. STG = superior temporal gyrus. TTG = transverse temporal gyrus. Volumes of activity change >150 L and p < 0.005.
82 Temporal region between-groups analyses. On the pseudoword comparison tasthe dyslexic group show k ed greater activity in the left superior temporal lobe (BA 42, 22) 20) (Figure 7-7). However, from the between-groups comparison it is difficult to know the groups have differential activity. For example, one of the following three scenariosae 3. Bovih wasig ttt. on and the other group xhibited increased activity or no activity. The hemodynamic response function (HRF), a graphical depiction of the time course of activity changes, can show where and when activity, deactivation, or no activity occurs. Figure 7-7. Significant Temporal Activity (red) for Dyslexic Group (yellow) greater than cluster > 150L, radiologic image). The HRF was derived for this specific region of the superior temporal gyrus by creating a mask from the between-groups analysis (which identified left BA 42 and 22 as a region of significantly different activity between these two groups, red region in Figure than the control group (volume = 295 L, maximum t-value = 4.93, x = -58, y = -38, z = just why could cause a signific nt diff rence in activity. t h groups showed actithan ty in t is region, but activhe ity of one group s n ificantly greater hat of other. 4. One group produced activity in this cortical region and the other group did no 5. One group produced deactivation in this cortical regi e Control Group (blue) During Tone Comparison (overlap = green, p < 0.005,
83 7-7). Next this mask file was applied to all subjects in both groups and an average HRF for each subject was calculated. Lastly, the individual HRFâ€™s were averaged within each group and then plotted to show the average HRF time course for each group (Figure 7-8). From the HRF time course (Figure 7-8), the nearly flat line of the control group indicated that the controls showed little or no activity in this specific portion of the superior temporal gyrus; the controls did show activity in other areas of the superior temporal gyrus on within-groups analyses. Thus, this very small region of the superior the controlence in activity existed because of the preseis d Comparison. Frontal region within-groups analyses. On the pseudoword comparison task, the control group evidenced activity in bilateral dorsal premotor (BA 6), right medial frontal the dyslexic group demonstrated bilateral inferior and middle frontal gyral (BA 8, 9, and temporal gyrus represents an area where the activity was significantly different between s and dyslexics. The significant differ nce of cortical activity by the dyslexics and a lack of activity by the controls in thcircumscribed region, during the pseudoword comparison. Figure 7-8. Temporal Region HRF for Dyslexic greater than Control During Pseudowor(BA 8) and rostral cingulate (BA 32) regions (see Table 7-15 and Figure 7-9). In contrast,
84 44), and right inferior frontal gyrus (BA 45 and 47) regions of activity (see Table 7-15 and Figure 7-9) Table 7-15. Frontal Region Activity per Group for Pseudoword Comparison Versus Baseline White Noise. Vol Anatomic Maximum Group Region L Areas Local Max t test valueControl R/L-PrCG, dFG, SFG 1188 BA 6, 8, 32 (3, 21, 42) 7.78 R-MFG, IFG 328 BA 8 (41, 14, 35) 5.40 Dyslexic R-IFG, MFG, 1028 BA 44, 8, 9, 45 (51, 12, 32) 7.22 L-MFG, PrCG, IFG 442 BA 8, 6, 4, 9, 44 (-37, 2, 33) 4.49 L-PrCG, MFG 1244 BA 6, 4 (-43, -2, 46) 6.02 R-IFG 248 BA 47, 44, 45 (45, 20, -2) 6.78 Note. BA = Brodmann's Area. dFG = dorsal frontal gyrus. IFG = inferior frontal gyrus. L = left. PrCG = precentral gyrus. MFG = middle frontal gyrus. MTG = middle temporal gyrus. R = right. SFG = superior frontal gyrus. Volumes of activity change >150 L and p < 0.005. igure 7-9n Activitye Dyslexic Group (yellow) and Control Group (blue) During Pseudoword Comparison (overlap = green, p 0.005, clusr > 150L, radiologic images)Froeen-groalygroindifferent rf activity for eachp. Hifict clu F . Frontal Regio for th < te . ntal region betw ups an ses. The withinups analyses cated di egions o grou owever, no sign antly differen sters of activity between the control and dyslexic groups were evident in the frontal region.
85 Parietal region within-groups analyses. The control group exhibited bilateral posterior cingulate gyrus (BA 31) activity on the pseudoword comparison task. Also, only the activity in the left hemisphere cluster extended into the superior parietal region (BA 7). In comparison, the dyslexic group produced bilateral activity in the supramarginal gyrus of the superior parietal region (BA 40) (Table 7-16). Table 7-16. Parietal Region Activity per Group for Pseudoword Comparison Versus Baseline White Noise. Vol Anatomic Maximum Group Region L Areas Local Max t test value Control L/R PCu, CG 170 BA 31, 7 (-5, -46, 42) -4.192 Dyslexic Note. BA = Brodmann's Area. CG = Cingulate Gyrus. dFG = dorsal frontal gyrus. IPL = L-STG, TTG, IPL 3929 BA 22, 42, 41, 40(-57, -39, 21) 8.01 R-STG, PoCG 3092 BA 42, 22, 40 (55, -14, 9) 7.122 inferioCentral cingulage gyrus. STG = superior temporal gyrus. TTG = transverse temporal gyrus. c groups, one cantly more gion waa small portion of a cluster that spanned the left STG and into the left SMG. r parietal lobe. MTG = middle temporal gyrus. PCu = precuneus. PoCG = postVolumes of activity change >150 L and p < 0.005. Parietal region between-groups analyses. When the cortical activity on the pseudoword comparison task was compared between the control and dyslexiregion of differential activity was noted. The dyslexic group showed signifi activity in the left supramarginal gyrus (BA 40) than the controls (Figure 7-10). This re s
86 Figure 7-10. Significant Parietal Region Activity (red) for Dyslexic Group (yellow) parison (overlap = green, pAn overall compthe within and between-groups analyses is presented in Figure 7-11. This 3-dimspecific to each group (yellow for dyslexic acortical activity thslexic and control subjects were displayed in green. Lastly, based on between-groupsregions that were significantly more activFigure 7-11. Movie of Whole Brain Activity per Group During Pseudoword Comparison control, green = overlap, p < 0.005, cluster > 150L, left to right). greater than Control Group (blue) During Pseudoword Com < 0.005, cluster > 150L, radiologic image). arison of the activity from ensional movie showed the activity that was nd blue for controls). Also, regions of at overlapped between the dy analyses, only the e (dyslexic greater than control) were represented in red. (red = significant between-groups difference, yellow = dyslexic, blue =
87 CHAPTER 8 DISCUSSION This study explored a functional ne uroanatomical model of phonological processing referred to as the Articulatory Awareness Model of Speech Perception (AAMSP) (Figure 5-1). The AAMSP presented a network of cortical regions believed to facilitate the segmentation of speech into the smallest units of language, individual phonemes. Segmentation skill was chosen as the target skill to model, as it strongly predicts the development of r eading skills. Since individual s with developmental dyslexia have a primary deficit in phonological proc essing, including phonol ogical segmentation abilities, a group of individuals with devel opmental dyslexia and a group with no history of reading difficulties were used to evalua te the AAMSP. FMRI was chosen as a noninvasive mechanism for evaluating the AAMSP. Based on a literature review of pure tone and phonological processi ng in individuals with norma l development of reading skills and those with developmental dyslexia, predictions of cortical activity were developed for each group. These predictions were then fitted to the AAMSP through functional neuropsychological predictions. Specifically, it was hypothesized that differential abilities in phoneme segmentati on skill would lead to differential cortical activity in downstream regions (regions fu rther along a pathway of cognitive processing) of cortex (inferior and middle frontal gyrus, somatosensory cortex, and the supplementary motor area). Thus, this study se t out to test the validity of the AAMSP, by selecting a priori hypotheses regarding pattern s of predicted cortical activity for two groups of adult men while they performed specific behavioral tasks.
88 Tenets of the AAMSP are described in the Motor Articulatory Feedback Hypothesis of Developmental Dyslexia (MAFHDD) (Heilman, Voeller, and Alexander, 1996) and Liberman and Mattinglyâ€™s Mf Speech Perception (1985). Heilman et al.â€™s (1996) motor feedback hypothesis presented a probable etiology for phonological processing deficits in individuals with developmental dyslexia. This hypothesis proposed that decreased sensory awareness of the articulatory characteristics of phonemes could contribute to the development of deficient phonological processing skills. The articulators (tongue, lips, and jaw) provide consistent motor movements for each phoneme, regardless of its position in a word or neighboring sounds. Therefore, it is likely that the awareness of the motoric movements of the lips, tongue and jaw can aid the perception and categorization of individual speech sounds within words. This hypothesis presented the importance of articulatory tactile-kinesthetic awareness, because unlike the reliable and specific articulator movement(s) for each phoneme, the acoustic properties of phonemes are variable within English words (Liberman & Mattingly, 1985). Liberman and Mattinglyâ€™s (1985) spectral analysis of phonemesâ€™ acoustic properties indicated that an individual phonemeâ€™s acoustic properties varied according to the phonemeâ€™s position in a word (initial, medial or final sound). Additionally, a phonemeâ€™s acoustic properties changed depending on preceding or following sounds. Therefore, it was concluded that speech perception was highly unlikely to be a purely auditory process. Additional support for the AAMSP comes from treatment studies of acquired and developmental reading problems. The successful treatment of phonological processing deficits in developmental dyslexia and acquired phonological alexia indicate that otor Theory o
89 articu latory tactical-kinesthetic input contributes to the remediation of deficientperception and phonological processing. Many reports of the successful treatment of acquired and developmental reading problems have included training in the multi-s(articulatory tactile-kinesthetic, auditory, visual) features of phonemes and their application to phonological processing (e.g. segmenting phonemes, reading and manipulating phonemes in pseudowords and real words) (Torgesen, Wagner, Rashotte, Alexander, & Conway, 1999; Conway et al., 1998; Kendall, Conway, Rosenbeck, Gonzalez-Rothi, 2000; Alexander, Andersen, Heilman, Voeller, & Torgesen, 1991; Adaet al., 2000). Also, individuals with developmental dyslexia are reported to have decreased articulatory tactile-kinesthetic awareness (Montgomery, 1981). Therefore, theAAMSP incorporated both auditory and articulatory tactile-kinesthetic features of phonemes when modeling one aspect of phonological processing, segmenting speech into its individual phonemes. The AAMSP proposed that speech perception included both auditory and articulatory tactile-kinesthetic sensory representations in specific regions of cortex. Specifically, the model described how auditory speech input (superior temporal gyrwould trigger both phonological (supramarginal gyrus) and articulatory tactile-kinesthetic(Brocaâ€™s area and somatosensory cortex) representations of phonemes in a word. Consequently, the neural instantiation of a word and its phonemes would be multi-sensory (a speech ensory ir us) uditory and tactile kinesthetic), as well as multi-modal (sensory and motor). A benefnsory it of this multi-sensory and multi-modal representation was proposed to be an enhanced perception of the phonological features of words and an improved ability to remember a word in auditory memory. Consequently, the combination of multi-se
90 and multi-modal representations and their enhanced preservation in memory was believed to improve the ability to segment speech syllables into phonemes. The AAMSP postulated a functional cortical network that could account for differing abilities to segment speech into its most basic unit (phonemes). To test tAAMSP, cortical activity was contrasted between a group of adults with intact phonological processing (phoneme segmentation skills) and a group of adults with a disorder (developmental phonological dyslexia) characterized by deficient phonoloprocessing (poor phoneme segmentation skills). The tasks in an initial experiment contrasted counting segmented pure tones, counting segmented phonemes and counsegmented phonemes in pseudowords. Individuals with difficulty segmenting speech into phonemes (developmental dyslexics) were postulated to show little or no activsome downstream regions (regions further along a pathway of cognitive processing) of the model believed to be in he gical unting ity in volved in articulatory tactile-kinesthetic perception and auditootor ical ry working memory (Brocaâ€™s area, somatosensory cortex, supplementary marea, and dorsolateral prefrontal cortex). In contrast, individuals with intact phonologprocessing skills were hypothesized to show activation in all the AAMSPâ€™s regions, during the most cognitively demanding task of holding, segmenting, and comparing pseudowords in auditory working memory. With stringent probability and volumetric thresholds (p < 0.005, volume > 15experiment one found limited cortical activity across all tasks and groups. For the control group, the initial findings included bilateral activity in the superior temporal gyrus dphoneme counting and no other regions of activity, when compared to a baseline of passively listening to white noise. The dyslexic group initially showed left superior 0 L), uring
91 temporal gyrus activity for both phoneme counting and pseudoword segmenting, as well as left middle frontal gyrus activity for pseudoword segmenting. H owever, the limited regionion. ls with ws or a old sis y s of at the limited activity observed with the more ality s of activity detected in experiment one when compared to previous studies (Zatorre et al., 1992, 1996, and Fiez et al., 1995) indicated a need for further exploratWhile many FMRI studies have reported decreased cortical activity in individuadevelopmental dyslexia and other studies have reported increased activity, the lack ofactivity in experiment one was striking and it was observed for both dyslexic and control subject groups. It was decided this lack of activity could be due either to design flalack of experimental power (i.e., small numbers of subjects per group) for experiment one. In an exploratory analysis, to determine if low power and sensitivity contributed to the scarcity of cortical activity in experiment one, the statistical probability threshold wasrelaxed. Experiment oneâ€™s data were reanalyzed with a less stringent probability thresh(p < 0.01) to rule out the presence of low power and sensitivity. This exploratory analyyielded significant activity across all tasks and groups. These regions of activity included those predicted by the AAMSP (see Appendix B) and were consistent with previouslpublished findings. Because the lower p-value reanalysis resulted in increased regionactivity in expected regions, it is highly likely th stringent criteria (p < 0.005) was due to low power and low sensitivity within the experimental design. Nevertheless, it is noteworthy that the exploratory analyses indicated the dyslexic and control groups produced broad patterns of differential cortical activity for all tasks. Overall, the exploratory analysis found differences in both later
92 and regions of cortical activity; detailed descriptions of the exploratory within-groups analyses are presented in Appendix B. On several occasions, during both exploratory and stringent analyses, the conand dyslexic groups presented different patterns of activity following within-groups comparisons. However, on between-groups analyses, these differential patterns did not result in significant differences in cortical activity. Despite the problems with the sensitivity and power for experiment one, the exploratory findings suggest that differences beyond those found in the stringent analyses most likely existed between the control and dyslexic groups. However, due to low power, i.e., low sensitivity, the between-groups differences did not reach the statistical or volumetric thresholds. Therefore, the stringent analyses provided the best indication of consistent differences between t trol he groups in primary auditory cortex and secondary auditory association cortex. ed likely indicate poorly defined phonemic and acoustic representations or an inefficiently While the within-groups analyses were limited by poor sensitivity and power, the between-groups analyses yielded robust and reliable results with stringent thresholds (p < 0.005 and ). For the tone stimulus, it was hypothesized that both groups would exhibit equal regions and volumes of activity. However, a significant difference between the groups existed in the left superior temporal gyrus (BA 42). In this region the dyslexic group showed greater activity than the controls. It appeared that even with non-speech input, the auditory cortex of the dyslexics was more active than controls. Based on Hebbian principles of learning (Kolb and Wishaw, 1996), a repeatstimulus may lead to a more fine-grained representation or a more efficient network of neuronal connections in the cortex. Thus, an over-abundance of cortical activity would
93 organized network of neurons. Thus, from the findings of experiment one, the broader range of cortical activity among the dys lexics may have indicated a less efficient cortical systemtion, for a ficit in d originy, e for auditory processing. Interestingly, the AAMSP itself was not a fine-grained model of speech percepe.g. it did not take into account the efficiency or inefficiency within a cortical network. Nonetheless, converging evidence from recent studies supports the likelihood of inefficient cortical activity among developmental dyslexics (Helenius, Salmelin,Richardson, Leinonen, & Lyytinen, 2002; von Plessen et al., 2002). Similar to these recently reported findings, the dyslexic group exhibited inefficient cortical activitynon-language based, acoustic auditory task. However, based on the AAMSP, a deprimary auditory processing was not predicted. But the presence of such a deficit impliethat the differences in phonological processing between controls and dyslexics may ate from a more primary or upstream (initial region of sensory input to the cortex) level of cognitive processing. Such an outcome would contradict the AAMSPâ€™s predicted differences in downstream regions (later regions along a cortical pathway). Additionallprimary auditory cortex was not maximally engaged for controls during the basic non-language stimulus (pure tones); however, the dyslexicâ€™s inefficient auditory processingappeared to have driven their primary auditory cortex toward asymptotic levels of over-activity with the same basic non-language stimulus (pure tones). Overall, based on a probability level commonly accepted in neuroimaging experiments (p < 0.005), experiment one found predicted and unpredicted differences inregions of activity between the two groups. For the control group, as the stimulus becammore complex, increasing the task's level of difficulty, then an additional cortical region
94 (left SMG) became active. This is consistent with reported SMG activity for a simple ocomplex speech stimulus, possibly requiring more processing and or storage of phonological information (Paules r u, Frith, & Frackowiak, 1993; Demont, Price, Wise et al., 19 the 94; Thierry, G., Boulanouar, K., Kherif, F., Ranjeva, J-P., and Dmonet, J-F, 1999). Also, this is in agreement with the tenets of the AAMSP that pseudowords wouldactivate the SMG as part of a broader cortical network among the control group. However, it was not predicted that the dyslexics would exhibit over-activity frommost basic stimulus in the study (pure tones). With even the simplest stimulus (pure tones) the dyslexic group showed a significantly greater amplitude of cortical activity in the left superior temporal gyrus than controls. While the AAMSP predicted equal activityin the superior temporal gyrus for tone counting, it was apparent that the AAMSP did not take into account potential differences between the two groups in their efficiency or inefficiency of auditory processing. Interestingly, post-mortem studies (reviewed by Lambe, 1999) have documented that male dyslexics typically exhibit microdysgensis or groups of malpositioned neurons that are predominately located in the left perisylvian, left temporal lobe and left inferior frontal cortex. Such cellular anomalies could contribute to the presence of inefficiently â€œwiredâ€ cortex and lead to the recruitment of alarger neuronal network to perform tasks that controls can do with a much smaller or more efficient neuronal network. The findings in experiment one are compromised by the lack of power and sensitivity. Given exploratory findings at a probability level more relaxed than commonlyaccepted in FMRI experiments, we can conclude that a greater number of subjects would have provided more stable patterns of activity between the groups. However, despite the
95 limitations of experiment oneâ€™s design, it is noteworthy that the dyslexic group demonstrated greater activity in the left superior temporal lobe than controls for even a basic acoustic and non-language stimulus, i.e. pure tones. In contrast to the dyslexic groupâ€™s over activity in the superior temporal regionduring tone counting, the control group evidenced reliably greater activity in posterior language cortex when they segmented pseudowords. Specifically, the controls exhgreater activity than the dyslexics in the SMG, an area believed to process and temporarily store pseudowords (Hickok & Poeppel, 2000; Ruff, Cardebat, Dmonet, Dmonet, 2002). However, differences in accuracy of performances on the pseudoword segmenting task may account for the differences in cortical activity, as the control group performed more accurately (p < 0.016) on this task than the dyslexics (88 percent correct versus 77 percent, respectively). Hence, the unequal task performance on pseudowordsegmentation indicated that different cortical networks may have activated due to different functional abilities to perform the task. However, the relationship between performance accuracy and regions or networks of cortical activity does not appear to belinear. Quite the contrary, on the tone counting task the dyslexic group producedsignificantly greater activity while performing significantly worse than the controls (p < 0.008, 93 percent correct versus 99 percent, respectively). Therefore, the differences inaccuracy and cortical activity may counting ibited and indicate that both groups attempted to use the same corticork al network, but the dyslexic group's inefficient or problematical cortical netwfailed to produce as much activity in the posterior language region (SMG) as the control group. Also, while there was no direct measure of cognitive effort or degree of trying on these tasks, all subjects were monitored for responding during the trials and there were no
96 observable signs during scanning nor any reported comments during debriefing indicated a lack of full effort on all tasks. Therefore, the relationship between accuracy, effort and cortical activity is not clear from experiment one; this m that ay be due to the power and samely ison h emands ensitivity limitations in this experiment. Nevertheless, a second experiment was designed and implemented to test the remaining components of the AAMSP, central executive and working memory. Experiment two attempted to test a different component of the AAMSP, nauditory working memory. Working memory was postulated as an essential skill for self-correction of reading, spelling, and speech errors (Lindamood, Bell, & Lindamood, 1997). For example, after reading a word incorrectly, an individual with dyslexia may reread a word to compare the initial reading with the rereading, to determine which is correct. Thus, in some instances, individuals may be holding two words in memory andcomparing them to determine which pronunciation was correct. Such online comparof words or pseudowords may require analysis of the individual phonemes within eacword. Such a detailed comparison would place significant demands on auditory working memory (Lindamood, Bell, & Lindamood, 1997). When these working memory dare evident, then additional cortical regions are necessary to maintain the word in auditory working memory. Thus, the AAMSP includes a fourth component with corticalregions believed necessary for segmentation and a fine-grained comparison of the individual phonemes within two words (supplementary motor area and dorsolateral prefrontal cortex). Experiment two was devised to test the auditory working memory component of the AAMSP, as used in a task requiring phoneme segmentation to compare two
97 pseudowords. As in experiment one, two groups of subjects were utilized. The dyslexgroup had historical accounts and current testing which documented deficient phonological processing skills. The control group reported no history of phonological processing deficits and demonstrated intact abilities on current testing. To test the AAMSP, two tasks were designed that included auditory working memory skills and of these tasks also required phonem ic one e segmentation skills. The first task, tone comparison, requirs. If t t at ). retical model of developmental dyslexia (Heilman, Voeller, and Alexaas ed individuals to count the number of tones in two separate series of tones. Since stimuli were presented rapidly, it is likely that many subjects had to hold both the first and second series of tones in working memory while counting the tones in each serieboth series had exactly the same number of tones, then the participant made a response with a button press device. In contrast, pseudoword comparison required the participanto hold and compare two pseudowords. The participant held one pseudoword in workingmemory and counted its individual phonemes. Because of the presentation rate, the participantsâ€™ counting of the first pseudoword was likely interrupted by the presentationof the second word. This necessitated both words being held in memory, their componentphonemes being counted, and then the number of phonemes in the first pseudoword was compared to the number in the second word. Participants in the control group, adepsegmenting auditory speech input into phonemes, were hypothesized to show activity in all of the AAMSPâ€™s cortical regions (STG, SMG, DLPC, SMA, Brocaâ€™s and SSCtxDyslexic participants were expected to show activity in only the STG and SMG. As implicated in a theo nder, 1996) and a behavioral study of dyslexia (Montgomery, 1981) no activity w
98 expected outside these regions in the remaining portions of the AAMSP (DLPC, SMA, and Brocaâ€™s SSCtx) for the dyslexic group. Experiment two demonstrated different regions of robust activity (p < 0.005) for the control and dyslexic groups on tone comparison. Unlike experiment one, expertwo required both auditory working memory and segmentation skills. For the control group, within-groups analyses of tone comparison reported activity in bilateral STG (BA 21, 22, 41, 42), right temporal pole (BA 38), bilateral SMA (BA 6), and left SMG (BA 40). Similar to experiment one, the control group demonstrated activity in two of the three predicted regions (STG and SMA). Activity was not present in one region predicted by AAMSP (dorsolateral prefrontal cortex). In comparison, the dyslexic group exhibitedcortical activity in bilateral superior temporal gyrus (BA 22, 41, 20), right superior temporal gyrus (BA 42), left superior temporal gyrus (BA 38), left precentral gyrus and sensory motor cortex (BA 6, 4, 8), right middle and inferior frontal gyri (BA 47, 9) bilateral inferior frontal gyrus (BA 44), and right parietal association cortex (BA 7), rightangular gyrus (BA 39), and bilateral supramarginal gyrus (BA 40). As with experiment one, the dyslexic group evidenced more active regions than the AAMSP predicted (STG, SMA, and DLPC) and the two groups did not activate all the same regions. Thus, the dyslexic group recruited many more cortical regions (20 versus the control groupâ€™s 12 regions) to perform tone comparison. Although, one of the two reliable differences between the groups indicated that the dyslexic group showed a significant differenceupstream regions of primary and secondary auditory cortex for tone comparison. Consistent with experiment one, it appears that the dyslexic groupâ€™s primary auditory iment in cortex is driven to maximum potentiation by even a simple acoustic non-speech stimulus.
99 Similar to experiment oneâ€™s indications, diffuse cortical activity among the dyslexic group could indicate a poorly organized and inefficient auditory processing network. p ent l r tal e and gly, P Since increased effort can be related to increased activity in functional brain imaging paradigms (Dhankhar et al.1997), it is possible that the dyslexic group was exerting more effort on the tone comparison task. This is supported by the control groubeing significantly more accurate with tone comparison than the dyslexic group (p < 0.01, standard deviation = 4.21 and 19.95, respectively), even though the difference in mean percent correct between the groups was small (99 percent correct versus 91 perccorrect, respectively). Perhaps the dyslexic groupâ€™s greater difficulty on this task can account for their larger number of active cortical regions; however, strong evidence challenging this account is addressed later. Within-groups analyses of the pseudoword working memory task provided additional evidence that the dyslexic group recruits a much larger network of corticaregions for both speech and pure tone auditory tasks. On pseudoword comparison the dyslexic group exhibited activity in bilateral superior temporal gyrus (BA 42, 22), left superior temporal gyrus (41), left supplementary motor area (BA 6), left primary motor cortex (BA 4), right inferior frontal gyrus (BA 45 and 47), bilateral premotor cortex (oeye fields) (BA 8), bilateral dorsolateral prefrontal cortex (BA 9), bilateral inferior frongyrus (BA 44), and bilateral supramarginal gyrus (BA 40). Similar to experiment onthe tone comparison task in experiment two, the dyslexic group evidenced many more regions of activity than were predicted by the AAMSP (STG and SMG). Interestinthe dyslexic group showed activity in all the regions that were predicted by the AAMSfor the control subjects, not the dyslexics. Specifically, the dyslexic group evidenced
100 activity in Brocaâ€™s area as well as the AAMSPâ€™s other regions (STG, SMG, DLPCx, SMA, SSCx). In contrast, the control group did not evidence activity in all the AAMS Pâ€™s predicl., â€™s primary strategy used was to repeat the pseudre, d t not ted regions. The control group exhibited activity in superior temporal gyrus bilaterally (BA 22, 41, 21, 42), right premotor cortex/frontal eye fields (BA 8), right anterior cingulate cortex (BA 32), supplementary motor cortex bilaterally (BA 6), posterior cingulate cortex bilaterally (BA 31), and left superior parietal lobe (BA 7). Interestingly, only the control group evidenced activity in Brodmannâ€™s area 32, which appears to be active during self-monitoring of performance during a task (MacDonald, Cohen, Stenger, Carter, 2000) as well as preparation for word generation (Crosson et a1999). Based on task instructions prior to scanning and subject reports following scanning of strategy per task, the participant oword silently in their head and then try to segment the pseudoword. Therefodespite similarities in reported cognitive strategies, the two groups of subjects appeareto utilize different cortical networks or the controls used the network more efficiently. Between-groups analyses of pseudoword comparison showed that activity in onlyone cortical region was reliably different between the groups (p < 0.005). Identical to tone comparison, the dyslexic group demonstrated significantly greater activity in the lefsuperior temporal gyrus (BA 42, 22) than the controls. This region included both primary and first order auditory association cortex. Interestingly, for both groups this region was only one of many regions that were reliably different from baseline white noise on within-groups comparisons. Thus, the appearance of differences for cognitive tasks between the groups in such regions as inferior frontal gyrus and supramarginal gyrus did not materialize. The degree of activity between the dyslexic and control groups could
101 be distinguished in these regions, even in those for which within-groups analyses showsignificant activity only for the dyslexic group (SMG, DLPC, SMA, SSCx, and Brocaâ€™s). In summary, during pseudoword comparison two regions of left primary and seconauditory cortex (BA 22, 42) were reliably more active in the dyslexics than the controAcross experiments one and two, cognitive effort and inefficient cortical netwmay account for the dyslexic group exhibiting greater activity than the control group upstream (primary) auditory processing cortex (left BA 42, 22). Because the dyslexic group performed tone counting with reliably less accuracy than controls (p < 0.008), it ipossible that the dyslexic participants put forth greater effort to perform this task. Additional effort during tone counting could account for the dyslexic groupâ€™s largregion of activity in the left superior temporal lobe (BA 42). Similarly, more cognitive effort could account for the dyslexic groupâ€™s larger region of activity (left BA 42, 22) during tone comparison on experiment two, as Dyslexics were also significantly less accurate than controls (p < 0.01). Additionally, when both groups performed equally we ed dary ls. orks in s er ll on phater ring oneme counting in experiment one, there were no reliable differences in activity between the groups for both stringent (p < 0.005) and exploratory analyses (p < 0.01). However, effort alone cannot account for all the reliable activity differences between the groups. For example, both the dyslexic and control groups performed equally poorly but better than chance on pseudoword comparison (56% and 65%, respectively). Despite the similarity in difficulty for both groups, the dyslexic group still exhibited reliably greactivity that was once again in the left superior temporal gyrus (BA 22, 42). Also, dupseudoword segmenting in experiment one the dyslexic group performed less accurately than controls (p < 0.01). However, despite the greater difficulty and possibly more effort
102 by the dyslexic group, on pseudowords segmenting the dyslexic group did not evidence greater activity than the controls. Rather, during pseudoword segmenting the control group exhibited a significantly greater region of activity in the left parietal lobe (BA 40, 7, 39) than the dyslexic group. Therefore, the relationship between task difficulty (as impliation ned by accuracy) and intensity of cortical activity is not a simple monotonic relationship. The differences between the dyslexic and control groups manifested on the simplest task in this study (tone counting) and on the most complex task (tone and pseudoword comparison), but not on tasks intermediate in difficulty (phoneme counting, pseudoword segmentation). In any case, the primary and first order auditory associcortices (BA 22, 42) show the most robust between-groups differences, with greater activity for dyslexics, suggesting differences in efficiency of processing. Besides the reliable differences in temporal lobe activity between control and dyslexic groups, there is a striking difference in activity in the left parietal lobe. Withigroups comparisons for both the control group and the dyslexic group did not show a difference between pseudoword segmenting and baseline white noise. Yet, a reliable difference existed between the dyslexic and control groups on this same task, pseudoword segmenting, in the left parietal lobe (BA 40, 7, 39). It is possible that the dyslexic group produced a mild activity reduction in this cortical region while the controlgroup produced a mild increase in activity in this region. This implies that the dyslexic group relied on temporal and frontal language cortices to perform the pseudoword segmenting while the control group used temporal, frontal and parietal regions. This samescenario was depicted in figure 7-8, when the time course of the HRF for both groups was plotted relative to pseudoword comparison. However, in this instance, the control
103 group demonstrated a nearly flat HRF, while the dyslexic group demonstrated about a 1.5% change in signal over the thirty-second stimulus block. From this plot, it appearsthat only the dyslexic group produced activity in this circumscribed region of the parietalcortex (SMG) during pseudoword comparison. According to Paulesu, Frith, and Frackowiak (1993), the SMG region is involved in processing and temporary storphonological information. Therefore, similar to the temporal lobe findings, the dyslexic group exhibited a larger region of cortical activity in the supramarginal gyrus durinpseudoword comparison. The limitations of this study are highlighted by several recent findings regarding selection of baseline tasks, and the effect of scanner background noise on FMRI of auditory processing, as well as some limitations in the implementation of t age of g he design and the ch n the e present ) to osen equipment. Early assumptions by many FMRI investigators held that restprovided a cortically inactive task for a baseline condition. However, this has been showto be problematic and far from accurate by Binder, Frost, Hammeke, Belgowan, Raod,and Cox (1999). Binder et al. reported that during â€œrestâ€ baseline activities, activity insemantic network can be evident. Thus, the use of a white noise baseline in thstudy may have added some â€œnoiseâ€ to the data analysis. Similarly, Binder et al. (1999also pointed out that allowing subjects to keep their eyes closed during FMRI experiments may lead to increased activity in the frontal lobes. This was attributed to theâ€œalways onâ€ nature of the human brain and that decrease visual stimulus could leadincreased self-monitoring or other sensory awareness. Potential implications for theoretically driven approaches to remediating or preventing phonological processing deficits and subsequently developmental
104 phonological dyslexia are also evident from these experiments. Fitch, Miller, and Talla(1997) have developed a substantial body of research corroborating their view that the judgment of rapid and fine-grained changes (such as rapid formant transitions) in phonological structures (words, phonemes and subcomponents of phonemes) is thsubstrate of phonological awareness deficits. The current findings offer additional support for the approach of Tallal et al.â€™s treatment programs targeted at enhancing fine-grained auditory processing of acoustic and speech representations. However, Share, Jorm, Maclean and Matthe l, e w (2002) have reported that these fast temporal judgments may brory. The overall finding from expers that, ditory e associated with language-based deficits, but that temporal processing is not the cause of phonological awareness impairments. Therefore, it is unlikely that such a temporal based treatment alone will be effective for the majority of individuals with developmental dyslexia. Nevertheless, it is likely that the connections between the functional systems involved in speech perception (motor, sensory, auditory, visual) as well as the systems themselves should be the target of treatment. Working towards bettedefined representations in the language systems as well as better integration of these representations across the components of the various systems (motor, language, sensory,and memory) could be the best approach to remediate acquired and developmental phonological awareness and reading difficulties. A summary of the findings from this study provides a framework for determining possible implications and potential future directions for neuroimaging research of phonological awareness and phonological working mem iment one (tone counting, phoneme counting, and pseudoword segmenting) idespite the power issues, the dyslexics appear to have a much less efficient au
105 cortex. Even with primary acoustic input, the dyslexicâ€™s primary auditory cortex (left superior temporal gyrus) seems to be driven to maximum potentiation. Thus, it appears that the dyslexics not only have difficulty with phonologic perception and processialso basic acoustic perception and processing difficulties as well. This finding was replicated by experiment two, which had much better power. In experiment two, ovactivity and inefficiency were present for not only the acoustic stimulus, but also for the pseudoword comparisons. Also, it appeared that the more demanding nature of the working memory tasks led to the recruitment of a portion of the supramarginal gyrus bthe dyslexics. As with the primary and first order auditory association cortex, activity inthe supramarginal gyrus was greater for the dyslexic individuals than the controls. Therefore, the major regions of activity and the extent of the activity were in direct opposition with the predictions of the AAMSP model. While the regions of the cortidentified by AAMSP are known to be involved in the perception of speech, the AAMSdid not take into consideration the various forms of compensatory activity (e.g. Brocaâ€™s or SMG) that the dyslexics might use during the scanning tasks. Nevertheless, based othe findings from this study, Hebbian principles of learning, and reports of successful treatment of developmental and acquired phonological dyslexia (Conway et al., 199Torgesen, et al, 1999., Alexander et al., 1991., Ken ng, but er-y ex P n 8, dall et al., 2001), the remediation or es of rehabilitation of impaired phonological processing should focus on refining the speech perception and processing abilities of individuals with dyslexia. Future FMRI studilanguage processing and speech perception may help elucidate how the successful remediation of developmental and acquired phonological dyslexia impacts the efficiency and neuroanatomical membership of the cortical network.
APPENDIX A NEUROPSYCHOLOGICAL ASSESSMENT General Assessment Instruments Spielberger State Anxiety Scale (Spielberger, 1983) â€“ a measure of state anxiety to rule out emotional reactivity as contributing to functional image activity (Rumsey et al. 1994). Hollingshead Scale of Socioeconomic Status â€“ a measure of socio-economic status. Intellectual Functioning Wechsler Adult Intelligence Scale-III (WAIS-III) (Wechsler, 1997) Vocabulary and Block Design two subtest short-form estimate of intelligenceAcademic Achievement â€“. Gray Oral Reading Test-3 (GORT-3) (Weiderholt, & Bryant, 1992) â€“ a measure of individual word reading ability (word identification subtest) and pseudoword reading (word attack subtests). Lindamood, 1975) â€“ working memory) for one-syllable pseudowords. Lindamood Auditory Conceptualization Test-Multisyllable Auditory Conceptualization (LAC-MAC) (unpublished research version) â€“ an experimental measure of comparator function for two to five syllable a measure of phonological awareness involving del a measure of passage reading accuracy and speed. Woodcock-Reading Mastery Test-III (WRMT-III) (Woodcock, 1987) â€“ Language and Memory Lindamood Auditory Conceptualization Test-Revised (LAC-R; Lindamood & a measure of comparator function (phonological awareness and phonological pseudowords. Comprehensive Test of Phonological Processing (Torgesen, Wagner, & Rashotte, 1998)â€“ etion of phonemes (Nonword Phoneme Elision subtest), segmentation of phonemes (Nonword Phoneme 106
107 Segmentation subtest), and a measure of rapid automatized naming (Rapid Naming of Numbers, Letters, and Colors subtests). Auditory Consonant Triagrams â€“ a measure of sustained and se lective attention,y working memo and susceptibility to cognitive interference during an auditorry task. Motor and Handedness Edinburgh Handedness Scale â€“ a measure of the degree of handedness (Oldfield, 1971). tory Awareness Test â€“ Articultory awareness adapted from a an experimental measure of oral-articula Montgomery (1981).
108 APPENDIX B EXPERIMENT ONEâ€“EXPLORATORY ANALYSIS Since experiment one appeared limited by low power and low sensitivity, detailed descriptions of exploratory analyses with a less stringen t probability (p < 0.01) are presented. Exploratory within-groups analyses for tone counting indicated that the two groups had cortical activity in different regi ons of the temporal lobe and in different hemispheres. The control group produced right middle temporal gyrus activity in auditory association cortex (BA 21). This is consis tent with previous reports of the right hemisphere being predominately involved in non-speech acoustic processing. Surprisingly, the dyslexic groupâ€™s activity was in the primary and first order association auditory cortex of the left superior te mporal gyrus (BA 42 and 22). A between-groups analysis of tone counting f ound that only one area of thes e active regions was reliably different (left primary aud itory cortex, BA 42), as the dyslexic group showed more activity in this region than the control gr oup. This finding of great er left superior temporal gyrus (BA 42) activity in the dyslexic group was iden tical to the results of the more stringent (p < 0.005) between-groups co mparison. Also in the exploratory betweengroups comparisons (p < 0.01), the control grou pâ€™s activity in the right hemisphere was not reliably different from that of the dyslexic subjects. S o, despite the fact that the dyslexic group did not show significant ac tivity in the right hemisphere (superior temporal gyrus) on within-groups analys es, the dyslexic group likely had enough variance of activity in the right temporal l obe (superior temporal gyrus) to prevent the control groupâ€™s pattern of right hemis phere activity from producing a significant
109 between-groups difference. Th e likelihood of right hemisphe re activity for the dyslexic group was also supported by exploratory within -groups comparisons between active tasks [phoneme counting vs tone counting and ps eudoword comparison vs tone counting]. While these comparisons will be discussed in detail later, it is noteworthy that the dyslexic group exhibited right hemisphere activity when comparing phoneme counting with tone counting and pseudoword segmenti ng versus tone counting. Even with limited power and sensitivity in the within-groups comparisons wi th tone counting, the robust findings from the stringent (p < 0.005) be tween-groups analyses indicated that the dyslexic group exhibited signifi cantly more activity in primary auditory cortex (BA 42) of the left hemisphere for a non-speech stimulus, pure tones. For phoneme counting, exploratory within-g roups comparisons (p < 0.01) implied patterns of activity that differed among the groups according to which hemisphere displayed the activity. Controls demonstrated bilateral superior temporal gyrus activity (right and left BA 22 and 42). In contrast, th e dyslexic group exhibite d a broad region of left superior temporal gyrus activity (BA 42, 22). However, comparisons between-groups still did not find any reliable differences, even with the relaxed probability threshold. The absence of between-groups differences was ev ident with both exploratory (p < 0.01) and stringent (p < 0.005) probab ility and volumetric (150 L) thresholds. Consequently, consistent with predictions, both groups ev idenced bilateral superior temporal gyrus activity for phoneme counting (p < 0.005). For pseudoword segmenting, the explorat ory within-groups comparison (p < 0.01) indicated activity beyond the temporal lobe and implied more active cortical regions for the control group than the dyslexic group. C ontrols showed activity in auditory
110 association cortex (BA 22) bila terally, in the dorsal premotor region bilaterally (BA 8 and 6), in the left superior pariet al lobe (BA 7), in the left a ngular gyrus (BA 39), and in the left supramarginal gyrus (40). The dyslexic group exhibited activity in left primary and first order association audito ry cortices (BA 42 and 22), a nd in the left middle frontal gyrus (BA 9). However, between-groups comp arisons, which were equally sensitive and powerful again for exploratory (p < 0.01) a nd stringent (p < 0.005) probability values, indicated that the left inferi or parietal lobe was the only region with reliable differences. Thus, the control group revealed more activity than the dyslexic group in the left angular gyrus and the left supramarginal gyrus (B A 39 and 40). Overall, while the controls demonstrated a larger network of active cortical regions during the segmenting of pseudowords (bilateral superior temporal gyrus, bilateral dor sal frontal gyrus, and left parietal lobe), this activity was reliably different in the parietal lobe only. Within-groups exploratory comparisons (p < 0.01) of phoneme counting, tone counting, and pseudoword segmenting, as well as comparisons of pseudoword segmenting with tone counting, and pse udoword segmenting versus phoneme counting, indicated a broader range of regions of activit y than indicated by the stringent analyses (p < 0.005). The control and dyslexic groups showed different patterns of activity when two active tasks were compared (phoneme c ounting versus tone counting, pseudoword segmenting versus tone counting, and pseudoword segmenting versus phoneme counting). Similar to the within-groups comparisons (tasks versus baseline white noise], both groups evidenced several more regions of activity for pseudow ord segmenting when compared to either tone or phoneme c ounting. The control group indicated that pseudoword segmenting resulted in more activity than tone counting in the left precentral
111 gyrus and middle frontal gyrus (BA 6), as well as right inferior fr ontal gyrus (BA 45 and 9). Likewise, the controls generated more activity for pseudoword segmenting than phoneme counting in the superior frontal gyrus bilaterally (B A 8 and 6) and in the left middle frontal gyrus and precentral gyrus (B A 6, 8, and 4). It is noteworthy that the Brocaâ€™s area homologue of th e right hemisphere was only active for the controls when pseudoword segmenting was compared to t one counting and not when compared to phoneme counting. Thus, it appears that for phoneme counting and pseudoword segmenting the controls show similar activity levels in the Brocaâ€™s area homologue of the right hemisphere. In contrast, the dyslexic group exhibited more activity for pseudoword segmenting than tone counting in the left in ferior and middle front al gyri (BA 8 and 9) and left precentral gyrus (B A 6). Also, the dyslexic group showed more activity in Brocaâ€™s area (left BA 44 and 45), the left middle frontal gyrus (BA 9), and the left precentral gyrus (BA 4 and 6) when comparing pseudoword segmenting with phoneme counting; unlike the controls who showed si milar right inferior frontal gyrus activity. This may reflect increased effort for the dyslexics when performing pseudoword segmenting, as only on pseudoword segmenti ng did the dyslexic group evidence greater difficulty than the controls with task accuracy. Overall, the exploratory analyses have some potential implications for the AAMSP. With a less stringent probability threshold, activity was evident in most of the AAMSPâ€™s predicted regions for the control group (superior temporal gyrus, Brocaâ€™s area, somatosensory cortex, supplementary motor ar ea, and supramarginal gyrus). In contrast, the dyslexic group showed activity in only one region predicted by the AAMSP (superior temporal gyrus) for this group. Surprisingly, several regions were also active for the
112 dyslexic group (Brocaâ€™s area, left IFG and MF G [BA 8, 9], and left precentral gyrus [BA 6, 4]), although activity in th ese regions was only predicted for the control group, not the dyslexic group. Therefore, thes e exploratory analyses provide important considerations for future empirical assessments of the AAMSP.
NCES Adair, J. C., Nadeau, S. E., Conway, T. W., Gonzalez-Rothi, L. J., Heilman, P. C., Green, I. A., & Heilman, K. M. (2000). Alterations in the functional anatomy of reading induced by a phonologic articulatory rehabilitation strategy in an alexic patient. Neuropsychiatry, Neuropsychology, and Behavioral Neurology, 13, 303-311. Alexander, A. W. Developmental language disorders. Manuscript in preparation. Alexander, A. W., Andersen, H. G., Heilman, P. C., Voeller, K. K. S., & Torgesen, J. K. (1991). Phonological awareness training and remediation of analytic decoding deficits in a group of severe dyslexics. Annals of Dyslexia, 41, 193-206. Alexander, A. W., & Heilman, K. M. (in press). Treatment of Developmental Language Disorders. In N. H. Noseworthy (Ed.), Neurological Therapeutics Principles and Practice. London: Dunitz. Baddeley, A. (1986). Working memory. New York: Oxford University Press. Bakker, D. J. (1992). Neuropsychological classification and treatment of dyslexia, Journal of Learning Disabilities, 25, 102-109. Bakker, D. J., Moerland R., & Goekoop-Hoetkens M. (1981). Effects of hemisphere-specific stimulation on the reading performance of dyslexic boys: A pilot study. Journal of Clinical Neuropsychology, 3, 155-159. Bakker, D. J., Smink, T., & Reitsma, P. (1973). Ear dominance and reading ability. Cortex, 9, 301-312. Ball, E. W., & Blachman, B. A. (1988). Phoneme segmentation training: Effect on reading readiness. Annals of Dyslexia, 38, 208-225. Beauvois, M. F., & Drouesn, J. (1979). Phonological alexia: Three dissociations. Journal of Neurology, Neurosurgery, and Psychiatry, 42, 1115-1124. Benson, D.F. (1993). The history of behavioral neurology. Neurologic Clinics, 11, 1-8. Benson, D. F. & Ardila, A. (1996). Aphasia: A clinical perspective. New York: Oxford University Press. LIST OF REFERE 113
114 Berndt, R. S., Dâ€™Autrechy, C. C., & Reggia, J. A. (1995). Empirically derived probabilities for grapheme-to-phoneme correspondences in English. Behavior Research Methods, Instruments & Computers, 19,1-9. Besner, D., Twilley, L., McCann, R. S., & Seergobin, K. (1990). On the association between connectionism and data: are a few wordecessary? Psychological Review, 97, 432-446. Binde. W. (1999). Conceptual Processing during the Conscious Resting State: A Functional MRI anowicz, A., & Hyde, J. S. (1995). Lateralized human brain language systems demonstrated i, P. x. r: an Press. 292-302. eading disability and , od diagnosis of dyslexia. Developmental Review, 13, 258-268. Calfee, R. C., Lindamood, P., & Lindamood, C. (1973). Acoustic-phonetic skills and reading--kindergarten through twelfth grade. Educational Psychology, 64, 293-8. s n Bilecen, D., Seifritz, E., Scheffler, K., Henning, J., & Schulte, A-C. (2002) Amplitopicity of the Human Auditory Cortex: An FMRI Study. NeuroImage 17, 710. r, J. R., Frost, J. A., Hammeke, T. A., Bellgowan, P. S. F., Rao, S. M. and Cox, R Study Journal of Cognitive Neuroscience, 11, 80 Binder, J. R., Frost, J. A., Hammeke, T. A., Cox, R.W., Rao, S. M., & Prieto, T. (1997). Human brain areas identified by functional MRI. Journal of Neuroscience, 17, 353-362. Binder, J. R., Frost, J. A., Hammeke, T. A., Rao, S. M., & Cox, R.W. (1996). Function ofthe left planum temporale in auditory and linguistic processing. Brain, 119, 1239-1247. Binder, J. R., Rao, S. M., Hammeke, T. A., Frost, J. A., Bandettini, P. A., Jesm by task subtraction functional magnetic resonance imaging. Archives of Neurology,52, 593-601. Binder, J. R., Rao, S. M., Hammeke, T. A., Yetkin, F. Z., Jesmanowicz, A., BandettinA., et al. (1994). Functional magnetic resonance imaging of human auditory corteAnnals of Neurology, 35, 662-72. Bradley, L., & Bryant, P. (1985). Rhyme and reason in reading and spelling, Ann ArboUniversity of Michig Bradley, V. A., & Thomson, M. E. (1984). Residual ability to use grapheme-phoneme conversion rules in phonological dyslexia. Brain and Language, 22, Breier, J. I., Gray, L. C., Fletcher, J. M., Foorman, B., & Klaas, P. (2002). Perception of speech and nonspeech stimuli by children with and without r attention decit hyperactivity disorder. Journal of Experimental Child Psychology82, 226 Bruck, M. (1993). Word recognition and component phonological processing skills of adults with childho
115 Coltheart, M., Curtis, B., Atkins, P., & Haller, M. (1993). Models of reading alRoute and parallel-distributed-processing approaches. Psychological Review, 1589-608. oud: Dual-00, ry discrimination in depth (ADD) program. Journal of the International Neuropsychological Society, 4, 608-620. Corin. D., et al., (2001). FMRI auditory language differences between dyslexic and able reading Court., & Haxby, J. V. (1998). An area specialized for spatial working memory in human frontal cortex. Science, 279, Cox, R. W. (1996). AFNI: Software for analysis and visualization of functional resonance neuroimages. Computers in Biomedical Research, 29, 162-173. Crosson, B. Sadek, J. R., Bobholz, J. A., Gokcay, D., Mohr, C. M., Leonard, C. M., et al. . Activity in the paracingulate and cingulate sulci during word generation: An fmri study of functional anatomy. Cerebral Cortex, 9, 307-316. Dhan cortex response to increasing word presentation rates. Journal Of Neurophysiology, 77, 476-483. Democal and semantic processing in normal subjects. Brain, 115, 1753-1768. Demont, J. F., Fiez, J. A., Paulesu, E., Petersen, S. E., & Zatorre, R. J. (1996). PET Studies of phonological processing: a critical reply to Poeppel. Brain And Demont, J. F., Price, C., Wise, R., & Frackowiak, R. S. (1994). Differential activation of Denckla, M. B., & Rudel, R. G. (1976). Related articles, rapid "automatized" naming , Conway, T. W., Heilman, P., Rothi, L. J. G., Alexander, A. W., Adair, J., Crosson, B. A., et al. (1998). Treatment of a case of phonological alexia with agraphia using the audito a, D. P., Richards, T. L., Serafini, S., Richards, A., Steury, K., Abbott, R children. NeuroReport, 12, 1195-1201. ney, S. M., Petit, L,. Maisog, J. M., Ungerleider, L. G 1347-1351. (1999) khar, A., Wexler, B. E., Fulbright, R. K., Halwes, T., Blamire, A. M., & Shulman, R. G. (1997). Functional magnetic resonance imaging assessment of the human brain auditory nt, J. F., Chollet, F., Ramsay, S., Cardebat, D., Nespoulous, J. L., Wise, R., et al. (1992). The anatomy of phonologi Language, 55, 352-379. right and left posterior sylvian regions by semantic and phonological tasks: A positron-emission tomography study in normal human subjects. Neuroscience Letters, 182, 25-28. (r.a.n): Dyslexia differentiated from other learning disabilities. Neuropsychologia14, 471-479. de Partz, M. (1986). Re-education of a deep dyslexic patient: Rationale of the methodsand results. Cognitive Neuropsychology, 3, 149-177.
116 D'Esposito, M., Detre, J. A., Alsop, D. C., Shin, R. K., Atlas, S., & Grossman, M. (1995) The neural basis of the central executive system of working memory. Nature, 378, 279-81. DeYoe, E. A., Bandettini, P., Neitz, J., Miller, D., & Winans, P. (1994). Functional magnetic resonance imaging (FMRI) of the human brain. Journal of NeurosciencMethods, 54, 171-87. e Eimas, P., Siqueland, E., Jusczyk, P., & Vigorito, J. (1971). Speech perception in infants. Ehri, ledge is essential for learning to read words in English. In J. Metsala & L. Ehri (Eds.), Word recognition in beginning literacy Ellis,ung, A. W. (1988). Human cognitive neuropsychology. Hillsdale, NJ: Erlbaum. Feldm, et al. (1993). Adult familial dyslexia: A retrospective developmental and psychosocial profile. Feltonofile of adult dyslexics. Brain and Language, 39, 485-497. Fiez, x, 6, 1-10. emands. Journal of Cognitive Neuroscience, 7, 357-375. ts. and Child Neurology, 26, 143-153. gy, Funnell, E. (1983). Phonological processes in reading: New evidence from acquired Science, 171, 303-306. L. C. (1998). Grapheme-Phoneme know (pp. 108-117). Hillsdale, NJ: Erlbaum. A. W., & Yo an, E., Levin, B. E., Lubs, H., Rabin, M., Lubs, M. L., Jallad, B. Journal of Neuropsychiatry and Clinical Neurosciences, 5,195-199. , R. H., Naylor, C. E., & Wood, F. B. (1990). Neuropsychological pr J. A., Raichle, M. E., Balota, D. A., Tallal, P., & Petersen, S. E. (1996). PET Activation of posterior temporal regions during auditory word presentation and verb generation. Cerebral Corte Fiez, J. A., Raichle, M. E., Miezin, F. M., Petersen, S. E., Tallal, P., & Katz, W. F. (1995). PET studies of auditory and phonological processing: Effects of stimulus characteristics and task d Finucci, J. M., Whitehouse, C. C., Isaacs, S. D., & Childs, B. (1984). Derivation and validation of a quantitative definition of specific reading disability for adulDevelopmental Medicine Fletcher, J. M., Shaywitz, S. E., Shankweiler, D. P., Katz, L., Liberman, I. Y., Stuebing, K. K., et al. (1994). Cognitive profiles of reading disability: Comparisons of discrepancy and low achievement definitions. Journal of Educational Psycholo86, 6-23. Friedrich, F. J., Glenn, C. G., & Marin, O. S. M. (1984). Interruption of phonological coding in conduction aphasia. Brain and Language, 22, 266-291. dyslexia. British Journal of Psychology, 74, 159-180.
117 Galaburda, A. M., Sherman, G. F., Rosen, G. D., Aboitiz, F., & Geschwind, N. (1985). Developmental dyslexia: Four consecutive patients with cortical anomalies. Annals Neurolog y,18, 222-233. SA, 91, 8010-8013. Hagman, J. O., Wood, F., Buchsbaum, M. S., Tallal, P., Flowers, L., & Katz, W. (1992). ulatory feedback hypothesis. Annals of Neurology, 39, 407-412. 1-138. cript, Hulme, C., & Roodenrys, S. (1995). Practitioner review: Verbal working memory 6, Hynd, G. W. (1992). Neurological aspects of dyslexia: Comment on the balance model. Johnsrude, I. S., Zatorre, R. J., Milner, B. A., & Evans, A. C. (1997). Left-hemisphere specialization for the processing of acoustic transients. NeuroReport, 8, 1761-1765. Kaas,rtex. Kendall, D., Conway, T., Rosenbek, J., & Gonzalez Rothi, L. (2001, February). Galaburda, A. M., Menard, M. T., & Rosen, G. D. (1994). Evidence for Aberrant Auditory Anatomy in Developmental Dyslexia. Proceedings of the National Academy of Science U Goodglass, H. (1993). Understanding aphasia. San Diego, CA: Academic Press. Cerebral brain metabolism in adult dyslexic subjects assessed with positron emission tomography during performance of an auditory task. Archives Of Neurology, 49, 734-739 Heilman, K. M., Alexander, A. W., & Voeller, K. (1996). Developmental dyslexia: A motor-artic Helenius, P., Salmelin, R., Richardson, U., Leinonen, S., & Lyytinen, H. (2002). Abnormal auditory cortical activation in dyslexia 100 msec after speech onset. Journal of Cognitive Neuroscience, 14, 603-17. Hickok, G. & Poeppel, D. (2000). Towards a functional neuroanatomy of speech perception. Trends in Cognitive Sciences, 4, 13 Hollingshead, A. B. (1975). Four factor index of social status. Unpublished ManusYale University at New Haven. development and its disorders. Journal of Child Psychology and Psychiatry, 3373-98. Journal Of Learning Disabilities, 25, 110-112,123. J. J., & Hackett, T. A. (1999). â€œWhatâ€ and â€œwhereâ€ processing in auditory coNature Neuroscience, 2, 1045-966. Phonological treatment for aphasia and apraxia of speech. Third Veteran's Administration Rehabilitation Research and Development Confere nce , Washington, DC. Kolb,europsychology. 4 Edition. New York: W. H. Freeman and Company. B. & Wishaw, I. Q., (1996). Fundamentals of human nth
118 Lambe, E. K. (1999). Dyslexia, gender, and brain imaging. Neuropsychologia, 37, 521-536). Lefly, D. L., & Pennington, B. F. (1991). Spelling errors and reading fluency in Liberr theory of speech perception revised. Cognition, 21, 1-36. Liberman, I. Y. (1973). Segmentation of the spoken word and reading acquisition. Libertion. In acy with language. (pp. 105-125). Baltimore: Orton Dyslexia Society. Liberiple and learning to read. In D. Shankweiler & I. Y. Liberman (Eds.), Phonology and Lindamood, C. H., & Lindamood, P. C. (1975). Auditory discrimination in depth. Austin, Lindamood, C. H. & Lindamood, P. C. (1979). Lindamood Auditory Conceptualization Lindamood, P. C. (1994). Issues in researching the link between phonological awareness, r the ltimore: Paul H. Brookes. Disorders, 1, 143-182, LivinGalaburda AM. (1991). Physiological and anatomical evidence for a magnocellular defect in developmental dyslexia. Logoure of roscience, 3, 413. compensated adult dyslexics. Annals of Dyslexia, 41, 143-162. man, A. L. & Mattingly, I. G. (1985). Moto Bulletin of the Orton Society, 23, 65-77. man, I. Y. (1987). Language and literacy: The obligation of schools of educaR. F. Bowler (Ed.), Intim man, I. Y., Shankweiler, D., & Liberman, A. M. (1989). The alphabetic princ reading disability: Solving the reading puzzle (pp. 201-212). Ann Arbor: University of Michigan Press. TX: PRO-ED Publishing. Test. Austin, TX: PRO-ED Publishing. learning disabilities, and spelling. In G. R. Lyon (Ed.), Frames of reference foassessment of learning disabilities: New views on measurement issues (pp. 351-373). Ba Lindamood, P. C., Bell, N., & Lindamood, P. (1992). Issues in phonological awareness assessment. Annals of Dyslexia, 42, 242-259. Lindamood, P. C., Bell, N., & Lindamood, P. (1997). Sensory-cognitive factors in the controversy over reading instruction. The Journal of Developmental and Learning gstone MS, Rosen GD, Drislane FW, Proceedings of the National Academy of Science USA, 88, 7943-7947. thetis, N. (2000). Can current FMRI techniques reveal the microarchite ctcortex? Nature Neu
119 Lovegrove WJ, Bowling A, Badcock B, Blackwood M. (1980). Specific reading disabili ty: differences in contrast sensitivity as a function of spatial frequency. Science, 210, 439-440. Lovett, M. W., Borden, S. L., Deluca, T., Lacerenza, L., Benson, N. J., & Brackstone, D. (1994). Treating the core deficits of developmental dyslexia: Evidence of transfer ology, 30, 805-822. he erior cingulate cortex in cognitive control. Science, 288, 1835-1838. MacDsses. hophysics,24, 253-257. Maughan, B. (1995). Annotation: Long-term outcomes of developmental reading problems. Journal of Child Psychology and Psychiatry, 36, 357-371. McGowan, R. S., & Faber, A. (1996). Introduction to papers on speech recognition and perception from an articulatory point of view. Journal of the Acoustical Society of McGurk, H., & McDonald, J. (1976). Hearing lips and seeing voices. Nature, 264, of learning after phonologically and strategy-based reading training programs. Developmental Psych MacDonald, A. W., Cohen, J.D., Stenger, V. A., Carter, C. S. (2000). Dissociating trole of the dorsolateral prefrontal and ant onald, J., & McGurk, H. (1978). Visual influences on speech perception procePerception in Psyc Marshall, J. C., & Newcombe, F. (1973). Patterns of paralexia: A psycholinguistic approach. Journal of Psycholinguistic Research, 2, 175-199. America, 99, 1680-1682. 746-Milleway following speech and pure-tone stimuli. Laryngoscope, 105, 1305-1310. Mody7). Speech perception deficits in poor readers: Auditory processing or phonological coding? Journal of Experimental Montyslexics have difficulty accessing articulatory information? Psychological Research, 43, 235-243. Morgence, chology And Behavioral Neurology, 8, 143-162 748. n, S. J., Haughton, V. M., & Yetkin, Z. (1995). Functional magnetic-resonance imaging of the central auditory path , M., Studdert-Kennedy, M., & Brady, S. (199 Child Psychology, 64, 199. gomery, D. (1981). Do d an, C. B. (1897). Reading. Lipincottâ€™s magazine of popular literature and sci60, 139. Nadeau, S., & Crosson, B. (1995). A guide to the functional imaging of cognitive-processes. Neuropsychiatry, Neuropsy
120 Newman, S., Twieg, D. Carpenter, P. (2001). Baseline conditions and subtractiveneuroimaging. Human Brain Mapping, 14, 228-235. logic in Nickels, L. (1992). The Autocue? Self-generated phonemic cues in the treatment of a Noll, agnetic Resonance Imaging, 5, 49-56 il, Proceedings of the National Academy of Science USA, 89, 5951. Ogdehonological dysgraphia following left and right hemispherectomy. Neuropsychologia,34, 905-918. Ojemalization of memory, syntax, and sequential motor-phoneme identification systems. Science, 205, 1401-Oldfield, R. C. (1971). The assessment and analysis of handedness: The Edinburgh Patterson, K. E. (1982). The relation between reading and phonological coding. In A.W. Ellis (Ed.), Normality and pathology in cognitive functions (pp. 77-112). London: Patterson, K. E., Marshall, J. C., & Coltheart, M. (1985). Surface dyslexia. London: Pauleon KJ, Heather J, Myers R, Gadian DG, Frackowiak RS. (1995). Functional MR imaging correlations with positron Paulesu, E., Frith, C. D., & Frackowiak, R. S .J. (1993). The neural correlates of the Pennington, B. F. (1990). The genetics of dyslexia. Journal of Child Psychology & Pugh, K. R., Shaywitz, B. A., Shaywitz, S. E., Constable, R. T., Skudlarski, P., Fulbright, Roberts, A. H. (1965). A statistical linguistic analysis of American English. The Hague, The Netherlands: Mouton & Co. disorder of reading and naming. Cognitive Neuropsychology, 9, 155-182. D. C., Cohen, J. D., Meyer, C. H., & Schneider, W. (1995). Spiral Kâ€“Space MRimaging of cortical activation. Journal of M Ogawa, S., Tank, D. W., Menon, R., Ellermann, J.M., Kim, S-G., Merkle, H., & UgurbK. (1992) Intrinsic signal changes accompanying sensory stimulation: functional brain mapping using MRI. n, J. A. (1996). Phonological dyslexia and p ann, G. A., & Mateer, C. (1979). Human language cortex: Loc 1403. inventory. Neuropsychologia, 9, 97-113. Academic Press. Erlbaum. su E, Connelly A, Frith CD, Frist emission tomography. Initial experience using a cognitive activation paradigm on verbal working memory. Neuroimaging Clinics of North America, 5,207-25. verbal component of working memory. Nature, 362, 342-345. Psychiatry & Allied Disciplines, 31, 193-201. R. K., et al. (1996). Cerebral organization of component processesin reading.Brain, 119, 1221-1238.
121 Ruff, S., Cardebat, D., Dmonet, N. M., Dmonet, J. F. (2002). Enhanced response of the left frontal cortex to slowed down speech in dyslexia: an FMRI study. NeuroReport, 13, 1285-1289. Rumsger, S. D., et al. (1992).Failure to activate the left temporoparietal cortex in dyslexia. An . Rumsey, J. M., Andreason, P., Zametkin, A. J., King, A. C., Hamburger, S. D., Aquino, Rums., & Andreason, P. (1995). Phonologic and orthographic components of reading imaged with PET. Journal of the International Salmon, E., Van der Linden, M., Collette, F., Delfiore, G., Maquet, P.,Degueldre, A., et Sams, M., Aulanko, R., Hamalainen, M., Hari, R., Lounasmaa, O. V., Lu, S. T., Simola, ivity ce Letters, 127, 141-145. Sattlement of Children: Cognitive Applications Fourth Edition. San Diego: J. M. Sattler. Seidengberg, M. S. & McClelland, J. L. (1989). A distributed, developmental model of word recognition and naming. Psychological Review, 96,523-568. Shalliyslexic and ournal, 15, 151-178. ght, iler, D. P., Katz, L., et al. (1995). Sex differences in the functional organization of the brain for language. Nature, Shaywitz, S. E., Shaywitz, B. A., Pugh, K. R., Fulbright, R. K., Constable, R. T., Mencl, W. E., et. al. (1998). Functional disruption in the organization of the brain for ey, J. M., Andreason, P., Zametkin, A. J., Aquino, T., King, A. C., Hambur oxygen 15 positron emission tomographic study. Archives Neurology, 49, 527-34 T., et al. (1994). Right frontotemporal activation by tonal memory in dyslexia, An 015 PET study. Biological Psychiatry, 36, 171-80. ey, J. M., Nace, K Neuropsychological Society, 1, 180. al. (1996). Regional brain activity during working memory tasks. Brain, 119, 1617-1625. J. (1991). Seeing speech: Visual information from lip movements modifies actin the human auditory cortex. Neuroscien Sattler, J. (1993). Assessment of children. San Diego: J. M. Sattler. r, J. (2001). Assess ce, T., & Warrington, E. K. (1980). Single and multiple component central dsyndromes. In M. Coltheart, K. E. Patterson, & J. C. Marshall (Eds.), Surface dyslexia (pp. 119-145). London: Routledge & Kegan Paul. Share, D. L., Jorm, A. F., Maclean, R., & Matthews, R. (2002). Temporal processingreading disability. Reading and Writing: An Interdisciplinary J Shaywitz, B.A., Shaywitz, S. E , Pugh, K. R., Constable, R.T., Skudlarski, P., FulbriR. K., Bronen, R. A., Fletcher, J. M., Shankwe 373, 607-609. reading in dyslexia. Proceedings of the National Academy of Sciences, 95, 2636-2641.
122 Sokolov, A. N. (1972). Inner speech and thought. New York: Plenum. erger, C. D. (1983). Manual for Spelb the State-Trait Anxiety Inventory. Palo Alto, CA: Consulting Psychologists Press, Inc. Stano21, 360-406. Stanohonological-core variable-difference model. Journal of Educational Psychology, 86, 24-53. Stein,Talairach, J., & Tournoux, P. (1988). Co-planar stereotaxic atlas of the human brain. 3-Tallal P. (1980). Auditory temporal perception, phonics, and reading disabilities in poral processing. Annals of the New York Academy of Sciences, 682, 27. Thier999). Temporal sorting of neural components underlying phonological processing. NeuroReport, Torgeo types of phonological awareness training on word learning in kindergarten children. Journal of Torgesen, J. K., Wagner, R. K., & Rashotte, C. A. (1994). Longitudinal studies of Torgerocessing (CTOPP). Austin, TX: Pro-Ed Inc. ties. vich, K. E. (1986). Matthew effects in reading: Some consequences of individualdifferences in the acquisition of literacy. Reading Research Quarterly, vich, K. E., & Siegel, L. (1994). Phenotypic performance profile of children with reading disabilities: A regression-based test of the p J. (2001). The sensory basis of reading problems. Developmental Neuropsychology, 20, 509-534. Stein J, Walsh V. (1997). To see but not to read: the magnocellular theory of dyslexia. Trends in Neuroscience, 20, 147-152. dimensional proportional system: an approach to cerebral imaging. New York: Thieme Medical Publishers. children. Brain and Language 9, 182-198. Tallal, P., Miller, S., & Fitch, R. H. (1993). Neurobiological basis of speech: A case for the preeminence of tem ry, G., Boulanouar, K., Kherif, F., Ranjeva, J-P., & Dmonet, J-F. (1 10, 2599-2603. sen, J. K., Morgan, S., & Davis, C. (1992). The effects of tw Educational Psychology, 84, 364-370. phonological processing and reading. Journal of Learning Disabilities, 5, 276-291.sen, J., Wagner, R. K., & Rashotte, C.A. (1998). Comprehensive Test of Phonological P Torgesen, J. K., Wagner, R. K., Rashotte, C. A., Alexander, A. W., & Conway, T. (1998).Preventive and remedial interventions for children with severe reading disabiliLearning Disabilities: An interdisciplinary Journal, 8, 51-61.
123 Torgesen, J. K., Wagner, R. K., Rashotte, C. A., Burgess, S., & Hecht, S . (1997). Contributions of phonological awareness and rapid automatic naming ability to the nway, T. roup and individual responses to instruction. Journal of Educational growth of word-reading skills in secondto fifth-grade children. Scientific Studies of Reading, 1, 161-185. Torgesen, J.K., Wagner, R.K., Rashotte, C.A., Rose, E., Lindamood, P.C., Co(1999). Preventing reading failure in young children with phonological processingdisabilities: G Psychology, 91, 579-593. Truch Smievoll, A .I. pus callosum in dyslexic subjects--a structural MRI study. Neuropsychologia, 40, 1035-44. Wagnrole in the acquisition of reading-skills. Psychological Bulletin 101, 2, 192-212. Wech: n. ustin, Woodcock, R. (1987). Woodcock Reading Mastery Test-Revised. Circle Pines, MN: Zatorre, R. J., & Binder, J. R. (2000). Functional and structural imaging of the human auditory system. In A. W. Toga & J. C. Mazziotta (Eds.) Brain mapping: The Zatordde, A. (1992). Lateralization of phonetic and pitch discrimination in speech processing. Science, 256, 846-849. Zatorprocessing of speech: Review, replication, and reanalysis. Cerebral Cortex, 6, 21, S. (1994). Stimulating basic reading processes using auditory discrimination indepth. Annals of Dyslexia, 44, 60-80. von Plessen, K., Lundervold, A., Duta, N., Heiervang, E., Klauschen, F.,et al. (2002). Less developed cor er, R., Torgesen, J. (1987). The nature of phonological processing and its causal sler, D. (1997). Wechsler Adult Intelligence Scale-Third Edition. San Antonio, TXThe Psychological Corporatio Weiderholt, J. L., & Bryant, B. R. (1992). Gray Oral Reading Test: Third Edition. ATX: Pro-Ed. Wolf, M., & Bowers, P. (1999). The double-deficit hypothesis for the developmental dyslexias. Journal of Educational Psychology, 91, 1-24. American Guidance Service. systems. (pp. 365-402). San Diego, CA: Academic press, Inc. re, R. J., Evans, A. C., Meyer, E., & Gje re, R. J., Meyer, E., Gjedde, A., Evans, A. C. (1996). PET studies of phonetic 30.
BIOGRAPHICAL SKETCH Tim Conway ha s spent th e majority of his academic career studying about the brain sponsored research university, The Union Institute. This nontraditional educational experthroughout his undergraduate training. Since attending the University of Florida he has becom and its functions. He completed his undergraduate degree through a fully accredited ience allowed Tim to merge his occupational, research, and academic pursuits e a true blue Gator fan and will always strive towards the levels of academic and research excellence by which he was surrounded at the University of Florida. 124