LIST OF TABLES

(1)

[TONE PATTERN PERCEPTION OF INFANTS AT FAMILIAL RISK FOR DYSLEXIA AND CONTROLS DIFFER AT 17 MONTHS AS DEMONSTRATED BY

EVENT-RELATED POTENTIALS (ERPS)]

by [Jingying Lu]

A Master’s thesis submitted in partial fulfillment of the requirements for the degree of

Master of Arts Master of Science

(Clinical Linguistics)

at the Joint European Erasmus Mundus Master’s Programme in Clinical Linguistics (EMCL)

Rijksuniversiteit Groningen Universitat Potsdam

[August, 2010]

(2)

[TONE PATTERN PERCEPTION OF INFANTS AT FAMILIAL RISK FOR DYSLEXIA AND CONTROLS DIFFER AT 17 MONTHS AS DEMONSTRATED BY

EVENT-RELATED POTENTIALS (ERPS)]

[Jingying Lu]

Under the supervision of Professor [Ben Maassen] and [Laurie Stowe] at University of [Groningen]

(3)

ABSTRACT

Event-related potentials (ERPs) to tone patterns were obtained from 17-month-olds from families with dyslexic history and non-dyslexic families.

The main finding was a distinct P150-N250-P350-N450 peak complex, the P150 being the most prominent. A longer P150 latency was observed over the left hemisphere for the at-risk infants but not for the controls. An overall delay of N250 in the at-risks was found. The control group had a weak P150-N250 latency correlation while P150 and N250 latencies are not related in the at-risk group.

An MMN at 100-150 ms window at Fz and FCz was found in both groups. An MMN was found at the 350-500 ms window with the at-risks and an MMP in the same time window with the controls. The LR by Group interaction at the 500-650 ms window suggests that the controls had a larger right-than-left negativity while the at-risks had a larger left-than-right negativity.

The ERP differences found between at-risk infants and infants from non-dyslexic families suggest that difference between these two populations is not restricted to linguistic processing.

The obligatory auditory ERP peak latency differences and MMR differences might be additional risk factors besides the genetic component.

(4)

ACKNOWLEDGMENTS

This work would not have been possible without the support of many people.

I wish to express my deepest gratitude to my supervisors, Professor Ben Maassen and Dr.

Laurie Stowe who have been a continuous guidance from the initial stage of data screening to the final stage of data interpretation. Working with them has been extremely enlightening.

Special thanks to Professor Pieter Been who provided good suggestions for the initial step of data screening.

I’m thankful to EMCL colleagues Claudia Teickner and David Guise who share ideas and gave me encouragement.

I would like to acknowledge my thanks to the Neuroimaging Center of the University of Groningen for giving me the opportunity to access the EEG lab facilities.

I am indebted to all the participating babies and their families though I have not got the chance to meet them.

Finally, I am grateful to my husband and my parents for their support and understanding.

(5)

LIST OF TABLES

3.1 Mean latencies (in millisecond) of the ERP peaks for the standard stimuli for midline, left hemisphere, and right hemisphere electrodes. Standard errors are provided between brackets ...20 3.2 Mean amplitudes (in microvolt) of the ERP peaks for the standard stimuli for midline,

left hemisphere, and right hemisphere electrodes. Standard errors are provided between brackets ...21

(8)

LIST OF FIGURES

3.1 Grand average ERP forms elicited by the standard stimuli for the control infants (green), at-risk infants (red), and the difference waves (black) at Fz, FCz, F1, FC1, F3, FC3, F7, FT7, F2, FC2, F4, FC4, F8 and FT8 ...20 3.2 P150 latency for the left hemisphere (LH) and the right hemisphere (RH) for the control and at-risk infants ...22 3.3 Scatter plot of the Laterality index of P150 latency for the control and the at-risk group

...23 3.4 N250 latency for the left hemisphere (LH) and the right hemisphere (RH) for the control

and at-risk infants ...24 3.5 Scatter plot of the Laterality index of N250 latency for the control and the at-risk group

...25 3.6 P150-N250 latency correlation (statistics and scatter plot) for the control and at-risk group at all 14 electrodes (Fz, FCz, F1, FC1, F3, FC3, F7, FT7, F2, FC2, F4, FC4, F8, and FT8) ...25 3.7 MMR at fronto-central sites Fz and FCz at 100-150 ms. Standard stimuli is shown as std, deviant as dev. MMR is std subtracted from dev...26

(9)

3.8 Difference waves (responses to the standard subtracted from those to the deviant) for the controls (black waveform) and the at-risks (red waveform) at Fz and FCz. Negativity is plotted upwards...27 3.9 MMR at fronto-central sites Fz and FCz at 200-250 ms. Standard stimuli is shown as std, deviant as dev. MMR is std subtracted from dev...27 3.10 MMR at fronto-central sites Fz and FCz at 350-400 ms. Standard stimuli is shown as std,

deviant as dev. MMR is std subtracted from dev...28 3.11 MMR at fronto-central sites Fz and FCz at 400-450 ms. Standard stimuli is shown as std,

deviant as dev. MMR is std subtracted from dev...28 3.12 MMR at fronto-central sites Fz and FCz at 450-500 ms. Standard stimuli is shown as std,

deviant as dev. MMR is std subtracted from dev...29 3.13 MMR at 14 fronto-central sites at 100-150 ms. Standard stimuli is shown as std, deviant as dev. MMR is std subtracted from dev ...29 3.14 MMR at 14 fronto-central sites at 500-650 ms. Standard stimuli is shown as std, deviant as dev. MMR is std subtracted from dev ...30 3.15 MMR at left hemisphere (LH) and right hemisphere (RH) for controls and at-risks at

200-250 ms. MMR is obtained by subtracting the response to the standard stimuli from the deviant stimuli...30 3.16 MMR at left hemisphere (LH) and right hemisphere (RH) for controls and at-risks at

300-350 ms. MMR is obtained by subtracting the response to the standard stimuli from the deviant stimuli...31 3.17 MMR at left hemisphere (LH) and right hemisphere (RH) for controls and at-risks at

500-650 ms. MMR is obtained by subtracting the response to the standard stimuli from the deviant stimuli...31

(10)

(11)

Chapter 1 Introduction

A considerable part of the population is affected by reading difficulties or developmental dyslexia (hereafter referred to as dyslexia), prevalence estimates ranging from 4% to 9%

(Shaywitz, Shaywitz, Fletcher, & Escobar, 1990, as cited in Schulte-Körne, Deimel, Bartling,

& Remschmidt, 2001).

Research has been going on for decades for the clinical purpose of diagnosis and intervention or for the theoretical goal of relating brain functions (Kujala, Myllyviita, Tervaniemi, Alho, Kallio, & Naatanen, 2000; Tallal, 1980; Van Leeuwen, Been, Van Herten, Zwarts, Maassen, & Van der Leij, 2007).

According to a report published by a committee of the Health Council of the Netherlands (1995), “dyslexia is present when the automatization of word identification (reading) and/or word spelling does not develop or does so very incompletely or with great difficulty”. Following this definition, an estimate of 3% of Dutch children can be considered dyslexic. The reading and spelling problems they experience as a child usually persist into adulthood (Dhar, 2009; Eleveld, 2005). Difficulty with reading can have devastating consequences on cognitive, social, and academic development, and create behavioral problems. A systematic approach in diagnosis and treatment of the disorder is urgent.

(12)

Following this proposal the committee, the Dutch dyslexia association “Stichting Dyslxie Nederland” (SDN) developed a brochure where dyslexia is re-termed as a persistent impairment in reading and spelling at word-level. Dyslexia should be diagnosed on two criteria:

A. The level of reading and/or spelling is significantly below the person’s expected level, given his age and circumstances.

B. Difficulties in the acquisition and the accurate and/or fluent application of reading and/or spelling skills at word-level continue to exist after adequate remedial instruction and practice (Eleveld, 2005: 15)

Based on these criteria, however, children cannot be assessed until they reach school age or even later. Since the recognition that developmental dyslexia is a genetically transmitted learning disorder related to neurobiology (Snowling, 1995; Van Leeuwen et al., 2007), researchers have been probing into neurobiological predictors of dyslexia, even as early as 36 hours from birth (Molfese, 2000). The launch of a large-scale prospective program

“Identifying the core features of developmental dyslexia: A multidisciplinary approach”, demonstrates such an endeavor (Brochure, 2000).

As stated in the Brochure (p.7), one of the aims of the prospective study was to unveil the biological predictors that may allow for early identification and possible intervention. To achieve this goal, infants from 180 dyslexic families and from 120 control families were recruited from the Dutch population. Starting from 2 months of age, the infants, who have been followed over a period of 10 years, have been regularly assessed on auditory and visual perception, word recognition, and speech production (Project manual, 2003). For the first 5 years of age, data were obtained by assessing cortical responses using electrophysiological methods. More data were collected on different aspects of behavioral and cognitive development. When the children were older, they were tested for dyslexia based on the

(13)

previously mentioned criteria. The project goes on until 2011, when the last cohort of children will be tested. After all the infants have been tested we will regard the infants from dyslexic families as at-risk (Dhar, 2009). Then the following 4 groups are distinguished:

controls plus or minus dyslexia, and at-risks plus or minus dyslexia.

The present experiment was carried out as part of the large-scale study. Cortical sound discrimination was investigated by recording event related potentials (ERPs). The four-toned non-speech paradigm used in this experiment was selected based on the evidence that dyslexics have a deficit in processing the temporal aspects of non-speech sounds (Kujala et al., 2000). It is used as a supplementary one to the consonant discrimination paradigm (Project manual, 2003: 9). By measuring neural activation elicited by non-speech sound patterns and comparing obligatory auditory ERPs and mismatch responses (MMRs), we attempt to discover differences that might be factors for developing dyslexia in 17-month-old infants from dyslexic families. If any difference were found between at-risk infants and infants from non-dyslexic families, we would assume that difference between these two groups is not restricted to phonetic/phonological/linguistic processing. There might be other factors such as a rhythm perception disorder that leads to reading difficulties. It, however, would not disapprove the hypothesis that dyslexics/at-risks might have a deficiency in the linguistic domain.

Chapter 2 discusses two aspects of relevant literature, the long-latency peak pattern (P150-N250-P350-N450) and MMRs (the mismatch negativity or MMN, the mismatch positivity or MMP). Relating to MMRs, we discuss the theoretical background on basic auditory dysfunction in dyslexics/individuals at risk for dyslexia in comparison to normal controls.

(14)

Chapter 3, the centerpiece of the present thesis, focuses on the method including participants, stimuli and paradigm, experimental procedure, electrophysiological recording and averaging, data screening and analysis, and results.

The last chapter discusses the results, discovers the theoretical implications, proposes clinical suggestions, and faces the drawbacks of the present study.

(15)

Chapter 2 MMR and P150-N250-P350-N450 Related to Auditory Processing in Dyslexia

Brain imaging methods such as ERP indexes have been used to investigate the basic auditory sensory processing functions/dysfunctions in human subjects. ERPs are a suitable measure into the cortical auditory processing functions for infants at an early age for several reasons.

First, they do not require overt responses or active cooperation from participants in comparison to the measurement of task performance which requires behavioral responses (Kujala & Naatanen, 2001). Second, ERPs are a non-invasive measure, making it safe even for participants like young children who have a thin scalp (Van Leeuwen, Been, Van Herten, Zwarts, Maassen, & Van der Leij, 2008). Finally, ERPs of an early age may be valid precursors and predictors of later linguistic and neurocognitive performances (Guttorm, Leppanen, Poikkeus, Eklund, Lyytinen, & Lyytinen, 2005).

The purpose of this chapter is to provide a literature overview of ERP studies on dyslexic individuals/individuals at-risk for dyslexia: the MMR (mainly MMN) and the obligatory auditory ERP peak pattern (P150- N250-P350-N450).

2.1 The MMR

(16)

With MMR in mind, the major body of the review will be the MMN. The MMN is an extensively used index of ERP for assessing the temporal discrimination ability of the human auditory processing functions (Naatanen, 1992, 2000). It is usually defined as a fronto- centrally distributed negative response reflecting a pre-attentive neuronal change-detection mechanism, occurring when an infrequent physically variant/deviant stimuli encounters a well-established sensory memory trace of a frequently presented constant/standard stimuli (Naatanen, 2001, as cited in Shaul, 2008: 77). The concept entails several important ideas.

First, the MMN is independent of attention and is incurred in an automatic process. This advantage allows the exploration of brain activities of unconscious individuals and young children who do not understand verbal instructions and perform accordingly. It is, therefore, a clinically valuable tool to detect possible impaired auditory processing system in pre-verbal children born to a family with a history of reading difficulty (Cheour, Leppanen, & Kraus, 2000). Second, the MMN is sensitive to any perceptible change such as frequency, duration, intensity, tone order reversal, rhythm and even abstract rule violations (Naatanen, 2000;

Naatanen, Tervaniemi, Sussman, Paavilainen, & Winkler, 2001). Third, the MMN is distributed over the fronto-central scalp electrodes. The MMN is a negativity obtained by subtracting the brain’s response to the standard stimuli from that of the deviant stimuli (Shaul, 2008: 77).

The most commonly described MMN occurs at 100-300 ms post-stimulus onset (Naatanen, 1992, 2000, 2001; Schulte-Körne et al., 2001), while others have found later MMNs between 300-600 ms (Kraus, McGee, Carrell, Zecker, Nicol, & Koch, 1996; as cited in Schulte-Körne, Deimel, Bartling, & Remschmidt, 2001).

There is a strong genetic component in dyslexia (Pennington, 1995, as cited in Naatanen, 2003). Early indicators of dyslexia can be detected even in infancy. ERPs recorded at birth to speech and non-speech stimuli differentiate with high discrimination accuracy between

(17)

infants who become poor-reading and normal-reading at the age of 8 (Molfese, 2000). The auditory processing skills of school-aged children as well as adults with reading difficulties differ from those of normal reading individuals (Richardson, Leppanen, Leiwo, & Lyytinen, 2003: 386). Pennington (as cited in Shaul, 2008: 79) found an attenuated MMN in children from families with a history of reading difficulties or dyslexia compared to children from families without dyslexic history. Behavioral evidence such as Tallal’s (1980) has demonstrated that school-aged dyslexics performed significantly poorer than non-dyslexic controls in a temporal order judgment task.

The difficulties with dyslexics are suspected to concern more general deficiency rather than deficits confined to phonetics/phonology/linguistics in auditory perception, such as discriminating the temporal aspects of sounds (such as silent gaps between sounds or rhythm) (Tallal, 1980; Kujala et al., 2000). Benasich & Tallal (2002) found a significantly lower threshold of rapid auditory processing ability for infants born to families with a history of reading difficulty. They propose that early deficits in rapid auditory processing abilities both precede and predict later language delays, suggesting the essential role of basic non-linguistic, rapid spectro-temporal processing in early language development. The temporal features of sounds provide key information for speech perception (Kujala, Kallio, Tervaniemi, &

Naatanen, 2001). Rosen (1992) has even proposed that temporal aspects of speech signals serve as the primary cue to speech perception. The impairment in perceiving and discriminating temporal sound features could lead to severe linguistic disturbances (Kujala et al., 2001). In addition, speech rhythm is considered as one of the earliest cues used by infants to discriminate syllables (Goswami, Thomson, Richardson, Stainthorp, Hughes, Rosen, &

Scott, 2002). A deficit in the perception of rhythmic timing is suspected to be the cause of the speech perception difficulty (Stein, 1994; Goswami et al., 2002).

(18)

Schulte-Körne, Deimel, Bartling, & Remschmidt (1999), using a complex tonal pattern, found a diminished MMN in their dyslexic subjects. The authors suspect that it may be the temporal information embedded in speech sounds, rather than the phonetic information per se, that resulted in the reduced MMN. Schulte-Körne, Bartling, Deimel, & Remschmidt (as cited in Shaul, 2008: 79) observed that changes in the temporal order of patterned tones generated a smaller MMN in dyslexics in comparison to non-dyslexic controls. Similarly, Kujala, Belitz, Tervaniemi, & Naatanen (2003) found in dyslexic subjects reduced MMN amplitudes for tone-pair order reversal where an additional sound followed a tone pair but not when the third tone preceded the tone pair. In both cases, rapidly following sounds or sound sequences were suspected to mask the preceding sound pattern. This is backward masking effects that were reported to occur with dyslexic individuals only (Schulte-Körne et al., 1999; Kujala et al., 2003). The processing of the preceding sounds were interfered by the contextual sounds that come immediately afterwards. The information flow is usually fast in speech and certain acoustic features become “muffled” phonetic representations (Merzenich, Jenkins, Johnston, Schreiner, Miller, & Tallal, 1996; Kujala et al., 2000; Stein, 1994). Impaired temporal discrimination and vulnerability to masking effects caused by fast incoming sounds might underlie the phonological difficulties in dyslexics (Kujala et al., 2003). For example, normal MMN to temporal interval shortening between paired tones was found in both dyslexics and control adults (Kujala et al., 2000) whereas no MMN was elicited in the dyslexics when the same shortening of the inter-tone interval occurred in the same tone pair except that it was both preceded and followed by an extra tone with a short interval. In the control group, however, a distinct MMN was elicited even with this more complex rhythmic pattern stimulus.

Kujala and colleagues (2000) found that deviant patterns elicited two consecutive MMNs in the control subjects in comparison to the dyslexic subjects. The ERP to deviants was

(19)

significantly more negative than that to standards at 50-100 ms, 200-300 ms, and 350-450 ms from the onset of the third tone in the deviant pattern. In dyslexic subjects no significant differences between these responses were found at windows earlier than 400-450 ms. The MMN amplitude for deviant stimulus patterns differed significantly between groups at 200- 250 ms. In addition, there was a significant Group by Hemisphere interaction for the response at 400-450 ms, the response being distributed over the cerebral hemisphere in the dyslexics and right-hemisphere dominant in the controls.

In a longitudinal study, Kushnerenko (2003) reported an MMP at 250-350 ms window with infants. This difference response partly overlapped the MMN. She argued that this positive difference might be an infant analogue of the early phase of the adult P3a component which indexes the automatic attention orientation to novel and distracting stimuli.

2.1.1 The generation process of MMN related to patterned sounds

The central auditory system for event perception involves the integrating mechanism of sequential information. The central auditory system can essentially integrate successive information into meaningful entities for auditory event perception (Yabe, Tervaniemi, Reinikainen, & Naatanen, 1997).

No matter the stimuli are tone order reversals (Schulte-Körne et al., 1999; Kujala et al., 2003) or changes in rhythmic patterns (Kujala et al., 2000), the generation process of MMN involves a mismatch of a deviant stimulation and transient abstract sound-pattern neural memory traces in the auditory cortex. To understand how the standard sound pattern is processed is crucial to understanding MMN as an index of stimulus change detection mechanism. The standard stimulus is first perceived and discriminated. Any change to the standard stimulus produces the so-called deviant stimulus, which is then detected. The response at the point of low probability deviant detection will be different from that to the high probability standard stimulus. Thus, a mismatch with the neural representation of the

(20)

standard stimulus stored in the auditory sensory memory is produced (Naatanen, 1992, as cited in Yabe, 2002). The process happens all in a few milliseconds.

How the sound stimuli are presented plays a crucial role in eliciting MMN. The differentiation between what is standard and what is deviant is not always clear-cut, especially when rhythm is involved (Picton, Alain, Otten, Ritter, & Achim, 2000: 117). In fact, it can be dynamic depending on the way the stimuli are presented. For instance, in the four-toned pattern used by Winkler & Schröger (1995), the standard may be a simple event if the four-tone-sequence is considered as a single stimulus and a patterned standard if the tones comprising the stimulus are considered separately. The difference would largely depend on the rapidity at which the successive tones are presented. It could also depend on the development of brain functions.

2.1.2 Auditory sensory memory and temporal window of integration (TWI)

To understand language, one needs to perceive and process constant inflow of acoustic information efficiently. Yet, the brain does not cope with each time point of the auditory event as discrete. Rather, it selectively assigns different types of auditory information to two types of sensory memory: short-form and long-form. The short-form sensory memory has the duration of about 200 ms and the long-form 10-20 s (Naatanen, 1992, as cited in Yabe, 2002).

The MMN depends on the presence of a memory trace formed by the frequently presented stimuli (Naatanen, 2000). The existence of MMN presupposes the existence of the auditory sensory memory (Naatanen, 1992, 2000).

Acoustically and temporally integrated sound representations are formed at an early stage of central sound processing (Naatanen, 1992, as cited in Ceponiene, Yaguchi, Shestakova, Alku, Suominen, & Naatanen, 2002). The short-form auditory sensory memory explains the existence of the brain’s mechanism called the temporal window of integration (TWI), which buffers and integrates the closely presented auditory information into single information units

(21)

(Yabe, 2002). The duration of the TWI is estimated to be 160-170 ms (Yabe, 2002) and 150- 200 ms (Naatanen, 2001: 2).

In an adult study, Yabe et al. (1997) discovered an MMN to an occasional omission of the second tone of a closely spaced tone pair (tone duration of 50 ms) with stimulus onset asynchrony (SOA) shorter than 150 ms. They found no MMN to the omission of the second tone in the pairs with SOA longer than 200 ms. Based on the results, their speculation was that the TWI is about 150 ms to 170 ms. In other words, the second of two consecutive sounds should come into the central auditory system no later than 150-170 ms to be considered by the auditory sensory memory as one unitary event. As soon as the memory representation of invariance has been established, any alteration to the unit in any feature, in theory, should incur an MMN (Picton et al., 2000).

Kujala and colleagues (2000) found a large MMN to an occasional addition of the second tone in a repetitive train of tones with SOA of 120 ms in the adult subjects. The results were in consistence with Yabe and colleagues’ (1997) postulation of the existence of TWI and its duration is shorter than 150-170 ms.

2.1.3 Hemispheric lateralization

According to Jamison, Watkins, Bishop, & Matthews (2006), hemispheres are functionally asymmetrical. The left hemisphere is specialized in rapid temporal processing.

Using fMRI, the authors found that temporal variation of pure tone stimuli elicited left- hemisphere lateralized responses in healthy adults. They argue that the left-hemispheric specialization in basic auditory processing even without linguistic information may lead to the left-hemispheric specialization in speech processing.

The left-hemisphere dysfunction in dyslexic individuals has been evidenced by two studies. A reduced lateralization of the MMN over the left hemisphere in children at risk for dyslexia during phoneme processing (Lyytinen, Guttorm, Huttunen, Hamalainen, Leppanen,

(22)

& Vesterinen, 2005) points to the left-hemisphere disadvantage in the population. Similarly, in an adult study, Kujala et al. (2003), using pitch change in the stimuli, found a diminished MMN to the left hemisphere.

It is possible that temporal integration windows may be better thought of as relative rather than absolute and might depend on linguistic experience. The nature of functional asymmetry and lateralization in auditory cortex is a complex issue that remains unresolved (Jamison et al., 2006). It is, therefore, a better way to examine lateralization individually.

2.2 P150-N250-P350-N450 peak pattern

The P150-N250-P350-N450 peak pattern reflects basic or exogenous auditory processing (Hamalainen, Leppanen, Guttorm, & Lyytinen, 2007). Auditory stimuli typically evoke a P1- N1-P2-N2 peak pattern (Been, Van Leeuwen, Van Herten, Maassen, Van der Leij, & Zwarts, 2008) in both children and adults. The auditory ERP is particularly suitable for studying auditory processing in children (Kabel, Mesallam, & Ghandour, 2009). The P150, N250, P350, and N450 waveform pattern is already identifiable at birth. From the age of 12 months on, the infantile ERP morphology stays stable throughout the next 10 years of life (Kushnerenko, Ceponiene, Balan, Fellman, Huotilainen, & Naatanen, 2002). Therefore, a similar peak complex is expected from the 17-month olds of the present experiment.

Furthermore, children’s long-latency obligatory auditory ERPs are dominated by the P150 and N250 peaks. During early childhood (1-4 years), the P150 is the most predominant peak (Kushnerenko et al., 2002; Pang & Taylor, 2000).

The N2 (with the adult latency of 220-270 ms) elicited by frequent repetitive stimuli was reported mostly in children. In language-impaired children, the N2 was smaller in amplitude and longer in latency than in their age-matched controls (Ceponiene, Cheour, & Naatanen, 1998; Kushnerenko, 2003).

(23)

Chapter 3 Experiment

3.1 Method

3.1.1 Participants

The children of the current study were a subset of the total cohort of children included in the longitudinal study of the Dutch Dyslexia Program. In total, 25 children participated in the current study of whom 14 (6 boys and 8 girls) were at genetic risk for developmental dyslexia and 11 (7 boys and 4 girls) were typically-developing controls. All children were 17 months of age and none had hearing loss or neurological impairment as reported by the accompanying parent.

3.1.2 Preliminary inclusion criteria

Dyslexia in the parents and family members was assessed by a test battery including speed tests on real word and pseudo-word reading, and verbal IQ. Criteria for dyslexia were a reported history of reading difficulty and either a score on the single-word reading test or pseudo-word reading test below 10% or both reading scores below 25%. Also included was a discrepancy criterion of a more than 60% difference between a high score on the WAIS Comprehension subscale and the score on one of the reading tests (Van Herten, Pasman, Van

(24)

Leeuwen, Been, Van der Leij, Zwarts, & Maassen, 2008) Inclusion of the children in the at- risk group was only possible if at least one parent and one first-degree relative fulfilled these criteria. To exclude familial risk of dyslexia in the control group, both parents of the control children had to score above criterion on the dyslexia tests and were interviewed about their own and their relatives’ literacy history. All participating families received oral and written information about the study. Informed consent was obtained from all parents of the participating children. The study protocol was approved by a medical ethics committee. The accompanying parent was paid travel expenses and the children received a small present (Van Herten et al., 2008).

3.1.3 Stimuli and paradigm

Stimuli were non-word four-tone patterns adapted from Kujala et al. (2000). The tones were identical in terms of frequency (500 Hz), duration (30 ms with 5-ms rise and fall times), and intensity (75 dB, SPL). The within-pattern tone onset asynchronies (SOA) were 200, 150, and 50 ms in the standard pattern (83% of the total 420 trials, n=350) and 200, 50, and 150 ms in the deviant pattern (17% of the total 420 trials, n=70). The standard and deviant stimuli were randomized with at least three standard stimuli in-between two deviant stimuli. Inter- stimulus interval (ISI) was kept at 800 ms. Each experimental session lasted approximately 11 minutes.

3.1.4 Procedure

The children were seated in a comfortable car-chair during the experiment. A silent video was played to attract their attention because the mismatch negativity (MMN) is best recorded when the participant is not attending to the auditory stimuli. When the participant attends to the stimuli, the MMN is difficult to recognize (Picton, Bentin, Berg, Donchin, Hillyard, Johnson Jr., Miller, Ritter, Ruchkin, Rugg, & Taylor, 2000: 129). The stimuli were

(25)

presented via two speakers, one at each side of the child at a distance of about one meter. The child was monitored by a hidden camera.

3.1.5 Electrophysiological recording and averaging

EEG was recorded with 64 Ag/AgCl-sintered electrodes mounted in an elastic cap (EasyCap), placed according to the 10-20 standard system and referred to the right mastoid (M2). Electrode impedance was kept below 15 kΩ. The electro-oculogram (EOG) was recorded bipolarly: vertical EOG by an electrode above and below the left eye and the horizontal EOG from the outer canthus of each eye (Project Manual, 2003). The signals were amplified with a SynAmp Model 5083 amplifier (Neuroscan Inc., ElPaso, Tx, USA) using a band-pass filter of 0.1 to 100 Hz, and digitally sampled online at 500 Hz.

The signals were re-referenced offline to the left mastoid (M1) electrode. The implicit reference (M2) was included in the calculation of the new reference, under the assumption that bias created by referencing all electrodes to M2 was evened out.

We then went through a procedure to detect evidence of awake and sleep state EEG respectively so that we could include awake children only in the final analyses. Raw EEG was examined in stretches of 60 seconds. The awake EEG criteria were (1) the posterior basic rhythm, varying from 6 to 8 Hz, (2) widely-scattered frequencies in the 2-5 Hz range, and (3) dramatic EOG deflections and accompanying head-movement-caused occipital activities. The sleeping EEG criteria were (1) diffusely preponderant irregular high voltage of 1-3 Hz with the maximum amplitude within 0.75-2 Hz, and medium to high voltage of 4-6 Hz, (2) spindle activities of 12-14 Hz, which are more stretched than multi-activity signals and usually with no EOG activity, (3) K-complex and the vertex waveform of 14-15 Hz, large with an abrupt sharp morphology and prominent over central region, Cz electrode in particular, and (4) general slow and synchronized waveforms (Niedermeyer & Lopes da Silva, 2005:144-145).

REM (rapid eye movements) sleep, though rare, and drowsiness were also inspected.

(26)

According to the first criterion, we compared the amplitude of the 0.75-2 Hz frequency range of EOG and the other channels. FFT (Fast Fourier Transformation) was used to this end. The signals with much higher voltage of the EOG than the overall EEG voltage were classified as a possible sleeping piece of recording. If this and at least one of the other criteria were met the child was classified as being asleep. All recordings with partial sleep episodes were considered as having incomplete waking time, thus were excluded. Finally, the data were carefully examined by an independent experienced EEG expert. In case of disagreement with the first examiners, the EEG signals were re-checked.

Prior to segmenting the EEG files, a band-pass filter of 0.5 to 15 Hz was applied (Butterworth zero phase shift, slope of 48 dB/oct). Also, a notch filter of 50 Hz was employed to attenuate the line frequency/noise because filters are not perfect and do not completely eliminate frequencies beyond the cutoff frequency (Picton et al., 2000). The continuous EEG was segmented into 1200-ms intervals including 200 ms pre- and 1000 ms post-stimulus onset (tone-pattern) time.

We then corrected ocular EEG. It was carried out by the Gratton and Coles algorithm (Gratton, Coles, & Donchin, 1983). Blink detection was done by algorithm. Overlapped segments were allowed. All epochs were aligned over a 200-ms pre-stimulus onset (baseline).

Epochs in which EEG and EOG amplitudes exceeded ±100 µV were automatically excluded from averaging and further analysis. A second -200-ms baseline was applied. All stimuli were then divided into standard, pre-deviant standard, deviant, and post-deviant standard epochs of 1200 ms with -200 ms pre-baseline and 1000 ms post-baseline window. Each type of epochs was averaged with standard deviation.

Only those children who were awake, with complete recording time, and whose data had over 143 artifact-free epochs for all-but-post-deviant standards and 36 artifact-free epochs for the deviants entered the final analysis. We started off with 38 children. 11 control children (7

(27)

boys and 4 girls) and 14 at-risk children (6 boys and 8 girls) were included in the final analyses. Nine children were excluded because they were in sleeping (or partly sleeping) status; one child had incomplete recording time; and the other three were not included due to artifact rejection.

3.1.6 Data measurement and analysis

Responses from 14 electrodes (Fz, FCz, F1, FC1, F3, FC3, F7, FT7, F2, FC2, F4, FC4, F8, and FT8) were analyzed although ERPs were recorded from all 64 electrodes over the scalp. These electrodes cover the frontocentral region of the scalp and can provide data that are representative of the activation in each hemisphere during the recording, thus allowing investigation of hemispheric dominance (lateralized to the left or right hemisphere) in processing the non-speech acoustic stimuli.

Two sets of analyses were performed, one on the responses to the standard, the pre- deviant standard, and post-deviant standard stimuli added together and the other on the responses to the deviant stimuli compared to the all-but-post-deviant standards.

3.1.7 Peak pattern analyses

The first set of analyses focused on the P150-N250-P350-N450 peak pattern. Under the assumption that the same stimuli would yield the same brain responses, we averaged over the signals to the standard stimuli, including the pre-deviant standard stimuli, and the post- deviant standard stimuli altogether for the purpose of maximizing the signal-to-noise ratio as compared to averaging the responses to the standards excluding pre- and post- deviant standards. The peaks were automatically detected using Brain Vision Analyzer. P150 was searched within the time window of 50-200 ms, N250 in window 150-350 ms, P350 in window 250-450 ms, and N450 in window 350-550 ms (Brain Vision Analyzer, Brain Products, GmbH, Munich, Germany). The presence of peaks in each individual infant was identified and checked visually by the examiner afterwards. For all infants, peaks were

(28)

detected. To find out whether the control and at-risk infants differ in processing of the standard stimuli, a repeated measures MANOVA was performed to compare the amplitudes around the peak and the latencies of the peak over the left hemisphere electrodes (F1, FC1, F3, FC3, F7, FT7) and the right hemisphere electrodes (F2, FC2, F4, FC4, F8, FT8), with Laterality (two levels: left hemisphere vs. right hemisphere) as the within-subjects factor and Group (two levels: control vs. at-risk) as the between-subjects factor. The emphasis was on group differences and hemispheric specialization. The responses at the midline sites (Fz, FCz) were analyzed with a one-way ANOVA, focusing on group differences only.

3.1.8 MMR analyses

The second set of analyses was on MMR (MMN and MMP). As in the peak analyses, we averaged together the standard and pre-deviant standard stimuli so that the signal-to-noise ratio would be higher as compared to averaging standard stimuli only. We did not include the post-deviant standard stimulus because the brain activation might be influenced by the deviant stimulus before it. Assuming that there is no activation difference to the first two tones in each stimulus, the mismatch responses are not expected to be generated until at least 100 ms from the stimulus-change onset at 250 ms, i.e. the third tone-onset in the deviant stimulus (Naatanen, 1992, 2000, 2001) or even later (Kraus, McGee, Carrell, Zecker, Nicol,

& Koch, 1996). We, therefore, set up a new baseline which begins at 50 ms after the stimulus onset or -200 ms before the stimulus-change onset. The averaged standards and deviants were adjusted to this baseline accordingly. Averages consisting of over 143 trials of standards and pre-deviant standards and over 36 trials of deviants entered statistical analysis. An independent sample t-test showed that the control and at-risk groups did not differ in the mean number of epochs included in the average (pre-deviant-standard: control: 63, standard deviation 4.5, at-risk: 64, standard deviation 3.2, p=.74 (2-tailed); deviant: control: 63, standard deviation 5.0, at-risk: 64, standard deviation 3.1, p=.48 (2-tailed)). Following Kujala

(29)

et al. (2000), the possible MMR amplitude was measured as the mean amplitude over the consecutive 50-ms periods starting 50 ms from stimulus-change onset for a 450-ms interval.

The mean amplitude values for these time windows were entered into a repeated measures MANOVA. As in the peak pattern inspection, only data from Fz, FCz, F1, FC1, F3, FC3, F7, FT7, F2, FC2, F4, FC4, F8, and FT8 entered the analyses as the largest mismatch effects were expected from the frontal and central sites (Cheour et al., 2000). Midline and lateral electrodes were analyzed separately so that laterality effects could be analyzed. The midline analysis included two electrodes (Fz, FCz) with Group (two levels: control vs. at-risk) as the between-subjects factor and Stimulus (two levels: standard vs. deviant) as the within-subjects factor. The laterality analysis included the averaged amplitude of 6 electrodes over each hemisphere (F1/F2, F3/F4, F7/F8, FC1/FC2, FC3/FC4, FT7/FT8) with Group (two levels:

control vs. at-risk) as the between-subjects factor, Stimulus (two levels: standard vs. deviant) and Laterality (two levels: left vs. right) as the within-subjects factors.

3.2 Results

3.2.1 Peak pattern

Visual observation revealed that the long-latency ERP waveforms to the standard stimuli formed a consistent peak pattern for both control and at-risk infants, as demonstrated in FIGURE 3.1. In both groups, the waveforms showed a distinct P150-N250-P350-N450 peak complex, most prominent over the frontocentral electrodes. The P150 appeared to be the largest peak. This ERP peak pattern is comparable to those reported in previous studies with children (Ceponiene, Shestakova, Balan, Alku, Yiaguchi, & Naatanen, 2001; Kushnerenko, 2003). The averaged peak amplitudes for the 14 electrodes (Fz, FCz, F1, FC1, F3, FC3, F7, FT7, F2, FC2, F4, FC4, F8 and FT8) were calculated. For the control group, P150 was 7.05 microvolts (mV) being the largest of the four. N250 was -.13 mV. P350 was 3.80 mV. N450

(30)

was -2.88 mV. Compared to the control children, the at-risk group showed an averaged 6.73 mV P150, -.93 mV N250, 3.81 mV P350 and -3.08 mV N450.

FIGURE 3.1: Grand average ERP forms elicited by the standard stimuli for the control infants (green), at-risk infants (red), and the difference waves (black) at Fz, FCz, F1, FC1, F3, FC3, F7, FT7, F2, FC2, F4, FC4, F8 and FT8

We then calculated the average amplitude and latency for the midline, the left hemisphere, and the right hemisphere electrodes, separately. The results are presented in Table 3.1 (for the mean peak amplitudes) and Table 3.2 (for the mean peak latencies).

Table 3.1: Mean amplitudes (in microvolt) of the ERP peaks for the standard stimuli for midline, left hemisphere, and right hemisphere electrodes. Standard errors are provided between brackets.

Midline LH RH

(31)

Control At-risk Control At-risk Control At-risk P150 7.00 (1.35) 6.97 (0.71) 6.86 (1.31) 6.63 (0.60) 7.29 (1.21) 6.60 (0.75) N250 -0.00 (0.80) -0.46 (0.46) -0.39 (0.91) -1.30 (0.48) -0.01 (0.78) -1.03 (0.47) P350 3.32 (0.84) 3.97 (0.56) 3.93 (0.74) 3.86 (0.50) 4.14 (0.61) 3.59 (0.59) N450 -2.89 (0.75) -2.64 (0.46) -2.95 (0.78) -3.21 (0.44) -2.79 (0.81) -3.38 (0.47)

Table 3.2: Mean latencies (in millisecond) of the ERP peaks for the standard stimuli for midline, left hemisphere, and right hemisphere electrodes. Standard errors are provided between brackets.

Midline LH RH

Control At-risk Control At-risk Control At-risk P150 150.45

(6.56)

155.86 (4.05) 155.15 (4.86)

156.43 (3.66)

157.79 (4.65)

154.57 (4.11) N250 257.45

(10.11)

273.36 (8.27) 250.79 (8.73)

268.93 (7.26)

258.03 (8.14)

273.62 (7.00) P350 356.55

(7.82)

356.36 (7.37) 361.76 (5.48)

358.26 (6.07)

362.12 (4.87)

354.36 (5.65) N450 466.91

(6.19)

462.79 (5.78) 466.27 (5.83)

466.74 (5.09)

468 (6.47) 466.45 (4.95)

3.2.1.1 Peak amplitude

The MANOVA (factors: Laterality, Group) showed no Group difference in amplitude for any of the four peaks. Nor did it reveal any significant Laterality by Group interaction.

3.2.1.2 Peak latency

(32)

Laterality by Group interaction tended to be significant (F(1, 23)=4.14, p=.054) for the P150 latency, suggesting that the P150 peak appeared later in the left hemisphere than in the right hemisphere in the at-risk infants (mean latency value of 156.43 ms and 154.57 ms respectively), while in the control group the peak occurred earlier in the left hemisphere than in the right hemisphere (mean latency values of 155.15 ms and 157.79 ms respectively), shown in FIGURE 3.2.

P150 latency

152 153 154 155 156 157 158 159

control risk

Mean latency (ms)

LH RH

LH 155.15 156.43

RH 157.79 154.57

FIGURE 3.2: P150 latency for the left hemisphere (LH) and the right hemisphere (RH) for the control and at-risk infants

A bivariate correlation of the P150 latency over the left and right hemisphere was calculated. We found a significant positive correlation (p<.001; r=.956 for the controls, r=.921 for the at-risks) between the left and right hemisphere for both the controls and the at- risks. The difference in latency laterality is further demonstrated by a scatter plot in FIGURE 3.3, showing a “laterality index” for all infants separately for each Group. The “laterality index” was calculated by subtracting the mean value of the right hemisphere responses from that of the left hemisphere responses (Van Herten, Pasman, Van leeuwen, Been, Van der Leij, Zwarts, & Maassen, manuscript). The index is shown by the Y-axis and infant identity by the X-axis. The infants labeled as 3** belong to the control group and 4** infants were labeled

(33)

as at-risk. From the plot, we observed that 6 out of 11 control and 6 out of 14 at-risk infants had a laterality index that falls below the zero line (indicating a longer right-than-left hemisphere latency) and 4 out of 11 control and 7 out of 14 at-risk infants had a laterality index that falls above the zero line (indicating a longer left-than-right hemisphere latency).

One infant of each group had a laterality index close to zero. Among the 6 controls who had a right-longer-than-left latency, 5 had an overriding longer latency in the right hemisphere (Right-Left>5 ms), indicating that about half of the control infants processed the auditory information faster in the left hemisphere than in the right hemisphere. Among the 7 at-risk infants who had a left-longer-than-right latency, 2 had a much longer latency in the left hemisphere (Left-Right>10 ms), suggesting that 2 of the at-risk infants were much slower in auditory processing in the left hemisphere than in the right hemisphere and might have a left hemisphere dysfunction.

FIGURE 3.3: Scatter plot of the Laterality index of P150 latency for the control and the at- risk group

(34)

The N250 latency analysis includes factors of Laterality by Group (2*2). There is a close-to-significant main effect of Laterality (F(1, 23)=4.07, p=.055) shown in FIGURE 3.4.

Surprisingly, there is no significant difference between the Groups (F(1, 23)=2.54, p=.125).

There is no Laterality by Group interaction (F(1, 23)=.19, p=.67).

We calculated a bivariate correlation between the left and right hemisphere latency. The result revealed a significant correlation between the left and right hemisphere latency for the control group (r=.714, p=.014) as well as the at-risk group (r=.976, p<.001).

N250 latency

230 240 250 260 270 280

control risk

Mean latency (ms)

LH RH

LH 250.79 268.93

RH 258.03 273.62

FIGURE 3.4: N250 latency for the left hemisphere (LH) and the right hemisphere (RH) for the control and at-risk infants

As in the P150, the laterality index of N250 was also calculated and is depicted in FIGURE 3.5. A further analysis of the data revealed that 5 of the 11 controls showed longer right-than-left latency meaning they had faster left hemispheric response, 1 had equal right- left latency, and 5 had shorter right-than-left latency. In contrast to the results for P150, 12 out of the 14 at-risks showed longer right-than-left latency with less variability than the control group as in FIGURE 3.5, the triangles (representing the at-risk children) are more clustered than the circles (representing the control children).

(35)

FIGURE 3.5: Scatter plot of the Laterality index of N250 latency for the control and the at- risk group

FIGURE 3.6: P150-N250 latency correlation (statistics and scatter plot) for the control and at-risk group at all 14 electrodes (Fz, FCz, F1, FC1, F3, FC3, F7, FT7, F2, FC2, F4, FC4, F8, and FT8)

(36)

Furthermore, a correlation test of the averaged P150 and N250 latency over all 14 frontocentral electrodes in each group yielded that P150 had a weak positive correlation with N250 in the controls (r=.545, p=.083) while no such correlation was found in the at-risk children (r=.416, p=.139), presented in FIGURE 3.6.

Neither the P350 nor the N450 peak latency differed between the control and the at-risk infants (F (1, 23)=.51, p=.48; F(1, 23)=.01, p=.94). Also, there was no Laterality effect (F(1, 23)=1.33, p=.26; F(1, 23)=.154, p=.70) or Laterality by Group interaction (F(1, 23)=1.93, p=.17; F(1, 23)=.30, p=.59).

The one way ANOVA for the midline electrodes (Fz, FCz) did not differentiate the groups in either amplitude (F<1) or latency (F<1) for any of the four peaks.

3.2.2 MMR

The repeated measures MANOVA (Stimulus×Group) for the midline electrodes (Fz, FCz) revealed an early MMN (100-150 ms, F(1, 23)=4.048, p=.056) in both groups but there is no difference between the groups, shown in FIGURE 3.7. The at-risk infants showed a slightly larger negativity than the controls. The waveforms were depicted in FIGURE 3.8.

M M R 100-150 ms

-2.5 -2 -1.5 -1 -0.5 0

control risk

Mean amplitude std

dev

std -1.17 -0.55

dev -2.34 -2.05

MMR -1.17 -1.5

FIGURE 3.7: MMR at fronto-central sites Fz and FCz at 100-150 ms. Standard stimuli is shown as std, deviant as dev. MMR is std subtracted from dev.

(37)

FIGURE 3.8: Difference waves (responses to the standard subtracted from those to the deviant) for the controls (black waveform) and the at-risks (red waveform) at Fz and FCz.

Negativity is plotted upwards.

MMR 200-250 ms

-8 -6 -4 -2 0

control risk

Mean amplitude std

dev

FC

fronto-central scalp

-100 -50 0 50 100 150 200 250 300 350 400 450 500 550 600 650 700 750

Fz

Standard stimuli Deviant stimuli

std -5.81 -5.23

dev -3.66 -4.1

MMR 2.15 1.13

(38)

At the 200-250 ms window, we found a close-to-significant MMP for both groups (F(1, 23)=4.215, p=.052) shown in FIGURE 3.9.

In the three successive windows from 350 to 500 ms we found an MMN for the at-risk infants and an MMP for the control infants (350-400 ms, F(1,23)=4.054, p=.056; 400-450 ms, F(1, 23)=6.030, p=.022; 450-500 ms, F(1, 23)=7.060, p=.014). The three windows are shown in FIGURE 3.10, 3.11 and 3.12.

MMR 350-400 ms

-5 -4 -3 -2 -1 0

control risk

Mean amplitude std

dev

M M R 400-450ms

-5 -4 -3 -2 -1 0

control risk

Mean amplitude

std dev

std -4.04 -3.24

dev -1.86 -4.62

MMR 2.18 -1.38

std -3.01 -1.49

dev -1.32 -4.51

MMR 1.69 -3.02

(39)

M M R 450-500ms

-4 -3 -2 -1 0

control risk

Mean amplitude std

dev

The MANOVA (Stimulus×Group) for all 14 electrodes over the frontocentral scalp averaged found an early MMR at 100-150 ms (F(1, 23)=3.869, p=.061) (FIGURE 3.13) and a late MMR at 500-650 ms (F(1, 23)=3.747, p=.065) (FIGURE 3.14)in both groups but it did not differentiate the groups.

M M R 100-150 ms

-2 -1.5 -1 -0.5 0

control risk

Mean amplitude std

dev std -2.49 -0.66

dev -0.37 -2.89 MMR 2.12 -2.23

std -0.41 -0.26

dev -1.56 -1.29

MMR -1.15 -1.03

FIGURE 3.13: MMR at 14 fronto-central sites at 100-150 ms. Standard stimuli is shown as std, deviant as dev. MMR is std subtracted from dev.

(40)

MMR 500-650 ms

-5 -4 -3 -2 -1 0

control risk

Mean amplitude std

dev

std -2.55 -2.29

dev -4.02 -3.58

MMR -1.47 -1.29

FIGURE 3.14: MMR at 14 fronto-central sites at 500-650 ms. Standard stimuli is shown as std, deviant as dev. MMR is std subtracted from dev.

M M R 200-250ms

0 0.5 1 1.5 2 2.5

control risk

Mean amplitude LH

RH

LH 1.69 0.71

RH 2.18 1.52

FIGURE 3.15: MMR at left hemisphere (LH) and right hemisphere (RH) for controls and at- risks at 200-250 ms. MMR is obtained by subtracting the response to the standard stimuli from the deviant stimuli.

The MANOVA (Stimulus×Laterality×Group) for the 12 lateral electrodes, 6 over each hemisphere found there were three MMRs in three time windows. The 200-250 ms window shows a significant MMP in both groups at both hemispheres (F(1, 23)=6.658, p=.017) (FIGURE 3.15). A close-significant MMP was found for both groups at the same time window at Fz and FCz. At the 300-350 ms, a significant Stimulus by LR interaction was

(41)

observed (F(1, 23)=4.588, p=.043) (FIGURE 3.16). At 500-650 ms, a LR by Group interaction ((F(1, 23)=6.748, p=.016) suggests a larger right hemisphere negativity than the left hemisphere negativity for the controls while the at-risks show the opposite left-right pattern (FIGURE 3.17).

MMR 300-350ms

-0.5 0 0.5 1 1.5 2 2.5

control risk

Mean amplitude

LH RH

M M R 500-650ms

-2.5 -2 -1.5 -1 -0.5 0

control risk

Mean amplitude LH

RH

LH 1.66 -0.25

RH 2.02 1.19

LH -1.07 -1.78

RH -2.12 -0.88

(42)

Chapter 4 Discussion

4.1 Peak pattern

Regarding the ERPs to the standard stimuli, our main finding was a distinct P150-N250- P350-N450 peak complex in early infancy, corroborating the findings of Ceponiene et al.

(2001) and Kushnerenko et al. (2002). The P150 was the most prominent peak among the four peaks, supporting the findings of Kushnerenko et al. (2002) and Pang & Taylor (2000).

We did not find any amplitude difference of any of the four peaks between the control and the at-risk infants.

We found a longer P150 latency over the left hemisphere for the at-risk infants. This finding indicates that they might have a left-hemisphere impairment processing the non- speech tone stimuli, especially responses to the first acoustic clues. The control children, on the other hand, had a slightly longer mean latency over the right hemisphere than the left hemisphere. The hemispheric difference regarding P150 latency between the groups could play a crucial role in determining if the at-risk infant becomes dyslexic later.

The N250 peak tends to be delayed in both hemispheres in the at-risk infants compared to the control group, indicating an overall delay in the at-risks irrespective of hemisphere. It

LIST OF TABLES

ABSTRACT

ACKNOWLEDGMENTS

TABLE OF CONTENTS

LIST OF TABLES

LIST OF FIGURES

Chapter 1 Introduction

Chapter 2

MMR and P150-N250-P350-N450 Related to Auditory Processing in Dyslexia

Chapter 3 Experiment

Chapter 4 Discussion