The effect of development on cortical auditory evoked potentials in normal hearing listeners and cochlear implant users

Transcription

1 University of Iowa Iowa Research Online Theses and Dissertations Spring 2016 The effect of development on cortical auditory evoked potentials in normal hearing listeners and cochlear implant users Eun Kyung Jeon University of Iowa Copyright 2016 Eun Kyung Jeon This dissertation is available at Iowa Research Online: Recommended Citation Jeon, Eun Kyung. "The effect of development on cortical auditory evoked potentials in normal hearing listeners and cochlear implant users." PhD (Doctor of Philosophy) thesis, University of Iowa, Follow this and additional works at: Part of the Speech and Hearing Science Commons

2 THE EFFECT OF DEVELOPMENT ON CORTICAL AUDITORY EVOKED POTENTIALS IN NORMAL HEARING LISTENERS AND COCHLEAR IMPLANT USERS by Eun Kyung Jeon A thesis submitted in partial fulfillment of the requirements for the Doctor of Philosophy degree in Speech and Hearing Science in the Graduate College of the University of Iowa May 2016 Thesis Supervisor: Professor Paul J. Abbas

3 Copyright by EUN KYUNG JEON 2016 All Rights Reserved

4 Graduate College The University of Iowa Iowa City, Iowa CERTIFICATE OF APPROVAL This is to certify that the Ph.D. thesis of PH.D. THESIS Eun Kyung Jeon has been approved by the Examining Committee for the thesis requirement for the Doctor of Philosophy degree in Speech and Hearing Science at the May 2016 graduation. Thesis Committee: Paul J. Abbas, Thesis Supervisor Carolyn J. Brown Christopher W. Turner Kate Gfeller Jacob J. Oleson

5 To my father, who taught me to have a thankful heart in all circumstances and to put all my hopes in heaven where we will meet soon. To my mother, who encouraged and supported me all the way. To Jay and Ellie, my husband and daughter, who sacrificed time with me and gave me unforgettable laughs to help me forget the hardships along the way! ii

6 Delight yourself in the LORD, and he will give you the desires of your heart. Psalm 37:4 The heart of man plans his way, but the LORD establishes his steps. Proverbs 16:9 iii

7 ACKNOWLEDGEMENTS I would like to first thank the members of my dissertation committee for their time and thoughtful comments. I would like to express deepest gratitude to my advisor Dr. Paul Abbas for his full support and expert guidance throughout my research. He also showed me how to be a humble researcher, welcoming all ideas from students and young researchers. Also, I express my appreciation to Dr. Carolyn Brown, who is my Iowa mom, for her incredible patience and help throughout Au.D. and Ph.D. programs. I appreciate Dr. Christopher Turner for sharing his wisdom for life as well as research, providing me an opportunity to work in his speech perception laboratory, and encouraging me whenever I felt small in academics. I would like to thank all my teachers and supervisors at the University of Iowa. I thank Drs. Bentler, Finnegan, Wu, Walker, Goodman, and Diane Niebuhr, a clinical professor, for their continuous support. Also, I thank Dr. Holly Teagle and the pediatric cochlear implant (CI) team at the University of North Carolina for teaching me a caring heart for children with a CI. Thanks also go to my fellow at graduate school, Li, Aayesha, Rachel, Bruna, Eli, Viral, Nick, Marcin, and Ben, and the CI research team, especially, Sue, Christine, and Ginny, at the University of Iowa Hospitals and Clinics. Finally but most importantly, I would like to thank God, who gives me the desires of my heart and directs my steps. I thank my parents for their faith and consistent prayers. I thank my brother for sending me many packages, filled with my favorite snacks, from Korea. I appreciate my husband, Jay, who moved from Canada to Iowa and became a huge blessing to me. I thank hugs from my daughter, Ellie. Without her, I couldn t have tasted the extreme joy of being a mom. iv

8 ABSTRACT This study investigates the developmental effects on two types of cortical auditory evoked potentials. One is the P1-N1-P2 complex, which occurs within 50 to 250 ms after a stimulus onset. It is also called an onset response and is related to sound detection. Another is the acoustic change complex (ACC), elicited by any acoustic change in ongoing sounds. It is also called a change response and is related to sound discrimination. The aim of the study was to document developmental effects on the P1-N1-P2 and the ACC in normal hearing (NH) listeners and cochlear implant (CI) users in quiet and noise conditions. For NH listeners, ninety-one children aged 3-19 years and eleven young adults participated. A total of fifty-nine CI users participated: forty-eight pre-lingually deafened children aged 3-19 years and young adults, and eleven post-lingually deafened adults. All pre-lingually deafened subjects were implanted with their CI(s) before 3.5 years of age. Speech-like stimuli were presented once in quiet and once in noise conditions to elicit the cortical auditory evoked potentials. Results show that the morphology, latency, and amplitude of both onset and ACC responses showed similar developmental effects. However, the ACC matured later than the onset response in both quiet and noise conditions. With background noise, the ACC was affected by noise more than the onset response, which led to a longer developmental trajectory for the ACC. The findings were similar between NH listeners and CI users, suggesting that a CI facilitates typical development of the two cortical responses. However, the effect of background noise was prominent in the ACC of CI users. This may indicate the perceptual difficulties of discriminating sounds in noise. v

9 PUBLIC ABSTRACT When a baby is born deaf, a cochlear implant is often recommended as a medical habilitation tool to the parents. A cochlear implant is designed to bypass a damaged cochlea and stimulates auditory nerve directly, from where signals are sent all the way to the auditory cortex where sounds are perceived. We expect that a deaf child can detect and discriminate speech sounds with this device. With continuous auditory experiences, we hope that the auditory cortex of the deaf child can be developed as children with normal hearing do. Can a cochlear implant facilitate the development of the auditory brain? This study attempts to answer this question, exploring developmental effects on evoked potentials measured at the cortical level. Early-implanted, pre-lingually deafened cochlear implant users showed similar developmental patterns of cortical auditory evoked potentials to those of normal hearing listeners. However, the responses, related to sound discrimination, were affected by noise more in cochlear implant users. This may be related to perceptual abilities of cochlear implant users in harder listening conditions. The findings indicate that cortical auditory evoked potentials, related to both detection and discrimination, can be used to document the long developmental trajectory of the central auditory system in both normal hearing listeners and cochlear implant users. This study suggests that these responses can be used as a tool for estimating behavioral performance in cochlear implant users. vi

10 TABLE OF CONTENTS LIST OF TABLES... ix LIST OF FIGURES... x LIST OF ABBREVIATIONS... xiii CHAPTER 1 INTRODUCTION Cochlear Implants in Children Development of Auditory Pathways Sound deprivation and electrical stimulation Neuroplasticity / Critical period Critical period for cochlear implants Cortical Auditory Evoked Potentials The P1-N1-P2 complex The acoustic change complex (ACC) Proposed Study Pilot experiment Aims and rationale of the proposed study Research questions and hypotheses CHAPTER 2 LITERATURE REVIEW Development of the Postnatal Brain Myelination Synaptogenesis Development of the Central Auditory System Auditory nervous pathways stimulated via acoustic or electric sounds Auditory deprivation and electrical stimulation Neuroplasticity: stimulating auditory pathways using CIs Methods to study the development of central auditory system The P1-N1-P2 Complex Generators General morphology Effects of stimulus, recording, and subject variables on the P1-N1-P Effects of development on the P1-N1-P2 complex Cortical Auditory Evoked Potentials to Assess Discrimination Mismatch negativity and P The ACC Summary and Limitations of the Literature vii

11 CHAPTER 3 METHODS Study Participants General Procedures Electrophysiologic Measures Stimuli Sound presentation: in quiet and noise conditions Recording parameters Analysis Grand average waveforms Peak latencies and N1-P2 peak-to-peak amplitudes Correlations between adults' and children's grand average waveforms CHAPTER 4 RESULTS Changes of CAEPs with Age in NH listeners in Quiet Changes of CAEPs with Age in NH listeners in Noise Changes of CAEPs with Age in CI users in Quiet Changes of CAEPs with Age in CI users in Noise Summary CHAPTER 5 DISCUSSION AND CONCLUSIONS Feasibility of Recording CAEPs using Long Duration of Complex Stimuli Developmental Patterns of the ACC in NH listeners and CI users in Quiet Effects of a Challenging Condition on ACC Responses Implications of the Findings Limitations of the Study Future Directions Conclusions APPENDIX A: A PILOT STUDY A.1. Overview A.2. Methods A.2.1. Participants A.2.2. Stimuli A.2.3. Procedures A.2.4. Statistical Analysis A.3. Results A.4. Discussion APPENDIX B REFERENCES viii

12 LIST OF TABLES Table 1. Normal hearing subjects Table 2. Cochlear implant subjects ix

13 LIST OF FIGURES Figure 1. Age-related changes in P1 latency of the P1-N1-P2 complex in children with normal hearing Figure 2. Age-related changes in P1 latency of the P1-N1-P2 complex in children with cochlear implants Figure 3. Neuronal processes in the pre- and post-natal human brain Figure 4. Brain myelination in 2 weeks, 1 year, 2 year, and adults Figure 5. Development of neurofilament in axons from 40 weeks to 11 years Figure 6. Development of synapses in human primary auditory cortex Figure 7. Development of gray matter from 4 to 21 years Figure 8. Changes in white and gray matters and total brain volumes with age Figure 9. Animated figures of the cochlea and an action potential Figure 10. Animated figures of a cochlear implant system Figure 11. Animated figure of central auditory pathways Figure 12. Grand average waveforms obtained using different inter-stimulus intervals.. 57 Figure 13. The P1-N1-P2 complex in adults and infants Figure 14. Maturation of the P1-N1-P2 complex in normal hearing children Figure 15. Comparisons of the P1-N1-P2 complex among different age groups Figure 16. The P1-N1-P2 and ACC responses elicited by [sei] speech sounds x

14 Figure 17. The P1-N1-P2 and ACC responses elicited by a spectral change Figure 18. Comparison of grand mean waveforms between adults and children Figure 19. Waveforms and spectrograms for stimuli Figure 20. NH adult listeners CAEPs in quiet and noise conditions Figure 21. Waveforms of CAEPs in NH listeners in quiet Figure 22. Changes of P1 latency with age in NH listeners in quiet Figure 23. Changes of N1 component with age in NH listeners in quiet Figure 24. Correlations between NH children and NH adults' CAEPs in quiet Figure 25. Waveforms of CAEPs in NH listeners in noise Figure 26. Correlations between NH children and NH adults' CAEPs in noise Figure 27. Comparisons of grand average waveforms between quiet and noise conditions in NH listeners Figure 28. Comparisons of P1 latency and N1-P2 response between quiet and noise conditions in NH listeners Figure 29. Comparisons of correlations between quiet and noise conditions in NH listeners Figure 30. Adult CI users' CAEPs in quiet and noise conditions Figure 31. Comparisons of CAEPs among the three adult listening groups Figure 32. Waveforms of CAEPs in pre-lingually deafened CI users in quiet xi

15 Figure 33. Changes of P1 latency with age in CI users in quiet Figure 34. Changes of N1 component with age in CI users in quiet Figure 35. Correlations between CI children and NH adults CAEPs in quiet Figure 36. Waveforms of CAEPs in CI users in noise Figure 37. Correlations between CI users and NH adults' CAEPs in noise Figure 38. Comparisons of grand average waveforms between quiet and noise conditions in CI users Figure 39. Comparisons of P1 latency and N1-P2 response between quiet and noise conditions in CI users Figure 40. Comparisons of correlations between quiet and noise conditions in CI users Figure 41. Examples of changes with age within individuals Figure 42. Comparisons of onset P1 latency between previous and current studies xii

16 LIST OF ABBREVIATIONS ABR AEP ACC ANSD CAEP CI CN CT EABR EACC ECAP EEG fmri HA IC ICA K-CID MGB MMN MRI NH ODR auditory brainstem response auditory evoked potential acoustic change complex auditory neuropathy spectrum disorder cortical auditory evoked potential cochlear implant cochlear nucleus computed tomography electrically evoked auditory brainstem response electrically evoked auditory change complex electrically evoked compound action potentials electroencephalography functional magnetic resonance imaging hearing aid inferior colliculus independent component analysis Korean version of the Central Institute of Deafness medial geniculate body mismatch negativity magnetic resonance imaging normal hearing optimized differential reference xiii

17 PCA PBK PET REM RMS SPECT SNHL SNR SON principal component analysis phonetically balanced kindergarten positron emission tomography rapid eye movement root mean square single photon emission computed tomography sensorineural hearing loss signal to noise ratio superior olivary nuclei xiv

18 CHAPTER 1 INTRODUCTION 1.1. Cochlear Implants in Children A cochlear implant (CI) is an electronic device that bypasses the peripheral damage in the cochlea and stimulates the auditory nerve directly using electrical impulses. It can restore some sense of hearing for individuals with impaired cochlear function. The first child to receive a multichannel CI in the US was implanted in Since then, CIs have proven to be one of the most successful (re)habilitation tools for establishing good speech perception in children who are born deaf (Tomblin et al., 1999; Kral & Pallas, 2010). According to the U.S. Food and Drug Administration, as of December 2012, more than 324,000 people worldwide have received a CI. This figure includes more than 80,000 children worldwide (Kral & Pallas, 2010) and 38,000 children in the United States alone (National Institute on Deafness and Other Communication Disorders). In the 1980s, most children who received a CI were 4-5 years of age and older (Tomblin et al., 1999; 2007). Today, thanks to newborn hearing screening we can identify congenitally deaf children shortly after birth and start rehabilitation by providing access to sounds with a CI within the first year of life Development of Auditory Pathways A sense of hearing begins to develop before birth. A fetus at 24 weeks shows blink and startle responses to vibroacoustic sounds. These responses become consistent by 29 to 32 weeks (Birnholz & Benacerraf, 1983). When babies are born, they can be soothed or irritated by sounds that they hear. The central auditory nervous pathways, 1

19 however, are far from maturity at birth. In fact, these pathways undergo developmental changes throughout the first two decades of life, parallel to the development of postnatal brain (Moore, 2000a; Moore 2002b). Many of the changes in the postnatal brain are experience-dependent; e.g., synapses used frequently will be strengthened and those used rarely will be lost. The central auditory system requires auditory experiences to develop normally (Kral et al., 2000, 2001; Moore & Guan, 2001; Moore & Linthicum, 2007) Sound deprivation and electrical stimulation The most common cause of sensorineural hearing loss (SNHL) is damage to the cochlea, including missing hair cells. Because Type I afferent auditory nerves innervate inner hair cells and send signals to the brain, the missing or damaged hair cells will block sound transmission to the upper auditory pathways. Studies show that if auditory stimulation is absent for a prolonged period of time, the result can be death of spiral ganglion cells (Hardie & Shepherd, 1999) and cochlear nucleus cells (Born & Rubel, 1988). The loss of auditory neurons in the peripheral auditory pathways can arrest the development of the central auditory nervous system (Tremblay & Cunningham, 2002; Turner, 2006). Animal studies show that using electrical stimulation to evoke action potentials in the auditory nerves prevents these adverse changes; in fact, electrical stimulation increased the number of surviving spiral ganglion cells and neural responses in the inferior colliculus (e.g., Hyson & Rubel, 1989; Shepherd et al., 1999) Neuroplasticity / Critical period A baby with normal hearing (NH) begins to recognize environmental sounds and familiar voices shortly after birth. Before long, typically developing children learn to further associate complex patterns of neural activity with speech sounds and contents; 2

20 therefore, they can successfully communicate with speech and language only a few years after birth. Individuals do not lose these aural communication skills, even if they acquire bilateral hearing loss later in life. These post-lingually deafened adults are considered to be optimal candidates for CI implantation. They may be able to quickly relearn to recognize the electrical signals of speech that they knew before their hearing loss. For this reason, cochlear implantation was initially provided to post-lingually deafened adults alone. Many of them attained high levels of performance on word recognition tasks, especially in quiet environments (Zeng et al., 2008). Their success is likely due, in part, to their previous normal auditory experiences during the developmental periods of speech and language acquisition, which likely led to a typical central auditory system development. Unlike post-lingually deafened adults, children who were born deaf, also referred to as congenitally and pre-lingually deafened, lack the experiences of hearing typical acoustic sounds. When they receive a CI they must use the electrical signal to drive development of their central auditory pathways and to acquire speech and language skills. Neuroplasticity can help to achieve these goals. Neuroplasticity refers to alterations in neuronal networks and brain structures that occur in response to changes in input from the environment or damage to a peripheral system (Pallas 2001; Pascual-Leone et al, 2011). While there is evidence that neuroplasticity can occur throughout life, many studies have shown that the cortex is most plastic to change during the first few years of life when synaptogenesis and synaptic pruning processes occur dynamically (e.g., Kral et al., 2001; Zhang et al., 2002; Nakahara, et al., 2004; Sale et al., 2009). 3

21 Numerous researchers in developmental psychology and biology have studied the critical periods for many different tasks. A critical period refers to the period in which a person can master or relearn a certain skill during developmental stages or rehabilitation. The critical period is task-dependent. For example, a speaker of English as a second language cannot master English at the level of native speakers when they are exposed to English after puberty (Kuhl, 2011). Another example of the critical period can be found in spontaneous recovery after a stroke. Researchers (e.g., Nicholas et al., 1993) concluded that spontaneous recovery occurs within hours, days, and up to 6 months after onset of a stroke. This spontaneous recovery time is also a critical period for a person to receive rehabilitation, including speech and language therapy. Similarly, a critical period of demanding auditory experiences can exist to develop speech perception fully. A critical period of providing electrical stimulation to a deaf ear can use the plasticity mechanism of the central auditory nervous systems Critical period for cochlear implants Behavioral studies have shown that acquisition of spoken language was easier for children who were implanted earlier in life (Fryauf-Bertschy et al., 1997; Hassanzadeh, et al., 2002; Harrison et al., 2005; Tomblin et al., 2005; Holt & Svirsky, 2008; Cole & Flexer, 2011). For example, Fryauf-Bertschy et al. (1997) compared open-set word recognition scores after 3 and 4 years of CI use in two pre-lingually deafened groups of children. One group of children received their CIs between 2.5 and 4 years of age and another group of children received their CIs at 5 years old or older. Results show that early-implanted children obtain higher scores than late-implanted children. Hassanzadeh, et al. (2002) measured closed-set and open-set speech perception in 119 CI children, 4

22 divided into 6 groups based on the age at implantation: 0-3, 4-5, 6-7, 8-9, and >12 years old. At the 2-year post-stimulation mark, children who were implanted between 0-3 years obtained the best average scores in speech perception compared to children in other groups. Similarly, Holt & Svirsky (2008) reported that children implanted before 4 years old progressed faster in oral language development than children implanted later in life. Tajudeen et al. (2010) divided 117 early-implanted children into three groups depending on when the child received a CI (i.e., in the first, second, and third year of life). There were significant differences in word recognition scores among the three groups when a chronological age was used to compare the group performances; however, no significant differences were found when a hearing age was used. The hearing age counts years after CI activation. All of the early-implanted children in the three groups scored ~ 80% word recognition after using their CI for three years. Recently, Dunn et al. (2014) studied whether the age at implantation impacts the long-term outputs of speech, language, and reading skills. Thirty-eight children implanted before 2 years of age and 45 children implanted between 2 and 3.9 years of age were examined for speech, language, and reading abilities. All measures were collected with over 7 years of CI experience. Results did not show significant differences in scores of speech, language, and reading tests in the two CI groups. This may indicate that children implanted between 2 and 3.9 years have capabilities to develop performance as successfully as children implanted before 2 years old. It can be concluded from the above behavioral literature that the critical period for developing these skills using electrical stimulation to a deaf ear might extend to 4 years of life; therefore, children implanted before 4 years of age have the capacity to develop speech and language skills more easily than children implanted after 4 years old. 5

23 1.3. Cortical Auditory Evoked Potentials Using behavioral measures, studying how the central auditory nervous system develops in children can be very challenging largely because of the lack of tests, which can include very young children with limited speech and language skills. Instead, researchers have relied on non-behavioral tools such as neuroimaging and electrophysiological recordings. Many neuroimaging techniques, however, are contraindicated for CI users because they can result in damage either to the implanted electronics or to the externally worn processor of the patient. Auditory evoked potentials (AEPs) have the most advantages over other neuroimaging techniques in that they safe for use with CI recipients. Additionally, the recording equipment is considerably less expensive compared to neuroimaging, in fact, electrophysiological recording methods are already available in many otolaryngology clinics The P1-N1-P2 complex The P1-N1-P2 complex is one of the cortical auditory evoked potentials (CAEPs). It is composed of three components; a positive peak (P1) as recorded at the vertex followed by a negative peak (N1), typically larger in amplitude than P1, and lastly followed by a second vertex positive peak (P2). This P1-N1-P2 complex is an obligatory evoked potential, which means that it can be recorded without a listener s active response to sounds. Also, its recording paradigm is safe for CI recipients. For these reasons, the P1-N1-P2 complex has been used to assess auditory processing or encoding sounds at a cortex level in young children and adults with NH or a CI (e.g., Novak et al., 1989; Pasman et al., 1991; Ponton et al., 1996a, b; Sharma et al., 1997; Wunderlich et al., 2006). The P1-N1-P2 complex is typically elicited using brief stimuli such as clicks, pure 6

24 tones, tone bursts, or short consonant-vowel (CV) syllables (for a review, see Hyde, 1997; Stapells, 2002). The presence of the P1-N1-P2 response elicited by the onset of any sound is interpreted as evidence that the listener can detect the sound (e.g., Davis, 1965; Davis & Zerlin, 1966; Cody et al., 1967; Stapells, 2002; Martin et al., 2008). For this reason, the P1-N1-P2 complex is also referred to as onset responses in this paper. The P1- N1-P2 complex has a relatively long latency reflecting the neural processing time from sound onset to the auditory brain. Neural generators are believed to be located in thalamocortical pathways and primary and secondary auditory cortices (e.g., Ponton et al., 2002). Each component of the P1-N1-P2 complex is likely generated in a slightly different location, which may lead to different time period of maturation (Paus et al., 1999; Moore, 2002b, for a review, see Stapells, 2002) The P1-N1-P2 complex in NH listeners In NH listeners, the developmental effects on the morphology, latency, and amplitude of the P1-N1-P2 complex have been well documented. In adults, all three components of the P1-N1-P2 response are present and their latencies occur between 50 and 250 ms after stimulus onset; latencies of P1, N1, and P2 are approximately 60 ms, 100 ms, and 175 to 200 ms, respectively. The N1-P2 amplitude is the most dominant component of the response in adults (e.g., Ponton et al., 1996 a,b, 2002; Sharma et al., 1997; Gilley et al., 2005; Purdy et al., 2005). In infants and young children, P1 is the dominant component of the P1-N1-P2 complex, and often it is the only peak that is identifiable, recorded at between 150 and 350 ms after stimulus onset (e.g., Kurtzberg et al., 1984; Novak et al., 1989; Pasman et al., 1991; Ponton et al., 1996a,b; 2002; Purdy et al., 2005). The amplitude of P1 recorded 7

25 in children is greater than the amplitude recorded from adults (e.g., Ceponien et al., 2002). The P1 is even present in about 95% of preterm babies at weeks postconceptual age (Pasman et al., 1991) and 97 % of infants before age of 2 months (Kurtzberg et al., 1984). The dominant P1 in young children is often followed by a distinguishable broad negativity at between 250 and 450 ms post-stimulus onset. The negativity measured in children is thought to be the immature late cortical response, because it is not recorded in adults and has a clearly different latency range from the N1 measured in adults (e.g., Kurtzberg et al., 1984; Pasman et al., 1991, 1999; Pang & Taylor, 2000; Kushnerenko et al., 2002; Sharma et al., 2002a,b; Purdy et al., 2005; Sussman et al., 2008). Recently, Small & Werker (2012) suggested that N1 and P2 components could be seen in infants when a slower stimulation rate (1 stimulus every 2.2 seconds) was used. However, the latencies of the three components were much longer than those obtained in adults; in fact, the latency of N1 in the study overlaps the range of the latency of what other researchers call immature negativity. Results from most developmental studies agree that adult-like N1 and P2 peaks are typically absent in infants and young children. N1 begins to emerge with small amplitude around 7-8 years, but is not consistently recorded until children are between 12 and 13 years old. P2 emerges after the N1 peak begins to develop, resulting in the growth in N1-P2 amplitudes with age (e.g., Ponton et al., 2000; Gilley et al., 2005, for a review, see Wunderlich & Cone-Wesson, 2006). While general morphology changes with age, peak latencies have been widely studied to quantify maturational changes in central auditory system. Postnatal processes in the brain, such as myelination of axons and synaptic pruning, reduce the travel time 8

26 from the sound onset to neural generators for the P1-N1-P2 complex. The reduced neural travel time is reflected in the decrease of peak latencies with development (e.g., Ponton et al., 2000; Moore & Guan, 2001; Moore & Linthicum, 2007). The peak latency is less variable than the amplitude of the responses (e.g., Wunderlich & Cone-Wesson, 2006). Comparison across studies reveals that the effects of development on P1 latency are less variable than either N1 or P2 peak latencies (e.g., Sharma et al., 1997, 2002a,b; Ponton et al., 2002; Ceponiene et al., 2002; Wunderlich & Cone-Wesson, 2006; Dorman et al., 2007; Moore & Linthicum, 2007). P1 latency has been one of the most popular measures to explain the impact of development on the P1-N1-P2 complex. Moreover, Sharma et al. (1997) refer to the latency of the P1 peak as a biologic maturation marker of the central auditory pathways. Figure 1 (from Sharma et al., 2002a) illustrates how P1 latency changes as a function of age in 187 NH listeners from 0.1 to 20 years. P1 latencies show a negative exponential relationship with age, showing more dramatic drops within the first 6 years. In infants, P1 latencies were as long as ~275 ms. By 3 years old, the average P1 latency was approximately 125 ms and by 6 years old it decreased further to approximately 100 ms. After 6 years, however, changes in P1 latencies vary only slightly. This finding on the changes of P1 latency with age is very consistent across many studies (for a review, see Dorman et al., 2007). 9

27 Figure 1. Age-related changes in P1 latency of the P1-N1-P2 complex in children with normal hearing P1 latency values for 187 NH children are plotted on the x-axis as a function of age on the y-axis from newborns to 20 years. The solid lines indicate the 95% confidence intervals for normal development of P1 latency described in Sharma et al. (2002a). This figure is taken from Sharma et al. (2002a), Figure The P1-N1-P2 complex in CI children Researchers have also successfully measured the P1-N1-P2 responses from pediatric CI users. Sharma et al. (2007) suggest that the onset P1-N1-P2 response can be used to determine whether the auditory nervous system is developing typically or not after (re)habilitation with a CI. They measured the P1-N1-P2 complex in 231 CI children (between 1 and 20 years) at various time after implantation. These children s data were divided into 6 different groups based on their age at implantation. Figure 2 from (Sharma et al., 2007) displays the obtained P1 latencies from 231 CI children applied to the exponential function of the P1 latency obtained from NH children from Sharma et al. (2002a). A solid line connects P1 latency data recorded from individual CI users for multiple occasions. The different colors indicate when a child received his/her implant: 10

28 green lines represent results from children who received a CI between 6 months and 1.5 years. Red lines represent data recorded from children who were implanted after 6.5 years. Results show that P1 latencies obtained from CI children decreased with increasing chronological age. This developmental trend of P1 latency was similar to that of NH children. Additionally, P1 latencies recorded from children implanted before about 3.5 years fit the exponential function after 3 to 6 months of CI use. However, P1 latencies measured from children who were born deaf but implanted at or after 7 years never reached the normal limits regardless of duration of CI use. P1 latencies obtained from children implanted between 3.5 and 7 years olds were highly variable. The authors interpret these results as an indication that children who are born deaf but implanted before 3.5 years of age can continue to develop the central auditory pathways normally. Unfortunately, this is less likely to occur for children implanted after 7 years old. These results support findings in speech and language development after cochlear implantation discussed previously in this chapter. P1 latency data and speech and language development after implantation suggest that neuroplasticity related to auditory neural development and language development is the greatest up to about 3 to 4 years of age. In other words, using electrical stimulation, early-implanted children (at least before 4 years of life) appear to overcome the auditory deprivation period and prepare to develop speech and language (e.g., Huttenlocher & Dabholkar, 1997; Sharma et al., 2002b, 2007; van Zundert, et al., 2004; Harrison et al., 2005; Holt & Svirsky, 2008; Sharma & Campbell, 2011). 11

29 Figure 2. Age-related changes in P1 latency of the P1-N1-P2 complex in children with cochlear implants Individual developmental trajectories for P1 latencies are plotted as a function of age for 231 CI children who ranged in age from ~1 to 20 years. These children were divided into seven groups, indicated by different colors, based on the age at implantation. The solid lines indicate the 95% confidence intervals for normal development of P1 latency (Sharma et al., 2002a). This figure is taken from Sharma et al. (2007), Figure The acoustic change complex (ACC) It is crucial to know whether CI recipients, especially very young children, can detect sounds using electrical stimulation. The P1-N1-P2 can help clinicians to make this decision (e.g., Martin et al., 2008). While sound detection is a prerequisite for further auditory perception, CI research over the decades show almost all CI users can detect sounds regardless of benefit they receive from their device. What separates high performing CI users from lower performing CI users is their ability to perceive changes in frequency, intensity, and duration of an ongoing signal. A simple metric like P1 latency or the presence or absence of the P1-N1-P2 is not likely to provide information regarding the listener s capacity of discriminating sounds (Hillyard & Kutas, 1983). 12

30 Another obligatory CAEP, which can help clinicians determine a listener s sound discrimination ability, is the acoustic change complex (ACC). In fact, it is often considered a derivative response from the onset P1-N1-P2 responses. Therefore, it shares many morphological characteristics with the P1-N1-P2 complex. What makes a clear distinction of the ACC from the onset P1-N1-P2 is the stimulus construct to evoke the responses. The ACC is evoked by any acoustic change within a stimulus (Ostroff et al., 1998). When present, the ACC is typically interpreted as an indication that the listener perceives the change in the evoking stimulus. As such, it may provide a metric that can be used to predict discrimination rather than simple detection (Martin et al., 2008) The ACC in NH listeners Previous studies have shown that the ACC can be recorded reliably from NH adults using various acoustic changes within a stimulus. For instance, Ostroff et al. (1998) demonstrated the ACC was elicited in NH adults by the transition from fricative [s] to vowel [ei] in a naturally produced syllable say. Martin and Boothroyd (1999), first used the ACC terminology. They obtained ACCs in NH adults using a spectral change from a tone to noise or vice versa. Later, Martin and Boothroyd (2000) constructed an 800-ms-long synthetic vowel stimulus with a particular acoustic change in the middle of the stimulus, i.e., at 400 ms. They were able to record both the onset P1- N1-P2 response, elicited by the onset of the vowel stimulus, and the ACC, elicited by the change at the halfway. For acoustic changes, Martin and Boothroyd (2000) used either a sound level change or a frequency change within the stimulus. Results show that the ACC was successfully elicited by both an increase and a decrease in sound level, when the formant frequency was not changed throughout the whole duration of the stimulus. 13

31 The ACC was also obtained in response to a change in formant frequency (e.g., a change from /u/ to /i/) when the intensity of the two vowel segments was equal. The ACC responses were greatest when evoked by a stimulus, which consisted of both a level change and a frequency change The ACC in adult CI users Several researchers have recorded ACC responses from CI users. Friesen and Tremblay (2006) recorded ACC responses from eight adult Nucleus CI users, presenting naturally produced speech syllables /si/ and / i/ through the subjects own speech processors in a soundfield. They showed that the ACC was reliably recorded from all CI users with test-retest coefficients ranging from 0.63 to 0.89 across subjects. Latencies were earlier when the ACC was elicited by / i/ than when it was elicited by /si/. These findings showed that the ACC recorded from CI users could reflect the distinction between two consonants. Brown et al. (2008) also obtained the ACC from adult CI users, but using a longer duration of stimulus, which was presented via direct input to the intracochlear electrodes. They created a 600-ms-long stimulus train using biphasic current pulses. Initially, they stimulated electrode 10 using this stimulus train, and then at the midpoint, i.e., 300 ms post stimulation, they varied the stimulating electrode. Thus, they did not create any acoustic change in the stimulus but varied the place of stimulation, i.e., the spatial separation, between the intracochlear electrodes to elicit the ACC. They found that the larger the spatial separation was between the two stimulating electrodes, the greater the ACC responses. Brown et al. (2008) referred to these responses measured from CI users using direct stimulation as electrically evoked auditory change complex (EACC). From now on, we will refer to the change responses as ACCs when 14

32 obtained in sound field and as EACC when obtained via direct input. From the same lab, Kim et al. (2009) created 600-ms-long biphasic pulse trains, with a current level change at the midpoint to elicit the EACC from 10 CI users. The researchers presented this stimulus via direct input, but they stimulated one electrode changing current levels sent to the electrode. For example, a biphasic pulse train was initially presented at 50% of the dynamic range of current levels to an individual intracochlear electrode and the stimulus level was either increased or decreased at the midpoint. They found that EACCs were reliably recorded from all 10 subjects in response to amplitude changes. Amplitude increases resulted in larger EACCs. These findings were similar to results obtained in NH listeners as reported by Martin and Boothroyd (1999). These previous studies demonstrate that despite electrical stimulus artifact, ACC can be measured reliably in adult CI users via either soundfield presentation or direct stimulation The ACC in pediatric CI users Recently there has been interest in recording ACC responses from pediatric populations (e.g., Martin et al., 2008, 2012; He et al., 2013; Martinez et al., 2013). These studies have shown that ACC (obtained via loudspeaker) or EACC (obtained via direct input) could be reliably recorded in NH children as well as hearing aid (HA), and CI users. He et al., (2013) created an 800-ms-long stimulus using biphasic pulses, this stimulus is similar to Brown et al., (2008), and Kim et al., (2009). At the middle of the sound, they inserted a gap with 5, 10, 20, 50, and 100ms. Fifteen CI children aged 5 to 17 years old participated in this study; all of these children were diagnosed with Auditory Neuropathy Spectrum Disorder (ANSD). The authors showed that EACC was obtained 15

33 from all subjects. The larger gap in the stimulus evoked larger EACC amplitudes. EACC thresholds were varied across subjects. EACC thresholds were compared with word recognition scores. A subject with better word recognition scores had better thresholds, i.e., the EACC was evoked using the shortest duration (5 ms) of the gap. While the authors tested EACC in CI children with a wide range of ages, they did not arrange the data in order of age or systemically look at the effect of age on the EACC. However, the authors clearly noted the EACC had different morphologies from those of the onset P1- N1-P2 responses, describing the EACC as having an immature morphology. Martinez et al. (2013) recorded ACCs from ten children between 2 and 6 years of age in NH and HA users. They used relatively short duration (40 ms) stimuli, changing from /u/ to /i/ or /u/ to /a/. Results show that both NH and HA children had ACCs, with an exception of one HA child. They reported poor ACC morphologies in 2 HA children. The results suggested poor morphology of the ACC to the spectral change might indicate limited audibility to a particular formant with HAs. This study may suggest that ACC measure could be used as a tool for estimating speech discrimination ability for HA children after amplification. However, they had a very limited number of subjects and disregarded possible developmental impacts on the ACC in their subjects whose ages could vary up to 4 years in the critical age range years in the critical age range. Unfortunately, ACC data from children either with NH or CIs are still very scarce. Furthermore, there are no published studies that describe developmental changes in the ACC recorded from children across different age groups with either NH or a CI. 16

34 1.4. Proposed Study Pilot experiment If the ACC is to be clinically applicable for children with hearing loss, e.g., to estimate their discrimination abilities, we should first understand ACC responses in developmentally normal children with good hearing sensitivity. Knowing how the ACC changes with age in these children will provide a normal developmental model of the ACC. Second, we should know if the ACC can be reliably measured in CI children of various ages. Once we know that the ACC can be recorded in almost all CI users regardless of age we can study the relationship between the ACC and behavioral measures, not only in a group but also on an individual level. Lastly, knowledge of how the ACC changes with age after CI activation will provide a foundation from which to study development of the central auditory system in CI users compared to NH listeners. Prior to the proposed study a pilot study was conducted (see Appendix A) to explore the feasibility of recording the ACC in both NH and CI children and to see the impact of development on the ACC as well as the onset responses. Nineteen NH children between the ages of 5 and 19 and five NH adults participated. Six pre-lingually deafened CI children participated. They received their CI before 3.5 years of age and ranged from 10 to 18 years of hearing age at the time of testing. Six post-lingually deafened CI adults participated. Two music-like stimuli, with 800 ms duration, were used to elicit CAEPs. A change in either pitch or timbre occurred at the midpoint (400 ms). The stimulus with a pitch change was a synthesized clarinet playing B3 to Gb6, considered to be an easy to distinguish condition. The stimulus with a timbre change was a synthesized clarinet to oboe playing the same note (B3). This was considered to be a more difficult to 17

35 distinguish condition because these two instruments have very similar spectral distributions. In NH listeners, results show that using long-duration complex stimuli, effects of development on the onset P1-N1-P2 responses had similar patterns to those described in previous studies using short duration stimuli (e.g., Ponton et al., 2000; Sharma et al., 2002a). P1 latencies decreased with age. For example, latencies measured about 100 ms at age 5 decreased to 65 ms in young adults. The N1 and P2 components emerged between 7 and 8 years of age and the P1-N1-P2 responses became adult-like in morphology at between 10 and 14 years of age. ACCs were also obtained in all age groups. The ACC followed similar developmental patterns to those of the onset P1-N1- P2 responses. However, the maturation time course of ACCs was longer than that of the onset P1-N1-P2 responses. The maturation of ACCs also appeared to be stimulusdependent. ACCs became adult-like earlier when the pitch change (i.e., easy contrast) stimulus was used than when the timbre change (i.e., difficult contrast) stimulus was used. The onset P1-N1-P2 responses obtained from CI users showed similar developmental trends to NH listeners. For example, the onset responses became adultlike around 13 years after CI activation. ACC was obtainable in all CI users. As seen in NH listeners, the developmental patterns of ACCs were similar to the onset P1-N1-P2 responses with later maturation. The effect of stimulus type (easy vs. difficult contrasts) also affected the developmental time course of the morphology of ACCs, showing a slower maturation with the timbre change stimulus. However, the timbre change stimulus had significantly degraded morphologies and small amplitudes in CI listeners for both 18

36 pre-lingually or post-lingually deafened individuals. This may suggest that the smaller amplitudes of ACC in CI users may relate to their perception of poor sound quality presented via the CI and may not be due to delayed development of central auditory pathways. The pilot study showed that the ACC could be obtained in both NH and CI children. The developmental patterns of the ACC were similar to those of the onset P1- N1-P2 responses. However, the time course of development of ACC was different from the onset P1-N1-P2 responses; the ACC became mature later. Interestingly, the ACC developmental pattern was affected by the difficulty of stimulus construct for both NH and CI users. CI users also had much smaller ACC amplitudes than NH listeners in the more challenging condition. The primary limitation of this pilot study was the small number of subject participants, and it also used music-like complex stimuli rather than speech to elicit CAEPs Aims and rationale of the proposed study The ultimate goal of this study is to investigate if the ACC can complement the onset P1-N1-P2 complex in documenting the development of the central auditory system in both NH and CI listeners. To achieve this the current study will measure onset P1-N1- P2 and ACC responses using speech stimuli in two listening conditions in both NH and CI listeners of various ages. The proposed study will provide information that is essential in using CAEPs, especially the ACC, in clinical applications. This study will investigate the impact of the development on the onset P1-N1-P2 and the ACC in NH and CI listeners using a long duration of speech-like stimuli. Previous studies in the P1-N1-P2 development literature used a very short duration 19

37 stimulus, such as /ba/ (90 ms) or clicks to elicit onset responses alone. Since the pilot study shows that both onset P1-N1-P2 and ACC responses can be elicited readily using complex stimuli, this study continues to use the same stimulus structure: a long duration (800 ms) of complex stimuli with a change halfway through. The pilot study used the complex musical stimuli; however, the current study will use complex speech stimuli to elicit the CAEPs. Speech-like stimuli can be easily applied in the clinic and may be more directly related to clinical questions. For example, if the ACC responses can be shown to correlate with speech discrimination capacity of the pediatric population, then they would be helpful in the CI testing process. Two vowel stimuli used in this study are /u/ and /i/. The intensity of these two vowels is matched; therefore, the change is primarily in spectral distribution. One stimulus is /u/ followed by /i/ and the other stimulus /i/ followed by /u/. In this way onset P1-N1-P2 and ACC responses will be elicited by either /u/ or /i/. Finally, these complex vowel stimuli will be presented via sound field. The pilot study shows that presenting sounds with a loudspeaker allows using identical stimulus, presenting, and recording equipment in both NH and CI listeners. It also better simulates everyday listening conditions than direct input sound presentation in CI listeners. The secondary aim of the current study is to investigate how a challenging listening condition impacts the CAEP responses, especially the ACC, in developing brains of NH and CI listeners. In the pilot study, differences in stimuli (perceptually easy vs. difficult-to-distinguish sounds) led to different rates of maturity in ACC responses in both NH and CI groups. For example, ACC responses matured earlier with a pitch change (i.e., the easy-to-distinguish stimulus in this study) than a timbre change (i.e., the difficult-to-distinguish stimuli in this study). These results suggest that the ACC 20

38 responses reflect a perceptual difficulty and the time course of ACC development may differ depending on the stimulus condition. The current study will use the same speech stimuli but with background noise to assess the effect of difficult listening conditions on the development of both onset P1-N1-P2 and ACC responses. It is well known that understanding speech in noise is challenging for individuals with hearing loss due to hair cell deficits in the cochlea, which results in broader tuning curves and deteriorates spectral resolution in turn (e.g., Liberman & Sossa, 1984, for a review, see Assmann & Summerfield, 2004). Therefore, we will measure the CAEPs in both NH and CI groups, once presenting speech stimuli in a quiet environment and once in a noisy environment. A speech-shaped noise of + 10 db SNR (signal to noise ratio) will be used as background noise. We assume that this amount of noise may still be able to elicit measurable CAEPs, but will affect the ACCs in both groups, especially in CI users Research questions and hypotheses 1. How do onset P1-N1-P2 and ACC responses change with age in NH listeners and CI users using long-duration speech-like stimuli in quiet listening conditions? I hypothesized that onset P1-N1-P2 responses obtained using long-duration complex stimuli in NH listeners would change with age similarly to patterns described in the current literature, which used clicks or short duration of speech sounds (e.g., Ponton et al., 2000; Sharma et al., 2002a). Regarding ACC responses, I hypothesized that the ACC would follow the similar developmental patterns of the onset P1-N1-P2 complex. However, maturation of the ACC would be delayed relative to the onset P1-N1-P2. In quiet conditions, I hypothesized that CI users implanted early in their lives (i.e., before 3.5 years in this study) would have similar developmental patterns of onset and 21

39 ACC responses to those observed in NH listeners. 2. How does a challenging listening condition, i.e., with background noise, impact the developmental patterns of the onset P1-N1-P2 and the ACC in NH listeners and CI users? I hypothesized that both onset P1-N1-P2 and ACC responses would be degraded (i.e., smaller amplitudes and longer latencies of both responses) with background noise in all age groups of NH listeners. This may lead to a delayed development of both onset P1- N1-P2 and ACC responses, compared to quiet listening conditions. Because ACC responses are less robust and smaller in amplitudes than the onset P1-N1-P2, it was expected that ACC responses would be affected more by background noise. I hypothesized that the developmental patterns of the onset P1-N1-P2 and the ACC in noise conditions would be similar in both NH listeners and CI users. ACC responses, however, would be degraded more in CI users in the presence of noise compared to NH listeners. 22

40 CHAPTER 2 LITERATURE REVIEW This chapter reviews background literature relevant to the current study. It will provide readers with more detailed information about (1) neural processes related to the development of the postnatal brain, (2) the general organization of the central auditory pathways and the methods used to assess development of these structures, (3) effects of maturation on the P1-N1-P2 complex for both NH and CI listeners, and (4) a summary of results from studies of the ACC measures as an index of the auditory discrimination tool. This section will show how the current study can add to the literature, expanding our understanding of how development impacts the ACC measures Development of the Postnatal Brain A newborn brain grows quickly. By 2-4 weeks it is approximately 36% of the size of an adult brain (Knickmeyer et al., 2008). By 2 years it has reached 80% of the adult size and by 5 years it is approximately 90% of its adult size (Dekaban, 1978; Knickmeyer et al., 2008). While the size of the brain increases rapidly during the first years of life, the neuronal structure and connections, based on both genetic and experiential factors, continue to mature well into late adolescence and some connections into the twenties (e.g., Huttenlocher & Dabholkar, 1997; Giedd et al., 1999; Giedd, 2004; Lebel & Beaulieu, 2011; Khundrakpam et al., 2012). Brain development trajectories are nonlinear; changes are greater in young childhood than in the late adolescence and young adults (e.g., Lebel & Beaulieu, 2011). It is during this relatively short period, young childhood, when the brain is most plastic and has the greatest capacity for reorganization (Zhang et al., 2002; Nakahara et al., 2004; Sale et al., 2009). Figure 3 shows pre- and post-natal 23

41 neuronal processes in the human brain. Shortly after conception neuronal precursor cells divide and create future neurons and glial cells, i.e., supporting cells, in the ventricular zone of the neural tube. This process is referred to as neuronal proliferation. Once formed, these neurons and glial cells migrate to their eventual locations in the cortex and brainstem. After birth, connections between cells are continuously remodeled via myelination, synaptogenesis, and to a lesser extent, apoptosis (Tau & Peterson, 2010). Figure 3. Neuronal processes in the pre- and post-natal human brain This figure shows neuronal processes related to human brain development including neurulation, neuronal proliferation, migration, apoptosis, synaptogenesis, and myelination. This figure is taken from Tau and Peterson (2010). Myelination and synaptogenesis are the two neural processes thought to be primarily responsible for the maturational changes observed in the auditory system after birth (Paus et al., 1999; Bock et al., 2009). In fact, they are responsible for much of the cortical (re)organization that results from experience. While myelination and synaptogenesis continues throughout an individual s lifespan, they are most active, proceeding at an explosive rate, during development from infancy to childhood (Giedd et al., 1999; Sowell et al., 2004; Lebel & Beaulieu, 2011). Synapse elimination also occurs throughout our lives, but it is most active before adolescence when about half of created synapses are pruned. From adolescence onward a person has a similar amount of 24

42 synapses for the rest of their lives (Shonkoff & Phillips, 2000; Shonkoff, 2010) Myelination Myelin is a white layer, made of protein and fatty substances, that covers around an axon. It acts, in many ways, like an insulator: increasing the speed of electrical transmission along the axon of the nerve fiber. While myelination starts during embryonic stages, newborn brains still contain very little myelin and, as a result, the speed of neural processing is slower and processing of information is less efficient than that in adult brains. Studies using neuroimaging techniques show that brain myelination develops rapidly in the first two years and slows down but continues to improve through childhood, adolescence, and perhaps well into a person s 20s (Paus et al., 1999; Tau & Peterson, 2010). Myelination occurs first in the primary motor and sensory cortices and then progresses to higher-order, or association regions which control the more complex processes responsible for integration of perceptions, thoughts, memories, and feelings (Paus et al., 1999). Maturational changes in white matter of the brain are a reflection of changes in the amount of myelination and/or axonal diameter. Figure 4 illustrates how brain myelination develops with age (Gilmore et al., 2006). The upper panels show T1- weighted axial MRI images acquired at 2 weeks, 1-year, and 2-years from one child. It shows age-related increases in brain size and white matter intensity that is the indication of myelination; a dramatic change in myelination is evident during these first two years of life. The lower panels display images of white matter tractography in a cross-sectional comparison. They show growth of myelination in the corpus callosum and major axonal fibers in infant and adult brains. Other studies show that it takes up to 10 to 16 years for 25

43 the corpus callosum white matter to become fully myelinated (Yakovlev & Lecours, 1967; Clayden et al., 2012; Muftuler et al., 2012). Figure 4. Brain myelination in 2 weeks, 1 year, 2 year, and adults. Top panels show the T1-weighted axial MRI images obtained from one child. White areas represent myelinated axons. Bottom panels show diffusion tensor images of white matter from three subjects. Higher values and yellow and red colors represent greater mylination. This figure is taken from Gilmore et al. (2006). Paus et al. (1999) showed an age-related increase (between 4 and 17 years) in white matter density in thalamocortical tracts and fronto-temporal pathways. Barnea- Goraly et al. (2005) found prominent white matter density increases with age (their subjects range from 6 to 19 years of age) in the internal capsule, inter-thalamic pathways, and the corpus callosum, and prefrontal regions. Consistent with these anatomic changes, the speed of neural processing also reaches its maximum during adolescence (Paus et al., 1999, 2001). Moore (2002b) studied human postmortem brain tissue data obtained from the 16 th -week-gestational fetus to 27-year-old young adults to infer the development stages of the human auditory cortex using the immunostaining of axonal neurofilaments procedure. 26

44 This method shows when the proliferation of axonal neurofilaments occurs, which is related to the maturation of axon size and function for rapid conduction. A newly generated axon is fragile, containing little cell framework elements (axoplasms or cytoskeletons); it does not yet conduct action potentials efficiently. With development, the axon increases in size, increasing neurofilaments within its axoplasm, followed by having a myelin sheath, i.e., myelination; then, it can rapidly conduct action potentials. Cortical development is classified into 3 periods of development of a different axonal system: the perinatal period (3 rd trimester to 4 th postnatal month), the early childhood (6 months to 5 years), and the later childhood (5 to 12 years). Figure 5 shows the neurofilament-immunostained brain tissue across different cortical layers (from Moore, 2000b). During the perinatal period, axons mature first in the most marginal layer I, which is not yet connected with other cortical layers but drives development of other layers. During early childhood the mature axons are present in the layer IV, which consists of the thalamocortical afferents, the first layer receiving input from the lower auditory pathways as well as from the deep layers V and VI. Finally, during school-aged years, mature axons emerge in cortical layers II and III, cortico-cortical connection layers, including commissural axons, which connect two hemispheres, and association fibers, which connect different areas within one hemisphere. By years, the mature axon density is similar to that of young adults. 27

45 Figure 5. Development of neurofilament in axons from 40 weeks to 11 years Each panel shows immunostained neurofilaments of axons obtained at four different ages. This figure is taken from Moore (2002b) Synaptogenesis Synapses are the communication points between neurons. The process of synaptic formation is referred to as synaptogenesis. Approximately over 100 billion neurons in the cortex are produced before birth; most are formed within the first five months of gestation (e.g., Shore, 1997; Shonkoff & Phillips, 2000). Synapses can be noted as early as 5 weeks of gestation within the preplate of cortical layer (Wood et al., 1992; Super et al., 1998). However, at birth only a relatively small number of neurons in the cortex have synaptic contacts and most of them are still poorly connected. After birth and throughout the lifespan neurons within the brain are constantly forming and re-forming connections with each other. Most importantly, the number of synapses in the cerebral cortex rises rapidly during the first few years of life. By two years of age there are well over a hundred trillion synapses in the human cerebral cortex (e.g., Shonkoff & Phillips, 2000). Shore (1997) reported that the brain of a 4-8-year-old child has 50% more synapses than the average adult brain. The period when the number of synapses increases dramatically 28

46 via synaptogenesis is referred to as synaptic overshoot. Huttenlocher and Dabholkar (1997), one of the early electron microscopy studies, counted synapses in the postmortem auditory cortex (Heschl s gyrus) and the prefrontal cortex from 8 brains age ranged from 28 weeks to 3.5 years and 6 brains from 12 to 19 years to study synaptogenesis and synapse elimination. They reported that synaptic density increases after birth and reaches the highest values at age three months in the auditory cortex and at 3.5 years in the prefrontal cortex, doubling adult numbers. Figure 6 shows an increase in synaptic densities in the human primary auditory cortex during the first six years of life (Kral, 2007). The dendritic complexity, corresponding to synaptic density, increases in the first years of life and reaches its peak between 2 and 4 years of age. The number of synapses decrease afterwards, generally by 10 to 15 years of age (for a review, see in Kral & Sharma, 2012). Figure 6. Development of synapses in human primary auditory cortex These figures shows development of synapses from newborn to 6 years old. Dendritic complexity increases during post-natal development, reaching its peak at 2-4 years. This figure is from Kral (2007). 29

47 While frequently activated synapses are fortified, those which are not used will be eliminated. This neuronal, or synaptic, pruning process is modulated by competition among neurons and synapses for nerve growth factors called neurotrophins (Boulanger & Poo, 1999). Experience can regulate both the amount of neurotrophic factors available and the number of receptors for those neurotrophins (Kral et al., 2006). Synaptic pruning, the decrease in the number of synapses, begins at 2-4 years of age when synaptic density reaches its peak and continues gradually until adolescence. Like myelination and dendritic development, this process of synaptic pruning continues throughout our lives but is most active in childhood and adolescence (Huttenlocher & Dabholkar, 1997; Luna et al., 2001, 2004). The number of synapses decreases by about 1/3 beginning in the middle elementary school years (Luna et al., 2001, 2004) and reduces by half, reaching adult values during adolescence in different areas of the cortex (Huttenlocher & Dabholkar, 1997). For example, the number of synapses reaches adult values by age 12 in the auditory cortex and age 15 in the prefrontal cortex (Huttenlocher & Dabholkar, 1997). The effects of maturation on the gray matter of the brain are evident in the loss of volume with age. This reduced volume of gray matter may be due to the myelination of intercortical axons which results in the increase in white matter and decrease in gray matter (Paus 2005; Paus et al., 2001, 2008; Tau & Peterson, 2010) or may represent the pruning of synaptic processes (Mata et al., 1980; Nudo & Masterton, 1986; Tau & Peterson, 2010; Lebel & Beaulieu, 2011). Neuroimaging studies suggest that volumes of cortical gray matter begin to decline during adolescence or later depending on the areas of brain (Gogtay et al., 2004; 30

48 Shaw et al., 2008; Tau & Peterson, 2010; Lebel & Beaulieu, 2011). For example, Figure 7 shows the sequence of gray matter maturation over the cortical surface from 13 healthy children between 4 and 21 years (Gogtay et al., 2004, reprinted in Tau & Peterson, 2010). The color scale represents the volume of gray matter; red indicates more gray matter and blue less gray matter. Therefore, areas in blue correspond to the specific cortices undergoing gray-matter loss, i.e., maturation. Loss of cortical gray-matter volume occurs earliest in the primary sensorimotor areas, followed by temporal and parietal association areas and latest in the dorsolateral prefrontal cortex. These maturational changes in the gray matter of the cortex parallel developmental cognitive milestones. The primary sensorimotor areas responsible for motor and sensory systems mature earliest. The temporal and parietal association cortices involved in language skills mature next. Lastly, the prefrontal cortex responsible for reasoning and executive functions matures last in the developing human brain (e.g., Gogtay et al., 2004). Figure 7. Development of gray matter from 4 to 21 years The gray matter volume are represented in a color scale. Dark purple represents the least and pink represents the most gray matter volume. This figure is taken Tau & Peterson (2010) where they reproduced the figure of Gogtay et al. (2004). 31

49 Lebel and Beaulieu (2011) summarized the long developmental trajectories of brain volumes (white matter, gray matter, and total brain volumes). They obtained MRI scans from 103 subjects whose age ranged from 5 to 32 years. For within-subject maturational changes, the MRI scan was taken at least twice within 3-5 years for 92 subjects. The results show that despite little changes in total brain volume, complex brain developmental process in changes in white and gray matter occurs in a nonlinear development trajectory (Figure 8, from Lebel & Beaulieu, 2011). While the white and gray matter volumes continued to change in the late adolescents, significant withinsubject changes in white and gray matter volumes were shown in children and early adolescents. Figure 8. Changes in white and gray matters and total brain volumes with age The three panels show volume changes in white matter (left), gray matter (middle), and total brain volume (right) 103 subjects aged from 5 to 32 years. This figure is from Lebel & Beaulieu (2011). Without hearing experiences, these developmental processes will be different in brains of deaf individuals compared to hearing individuals. Providing hearing sensation at an early age to congenitally deaf children could reduce the interference of deafness with these developmental postnatal brain processes and help the auditory neuronal networks develop normally. The next sections describe (1) the cochlea and the central auditory nervous pathways, transferring either acoustic or electric stimulation to the auditory 32

50 cortex, (2) the effect of auditory deprivation and electrical stimulation on the auditory pathways, (3) cortical plasticity, and (4) brain imaging methods which are used to study cortical development Development of the Central Auditory System Auditory nervous pathways stimulated via acoustic or electric sounds The ascending pathway of the auditory nervous system starts at the cochlea in the inner ear. Figure 9 (a)-(d) shows the structures of the cochlea and hair cells (Purves, 2007). A traveling wave along the basilar membrane sets movement of the hair cells, which sit on the basilar membrane and bends the stereocilia on the tip of the hair cells. The stereocilia of the outer hair cells are embedded in the tectorial membrane. The outer hair cells work as an active amplifier to enhance the basilar membrane movements. The stereocilia of inner hair cells are not attached to the tectorial membrane; therefore, they bend freely. One direction of bending of the stereocilia results in hyperpolarization, the mechanically gated ion channel remains closed. The other direction of bending of the stereocilia results in depolarization, stretching tip links open potassium channels. The inflow of potassium into the hair cells generates a receptor potential; the depolarized cell releases the excitatory neurotransmitter, glutamate, into the synaptic cleft between inner hair cells and spiral ganglion cells (SCGs). Action potentials start in the spiral ganglion cells, can be recorded 1ms after sound onset, and are sent to the central auditory system, can be recorded about 100 ms after stimulation (for a review, see Abbas, 1993; Kandel, 2000). 33

51 Figure 9. Animated figures of the cochlea and an action potential (a) the cochlea, (b) hair cells in the organ of Corti, (c) transduction in the hair cell, (d) an elicited action potential. It is taken from Neuroscience, Fourth Edition (Purves, 2007). Damage or loss of hair cells is considered to be the primary physiological reason for SNHL. Extensive damages to the inner ear can cause a severe to profound hearing loss. In the United States alone, there are more than one million people with this amount of hearing loss (Blanchfield et al., 2001). Severe to profound SNHL are not simply treated with conventional hearing aids (for a review, see Turner, 2006). Cochlear implantation has become a revolution used to restore sounds to these individuals. Those implanted with CIs can achieve high scores in open-set speech testing and even enjoy music (Zeng et al., 2008). CIs bypass the non-functioning receptor hair cells in the cochlea by using electrical impulses to stimulate the auditory nerve directly. Therefore, a significant number of spiral ganglion cells (SGCs), which connect the inner ear to the 34

52 brainstem, are necessary for successful CI rehabilitation. While the number of SGCs varies depending on the etiology and duration of deafness, human temporal bone studies show that some SGCs are still intact even in profound SNHL subjects (e.g., Fayad et al., 1991; Nadol, 1997). Figure 10 illustrates the electrode array inserted in the scala tympani of the cochlea (a) and a typical current CI system (b). (1) The speech processor picks up sound by using a microphone, converts the sound into a digital signal according to various speech processing strategies and sends it to the headpiece. The headpiece sends the information to the internal receiver using radio frequency signals. (2) The implanted receiver/stimulator decodes the signal and generates a series of electric pulses that are routed to (3) a series of electrodes spaced longitudinally along the cochlear duct. The number of intracochlear electrodes commonly used in current conventional CI systems varies between 12 and 22. (4) These electrical signals stimulate the auditory nerve fibers closest to the electrode contacts and the auditory nerves transmit the neural information to the cortex via relayed pathways. Figure 10. Animated figures of a cochlear implant system (a) shows the electrode array inserted in the cochlear and (b) shows the external and internal devices of a cochlear implant system. Sources: (a) Functional replacement of the ear by Gerald E. Loeb. Copyright 1985 by Scientific Abreican, Inc. (b) cohlearamericas.com 35

53 Once action potentials are generated in auditory neurons by either acoustic stimulation or electrical stimulation, action potentials travel common central auditory pathways through a series of relay stations. Figure 11 illustrates the major landmarks within the ascending pathways from the auditory nerve to the auditory cortex. Figure 11. Animated figure of central auditory pathways This image is taken from Neuroscience, Third Edition, Figure 12.12, (Purves, 2004). The afferent fibers of the auditory nerve project initially to the ipsilateral cochlear nucleus (CN) complex located in the rostral medulla. This structure includes the anteroventral CN, the posteroventral CN, and the dorsal CN. Both postsynaptic axons from the anteroventral CN and posteroventral CN project to the superior olivary complex (SOC) located in the pons; some axons synapse in the ipsilateral SOC and other axons cross and synapse in the contralateral SOC. Axons from the dorsal CN also synapse in the ipsilateral SOC or cross to the contralateral SOC. In fact, more axons decussate than 36

54 travel ipsilaterally. Binaural innervation at this level in the auditory brainstem is likely to be a necessary prerequisite for sound localization, allowing for neural encoding of differences in the sound arrival time or the differences in the sound intensity between the two ears (Rapparport & Provencal, 2002). All postsynaptic axons from the SOC pass along the lateral lemniscus bilaterally, but a majority of these axons ascend in the contralateral lateral lemniscus and project to the nucleus of inferior colliculus (IC) located in the midbrain. Postsynaptic axons from the IC project to the medial geniculate body (MGB). This is known as the auditory nucleus of the thalamus. It is located near the midbrain and at the tail of the thalamus. Axons of afferent neurons in the medial MGB project to the primary auditory cortex, also called the A1 region. The A1 region is located on the Heschl s gyrus in the superior temporal lobe (Brodmann area 41). Using cortico-cortical connections, the A1 relays the incoming information to the higher-order areas, including the secondary auditory cortex (Brodmann area 42), which surrounds the primary auditory cortex (Kandel, 2000; Rappaport & Provencal, 2002). The primary areas represent features of the auditory object that the cortex is processing (e.g., they identify the frequencies the sound is composed of), the higher-order areas classify the elements into objects; for example, neural activity in Wernicke s area is involved in perception of spoken words. The A1 region also sends neural feedback to the auditory thalamus. The human cortex has six layers, labeled as Layer I (superficial) to VI (deep). Functionally, they can be classified into the supragranular (lavers I to III), the internal granular (layer IV), or the infragranular layers (layers V and VI). The supragranular layers are responsible for both intrahemispheric (cortico-cortical) and interhemispheric 37

55 (association) neuronal communication. The supragranular layers form the foundation for more complex cortical processing of auditory input (Moore, 2002a,b). The internal granular layer receives information from all thalamocortical afferents, including thalamic input from peripheral auditory system; therefore, it is most prominent in the primary sensory cortices. Activity progressing from the internal granular layer is sent to the supragranular layers (mainly, layer II and III), and from the supragranular layer to the infragranular layers. These layers connect the cerebral cortex with subcortical structures and project feedback to subcortical auditory structures. Since these deep layers (V and VI) integrate ascending inputs from the auditory thalamus with descending inputs from higher-order areas, these output layers can modulate and direct cortical plasticity (for a review, see Kral, 2007; Kral & Eggermont, 2007). Tonotopic organization is preserved from the cochlea all the way to the auditory cortex. In the cochlea, neurons innervating hair cells located at specific locations along the basilar membrane respond best to a restricted frequency range. Neurons located near the basal end of the basilar membrane respond best to high frequencies. Neurons that located the apex of the basilar membrane responds best to the low frequencies. The design of the internal electrode array spaces the electrode contacts out along the longitudinal axis of the cochlear duct in order to use the tonotopic information of the cochlea. Auditory nerve neurons synapse at the CN where this tonotopic organization is maintained. At the CN, high-frequency information, conveyed from neurons near the base of the cochlea is carried to the deep part of the CN. The low frequency information from the apical turn of the cochlea, is carried to the superficial part of the CN. This 38

56 tonotopic frequency arrangement is maintained in organization at the SOC and carried through to the IC. High frequency information results in activation of neurons in the ventral portion of the IC and low frequency input results in activation of neurons located in the more dorsal regions of the IC. The axons from the IC project to the MGB and then to the A1 region of the cortex; both areas preserve the tonotopic arrangement. Particular regions of A1 called isofrequency bands, respond to specific frequencies Auditory deprivation and electrical stimulation Sensory hearing loss due to damage or loss of hair cells is irreversible and can later reduce the density of SGCs and CN volumes. Haride and Shepherd (1991) observed SGC density and CN volumes in 4 NH cats (controls) and 13 neonatally deafened cats (with an ototoxic aminoglycoside). At the time of testing, 12 cats had less than 2.5 years of deafness; one cat had longer auditory deprivation (> 8 years). In the short-term deaf cats, SGC density reached 17 % of that reached by NH cats, and total CN volumes were reduced by 46%. The long-term deaf cat had only 1.5% SGC density compared to NH cats and had a 60% reduction in CN volume. Animal studies have also shown that some detrimental changes due to cochlear damage can be reduced or prevented via electrical stimulation (Leake et al., 1991; Snyder et al., 1995). Leake et al. (1991) injected neomycin sulfate to 10 newborn kittens to induce profound cochlear deafness and implanted them unilaterally between 9 and 17 weeks of age. Six of the ten implanted kittens were stimulated over the course of 1-3 months while the other four implanted kittens were not stimulated. With electrical stimulation, however, the implanted ears of the six implanted kittens had more spiral ganglion neurons than the non-implanted ears. Snyder et al. (1995) recorded temporal 39

57 responses from single neurons in the IC of deafened cats: (1) neonatally deafened cats with electrical stimulation, (2) neonatally deafened cats without electrical stimulation, and (3) adult-deafened cats without electrical stimulation. The IC neurons in the neonatally deafened cats with electrical stimulation had responses with shorter latencies and phase locking to higher electrical pulse frequencies compared to neurons in cats in the other groups. This study suggests that reactivating the developing auditory system via chronic electrical stimulation can improve temporal processing mechanisms. Performance of CI children also can reflect the benefits of electrical stimulation for the development of cortical areas related to speech and language development. Despite variable performance, numerous studies show that children with severe-toprofound hearing loss can achieve good speech and language scores after implantation. In one early study, Tomblin et al. (1999) measured language comprehension and production scores in 29 profoundly deaf children with CIs implanted between 2 and 13 years. Their performance was compared to 19 deaf children with HAs (between from 3 and 14 years). CI children had significantly better performance than deaf children with HAs. More recently, Yoshinaga-Itano et al. (2010) showed that 87 children with severe-to-profound HL demonstrated the equivalent rate of language growth from 4 to 7 years (both in production and comprehension) as NH children. They further investigated language growth trajectories between children with HAs (n = 38) and children with CIs (n = 49), most implanted between 1 to 2 years. Children in both groups had a severe to profound hearing loss and were matched for age of intervention and nonverbal cognitive ability. Results show that children with CIs as a group were less deviant from the age equivalent language growth rate of children with NH than children with HAs. 40

58 Animal and human studies support the idea that much of the central auditory system development is experience dependent and that cochlear implantation (i.e., electrical stimulation) facilitates the development of the central auditory system in deaf ears. The importance of early and chronic electrical stimulation has been emphasized for children with profound hearing loss who received no developmental speech or language benefit from amplification (for a review, see Hardie, 1998). Because the developing brain has more plasticity, it is expected that auditory input via electrical stimulation would provide a positive effect on the development of the central auditory system (Moore, 2002a; Sale et al., 2009). However, there is a sensitive period for positive changes in neural reorganization after auditory deprivation Neuroplasticity: stimulating auditory pathways using CIs Kral et al. (2002) showed that cats have a critical period lasting up to 5 months. A total of 10 congenitally deaf cats were divided into 3 groups: 5 cats implanted at months (early implantation), 2 cats implanted at 5-6 months (late implantation), and 3 cats with no treatment. The early-implanted cats had larger activated cortical areas and more robust middle latency responses than the cats that were implanted later in life. In a subsequent study, Kral et al. (2005) show decreased activity in deeper layers of the primary auditory cortex in the late-implanted cats. This decreased activity may disrupt the feedback connection between the primary auditory cortex, association areas and other cortices including speech and language areas. These two studies suggest that implanting a CI when the brain is still reasonably plastic may help optimize outcomes. Moreover, children with late implantation (after the critical period) might demonstrate reduced multimodal integration abilities, poor speech perception, and late or limited language 41

59 development. Behavioral studies of CI children also have shown that late-implanted children performed worse in a multimodal integration task than early-implanted children. For example, Gilley et al. (2010) measured the reaction time to the detection of three types of inputs (auditory, visual, and auditory-visual) in NH children (7-12 years), earlyimplanted CI children (9-14 years, implanted before 4 years), late-implanted CI children (10 to 13 years, implanted between 5 and 9 years), and NH adults (23-26 years). All participants were required to press a button as soon as they detected any auditory, visual, or auditory-visual stimulus. NH adults had the fastest reaction time for all stimuli. Lateimplanted CI children had significantly slower reaction times not only for auditory or visual stimulus but also for auditory-visual stimulus. Kral et al. (2005) may explain this result by the fact that the performance, including multisensory integration ability, will be affected in late-implanted children due to near absent responses in supragranular and infragranular layers. Schorr et al. (2005) investigated the capacity for bimodal integration in children who were implanted before 30 months of age compared to children who were implanted after 30 months of age. They presented /pa/ and /ka/ as auditory and visual stimuli independently. They also presented a combined stimulus using auditory /pa/ and visual /ka/, in this case the stimulus is perceived as /ta/ (McGurk test). Results show that the bimodal fusion ability was affected by age of implantation. Children implanted early in life showed more consistent audiovisual fusion than children implanted later. This study supports the idea of a sensitive period for auditory development in the cortex which affects the bimodal fusion. Together with Kral et al. s observation of near absent activities in deeper layers responsible for connections with other higher cortices, these 42

60 results may help explain the poor speech and language performance for the lateimplanted CI children. Using a positron emission tomography (PET) scan, Lee et al. (2001) measured glucose metabolism in the superior temporal and inferior frontal regions (primary and secondary auditory cortex) in 15 deaf individuals ranging from 2 to 20 years. They measured glucose metabolism before cochlear implantation. They compared these metabolism results with speech perception testing after cochlear implantation. They expected good speech perception scores after implantation if auditory cortex areas had not been taken over any other modalities (shown by an increase in hypometabolic activity) and poorer speech perception scores if hypometabolic activity areas decrease compared to NH children. They observed that as the duration of deafness increased, the extent of the hypometabolic area decreased in the primary auditory cortex and associated areas, indicating that these areas were taken over by other modalities, such as vision or sign language. Also, researchers showed a positive correlation between the size of hypometabolic areas and the speech perception scores measured using the Korean version of the Central Institute of Deafness (K-CID) test. Oh et al. (2003) also measured hypometabolic activity using PET and followed 21 pre-lingually deafened children and 17 post-lingually deafened adults for up to four years post-implantation. They also showed negative correlations between the K-CID scores and age at implantation for both pre-lingually and post-lingually deafened groups. Children implanted between 5 and 7 years showed more variability in outcomes than children implanted before 5. They found similar results to Lee et al. (2001): the individuals with the widest hypometabolic area in the auditory cortex before implantation achieved the 43

61 best K-CID scores. Both studies indicate that, as duration of auditory deprivation extends, the regions of hypometabolism decrease since other modalities may use these areas. Therefore, even with implantation, children with prolonged auditory deprivation cannot have good speech perception. In fact, numerous previous behavioral studies have shown that children who were implanted at or before 3-4 years develop significantly better speech and language skills compared to children who received their cochlear implants after 6-7 years of age (e.g., Kirk et al., 2002; Geers, 2003; Harrison et al., 2005; Nicholas & Geers, 2006; Holt and Svirsky 2008; Sharma et al., 2009). Early-implanted children show the ability to overcome their auditory deprivation period and show better multimodal integration skills and higher speech perception scores than late-implanted children. In the next section, the brain imaging and auditory evoked potentials will be explored as efficient methods to study neural (re)organization of the cortex as a result of normal development, damages, or rehabilitation. The obligatory auditory evoked potentials at the cortical level will be our primary interest in this study to understand how the typical central auditory system develops in humans and how a CI, implanted early in a child s life, impacts the development of the impaired auditory system Methods to study the development of central auditory system Brain imaging Researchers have used several different methods to assess the effect of development on the central auditory nervous system in humans. One approach is to use behavioral assessment techniques. For example, studies using a conditioned head-turn procedure showed that even very young infants could discriminate different speech 44

62 sounds (Werker & Tees, 1984; Kuhl et al., 1997). However, behavioral measures may not be easy or accurate in infants and young children (Kuhl & Rivera-Gaxiola, 2008). Speech and language skills are limited and highly variable in young children; this complicates interpretation and may force reliance on parent surveys. Young subjects can be affected by non-auditory factors such as attention span and memory (Viemeister & Schlauch, 1992). In short, while behavioral measures of hearing and speech perception are necessary, they are not always practical. Auditory development has also been studied using non-invasive neuroimaging (or brain imaging) techniques. Studies using these structural and functional brain-imaging techniques have expanded our understanding of auditory cortex development and sound processing and can serve as an alternative to behavioral assessment. The neuroimaging techniques include computed tomography (CT), MRI, fmri, PET, single photon emission computed tomography (SPECT), and Electroencephalography (EEG). CT scans create 3-dimensional images taken using x-rays. While bone and hard tissue absorb X-rays well, soft tissues, air, and water do not. Another limitation is that the subject needs to remain still during scanning which means that sedation is often required if CTs are performed on very young children. Finally, many researchers have become less enthusiastic about the use of CTs to assess typical development because of the risk of unnecessary exposure to radiation (e.g., Gazdzinski et al., 2012; Pearce et al., 2012). Unlike CT scanning, MRI and fmri scanners do not use radiation but do require exposure to use high electromagnetic fields. The fmris allow researchers to obtain brain images with excellent spatial resolution of 2 mm or less (Gernsbacher & Kaschak, 2003; Tobey & Devous, 2007). The temporal resolution, the time taken to obtain images after 45

63 the stimulus onset, of fmri is one second or less (Tobey & Devous, 2007). MRIs result in good images of soft tissue, including muscles, ligaments, and organs; fmris measure changes in blood oxygenation levels in response to neural activity. MRI/fMRI scans have practical limitations for use with children. The participants need to remain very still during data acquisition. The MRI/fMRI scans also produce very loud noises; therefore, sometimes infants ears need to be shielded (Tobey & Devous, 2007; Kuhl & Rivera-Gaxiola, 2008). Additionally, fmris often require active participation by the subject, which can be unreliable in young children. Due to the strong electromagnetic fields created by MRI scans, especially fmris, they cannot be used with CI users. PET and SPECT images have one great advantage over the MRI scans: they are safe for cochlear implants (Tobey & Devous, 2007). PET and SPECT can also be used in a quiet environment (Rahmim & Zaidi, 2008). Both PET and SPECT measure cerebral blood flow or regional cerebral glucose metabolism to identify neuronal populations involved in activating tasks (Rahmim & Zaidi, 2008). The change of metabolic activity is consistent in time courses for myelination and synaptogenesis. These methods can be used to study metabolism in the auditory cortices of deaf and NH hearing listeners. Also, cortical metabolism appears to be responsive to electrical stimulation and alternations in cortical activations are observed in response to speech and other sounds (Tau & Peterson, 2010). While PET and SPECT can be used with CI users, they have the some of the same limitations as MRI/fMRI. They require a subject to remain still during testing. All functional brain-imaging studies need a contrast between cortical functions measured 46

64 under at least two test conditions. Therefore, depending on the cognitive task, obtaining functional brain images can be very challenging for very young children. Additionally, PET and SPECT studies involve the injection of a radioactive isotope of a chemical element (i.e., radioisotope) often via an IV or arterial line. This requirement alone often rules out participation by young children. PET scans also have relatively poor temporal resolution Auditory evoked potentials EEG records electrical activities using surface electrodes on the scalp. It is a noninvasive brain imaging tool. EEGs have several advantages over other imaging methods. They can detect changes in the ongoing electrical activity of the brain with relatively good time resolution (a millisecond or even less) compared to other brain imaging techniques (Gains & Kosslyn, 2002; Kuhl & Rivera-Gaxiola, 2008; Karmiloss-Smith, 2010). It is the least expensive method of neuroimaging and, most importantly, it can be easily used with all age groups of various hearing status, including very young children with HAs and/or CIs. Auditory evoked potentials (AEPs) are measures of the synchronized EEG activity using auditory stimuli presented repeatedly. AEPs can provide an objective (i.e., non-behavioral) indicator of detection and, in some cases, discrimination of sounds in ongoing auditory stimuli in people of various hearing statuses and ages. There is a range of different AEPs that can be recorded from humans. They are often classified based on the listener s involvement (processing contingent vs. obligatory potentials). While processing contingent AEPs require the use of an active listening paradigm, obligatory auditory evoked potentials can be recorded in a passive (or 47

65 unattended) listening paradigm making them ideal for use with pediatric populations. Obligatory AEPs are primarily dependent on the physical properties of the stimulus; i.e., intensity, frequency, and duration (Hyde 1997; Stapells, 2002). Obligatory AEPs can be also divided into early (<10 ms), middle (10-50 ms), and long ( 50 ms) latency potentials depending on the time between the presentation of the stimulus and resultant neural response. Response latency provides an indication of where the AEP is generated in the auditory pathways. The auditory brainstem response (ABR) is a short-latency evoked response. ABR responses are typically generated more peripherally, i.e., at the level of the auditory nerve and the low brainstem (e.g., for a review, see Moller, 1994). The ABR has become a very popular clinical tool for assessment of hearing sensitivity in infants and young children since the presence of the wave V corresponds well with behavioral thresholds and since the waveforms can be obtained regardless of sleeping stages. While the middle latency AEPs that occur between 10 and 50 ms also provide developmental information, this is not of interest in this study. The longer latency AEPs occur later than 50 ms post stimulation and have more central neural generators in or near the auditory cortex. These are also referred to as cortical auditory evoked potentials (CAEPs). Maturation of the central auditory system beyond the cochlea, characterized by myelination and synaptogenesis, is reflected in the time course of auditory responses. Therefore, obligatory AEPs that are characterized with different latencies have been used successfully to address questions of neural maturation of the human auditory system from cochlea to cortex for both NH and CI listeners (e.g., Moore et al., 1995; Ponton et al., 1996a,c; Sharma et al., 1997). 48

66 a. Auditory brainstem response The ABR is evoked in response to a brief auditory stimulation, such as clicks or tone bursts. The normal ABR is characterized by a series of five vertex positive peaks, labeled traditionally with Roman numerals I through V. Wave I has a latency of 1-2 ms. The auditory nerve is the generator of Wave I. Eggermont et al. (1991) show that Wave I was present in their all 40 full-term newborn babies in their study; this indicates that the auditory nerve is intact at birth. Wave V of the ABR reflects contributions of the inferior colliculus (midbrain) neurons and has a latency of 5-6 ms. The time interval from Wave I to V is interpreted as a measure of conduction time. The ABR is routinely used for evaluation of hearing in infants and otherwise difficult to test populations. The effects of development on the ABR have been well established (e.g., Reiman et al., 2009). Replicable ABRs have been recorded from premature infants as young as 28 weeks gestational age (Amin et al., 1999; Ponton et al., 1996c), and from 32 weeks gestational age onward; an ipsilateral and/or contralateral response is present in 90% of all the infants (Coenraard et al., 2011). In full-term babies, the amplitude of wave I is larger, relative to wave V, than is typical for adults. This relationship changes over the first few years of life (Hecox & Burkard, 1982; Starr et al., 1977; Eggermont & Salamy, 1988). The latency of wave I is longer in infants, but quickly decreases and becomes adult-like 3 to 6 weeks after birth. Morphology and latencies of waves III and V rapidly change during the first couple of years of life reaching adult values (Fria & Doyle, 1984; Moore et al., 1995; Coenraad, et al., 2010). This leads to a gradual decrease in the time interval between waves I and V between 11 and 18 months (Ponton et al., 1992, 1996). 49

67 b. ECAPs and EABRs When compound action potentials and ABRs are electrically elicited they are referred as to ECAPs (electrically evoked compound action potentials) and EABRs (electrically evoked auditory brainstem responses). The ECAP is essentially equivalent to Wave I of the EABR, representing responses from the auditory nerve. Gordon et al. (2003) reported that even at initial activation, 94% of children (47 out of 50) had wellformed EABRs. That shows the connections between the cochlea and brainstem are all mostly intact, even in a congenital ear. Researches also showed that in pediatric CI recipients, the early waves (wave I recorded using ECAPs and wave II recorded using EABRs) show little change in latency within the first couple of months of CI use. Later waves (III and V) show a decrease in latency over a longer period but reach adult values within the first year of CI use. Neither ECAP nor EABR latencies appear to be dependent on the age at implantation, i.e., the duration of deafness (Gordon et al., 2003, 2006; Sharma & Dorman, 2006). Studies using objective ABR and EABR/ECAP measures originating from the cochlear nerve and brainstem show more peripheral structures in the human ear at birth that fully mature in the first a couple of years of life. These findings support the idea that the auditory system develops first in the periphery (e.g., Rubel, et al., 1998; Moore, 2000a; Moore, 2000b). c. Cortical auditory evoked potentials (CAEPs) CAEPs are long-latency responses measured at least 50 ms after the onset of an auditory stimulus. CAEPs are believed to reflect excitatory post-synaptic neural activities at the precortical (thalamus) and cortical auditory areas. Their latencies are thought to 50

68 reflect the sum of delays in synaptic transmission throughout the central auditory pathways (Eggermont et al., 1997; Purdy et al., 2001; Ponton & Don, 2003; Sharma & Dorman, 2006; Wunderlich & Cone-Wesson, 2006). CAEPs include several different evoked responses, such as the P1-N1-P2 complex, the acoustic change complex (ACC), the mismatch negativity (MMN), and a P300 response. The P1-N1-P2 complex, first described by Davis (1939), is referred to as the obligatory onset response because it primarily relates to the physical characteristics of the stimulus and consistently occurs after stimulus onset. When it is present, it is thought that the sound is detectable by the subject (for a review, see Davis & Zerlin, 1966; Hyde 1997; Stapells, 2002). The effect of maturation on the P1-N1-P2 response has been widely reported, especially using the changes in P1 latency with development (e.g., Ponton et al., 2000; Ponton & Eggermont, 2001; Sharma et al., 2002a,b; Wunderlich & Cone-Wesson, 2006). The rest of the CAEPs (ACC, MMN, and P300 responses) can be used to estimate the discrimination capacity in the auditory cortex. Much less is known about how maturation of the central auditory pathways affects these neural responses. One of the overall goals of the current study is to better understand these processes. An overview of the P1-N1-P2 complex and the ACC is provided in the next section. Factors known to affect these recordings are reviewed as well as the differences among the ACC, the MMN, and P300 responses. 51

69 2.3. The P1-N1-P2 Complex Generators P1 is thought to be generated in the medial geniculate nucleus (i.e., auditory thalamus) and the primary auditory cortex, specifically in layer IV of Heschl s gyrus, located in the superior temporal gyrus (Buchwald et al., 1992; Liegeois-Chauvel et al., 1994; Ponton & Eggermont, 2001; Sharma et al., 2002 a, b; Sharma & Dorman, 2006). Layer IV is where the excitatory thalamic inputs enter the cortex. Recording pyramidal neuron activity in this layer results in a positive peak (i.e., P1) from the scalp (Eggermont & Ponton, 2003). P1 latency reflects the total delay of synaptic transmission from the periphery to these more centrally located generators (e.g., Cauller & Kulics, 1988; Steinschneider et al., 1992; Eggermont et al., 1997). P1 latency decreases with increasing age as myelination and synaptic formation and elimination in these areas progress. The latency change in P1 is therefore often considered to be a marker of how maturation affects the central auditory pathways (e.g., Sharma et al., 2005). N1 is thought to have multiple neural generators, reflecting neural activity in layers II and III of the primary auditory cortex and auditory association areas. These neural generators include Heschl s gyrus, the planum temporale, and the cingulate gyrus (Knight et al., 1980; Näätänen & Picton, 1987; Ponton et al., 2000; Sharma et al., 1997, 2007; Wunderlich et al., 2006; Kral & Eggermont, 2007; Martin et al., 2007). Layers II and III are responsible for the recurrent intra- and inter-cortical pathways, including primary bilateral sources and are mainly considered as a current source of the negative peak (Näätänen & Picton, 1987; Cauller & Kulics, 1988; Steinschneider et al., 1992; Eggermont & Ponton, 2003). These N1 generators may be immature in young children 52

70 and therefore, the N1 response is typically absent in young children, especially when the stimulus is presented at a fast rate. P2 is the least understood component of the complex. The P2 response is thought to reflect activation of neurons in the primary auditory cortex particularly near Heschl s gyrus and multiple other areas including the secondary auditory cortex (Knight et al., 1980; Hari et al., 1984; Crowley & Colrain, 2004; Martin et al., 2007) and the reticular activating system (Rif et al., 1991; Crowley & Colrain, 2004). N2, or immature negativity, follows the P1-N1-P2 responses and is prominent in young children. Some researchers think that N2 has neural generators in the frontal lobes and the subcortical structures including the limbic system (e.g., Kiehl et al., 2001), while other researchers think that it has generators in the supratemporal planes (e.g., Bruneau & Gomot, 1998; Ceponien et al., 2002). Gomot et al. (2000) found that the frontal lobe contributes to the N2 activities. Ceponien et al. (2002) found N2 generators located anteriorly to N1 generators, i.e. distinct from each other, but both located in the supratemporal lobes. Ceponien et al. (2002) believe that the sources of N2 is in the supratemporal lobes because N2 emerges early in life when the frontal lobe is the least active and because N2 changes according to acoustic features, encoded in the auditory cortex in children, General morphology A wide range of repeated stimuli can elicit the P1-N1-P2 complex, including pure tones, clicks, tone bursts, and speech sounds. In NH adults, three peaks of the P1-N1-P2 response are recorded between 50 and 300 ms post-stimulus onset: the prominent N1 peak measured at a latency of approximately 100 ms follows P1 at a latency of ms 53

71 and precedes P2 at a latency of approximately ms (e.g., Ponton et al., 1996 a, b; Sharma et al., 1997; Stapells, 2002; Gilley et al., 2005). The N1-P2 peak-to-peak amplitude is dominant in the response. Unlike the latencies, the N1-P2 amplitude varies considerably depending on the stimulus parameters; it is measured to be at least 5-10 µv and can be up to 25 µv using the suprathreshold stimuli (Lightfoot & Kennedy, 2006). In NH children, the response complex is strongly dominated by P1, which is consistently recorded with latencies between 300 and 400 ms post-stimulus onset during infancy and between 100 and 300 ms during childhood (e.g., Sharma et al., 1997, 2002; Ponton et al., 2002). P1 latency decreases with age reaching ms in adults (for a review, see Wunderlich & Cone-Wesson, 2006). The amplitude of P1 also decreases with age. For example, 4 µv in 3-4 years olds reduces to 1 µv in adults (Ceponien et al., 2002). With a typical stimulus presentation rate (1 or 2 stimuli per second), N1 and P2 are usually absent at birth and begin to emerge at around 7-8 years of age (Ponton et al., 1996) and are not consistently measured until 12 years of age (Gilley et al., 2005; Ponton et al., 2000). While Small and Werker (2012) report that adult-like P1-N1-P2 morphologies were obtained in 4-month-old infants using /daba/ stimulus (564 ms in duration) with an ISI of 2.2 second, only the P1 component was consistently present and its latency was, on average, 113 ms, 66 ms longer than average adult P1 values. Unlike adult responses, a large, and broad negativity following the dominant P1 occurs at ms or up to 450 ms post-stimulus onset for infants, toddlers and even in young school-aged children (e.g., Kushnerenko et al., 2002; Ponton et al, 2002). Because this late negativity is obtained in immature CAEP waveforms, some researchers refer to this as an immature negativity or the late negativity (Kurtzberg et al., 1984; 54

72 Pasman et al., 1991, 1999; Pang and Taylor, 2000; Kushnerenko et al., 2002; Sharma et al., 2002a,b). In fact, it has earned various names by different researchers: N1b (e.g., Sharma et al, 1997; Ponton et al., 2000), N2 (e.g., Pasman et al., 1999; Sussman et al., 2008; Small & Walker, 2012), N250 and N450 depending on its latency (e.g., Kushnerenko et al., 2002), N late (e.g., Purdy et al., 2005). The P1 component has often been obtained from infants and young children and is identifiable in adults as well. P1, particularly its latency, has been a focus used to describe the changes with development Effects of stimulus, recording, and subject variables on the P1-N1-P2 While the P1-N1-P2 complex is primarily obligatory, its morphology, latency, and amplitude can still be affected by not only stimulus and recording factors but also subject factors (Näätänen & Picton, 1987). The next two sections describe (1) the effect of stimulus and recording s exogenous factors on the CAEPs: stimulus rate, scalp topography, and stimulus/implant artifact, and (2) the effect of the subject s endogenous factors (sleep, attention, and hearing status) on the CAEPs Stimulus rate Stimulation rate, or inversely, the length of the inter-stimulus interval (ISI), can have a profound effect on the morphology and amplitude of the component peaks of the P1-N1-P2 complex (Ceponiene et al., 1998; Gilley et al., 2005; Wunderlich et al., 2006; Small & Werker, 2012). Traditionally, studies of the maturation of sound processing that have used CAEPs have used stimulation rates slower than 1 Hz (ISIs = 1 s) (Ceponiene et al., 1998; Ponton et al., 2000; Wunderlich et al., 2006). At very short ISI (below about 300 ms) the amplitude of N1 is often diminished so greatly that it may not be readily 55

73 detected (Näätänen & Picton, 1987; Teder et al., 1993). Long ISIs may enhance the amplitude and morphology of CAEP component peaks, improving the signal-to-noise ratio. Also, Cepoinene et al. (1998) reported the N1 peak latency was changed across different ISIs (350, 700, and 1400 ms) in children ages 7-9. The N1 peak shortened in latency from 250 to 160 ms with different ISIs (350 and 1400 ms, respectively). Several other researchers have shown the systematic effect of stimulation rate on the onset P1- N1-P2 response in children (Gilley et al., 2005; Sussman et al., 2008; Small & Werker, 2012). Gilley et al. (2005) studied morphological changes of the P1-N1-P2 responses recorded at varying stimulation rates. They used the vowel /uh/ with a duration of 23 ms. They presented this vowel 4 times in the stimulus train using 4 different ISIs varying from longest (2000 ms) to shortest (360 ms). Each stimulus train was separated by a sequentially decreasing ISI. Fifty NH children, whose age ranged from 3 to 12 years, and ten NH young adults in the twenties participated. Figure 12 shows the results (Gilley et al., 2005). P1 was detected in all subjects and in all ISI conditions. While P1 latencies were stable in all ISI conditions in the adult group, it was found to be shorter in the 2000 ms ISI condition than in the 360 ms ISI condition for both young children (3-4 years) and older children (11-12 years). Detectability of the N1 and P2 components increased with age. For example, N1 and P2 components were measured in 63% of the year old children when the shortest ISI was used but they were recorded for 100% of the same children in the same age group when the longest ISI was used. By years old the P1, N1, and P2 components are apparent for all ISIs but the N1 and P2 components have the largest amplitude in the slowest ISI conditions. 56

74 Figure 12. Grand average waveforms obtained using different inter-stimulus intervals Each sub-figure shows mean waveforms of four age groups obtained using 2000, 1000, 560, and 360 ms ISIs. This figure is taken from Gilley et al. (2005). Sussman et al. (2008) explored the effect of age and ISI on P1-N1-P2 responses using rapid ISIs, which they argue may be more predictive or realistic of speech. Fortynine children (8-16 years) and 12 adults (22-40 years) participated in this study. Children were grouped by age: 8 years (n=10), 9 years (n=7), 10 years (n=12), 11 years (n=10), and 16 years (n=10). They presented pure tones (88 Hz) binaurally via insert earphones and varied the ISIs (onset-to-onset), 200, 400, 600, and 800 ms. Results show a profound ISI effect on the CAEP morphology. The individual components were best identified when the slowest ISI (800 ms in this study) was used. P1 was observed for all age groups and all ISIs. Increasing the stimulation rate resulted in a loss of the N1 and P2 peaks. With the fastest ISI (i.e., 200 ms) only P1 was identified even in the adult and adolescent (16 years) groups. Both age and ISI also affected P1 latency. In addition, they reported 57

75 that the increased speed of presentation resulted in waveforms that appeared to reflect patterns observed in younger age groups; for example, the morphology of the waveforms elicited in the 400 ms condition in the 16-year-age group appeared similar to those obtained in the 800 ms condition in the 11-year-age group. The authors reported that this is because of the extra processing load put on the auditory system. The P1 component persists in the waveform regardless of age or stimulus rate. Sussman et al. (2008) hypothesized that P1 has a different refractory period and neural generators than the N1 and P2 components. While N1 disappeared with increasing rate in adults, N1 was absent even at longer ISIs in children, suggesting that the refractory periods for N1 generators are longer for children than they are for adults. Small and Werker (2012) used a speech stimulus /da/ with duration of 564 ms presented at a very slow rate (ISI = 2200 ms) to elicit a P1-N1-P2 response from infants. They found that the morphology of the grand mean responses of twenty-five 4-month-old infants was similar to those of adults, showing the three peaks, P1, N1, and P2. The latencies of each component in infants were reported to be 55 to 120 ms later than peak latencies of adults. These results are very different from other studies showing negative peaks rarely present in infants and young children. Small and Werker (2012) suggest that the adult-like morphology observed in infants could be due to long ISIs. To explore this further, Small and Werker (2012) can be compared to the Gilley et al (2005) and Wunderlich et al. (2006) where each used similar speech sounds with longer ISIs than other previous studies. Wunderlich et al. (2006) presented the word bad, with duration of 200 ms, and with an ISI of between 3100 and 6050 ms (median 4600 ms) to elicit CAEPs in newborns, children (1-3 years, 4-6 years), and adults. While their ISI is much 58

76 longer than the one used in Small and Werker (2012), Wunderlich et al. (2006) reported that the morphology of P1-N1-P2 responses were not adult-like in either group children. Gilley et al. (2005) (discussed earlier in this section) supports the results of Wunderlich et al. (2006). With an ISI of 2000 ms, two age groups of children (3-4 years and 5-6 years) did not show N Scalp topography There have been many studies exploring scalp topography for the P1-N1-P2 response. In adults, the largest responses for all three components are typically obtained using electrodes placed on the frontocentral recording sites (e.g., Tonnquist-Uhlen et al., 1995; Potts et al., 1998; Ceponiene et al., 2002; for a review, see Wunderlich & Cone- Wesson 2006). In studies using a simple bipolar electrode setup, researchers typically select the vertex (Cz) as the active electrode and a mastoid area as the reference electrode for maximal amplitudes of responses (e.g., Sharma et al., 2002a, 2002b, 2005; Nash et al., 2008). Fewer studies of full scalp recordings in infants and young children are reported. Available studies show that scalp topography affecting the amplitudes of evoked potentials change with age. In infants, the largest P1 response amplitude appears to be in relatively broad frontal, parietal, and temporal areas (Wunderlich & Cone-Wesson, 2006). At 6-7 years old, P1 amplitude is larger at the midline (frontally or at the vertex) than its amplitude measured at the posterior scalp or at the temporal scalp (Satterfield et al., 1988; Kurtzberg et al., 1995; Nelson et al., 1997; Ceponiene et al., 2002). In younger adolescents (10 to 14 years) the scalp distribution of N1 is wide. In older adolescents (15 years and older) it becomes more localized and adult-like with a frontocentral distribution 59

77 (Tonnquist-Uhle n et al., 1995; Oades et al., 1997; Ponton et al., 2000). There is inconsistency in results of scalp topography comparing the right versus left hemispheres. Using binaural stimulation, Satterfield et al. (1988) reported that a response is larger on the right hemisphere than the left hemisphere. However, Oades et al. (1997) did not find significant hemispheric differences Subject factors a. Sleep and attention Several different studies have shown that CAEPs are strongly affected by anesthesia (Plourde & Picton, 1991), coma (Fischer et al., 2000), sleep state (Campbell & Colrain, 2002), and attention (Näätänen, 1990). Sleep consists of rapid eye movement (REM), where most dreaming occurs, and non-rapid eye movement (non-rem) stages. Non-REM sleep stage can be divided into three stages: Stage 1 (drowsiness), Stage 2 (almost unconscious and unaware of the external environment) and Stage 3-4 (deep sleep). Campbell and Colrain (2002) reviewed previous studies about the effect of sleep onset period (Stage 1 and 2) on short-latency, mid-latency, and long-latency ERPs. While short-latency ERPs were not unaffected by sleep stages, they found a significant impact on long evoked potentials, such as the P1-N1-P2 complex. At stage 1, when subjects were not always conscious and failed to respond to external signals, the amplitude of N1 rapidly decreased, falling to near baseline levels, possibly because the attention level decreases. An additional negative peak appeared over central areas of the scalp at approximately 300 and 350 ms post stimulation, also referred to as N350. The amplitude of P2 increased at the onset of sleep. As the subject enters Stage 2 of sleep the amplitude of the N1 response continued to decrease and was sometimes completely absent. The 60

78 amplitude of P2 increased further during the Stage 2 (the transition to deep sleep) (Crowley & Colrain, 2004). These measured CAEPs are largely distinguishable from responses measured when the subject is awake. However, other researchers have reported that CAEPs were not significantly different between fully awake and lightly sleeping or drowsy individuals (Cody et al., 1967; Kushnerenko et al., 2002). Determining sleep stage objectively using EEG is difficult in a clinic. Therefore, standard practice of the P1- N1-P2 requires subjects to remain awake and to be monitored during recording sessions. b. Attention Recording the obligatory P1-N1-P2 response does not require active attention or participation in a listening task. In fact, most studies are conducted with the subject in a passive listening state. However, some studies have shown that actively attending to the auditory stimuli can have a measurable impact on the response morphology. Picton and Hillyard (1974) required subjects to detect and count the signals during the attend condition and to read a book during the ignore condition. They recorded a series of AEPs from early-latency to long-latency responses and measured latencies and amplitudes of the AEPs. They found no significant change in the latencies of any AEPs between the attend and ignore conditions. They observed no amplitude changes in the peripheral EEG recordings but in the attend condition, the amplitude of N1 and P2 components was slightly but significantly increased from 2.5 μv in the ignore condition to 2.8 μv in the attend condition. In pediatric practice it will be challenging to maintain the subject s attention to repeated stimuli during recording hours; therefore, it may be better to not require children to attend to sounds while recording the P1-N1-P2 response. 61

79 c. Hearing status Prior to the advent of ABR recording techniques, CAEPs was used to predict hearing sensitivity. Early studies of the P1-N1-P2 complex found that the presence and amplitude of N1-P2 complex correlate fairly closely with behavioral detection thresholds with approximately 10 db differences in a variety of audiometric configurations (Davis, 1965; Cody et al., 1967; Alberti et al., 1987; Ferraro & Durrant, 1994; Hyde, 1997; Stapells, 2002; Tsu et al., 2002). However, unlike ABRs, CAEPs are typically obtained most reliably when a subject is awake, due to the impact of sleep, arousal, and attention levels on the results. That fact alone contributed to ABR becoming more popular in clinics to estimate hearing sensitivity in infants and children (Hood, 1998). Recently, the P1-N1-P2 complex has been used more commonly in studying the development of central auditory pathway than in hearing assessment (Näätänen & Picton, 1987; Ponton et al., 2002; Wunderlich & Cone-Wesson, 2006; Nash et al., 2008) Effects of development on the P1-N1-P2 complex Normal hearing listeners A number of researchers have found that the obligatory cortical evoked potentials detectible in infants and these responses were the most dramatically different from adults responses (e.g., Notermans, 1991; Pasman et al., 1991; Kushnerenko et al., 2002). Kurtzberg et al. (1984) obtained the obligatory CAEPs in 100% of healthy infants (n=17) and 97% (n= 34 of a total of 35) of infants with very low birth weight. Pasman et al. (1991) showed that the cortical potentials were present in 95% of preterm babies after weeks of conception age. However, the P1-N1-P2 responses of the infants are the most distinct in morphology and latency from those of adults. Purdy et al. (2005) 62

80 compared the P1-N1-P2 responses obtained from 14 adults to those obtained from 20 infants from 3 to 7 months. For this study, four tonebursts (500 Hz., 1kHz, 2kHz, 4kHz) and four speech phonemes (/t/, /k/, /d/, /g/) were used to elicit CAEPs in adults and a subset of these stimuli (500 Hz, 2kHz, /t/, /g/, /m/) were used in infants. Tone bursts were 60 ms in duration. They presented stimuli with ISI of 750 ms in sound field at 65 db SPL. They found substantial differences in results between adults and infants shown as in Figure 13. In adults, three peaks were well defined across all stimuli, occurring at 57, 106, and 198 ms on average for P1, N1, and P2 responses. In infants, on average, a broad P1 was present with an average of 202 ms, followed by a late negativity at 367 ms. They found that 19 out of 20 infants had significantly different CAEP responses in /m/ vs. /g/ and /m/ vs. /t/. Infants also showed post auricular muscle response (PAMR) prior to the P1 response; the PAMR is a biphasic waveform typically occurring at about ms post-stimulus onset (O Beirne & Patuzzi, 1999; Purdy et al., 2005). Figure 13. The P1-N1-P2 complex in adults and infants The left panel displays the grand average waveforms from 14 adults showing three components of the P1-N1-P2 response. The right panel displays results from 20 infants showing one large positive peak. Various stimuli used for recordings are shown in different colors. This figure is taken from Purdy et al., (2005). 63

81 Unlike many studies showing that P1 is the dominant peak in infants and young children, one study recently showed that infants have very adult-like morphology. Small and Werker (2012) measured CAEPs in 6 adults and 31 infants from 16 weeks to 28 weeks. To elicit the onset P1-N1-P2 response they used /da/ with an ISI of 2200 ms. They found that P1 was present in all infants and also found that some infants had N1 and P2 responses as well. Authors explained that a long ISI (> 2 seconds) might lead to a better morphology of the P1-N1-P2 responses in infants. While Small and Werker (2012) attempted to explain their adult-like morphology of the responses obtained from infants using the longer ISIs, it cannot be adequately explained merely by use of longer ISIs (2200 ms used) since other studies (e.g., Gilley et al., 2005; Wunderlich et al., 2006) used similar or much longer ISIs and showed apparent differences in responses between young children and adults. Numerous studies have shown the effects of development on the P1-N1-P2 complex in NH listeners from young children to adults. Figure 14 (Ponton et al., 1996a) illustrates the general trends. They used stimuli consisting of ten 100 µs pulses with a rate of 500 pps (pulse per second). This figure shows P1-N1-P2 waveforms obtained from 14 NH listeners who ranged in age from 6 to 19 years (Ponton et al., 1996a). Also shown is a grand-mean waveform constructed from results obtained from 10 NH adults in their twenties. Solid vertical lines represent the peak latencies for the adult P1, N1, and P2. The dashed line connects peaks identified as P1 for each child. The trend that emerges is a systematic reduction in both P1 latency and amplitude with age. For example, at six years old, P1 latency is approximately 100 ms but by the time the subjects reach their twenties the latency of P1 is approximately 50 ms. The N1 component is prominent in the 64

82 adult waveform but is absent in children under 8 years old. N1 begins to emerge around 8 years but is recorded as a dip or indentation in what is a broader positive P1 potential. N1 becomes more robust over time and the general morphology of the P1-N1-P2 response becomes adult-like around 13 years of age. In a follow-up study, Ponton et al. (2000) used the same stimulus with the stimulation rate of 1.3 /sec at 65 db SL to the left ear via headphone. Results show, once again, that P1 and N1 peaks decrease in latency with age and do not reach asymptotic values until well into adolescence. P1 latencies ranged from 80 to 110 ms in 5-6 year olds and dropped to a range from 30 to 50 ms in year olds. N1 peaks decreased with age from approximately ms in 5-6 year olds to ms in year olds. In this same study, N1 was recorded inconsistently before 9 years of age and reached adult-like values by 5-16 years of age. No consistent age-related trends were apparent for P2 and few studies have attempted to quantify how development impacts the late negativity, also referred to N2, which is recorded at 200 to 250 ms and can dominate the P1-N1-P2 response complex in some young school-age children. 65

83 Figure 14. Maturation of the P1-N1-P2 complex in normal hearing children Grand average waveforms obtained from 10 normal hearing adults are shown in the top trace. Onset P1-N1-P2 responses from 14 children at different age are shown in the age order. Solid vertical lines indicate latencies of each component from the adults mean waveforms. Broken vertical lines indicate P1 in each waveform. This figure is taken from Ponton et al. (1996a). Fewer studies have focused on attempting to characterize the effects of development on the P1-N1-P2 response in infants or pre-school aged children. Both Sussman et al. (2008) and Wunderlich & Cone-Wesson (2006) tested infants and preschool aged children and reported measuring a single vertex positive peak (P1) with a latency of approximately ms. They reported a large negative peak at approximately ms followed the P1. This later negativity has been referred to as N2 or as immature negativity. Figure 15 (from Sussman et al., 2008) shows grand mean P1-N1-P2 waveforms recorded from groups of NH listeners of varying ages (n = 61). In 66

84 this study, the stimulus was pure tones with 50 ms duration at 880 Hz sinusoid gated with a 7.5 ms rise/fall time and presented via insert earphones binaurally at 75dB SPL. Figure 15 shows CAEPs from 8-year-old children to young adults. This figure showed that P1 does not fully mature until about 20 years of age. In fact, P1 latencies recorded from the adolescent group (16 years) were still significantly prolonged relative to the adult group (20-40 years of age). A precursor of N1 is apparent in the waveforms obtained from the 8-year-old group (n=10). N1-P2 amplitudes increase and P1 latencies shorten over this period from 8 years of age to adulthood. Examination of these grand mean waveforms suggests that the trend for P1 latency and amplitude to decrease over time may relate to the development of the N1 component, broadening its peak and increasing its negativity relative to baseline. By age 16 (n=10), the morphology of the P1-N1-P2 complex appears to be very similar to the adult response, with a dominant N1-P2 response, albeit with prolonged P1 latency. Figure 15. Comparisons of the P1-N1-P2 complex among different age groups A total of sixty-one normal hearing subjects was divided into 6 different age groups (8, 9, 10, 11, 16, and adults) This figure shows all recording are collected at the forehead (Fz) position. This figure is taken from Sussman et al. (2008). 67

85 Also illustrated in Figure 15 is the immature negativity that dominates responses from younger children. Here the immature negativity is recorded as a large, slow, negative peak with a latency of approximately 225 ms. In the Sussman et al. (2008) study, the immature negativity referred to it as N2 in the study is clearly evident in the grand mean waveforms recorded in children under 16. By 16 years of age, however, the amplitude of this component is markedly reduced and it is largely absent in waveforms recorded from the adult listeners. The developmental trends are relatively robust and independent of stimulus type. For example, Ponton and his colleagues (Ponton et al., 1996a,b, 2000) used trains of 10 clicks, each of which was 100 µs in duration. Sussman et al. (2008) used gated pure tones of 880 Hz also presented at a rate of 1.25 stimulus/sec. Sharma et al. (1997) used a short duration (90 ms) of synthesized consonant-vowel syllable /ba/, and presented at a rate of 1.6 stimulus/sec. Similarly, Cunningham et al. (2000) used a short duration (100 ms) of syllable /ga/. These previous studies used different stimuli but they all have relatively short duration and presented every 1-2 second. Findings on the developmental trends on the onset P1-N1-P2 response are similar among these studies. Some example studies are described in the below. Sharma et al. (1997) tested 96 NH listeners (86 school-aged children ranging from 6 to15 years and ten young adults ranging from 21 to 27 years). In adults, P1 was small in amplitude and was recorded with a mean latency of 59 ms (SD = 12). N1 was recorded with an average latency of 109 ms (SD = 9). In children, P1 was the dominant component of the response and had an average latency of 87 ms (SD = 14). At 6 years of age, P1 was followed by a broad negativity with an average latency of 221 ms (SD = 12). Sharma et 68

86 al. (1997) reported that the latency of P1 evoked using a speech stimulus decreased with age from 6 years to years. This is consistent with results reported by Ponton et al. (1996a, 2000). Using the same short speech stimulus /ba/ and recording procedures, Sharma and colleagues (2002a) revisited the issue of how maturation affects cortical auditory evoked potentials. Fifty-one normal hearing children from 0.1 to 20 years of age participated. The researchers combined those results with their previous data and reported P1 latencies obtained from a total of 136 children and young adults with normal hearing. Their data were best fit by a function based on the natural log of age from 0 to 20 years (R 2 =0.78, p < ). The function was P1 latency = ( Age). The 95% confidence interval derived from this data set was approximately 100 ms, suggesting substantial inter-subject variability. Results show a relatively rapid decrease in latency during the first decade of life, especially by 4 years of age. This is related to the period of synaptic overshoot in cortical layers that occurs dramatically during the first 4 years of life. A gradual decrease in latency during the second decade of life may relate to the continuous changes in synaptic pruning and myelination. The immature late negativity following the P1 in young children is also observed in Sharma et al. (2002a,b). However, researchers analyzed the data based on P1 latency only. Cunningham et al. (2000) also used speech stimuli to explore the effects of maturation on the P1-N1-P2 complex in 150 NH children and adults. One hundred thirty children ranging from 5 to 15 years and 10 young adults in twenties, and 10 senior adults between ages of 55 and 78 years participated. All participants had normal hearing sensitivity (< 25 db HL) from 500 to 4000 Hz at the time of testing. In this study, the 69

87 stimulus was a synthesized consonant-vowel (CV) syllable (/ga/) that was truncated to 100 ms. The stimulus rate was 1.7 per second. P1 was present in all groups and both P1 latency and amplitude decreased from a mean latency of 90 ms and mean amplitude of 2 µv at 5 years of age to a mean latency of 68 ms with mean amplitude of 1.3 µv at 78 years of age. P1 latency and amplitudes, however, were similar when results obtained from young adults (19 to 27 years) were compared with measures obtained from older adults (55 to 78 years) indicating that the effects of development on the P1 peak were mostly complete by the second decade of life. When P1 latency was compared between sub-groups of participants 5 to 15 years, they found the P1 latency of the 13 to 15 years olds to be shorter than P1 latencies of the 5-7 or 8-10 year old age groups. P1 is present in all NH listeners across all age groups. In young children, N1 is often absent but in many cases immature negativity is recorded. With normal development, the N1 peak initially emerges by splitting the large and broader P1 peak. Over time, the N1-P2 peaks become larger and more prominent, and P1 becomes smaller in amplitude but shorter in latency. The immature negativity, that is clearly visible in responses from younger children, shrinks in amplitude and disappears altogether around the time when the N1 component becomes adult-like (e.g years of age) The hearing aid and cochlear implant users a. Hearing aid users A few studies have been published where CAEPs have been recorded from pediatric HA users. In each case, the goal was to establish the effectiveness of the intervention strategy. In 2005, Sharma et al. published case studies from three hearing impaired children 70

88 where she measured CAEPs repeatedly with the goal of assessing efficacy of the intervention strategies used. P1 latencies obtained from HA children were plotted against the 95% confidence intervals for P1 latency values obtained from typically developing children (reported in Sharma et al., 2002b). The first child had a severe SNHL and was fitted with a HA at a young age. For this child, P1 latency changed rapidly and was within the normal range after five months of HA use. They interpreted this to mean that the HA provided a sufficient amount of stimulation for normal development of central auditory pathways to occur. The second child used a HA for three months before receiving a CI. For this child, P1 latency did not change during the HA trail period but decreased rapidly after he received his CI. In this case, lack of a change in P1 latency was interpreted as evidence that the signal provided by the hearing aid was insufficient to drive development. The third child was diagnosed with auditory dys-synchrony. He showed no P1 latency changes during the 10-month HA trial but within three months of receiving a CI his P1 latency was found to be within the normal range for his age. Based in part on these results, Sharma et al. (2005) suggest that P1 latency can serve as an objective evaluation tool for the decision of whether HA amplification is sufficient for a child to provide for regular development of central auditory pathways or not. Two years later, Dorman et al. (2007) published results from two additional subjects, both of whom received a CI before 3 years of age. One child received a CI at nine months. His P1 latency decreased by 200 ms over the following four months and was within the normal range for NH peers. Additionally, a speech and language evaluation indicated that the child was making rapid progress in his speech and language development. However, results from the second child were not as encouraging. This child 71

89 was diagnosed with Goldenhar s syndrome and was born with bilateral severe-toprofound hearing loss. He had a HA when he was 9 months old. Subsequent testing in the sound field revealed no aided benefit and P1 responses were not elicited at maximum output levels after 17 months of hearing aid use. He received a CI at 2 years 7 months. Post-implantation, a P1 peak was recorded but the latency was prolonged beyond the normal range even after 13 months of CI usage. The authors cited audiologists observation that the child wore the CI inconsistently and did not respond to sounds behaviorally. For this child, electrical stimulation at a young age did not lead to typical development of the central auditory system. Jang et al. (2010) measured P1 latency using a short /ba/ stimulus (a 90 ms speech sound /ba/, reportedly the stimulus provided by Sharma and colleagues) in a group of children who were candidates for CI but were still using HAs along with a group of pediatric CI users. The 13 HA users ranged in age from 2 to 12 years; three children did not have measurable P1 components. The P1 latencies from the other ten children in the HA group ranged from ms. These P1 latency values were compared to those obtained from 53 NH children from 1.7 to 17.5 years of age. In NH children the P1 latency decreased with increasing age showing a significant negative correlation (r = , P <0.001). The P1 latencies measured from 13 HA users were longer than the 95% confidence intervals for NH children. The 10 CI users ranged in age from 3 to 15 years. All had measurable P1 responses with peak latencies that ranged from ms. The P1 latencies from five of ten post-ci children were within the 95% confidence interval at their chronological age. The authors interpreted the delayed and absent P1 responses observed in the HA group to be an indication that the gain provided by the HA was 72

90 insufficient to support normal central auditory system development. All three of these studies are based on results from a limited number of study participants; however, they seem to provide evidence that CAEP testing could potentially play a role in the clinical management of children with severe-to-profound SNHL. Absent or abnormally prolonged P1 latencies appear to be common in children with severe-to-profound hearing loss and provision of adequate auditory input either through a well fit HA or a CI appears to drive development of the central auditory pathways as evidenced by changes in P1 latency. b. Cochlear implant users How auditory deprivation affects the developing auditory system in humans is a question that has been of interest to researchers for decades. The approach most commonly used to address that question has been to compare P1 latencies measured from children who were born deaf but received a CI later in life to P1 latencies measured in typically developing children (Ponton et al., 1996 a,b, 2000; Eggermont et al., 1997; Sharma et al., 1997, 2002a, b). Ponton and colleagues (1996a) recorded CAEPs from 8 NH adults and 31 NH children (6 to 19 years old) and compared those results with similar measures obtained from 6 adult CI users and from 12 pediatric CI users. The authors did not report the chronological age information for CI users but instead described these subjects based on the duration of auditory deprivation they had experienced prior to receiving a CI. For the pediatric CI group, the average age at which deafness was detected was 1 year 3 months (range 0-5 years) and the mean interval between detection of hearing loss and implantation was 4 years 5 months (range years). For the CI adult group, 73

91 researchers reported that five out of six CI adult users had hearing loss in adulthood, i.e., they were post-lingually deafened; however, the duration of hearing loss or CI use were not reported. The stimulus used to test the CI recipients was an electrical pulse train consisting of 10 biphasic current pulses that were 200 µs/phase with each pulse 2-ms apart within the train. For NH listeners, an acoustic stimulus consisting of 10 clicks with 100 µs duration was presented monaurally via headphones. For both CI and NH groups, the stimulus was presented at a rate of 1.3 per second. They found that in both groups P1 latency appeared to decrease exponentially as a function of age. They divided the CI children into 3 groups based on the duration of deafness prior to implantation: short deprivation (mean 1.1 years of deprivation), medium deprivation (mean 4.9 years of deprivation), and long deprivation (over 8 years of deprivation). They noted when P1 latencies reached adult norms for individuals in each of these groups. They found that for NH children, P1 latency reached mature values by 15 years of age. Using a log-linear function they calculated that P1 latencies would not reach adult values until approximately 17, 20, and 25.5 years of age for the implanted children with short, medium, or extended periods of deprivation, respectively. This study concludes that P1 does not mature without stimulation and that once stimulation is restored, the typical rate of maturation for the cortical activity resumes even after an extended period of sensory deprivation. However, this conclusion was based data from a very small number of CI users: there were only two children in the long deprivation group. In 2002, Sharma and colleagues measured CAEPs from 22 children between 1 and 5 years of age (mean = 3 years) at the time of testing. For these children, the initial fitting for their CI occurred at an average at 2.6 years of age. They were considered to be 74

92 early-implanted children. They were tested at one of the four different times: 1 week, 2 months, 5 months, or 8 months post initial stimulation. Results of this cross-sectional study showed that P1 latency decreases rapidly following implantation and reaches ageappropriate values within eight months after implantation. The authors conclude that early-implanted pre-lingually deafened children have plasticity in their central auditory pathways that allows for following typical development of P1 latency when using electrical stimulation. In a second study, Sharma et al. (2002b) studied a critical period for the development of the central auditory system. They compared P1 waveforms from 121 CI users (2-18 years) to P1 waveforms recorded from 136 normal hearing listeners (less than 1 to 20 years). In this study, they divided the CI users into three groups: those fitted with a CI at an early (<3.5 years), middle (between 3.5 and 7 years), or late age (>7 years). They report P1 latency values as a function of age at the time of testing. They showed that if children received CIs before 3.5 years old, P1 latencies were often recorded within the range of latencies exhibited by NH listeners. In contrast, P1 latencies obtained in children implanted later (>7 years at the time of implant) were significantly longer than those obtained in age-matched controls. Sharma et al. (2002b) concluded that a critical period of about 3.5 years exists and suggested considering implantation for children with aided P1 latency not in the typical range for the child s age group even after a HA trial. For bilateral CI children (Sharma et al., 2005), when a child received CIs sequentially, the ear with a CI implanted after 7 years of life showed responses similar to those of the late-implanted children; however, the other ear with a CI implanted before 3.5 years had responses changed its morphology quickly and become similar to those of 75

93 NH children. In 2006, Sharma and Dorman described the developmental morphological patterns in the P1-N1-P2 complex for 37 NH children and 41 early-implanted CI children (implanted before 3.5 years) aged from 0 to 12 years divided into groups of 0-3 years, 3-6 years, 6-9 years, and 9-12 years. In NH children, P1 developed first at age between 0 and 3 years old. A hint of emerging N1 showed by 6 years and was consistently measured by 9 years. N1 amplitudes become larger and P1 amplitudes become smaller between 9 and 12 years. CAEPs measured in early-implanted CI children paralleled those observed in NH children. A large P1 was present between 0 and 3 years and N1 began emerging at 6 years. This finding agrees with observations made by other investigators that children implanted before 3.5 years show emergence and the maturational changes of the N1 component similar to those in NH children (Eggermont & Ponton, 2002). However, children implanted after 3 to 7 years of auditory deprivation show variable results in the presence and latency of N1 peaks Children implanted after age 6-7 years do not have ageappropriate N1 responses even measured at the hearing age of 10 and 15 years (Eggermont & Ponton, 2003; Gilley et al., 2005). Dorman et al. (2007) reviewed P1 data from 245 pediatric CI users who participated in the previous studies and concluded that 3.5 years appeared to be a good estimate of the age where the auditory system was still plastic. The Dorman et al. (2007) study showed P1 latencies from 246 congenitally deaf children who used a CI for at least six months. The P1 latencies of the 245 CI children were plotted against the 95% confidence intervals of P1 latencies measured from 190 NH children. About one-half of the children who experienced shorter periods of deprivation (between 3.5 and 7 years) 76

94 had normal P1 latencies and almost every child who experienced fewer than 3.5 years of deprivation had normal P1 latencies after CI use. These results further suggest again that if a child receives implantation before 4 years old then this child will reach the typical development of Pl latency. This may be due to the neuroplasticity in the central auditory system especially during the early developmental stages. Even moderately prolonged periods of deafness (> 3.5 years) may reduce the chances for normal development following implantation. Children who have experienced more than 7 years of auditory deprivation are not likely to develop regular CAEPs. The effects of development on onset P1-N1-P2 responses have been described and measured in both NH and CI listeners. The onset P1-N1-P2 complex helps clinicians and researchers to infer how the central auditory system develops with rehabilitation using CIs after auditory deprivation Cortical Auditory Evoked Potentials to Assess Discrimination Traditionally, the MMN and P300 CAEPs were obtained using an oddball paradigm to study the discrimination capacity. More recently there has been an attention to the ACC because it has relatively large amplitudes compared to CAEPs using oddball paradigms, and it does not require participation by the user. In the next two sections, MMN, P300, and ACC techniques are reviewed Mismatch negativity and P300 Traditionally, there are two different auditory evoked potentials (AEPs) that have been used to assess discrimination between two acoustic stimuli (Kraus & Cheour, 2000). These are the auditory MMN and P300 (or P3) response. Sutton et al. (1965) first these responses in 1965 followed by Näätänen et al. in Both AEPs are long-latency 77

95 evoked potentials that are elicited using an oddball stimulation paradigm. In the oddball paradigm, one stimulus often labeled the oddball or deviant stimulus is presented infrequently and is embedded in a train of frequently occurring standard stimuli, for example, ba, ba, ba, ba, ba, da, ba, ba, ba, da, ba, ba... Surface electrodes are used to record the neural activity associated with the audition while responses evoked by the frequent and infrequent stimuli are recorded into separate buffers. The MMN is a negative component which is computed by subtracting waveforms recorded following presentation of the frequent or standard stimulus (in this example, /ba/) from waveforms obtained in response to presentation of the deviant stimuli (in this example, /da/) (for a review, see Näätänen, 2007). The MMN is usually measured at the frontocentral electrodes relative to a mastoid or nose electrode; its peak is recorded between 150 and 250 ms after a change occurs. The MMN latency decreases with increasing magnitude of stimulus change (Sams et al., 1985). The MMN is a neural response that indicates when a change within a train of similar sounds has been processed at a pre-attentive level (Näätänen et al., 2007). Studies show a relationship between presence/absence, and amplitude of this response and behavioral discrimination capabilities, such as frequency discrimination (Lang et al., 1990), and syllable distinction (Sharma & Dorman, 1999). This has been found to be true across different languages (Kuhl, 2004). Recently, MMN has been obtained from CI users to study pitch or timbre perception (Zhang et al., 2013). However, MMN has a critical limitation to being used as a clinical auditory function evaluation: it is not reliably recorded, even in NH children and adults using the easily discriminable stimuli (Kraus et al., 1993; Picton et al., 2000; Pettigrew et al., 2004; Sharma et al., 2004). This may be due to the subtracting waveform method used to obtain 78

96 MMNs, which results in small amplitudes, often unreliably recorded in individuals (see Bishop, 2007; Näätänen et al., 2007 for reviews). MMNs are not reliably measurable at an individual level; therefore, studies have used group data rather than individual data to correlate with behavioral measures (for a review, see Bishop, 2007). It is even more challenging to obtain individual MMNs in CI users and young children. Zhang et al. (2013) reported that MMN was present in only half of CI participants (Zhang et al., 2013). The small amplitudes of MMN also can be disrupted with very little background noise, which is problematic for assessing discrimination capacity in noise (Bishop, 2007). These observations have garnered limited enthusiasm for the use of this evoked potential in clinical studies assessing underlying auditory discrimination in young children, especially those with CIs. The P300 is a positive component of the long-latency cortically evoked potential. It occurs at approximately 300 ms post presentation of a deviant stimulus (Linden, 2005). The P300 response is typically recorded also using an oddball paradigm. While the MMN can be evoked using passive listening paradigms, the P300 is best elicited when the subject is engaged in an active discrimination task. For example, the subject is required to participate actively in the study counting or pressing a button in response to an infrequent deviant stimulus. Therefore, the P300 is an endogenous potential whose response depends greatly on the subject s the level of attention, arousal, and processing capacity of the task (Linden, 2005; Polich, 2007). The presence of P300 has been interpreted as an indication that a change in an acoustic stimulus has been consciously noted. P300 amplitudes are larger and latencies are shorter for easy discrimination tasks. Amplitudes decrease and latencies increase when the task becomes more difficult. P300 has moderate 79

97 to strong test-retest correlation coefficients for oddball tasks, these range from 0.50 to 0.80 for amplitude and from 0.40 to 0.70 for peak latency (Polich, 2007). However, the fact that recording this response requires active participation by the listener limits its usefulness as a tool for studying auditory development or discrimination skills in very young children The ACC An alternative to both the MMN and the P300 response is the ACC. Recording the ACC does not require the use of an oddball stimulation procedure nor does it require active participation in a listening task. It also has larger amplitudes than the MMN and requires less time to record. These features have led to a resurgence of interest in this method of assessing discrimination capacity using electrophysiological techniques. The ACC is a variant of the P1-N1-P2 complex, but rather than being recorded at the onset of an acoustic stimulus, it follows the introduction of a change in some aspect of the ongoing stimulus. While not called an ACC, this neural potential was described in studies as early at the 1970 s and 1980 s (e.g., Arlinger et al., 1976; Kohn et al., 1978; Maiste & Picton, 1989). These studies showed that it was possible to record an N1-P2 response, not only at the onset of an acoustic signal, but also after the introduction of a change in some aspect of that signal. For example, Arlinger et al. (1976) reported measuring N1-P2 responses to ramp changes in frequency in a continuous pure tone of 1 khz. They varied frequencies from 10 Hz to 500 Hz in ongoing sounds. They showed that the response was related to the magnitude of frequency change within the ongoing stimulus. When the P1-N1-P2 complex is evoked in this manner, it has been referred to as the ACC (Martin & Boothroyd, 1999). ACC is elicited by changes in periodicity, 80

98 frequency, or intensity of ongoing sounds; therefore, it can be a non-invasive and objective index of neural coding of auditory discrimination. A series of previous studies show that ACC can reliably recorded from NH and CI listeners. The ACC is also used interchangeably with the change response in this paper since the presence of these responses indicates neural detection of a change in a stimulus ACCs in NH listeners Ostroff et al. (1998) recorded CAEPs from 8 NH adults using the naturally spoken syllable say /sei/ with the sibilant /s/ and the vowel /ei/, which were extracted from the recording of the syllable. Figure 16 (Ostroff et al., 1998) shows that both the isolated sounds elicited the onset N1-P2 responses but they note that the change from /s/ to /ei/ in the syllable elicited a second N1-P2 response following the onset response to the sibilant /s/. Ostroff et al. (1998) suggested that this change in the N1-P2 response be a measure of neural activation resulting from changes in the amplitude and/or spectrum of the acoustic signal as it transitions from the consonant to the vowel and/or from aperiodic to periodic stimulation. Ostroff et al. (1998) interpreted the presence or absence of the N1-P2 acoustic change complex as evidence of the discrimination capacity for a change from a fricative to a vowel. However, based on results of this study, it was unclear if a distinct change in the stimulus or all features (e.g. intensity, periodicity, and/or frequency of the acoustic signal) were responsible for generating this second cortical evoked potential. Therefore, Martin and Boothroyd (1999) wanted to study the change response more systematically. They consistently used acoustic change complex terminology consistently for these change responses. 81

99 Figure 16. The P1-N1-P2 and ACC responses elicited by [sei] speech sounds The three panels shows grand average waveforms obtained using [s], [ei], and [sei], respectively, from eight normal hearing adults. The red arrow identifies the onset or the change response elicited by [s], [ei], or [sei]. This figure is taken from Ostroff et al. (1998). Martin and Boothroyd (1999) elicited the ACC using a tonal complex and a narrow-band (600 and 1400 Hz) noise. The spectral envelope and the RMS intensity of both the tone and noise bursts were matched. Both stimuli were 390 ms in duration. The difference between these two stimuli was the spectral detail. For example, the noise stimulus had a continuous spectrum whereas the periodic tonal stimulus had a line spectrum. The two stimuli were concatenated by overlapping the two signals for 10 ms creating a single stimulus that was 790 ms in duration and changed from a tonal complex to a noise at the acoustic midpoint. These stimuli were presented to 10 NH adults via headphones at 80 db SPL. While there was no evidence of the ACC at the midpoint of the response to tone-only or noise-only control conditions (Figure 17), the authors were 82

100 able to obtain the ACC at the midpoint where the tone changed to noise, and where noise changed to the tone stimulus. Martin and Boothroyd (1999) also used the same stimuli presented in an oddball paradigm to record the MMN from the same group of NH listeners. They noted that the ACC was, on average, about 2.5 times as large as the mismatch negativity response. Martin and Boothroyd (1999) concluded that while both MMN and ACC responses can be used as an index of changes in periodicity, because ACC has much larger amplitudes than the MMN, the ACC is more clinically useful. Figure 17. The P1-N1-P2 and ACC responses elicited by a spectral change Solid lines with a black arrow identifies when the onset, change, and offset of sounds occur. The red arrow identifies the acoustic change complex elicited by a change from tone to noise or vice versa. This figure is taken from Martin and Boothroyd (1999). Martin and Boothroyd (2000) published a follow-up study where they extended results to include synthetic vowels (as opposed to natural speech segments used in Ostroff et al., 1998). These investigators recorded the ACC response using a relatively long-duration synthetic vowel stimulus (/u/ and /i/) that included a change in level and/or 83

101 formant frequency to record both cortical onset and change responses. The stimulus consisted of two 400-ms segments of the synthesized vowels /u/ and /i/. The only difference between these two vowel segments was the F2 frequency (i.e. 900 Hz for /u/ and 2400 Hz for /i/). The two segments were combined sequentially to create a set of stimuli that were 800-ms in duration. In some cases, the stimuli included a simple change in F2 frequency labeled /ui/. In other cases, the F2 frequency was consistent (labeled /uu/) but the amplitude of the F2 frequency varied. The level of the second half of the 800 ms stimulus was varied from -5 db to +5 db in 1 db steps (i.e., 11 amplitude change conditions) relative to the first half of the acoustic stimulus that was fixed at a 70 db SPL level. These stimuli were presented from a loudspeaker at 0 degrees azimuth at a rate of approximate 0.33 Hz. They reported that the onset P1-N1-P2 responses were obtained in all conditions. Results also showed that for amplitude change alone, the ACC was measurable for intensity changes of 2 3 db or more; the ACC was elicited by amplitude changes as small as +2 db and -3dB RMS even in the absence of a change in the F2 frequency. Researchers compared these CAEP results to psychophysical discrimination thresholds of similar stimuli, which ranged from 0.2 to 0.5 db (Gelfand, 1990). The ACC was also recorded in response to a shift in the frequency of the second formant even when the RMS level was held constant indicating that the formant change alone can elicit the ACC. However, the largest ACC responses were recorded using stimuli that included both amplitude and spectral changes. Moreover, amplitude increments enhanced the effects of spectral change, whereas amplitude decrements had little to no effect. This research is one of the most critical studies indicating a potential use of the ACC to evaluate spectral resolution in the central auditory systems. The neural response at 84

102 thalamocortical levels may be the index of something closer to discrimination of intensity and spectrum than a simple detection of sound. Tremblay et al. (2003) used naturally spoken speech sounds such as /bi/, /pi/, /shi/, and /si/ to evoke ACC in 7 native English NH speakers They reported good testretest reliability (range 0.60 to 0.94) for individual subjects and found that different speech tokens, representing different acoustic cues, elicited distinct patterns of neural responses. In 2010, Martin et al., recognizing that very little ACC data had been collected from children, used their synthesized /u/-/i/ stimuli to record the ACC in a group of nine NH and typically developing children who ranged in age from 6 to 9 years. The stimuli were presented in the sound field at 75 db SPL. CAEPs were obtained from a total of nine participants. Responses evoked by the onset of the acoustic stimulus following a change in formant frequency consisted of a single, large amplitude P1 component. Figure 18 shows grand mean waveforms for CAEPs obtained from 9 children and 10 adults using the /u-i/ stimulus with 2 second ISI (Martin et al., 2010). Differences between adults and children were significant in amplitude and morphology of the onset P1-N1-P2 response and the ACC. The amplitude of the ACC was significantly greater for children than for adults. The P1 component alone followed by an immature negativity, N2, was prominent in both the onset and the ACC responses. They stated that there were subtle, but apparent developmental changes in ACC responses of children age from 6 to 9 years old. Unfortunately, this study did not display any individual data of children for either latencies or amplitudes; therefore, it is not possible to compare maturational effects on ACCs with maturational effects on the onset responses reported in previous literature. It 85

103 can be noted that N1 starts to emerge in children around 7-8 years old in onset responses (e.g., Ponton et al., 1996a). However, N1 in ACCs is not present in children (aged between 6 to 9 years) in this study. Figure 18. Comparison of grand mean waveforms between adults and children The onset P1-N1-P2 and the acoustic change complex were obtained using an /ui/ stimulus with 2 second inter-stimulus intervals in normal hearing adults (n=10) and children (n=9) The acoustic change complex is indicated by the gray inset boxes. This figure is taken from Martin et al. (2010) ACCs in HA and CI users Several researchers have recorded ACC responses from CI users. In 2006, Friesen and Tremblay reported ACC measures obtained from eight adult Nucleus CI users who were tested using the same natural speech tokens used by Tremblay et al. (2003). Subjects were tested in a sound field using their speech processors at their own volume and sensitivity settings. The speech tokens /si/ and / i/ were presented at 63dB pespl in the sound field. The stimuli had durations of approximately 756 ms for /si/ and 655 ms for / i/ with the vowel /i/ started at 250 ms following /s/ and 220 ms following / /. The grand mean CAEP waveforms evoked by /si/ were compared to those evoked by / i/. While waveforms evoked by /si/ and / i/ are not different in amplitude, the latencies of the onset components to the consonant (P1 latency 123 ms for / i/) are later for the CI users than for the NH adults (P1 latency approximately 80 ms for / i/). The authors 86

104 attribute these latency differences to delays in processing introduced by the speech processor of the CI. To compare waveforms based on performance, authors analyzed the data from classifying participants into two groups based on performance. They used /si/ and / i/ identification scores extracted from the Nonsense Syllable test and called those CI users who achieved identification scores of 0-20 % poor perceivers and CI users who obtained identification scores of 80% and better better perceivers. The authors observed different response characteristics in the two groups; poor onset responses (reduction in amplitudes) for /si/ stimulus and poor change responses (reduction in amplitudes) for / i/ stimulus for poorer performers compared to good perceivers. Unfortunately, this study did not report the number of CI participants for each subgroup. However, this research is important in showing that the ACCs were reliably evoked by natural speech sounds in individual CI users indicating that different speech sounds elicit different patterns of CAEPs and that the ACC amplitudes may be related to perception. One weak point of this study may be that the use of syllables, such as /si/ and / i/, can lead to overlap responses between the onset P1-N1-P2 and the ACC; therefore, interpretation of the ACC in these waveforms can be complicated. Other researchers also succeeded in recording the ACC from individual CI users using a more systematic stimulus set. For example, Martin (2007) created a series of the synthesized vowel stimulus /u/ (F0 =150 Hz, F1 = 300 Hz, F2 = 1050 Hz, and F3 = 3000 Hz) each of which were 400 ms long. Martin also created an additional set of 9 stimuli where they varied the F2 frequency from 1050 Hz (0 Hz change, control condition) to 2250 Hz (120 Hz change, where the native speaker perceived the sound as /ui/). One 87

105 young MedEl CI user, with a diagnosis of auditory dys-synchrony, participated in this study. The subject was tested in two listening conditions: ignore condition and attend condition. In the ignore condition, the subject watched a captioned video. In the attend condition, the subject had to decide whether she heard a change at the midpoint of the stimulus or not, and she pressed one button or the other based on her decision. Results show that the ACC was obtained in both the ignore and attend conditions. In the attend condition, the ACC was elicited by the F2 change of 150 Hz to 1200 Hz, and the amplitudes of the ACC were larger than those measured in the ignore condition. In the ignore condition, the ACC was elicited by the F2 change of 300 Hz to 1200 Hz. When the F2 change of 150 Hz was used, the ACC was not clear. The behavioral discrimination performance (% correct) as a function of F2 change was correlated with the ACC recordings. Perception scores were 27% correct for the 150 Hz F2 change, and improved to 44% correct for the 300 Hz F2 change, 90% correct for the 600 Hz F2 change, and up to 92% correct for the 1200 Hz F2 change. This agreement between the ACC and perception measures suggests that the ACC threshold may be useful as an estimation measure of the discrimination performance of CI users. In 2008, Brown et al. measured electrically evoked ACC responses from postlingually deafened adult Nucleus CI users. Instead of stimulating in the sound field, Nucleus Implant Communicator (NIC) subroutines were used to stimulate the implanted receiver/stimulator directly. The stimulus consisted of a 600 ms train of biphasic current pulses. The individual biphasic current pulses in the 600 ms pulse train were 25 µsec/phase, contained an interphase gap of 8 µsec, and were presented at a rate of 1000 pps. Each 600 ms stimulus burst was presented at a relatively slow rate: one every

106 seconds. The stimulating electrode changed midway through the stimulus. During the first 300 ms of stimulation, the electrical pulse train was applied to electrode 10. The second 300 ms stimulus burst was applied to a different intracochlear electrode. A series of evoked potentials, including both the onset and change responses, were recorded as the spatial separation of the two stimulating electrodes was systematically varied. Increasing the spatial separation between the two stimulating electrodes resulted in an increase in amplitude of the EACC; for example, researchers observed the smallest amplitude ACC responses when the stimulating electrode changed from electrode 10 to adjacent electrode 11. Larger response amplitudes were recorded when the stimulating electrode was switched from electrode 10 to electrode 15. The authors conclude the CAEPs can be supplemented with peripheral measures to better understand the function of the auditory systems of both NH and CI listeners. In 2009, Kim et al. extended the study of Brown et al. (2008) by eliciting the EACCs using the stimulation level change on one electrode. They also explored the relationship between the slope of growth function of the onset P1-N1-P2 response and the amplitudes of EACCs. They hypothesized that the slope of the onset response will be rapid for the CI user who has more surviving neurons than the CI user who does not; therefore, the amplitudes of EACC will be greater for the CI user with a rapid onset growth function. The CAEPs were obtained in 12 Nucleus CI participants. Both T- and C-levels were measured behaviorally for each subject to identify the current level corresponding to 0 % (equal to T-level), 25%, 50%, 75%, and 100% (equal to C-level) of the dynamic range (DR). To elicit CAEPs, the pulse train, with a total duration of 800 ms, was initially presented to the internal device at 50% of DR for the first 400 ms and then 89

107 systematically increased (50% to 75% of DR) or decreased (50% to 25 %) in current level for the next 400 ms. Results show that while the amplitudes of EACC responses varied across subjects and level changes, the EACC was reliably recorded from all participants. The amplitudes of EACCs were larger with a greater level change. More interestingly, they found that as the current level increases (for example, from 50% to 75% of DR) the result is more robust and larger EACC amplitudes than when the current level decreases (for example, from 50% to 25% of DR). This is similar to results obtained from NH listeners, where an increase level of 2 db or a decrease level of 3 db elicited the ACC in the ongoing steady state speech sound (Martin & Boothroyd, 2000). The author s hypothesis was found to be correct; a positive moderate correlation existed (r = 0.61) between the slope of the onset response (the N1-P2 amplitudes) and the EACC (the N1- P2 amplitudes). In other words, a CI user, with a steeper growth function of onset response (with more neurons recruited with a level change), has more robust EACCs. This study led to the conclusion that the ACC might serve as an objective tool to supplement some psychophysical measures of intensity discrimination of CI users. Recently, He et al. (2013) measured EACC in auditory neuropathy spectrum disorder (ANSD) children. He et al. (2013), using direct input, presented a biphasic pulse train for 800 ms to electrode 12 and introduced a gap (5, 10, 20, 50, and 100 ms) at the midpoint, i.e., at 400 ms. The ISI was 1200 ms. A total of 15 children with ANSD, from 5 to 17, participated. He et al. (2013) also measured a behavioral speech perception ability using the PBK (the phonetically balanced kindergarten) list, presented via a loudspeaker using live voice, to all subjects. Results show that the onset and EACC responses were recorded in all subjects. Authors reported a relationship between EACC 90

108 thresholds and the PBK word scores, showing that 5 subjects whose EACC were 20 ms or longer achieved lower than 70 % correct, while 5 subjects whose EACC were 10 ms or shorter achieved higher than 70% correct in the PBK word scores. In the following study, He et al. (2014) measured EACC in 15 CI children (between 5 and 17 years at the time of testing) who were diagnosed with ANSD, using the same method and stimuli. This time they presented the same biphasic pulses presented to the electrode 12 for 400 mses but then changed the second electrode, varying from electrode 13 to electrode 22. Speech perception was measured using the PBK word lists. The behavioral electrode-discrimination test was also administered, presenting two bursts of pulse trains to each electrode pair (e.g., electrode 12 vs. electrode 13). Subjects were required to inform the examiner whether they heard one sound or two sounds. Results show that EACC was recorded reliably in all subjects. And, the amplitude of EACC grew as the two electrodes were further separated. They reported that the PBK word scores are correlated with the behavioral electrode-discrimination thresholds (r = ), with good performers having significantly larger EACC than poor performers (p < 0.05). While these two studies (He et al., 2014) showed a broad range of variance in response morphologies of EACC in CI children with ANSD, authors did not discuss the changes in the EACC with age in this population. Martinez et al. (2013) recorded ACCs in children with NH (n=5) and bilateral SNHL (n=5) aged from 2 to 6 years. They used three vowels and created two vowel contrasts: /u/-/a/ and /u/-/i/. These vowels are presented at 65 dba via a loudspeaker. The authors did not use any ISI between stimuli. They used a trigger to start recording 100 ms prior to the change occur. In this way, we obtained the ACC waveforms only rather than 91

109 obtaining both the P1-N1-P2 and the ACC. The authors assume that this method is efficient to record the ACC in young children for whose attention span is really short. The authors obtained the ACC, characterized by a large, broad P1 followed by negativity, was recorded in both NH and HA children for both vowel contrasts. However, the ACC was absent in one (27-month-old) HA child. The authors assume that since the HA child had mild-moderate hearing loss the sounds were heard by the child. They mention that the absent ACC might be related to the young age of the child or the lack of examiners skills in testing CAEPs of young children. These preliminary data, while limited, show that ACC can be recorded from children. Further, it may serve as a tool for estimating speech discrimination ability, particularly for infants and young children for whom behavioral tests are not available. Unfortunately, the effect of development on this potential is not well studied in either NH children or children with hearing loss and CI Summary and Limitations of the Literature A review of the literature in Chapter 1 summarizes that we can use electrophysiological techniques to investigate how either acoustic or electric stimuli are encoded at a cortex level. While the onset P1-N1-P2 responses only indicate the detection of sound in higher auditory systems, the ACC responses can show ability discriminate any acoustic change in ongoing, long-duration, spectrally complex acoustic sounds. Both the onset and change of CAEP responses can be obtained objectively and recorded reliably from adults as well as children with any hearing status: NH, HA, and CI listeners. Finally, identical stimulating and recording systems can be used to obtain these CAEPs from listeners regardless of ages or hearing status. 92

110 Due to these advantages, there has been renewed interest in the onset P1-N1-P2 and ACC responses. Recently, using onset responses, researchers have studied typical development of the central auditory system and investigated the impact of auditory deprivation on the development of children with congenital hearing loss. The previous studies show that even if a child is born deaf, if they receive a CI at a young age (< 3.5 years), then the development of the onset P1-N1-P2 responses will be similar to those from children with NH, suggesting a plastic central auditory system at those ages. The ACC can be used as an objective electrophysiologic test of sound discrimination capacity, which is crucial for perceiving speech and music. While the impacts of development, auditory deprivation, and (re)habilitation on the onset P1-N1-P2 complex are fairly well understood, the effects of those on the ACC have not been studied. In fact, little is known about how development impacts the ACC response regardless of whether it is evoked using acoustic stimulation in NH or HA children or via electrical stimulation from pediatric CI users. For clinical applications, knowing the relationship between behavioral performance and ACC changes with development can be used to facilitate the management of pediatric CI recipients. These findings will undoubtedly help clinicians to program speech processors for young children who cannot give feedback, to understand the impact of CIs on the child s central auditory system development, and to counsel the children s family. 93

111 CHAPTER 3 METHODS 3.1. Study Participants Subjects were recruited from two listening groups: NH and CI. For NH listeners, ninety-one children and adolescents between the ages of 3 and 19 and eleven adults between the ages of 20 and 40 participated (mean =28 years old, SD = 6). All NH listeners were screened at 25 db HL from 250 to 8000 Hz in both ears. They were recruited in the local community near Iowa City, IA. Table 1 shows age groups and a number of NH subjects. Five children did not pass the hearing screening, had a cold, or did not complete the test; therefore, their data were excluded from further analysis. For CI users, a total of fifty-nine subjects participated. Forty-three pre-lingually deafened CI children and adolescents between the ages of 3 and 18 and five pre-lingually deafened CI adults between the ages of 20 and 30 participated (mean = 25, SD = 3). Also, eleven post-lingually deafened CI adults between the ages of 27 and 63 (mean = 50, SD = 14) participated in this study. All CI users were recruited from the University of Iowa Hospitals and Clinics (UIHC) research subject database. All pre-lingually deafened CI children received their CI(s) before 3.5 years of age with Cochlear CI system. All postlingually deafened adult participants had an acquired hearing loss after the age of 20, i.e. after they developed typical speech and language skills. All CI subjects had more than 1 year of CI experience at the time of testing. They had seen audiologists within 6 months of the testing session or saw audiologists on the same day of testing. All CI subjects were considered to be a successful full-time CI user by their audiologists and their devices were working appropriately for testing. Parents and audiologists did not express any 94

112 concerns about the development of the child; all children had not been diagnosed with other difficulties except a hearing loss. Table 2 shows age groups and a number of CI subjects. One 10-year-old subjects did not finish testing in noise due to time constraint. Recordings from two subjects (one 13-year-old and one 20-year-old subjects) had contaminations of CI artifact in all recording channels. For these reasons, data from the three subjects were excluded from further analysis. Table 1. Normal hearing subjects Table 2. Cochlear implant subjects 3.2. General Procedures Both NH and CI listeners were tested using the identical equipment and procedures to present stimuli and record CAEPs. A subject sat in a comfortable chair in a sound booth. Speech stimuli consisting of two vowels were presented at 65 dba via a loudspeaker located at 0-degree azimuth approximately 1 meter from the listener. All stimuli were presented once in quiet and once in noise conditions. A subject was encouraged to read, play with an ipad, or watch captioned videos in order to stay alert during testing. The subject was informed that he/she would be monitored using two video cameras with an audio system mounted in the sound booth to make sure the subject is awake and comfortable without much movement. For recording, a total of seven disposable surface electrodes were used to create 95

113 four channels for all NH and CI subjects. Three channels were used to record CAEPs, and one channel was used to record eye blinks. One electrode was served as a ground electrode. In order to reduce the CI artifact, bilateral CI users were asked to wear only one CI during testing. The better hearing ear, according to the participant or parents report, was selected for testing. If the person had CIs sequentially, the one implanted less than 3.5 years of life was selected. If there was no preferred test ear by the subject or if it was simultaneous implantation, then the right ear was selected and the CI on the left side was taken off. For CI users who used an HA in the contra ear, the hearing aid was also taken off to reduce any artifact caused by the HA and the ear was plugged. This study required one visit of 2 hours for all subjects. This includes preparation, recordings, and two short breaks. More breaks were provided to subjects, as needed Electrophysiologic Measures Stimuli Two 800-ms-long stimuli, consisting of 400-ms-long quasi-synthetic vowels were procedurally created. First, a male English speaker produced two vowels /u/ and /i/. These two naturally spoken vowels were recorded at a sampling rate of Hz in the WAV sound format. Adobe Audition 1.5 (Adobe Systems Inc., San Jose, USA) was used to make changes to the recorded waveforms. One cycle of each vowel was selected and this cycle was repeated for longer than 400 ms, by copying and pasting each vowel cycle at zero crossing. Using this method two vowel segments (/u/ and /i/) with steady amplitude were created. Each vowel segment was truncated to 400 ms. Formant frequencies of /u/ are: F1 = 316 Hz, F2 = 1178 Hz, and F3 = 2306 Hz. Formant frequencies of /i/ are: F1 = 254 Hz, F2 = 2270 Hz, and F3 = 2701 Hz. Prior to connecting 96

114 these two quasi-synthetic vowels, /u/ and /i/, were adjusted to match the average RMS amplitudes. The two vowels were concatenated at the zero crossing, i.e., the last sampling point of one vowel to the first sampling point of the following vowel. In this way, two 800-ms-long stimuli were created: one beginning with the quasi-synthetic vowel /u/ followed by /i/ and vice versa. Creating stimuli in pairs (/u/-/i/ and /i/-/u/) allows us to measure both onset the P1-N1-P2 and the ACC responses for each segment of the stimulus. No discontinuity was observed in waveforms, and spectral splatter was marginal in the spectra. No audible clicks or pops were present in the stimuli, /u/-/i/ and /i/-/u/. These two stimuli were tapered over 5 ms at the onset and the offset using fade-in and fade-out cosine functions equipped in the Adobe Audition. The final two stimuli, /u/- /i/ and /i/-/u/, were again equalized in terms of overall RMS. Figure 19 shows the waveforms and spectra for the two 800-ms-long stimuli. Figure 19. Waveforms and spectrograms for stimuli (a) This stimulus includes two vowel stimuli with a change halfway through (400 ms after the onset) from /i/ to /u/ and vice versa (b) /u/ to /i/. 97

115 Sound presentation: in quiet and noise conditions A PC-based MATLAB program read each 800-ms-long speech stimulus. This MATLAB program had a low-pass filter with a cutoff frequency of 8 khz to attenuate any energy above the cutoff frequency of speech stimuli. The MATLAB program added 300-ms-long baseline prior to each stimulus. It also added a time delay between stimulus presentations so that the stimulus can be presented at a slow rate, on average 1 presentation every 3.3 seconds. The program varied the length of the ISI (onset-to-onset of the stimulus in this study) randomly between 2.8 and 3.8 seconds to help minimize effects of adaptation. The presented stimulus via MATLAB program was routed to the Grason-Stadler GSI-61 audiometer. The sound level was controlled by GSI-61 to obtain 65 dba at the subject s ear level located about 1-meter away from the loudspeaker in a double-walled sound booth. All speech stimuli were presented once in quiet and once with background noise. In quiet conditions, each stimulus alone was presented at about 65 dba by loudspeaker located at 0 º azimuth. In noise conditions, each stimulus was presented with background noise of 55 dba, creating +10 db SNR by the same loudspeaker located at 0 º azimuth. We used the speech-weighted noise built into the audiometer for the background noise. This speech-weighted noise consists of equal energy per frequency from 0.25 to 1 khz and a roll-off (12 db per octave) of 1 to 6 khz. All stimuli with quiet and noise listening conditions were randomly presented for all subjects Recording parameters A total of seven surface electrodes were applied to the scalp and on the face. Impedances of each electrode were lower than 5000 ohms and impedances between 98

116 electrodes were within 2000 ohms. Impedances were checked before and in the middle of recordings. The locations of the seven electrodes were vertex, right mastoid, left mastoid, inion, forehead, and above and below an eye. The vertex (Cz) electrode was used as an active, non-inverted electrode (+). The right and left mastoid electrodes and the inion electrode were used as a reference, i.e., an inverted electrode (-). Using these electrodes, three differential recording channels were created to record the onset P1-N1-P2 and ACC responses. The first channel recorded activities between vertex (Cz) and the right mastoid. The second channel recorded activities between vertex (Cz) and the left mastoid. The third channel recorded activity between vertex (Cz) and inion (Oz). Additionally, one channel was created to record eye blinks using the two electrodes, above and below the eye contralateral to the CI. Traces of possible contamination from eye blinks were rejected on-line. A ground electrode was placed on the forehead. An optically isolated Intelligent Hearing System (IHS 8000) differential amplifier was used to record the ongoing EEG activity with a gain of 10,000 and band-pass filter between 1 and 30 Hz. A National Instruments Data Acquisition board (DAQCard-6062E) was used to sample the ongoing EEG activity using a sampling rate of 10,000 Hz. Custom-designed LabView software was designed to display each waveform after each stimulus and to generate averaged responses on-line in all three auditory evoked potential recording channels. This software also allowed for monitoring waveforms from eye movements on-line. The recording time window was 1400 ms, starting 300 ms prior to each 800-ms-long stimulus. For each subject, a total of 800 accepted recordings were obtained for the two speech stimuli in both listening conditions. This was achieved by combining 4 separate recordings of 50 sweeps per stimulus in each listening condition. 99

117 Each 50-sweep recording took about 4 minutes and all recording took slightly more than 1 hour Analysis Grand average waveforms The obtained evoked potentials were analyzed off-line using a custom MATLAB program. Each subject had a total of two average waveforms (400 sweeps each) for cortical auditory evoked potentials recorded using speech (recordings using /u-i/ and /i-u/ were combined) in quiet and background noise listening conditions. Grand average responses were calculated for subjects grouped by one year and three-year age intervals to illustrate how the morphology of the onset P1-N1-P2 and ACC responses change as a function of age Peak latencies and N1-P2 peak-to-peak amplitudes Latencies of any identified peaks (P1, N1, P2) and peak-to-peak amplitudes of N1-P2 for both onset and ACC responses were analyzed using customized peak picking program using MATLAB. Means and standard deviations of peak latencies and amplitudes are reported in Appendix B. P1 latency change with age is of particular interest to compare developmental patterns of the onset and the ACC, because the P1 component is present in all age groups and is known to be a biological marker for the central auditory system development from previous studies of the onset response. Linear regression was used to test the age effect on the P1 latency for the onset and the ACC. A two-sample t-test was conducted to test whether the slopes of the onset and the ACC regression lines are significantly different 100

118 from each other. Also, a paired t-test was used to compare P1 latency obtained between the onset and the ACC, and between quiet and noise conditions. Since N1 peak is known to be absent in young children and starts to emerge later in life, the detectability (in percentage) of the N1 component and the N1-P2 peak-to-peak amplitude of the onset and the ACC were compared in each age group. A linear regression and a test comparing two slopes were also used to test the age effect on N1-P2 amplitudes and to show if they are different between the onset and the ACC. Again, a paired t-test was used to compare N1-P2 amplitudes between the onset and the ACC and between quiet and noise conditions Correlations between adults and children s grand average waveforms To quantify the developmental changes, correlations between waveforms at each age group were calculated relative to NH adult responses. This was a way to assess the similarity of waveforms at each age relative to the normal adult responses. This allows us to quantify the pattern of change and to compare the pattern of development easily between the onset and the ACC in each listening condition. Correlations between adults and children were done for the recordings for 250 ms in duration after a stimulus onset and after a stimulus change. This 250 ms window was selected, because both onset P1- N1-P2 and ACC responses occur within this latency in mature cortical auditory systems. To calculate correlations among adults, grand average waveforms from the half of the adults were compared with grand average waveforms from another half of the adults. This serves as a comparison noise level of correlations in adults. When correlation coefficients are plotted as a function of age, they show changes in the morphology and 101

119 amplitude of the onset P1-N1-P2 and ACC responses with increasing age in a quantitative way. All analyses were done in the same way for quiet and noise listening conditions. To assess the effect of background noise on the developmental patterns of the onset and the ACC, the responses were compared between quiet and noise conditions using the latency, amplitude, and correlation data described above. Also, all analyses were done using the same methods for both NH subjects and CI users. In addition, onset and ACC responses were compared among three adult groups (NH listeners, pre-lingually and postlingually deafened CI users) in both quiet and noise listening conditions, using analysis of variance (ANOVA) with post-hoc tests. All statistical analysis was conducted using SAS (Statistical Analysis System. 9.3.). 102

120 CHAPTER 4 RESULTS While the morphology of responses varied with age, both onset P1-N1-P2 and ACC responses were present in all waveforms obtained using long-duration speech-like stimuli from NH and CI subjects in both quiet and background noise conditions. Before discussing NH and CI children s data, Figures 20(a)-(d) show these cortical responses in NH adults, i.e., those with mature central auditory systems. Figures 20(a) and (b) show onset and ACC responses obtained in quiet and background noise conditions, respectively. Individual data are shown in black and grand average waveforms from 11 NH adults are shown in blue for quiet and in red for background noise conditions. The variability across subjects was small in both listening conditions. Figure 20(c) compares grand average waveforms obtained from the two listening conditions and Figure 20(d) shows the comparison of listening conditions for individual waveforms. Adding background noise (in this case, speech-shaped noise at 10 db SNR) did not change the morphology for onset and ACC responses. However, it did increase latencies and decrease N1-P2 peak-to-peak amplitudes significantly for both onset and ACC responses at the α = 0.01 level. Table B.1 in Appendix B summarizes latencies and amplitudes of both onset and ACC responses for the two listening conditions. The remaining results show how these responses are affected by age and hearing loss with CI use. 103

121 Figure 20. NH adult listeners CAEPs in quiet and noise conditions All (a)-(b) figures show cortical responses obtained from eleven NH adults. (a) and (b) show onset and ACC responses obtained in quiet and background noise (+10 db SNR using speechshaped noise) conditions, respectively. Individual data are shown in black and grand average waveforms in blue and red for each listening condition. (c) and (d) compare the two listening conditions in grand average waveforms and individual subjects, respectively. 104

122 4.1. Changes of CAEPs with Age in NH listeners in Quiet Results of NH listeners CAEPs recorded in quiet are shown in Figures Figure 21 shows morphologic changes of the P1-N1-P2 and the ACC with age, recorded at electrode Cz. The top panel shows individual waveforms, shown with dashed lines, and grand average waveform, with solid lines, for each age group. To more easily compare waveforms across age, the bottom left panel shows grand average waveforms for each age group, from 3 to 19 years old, and adults. P1 is present in all age groups for both onset and ACC responses. The N1 of the onset response emerges at around 7 or 8 years old. The morphology of the onset P1-N1-P2 becomes adult-like at around 11 years old with dominant N1-P2 components. The N1 of the ACC emerges later than the N1 of the onset, at around 10 or 11 years old. The morphology of the ACC becomes adult-like later than the onset response, showing P2 amplitude larger than P1, at 14 years old. The bottom right panel shows grand average waveforms grouped in 3-year intervals for five groups (ages 3 to 5, 6 to 8, 9 to 11, 12 to 14, 15 to 17, and 18 to 19) and one adult group. For the onset P1-N1-P2, at 3 to 5 years, P1 is the dominant component with longer latencies than any other age groups. At 6 to 8 years, the N1 peak begins to emerge in the onset response. At 9 to 11 years, the N1 and P2 responses become dominant in the onset response and the N1-P2 peak-to-peak amplitudes grow larger in older teenage groups. For ACC responses, P1 is the only component that is large in amplitude from 3 to 8 years old. At 9 to 11 years, N1 and P2 peaks emerge, but P1 is still dominant. At 12 to 14 years, the N1 and P2 are dominant in the ACC and the N1-P2 peak-to-peak amplitude grows in late adolescence. 105

123 Figure 21. Waveforms of CAEPs in NH listeners in quiet Grand average waveforms measured at the electrode Cz are shown for normal hearing listeners. Responses using /u-i/ and /i-u/ are combined. The top panel shows individual waveforms and grand average waveforms for each age group. The bottom left panel shows grand average waveforms for each age group ranged from 3 to 19 years, and adults. The bottom right panel shows grand average waveforms for 5 groups, grouped in 3 year intervals, and adults. The three straight lines at 0, 400, and 800 ms indicate when the onset, change, and offset of stimulus occur, respectively. Dashed lines indicate when P1, N1, and P2 components are shown in adults onset and ACC responses. 106

124 Table B.2 in Appendix B summarizes peak latencies and N1-P2 amplitudes of cortical responses measured from all NH listeners in quiet conditions. Traditionally, P1 latency has been used to characterize developmental changes in the onset response, since it is observed in all age groups. Figure 22 illustrates the change of P1 latency with age for the ACC as well as the onset response. At age 3, P1 latency was measured at 105 ms for the onset and 118 ms for the ACC. In adults, it was measured at 54 ms and 79 ms for the onset and the ACC, respectively. Figure 22. Changes of P1 latency with age in NH listeners in quiet This graph shows changes of P1 latencies in both onset and ACC responses. Black dots indicate data from the onset response and red dots indicate data from the ACC. Error bars show standard errors. The P1 latency of the ACC was calculated based on the post stimulus change. 107

125 Linear regression analyses show that the P1 latency for both the onset (p < ) and ACC (p < ) decreases significantly with age. For the onset response, the P1 latency decreases on average by about 3 ms/year. For the ACC, the P1 latency decreases on average by about 2 ms/year. Age alone predicts about 60% of P1 latency variations in the onset and 58% of P1 latency variations in the ACC (Table B.3 in Appendix B). The slope of the onset P1 latency with age was significantly different from the slope of the ACC P1 latency with age, using a two-sample t-test (t = 2.93, df = 168, p = ). This indicates that the developmental pattern of P1 latency may be different between the onset and the ACC. Additionally, P1 latencies of the onset and ACC responses were compared within a subject, using a paired t-test. The P1 latency of the onset response was significantly shorter than the Pl latency of the ACC response (Mean = ms, SD = 14, t = , df = 85, p < ). Unlike P1, N1 was not observed in young children for either onset or ACC responses. But, it emerged in older children. Figure 23 shows the course of development in the detectability of N1 (in the left panel) and the peak-to-peak amplitude of the N1 and P2 (in the right panel). For the onset response, N1 was detectable at 5 to 6 years in about 20% of subjects, reached about 50% at 7 to 8 years old, and was present in all subjects who are over the age of 10. For the ACC, the pattern is similar to that of the onset response; however, it does not emerge until 9-10 year olds and is not present in all subjects until 14 years. 108

126 Linear regression analyses show that the N1-P2 peak-to-peak amplitude increases with age significantly for both onset (p = ) and ACC responses (p = ). The slope of the N1-P2 amplitude is about 0.4 µv/year and 0.3 µv/year for the onset and the ACC, respectively. However, age predicts only about 10% of the variability in the amplitude of both onset and ACC responses (Table B.3 in Appendix B). The slopes of regression for the onset and ACC N1-P2 amplitudes are not significantly different at the α = 0.05 level. Additionally, a paired t-test showed that onset N1-P2 amplitude was significantly larger than the ACC N1-P2 amplitude in all age groups (Mean = 7.69 µv, SD = 4.10, t =13.78, df = 53, p < ). Figure 23. Changes of N1 component with age in NH listeners in quiet The left panel shows detectability of N1 and the right panel shows the N1-P2 peak-to-peak amplitude with age for both onset and ACC responses. For both graphs, black dots indicate data from the onset and red dots indicate data from the ACC. Error bars show standard errors. 109

127 A comparison of CAEPs measured from children and adults are complicated because they have different morphologies, peak latencies, and amplitudes (Figures 21-23). To compare responses efficiently, a correlation analysis was conducted between the adults grand average waveform and the group grand average waveform for each age group. Since the onset and ACC occur during the first 250 ms after a stimulus onset and a stimulus change, data from this duration was used to calculate the correlation coefficient between adults and each age group, respectively. Figures 24(a)-(c) show waveforms and calculated correlation coefficients for example age groups: 3 years, 9 years, and 15 years (shown in red) compared with the adults grand average waveform (shown in black). At 3 years, onset and ACC responses have peaks in opposite directions from the adults grand average waveform. They have strong negative correlations for both onset (r = ) and ACC responses (r = ). At 9 years, the onset response has a moderate positive correlation with adult responses (r = 0.54), but the change response has a moderate negative correlation (r = ). This is when the onset morphology becomes more adult-like while the P1 is still a dominant peak in the ACC. At 15 years, both onset and ACC responses have strong positive correlations with adults responses (r = 0.97 and 0.96, respectively). The bottom panel of Figure 24 shows all calculated correlation coefficients for both onset (shown in blue dots) and ACC (shown in red dots) responses. The onset response correlation coefficient plateaus around 11 years old. However, the ACC response correlation coefficient does not reach that plateau until 14 years old. 110

128 Correlation Figure 24. Correlations between NH children and NH adults' CAEPs in quiet The top panel figures (a)-(c) show examples of correlation coefficients between adults and children s grand average waveforms obtained in quiet. The adults grand average waveform is in black solid lines and 3-, 9-, and 15-year-old s grand average waveforms are in red. The dashed blue and red boxes indicate the duration (250 ms) used for a correlation analyses for the onset and the ACC, respectively. The bottom figure shows calculated correlation coefficients for the onset, in a blue line with dots, and the ACC, in a red line with dots. For the adults noise level, the adult group was divided into two groups and a correlation coefficient was calculated between the two adult groups. 111

129 4.2. Changes of CAEPs with Age in NH listeners in Noise The secondary aim of the current study was to investigate how challenging listening conditions impact the developmental patterns of the onset and the ACC. Reliable CAEPs were measured with background noise for all subjects. Figures 25 and 26 show CAEPs recorded from all NH listeners with speech-shaped noise at 10 db SNR. Figure 25 shows morphologic changes of the P1-N1-P2 and the ACC with age, recorded at electrode Cz with background noise. This graph shows data plotted in the same way as Figure 21 which showed CAEPs recorded in quiet conditions. The bottom right panel shows grand average waveforms from the three year interval age groups. In onset responses, at 3 to 5 years, P1 is the dominant component with larger amplitude. At 6 to 8 years, while amplitude becomes smaller, P1 is still the only clear component. At 9 to 11 years, the N1 and P2 peaks are evident in the onset response. The left panel shows that N1 may emerge as early as 8 years. At 12 to 14 years, the N1and P2 responses become dominant in the onset response and the morphology of the onset response becomes adult-like. For ACC responses, P1 is the only clear component between 3 and 11 years. N1-P2 peaks are evident at ages 12 to 14. The left panel shows that N1 becomes clear at around 12 years old. At ages 15 to 17 years, the N1-P2 response becomes larger in amplitude. The left panel shows that the morphology of the ACC becomes adult-like as early as 15 years, showing P2 amplitudes larger than P1. 112

130 Figure 25. Waveforms of CAEPs in NH listeners in noise Grand average waveforms are shown for NH listeners in background noise. The top panel shows individual waveforms and grand average waveforms for each age group. The bottom left panel shows a series of grand average waveforms from 3 to 19 years, and adults. The bottom right panel shows grand average waveforms for five groups of children, grouped in three year increments, and adults. Three straight lines indicate when the onset, change, and offset of stimulus occur. Dashed lines indicate when P1, N1, and P2 components are shown in adults responses. 113

131 Table B.4 in Appendix B summarizes data in peak latencies and N1-P2 amplitudes of the onset and the ACC responses obtained from NH listeners in noise. Figure 26 shows calculated correlation coefficients for both onset (shown in blue) and ACC (shown in red) responses, using data obtained in noise. The onset response s correlation coefficients plateau around 13 years old. However, the ACC response s correlation coefficient continues to improve even in late teen years and reach adult levels at around 18 years old. Figure 26. Correlations between NH children and NH adults' CAEPs in noise Correlation coefficients are calculated between each age group grand average waveform obtained with background noise and adult s in noise. The blue line with dots represents the onset response and the red line with dots represents the ACC. The same time window (250 ms) after a stimulus onset and change was used for correlation calculation for onset and ACC responses, respectively. For the adults noise level, the adult group was divided into two groups and a correlation coefficient was calculated between the two adult groups. 114

132 Figures present the same data as in previous figures, but directly compare CAEPs in quiet and noise. Figure 27 shows CAEPs recorded in both quiet and noise for NH children and adults. Adding background noise clearly impacted the morphology in young children showing a delayed developmental pattern. At 6-8 years, N1 and P2 peaks are identifiable in the onset response in quiet; however, P1 is still dominant in noise. From 9 to 14 years, N1 and P2 peaks are identifiable, and their amplitude grows in the ACC. In background noise, N1-P2 peaks of the ACC are not identifiable at 9 to 11 years, and their amplitude is smaller compared to that in quiet at years. Figure 27. Comparisons of grand average waveforms between quiet and noise conditions in NH listeners Grand average waveforms measured in both quiet and noise conditions are shown for NH listeners. Gray lines indicate recordings in quiet and colored lines indicate recordings with background noise. The three straight lines at 0, 400, and 800 ms indicate the onset, change, and offset of stimulus respectively. Dashed lines indicate when P1, N1, and P2 components are shown in adults responses in a quiet listening condition. 115

133 Figure 28 further illustrates these changes, comparing (a) P1 latency, (b) N1 detectability, and (c) N1-P2 peak-to-peak amplitude between quiet and background noise for the onset and the ACC. Gray lines with dots represent data in quiet, and colored lines with dots represent data in noise. The left columns compare them for the onset, and the right columns for the ACC. In noise, the P1 latency for both the onset (p < ) and the ACC (p < ) decreased significantly with age (Figure 28(a)). Slopes of P1 latency functions are by about 3 ms/year and 2 ms/year for the onset and the ACC, respectively. Age explains about 58% of P1 latency variations in the onset and only 40 % in the ACC (Table B.5 in Appendix B). The slopes of the onset P1 latency and the ACC P1 latency with age are also significantly different (t = 2.62, df = 168, p = ). This indicates that the developmental pattern of P1 latency may be different between the onset and the ACC in noise as well as in quiet. In addition, P1 latencies of the ACC were significantly longer than the P1 latency of the onset in noise (Mean = ms, SD = 16.47, t = , df = 85, p < ). The slope of onset P1 latency with age was not significantly different for quiet and noise conditions at the α = 0.05 level. No significant difference was found in the slope of ACC P1 latency between quiet and noise at the α = 0.05 level. This may indicate that the developmental pattern of P1 latency may be same between quiet and noise listening conditions. The onset P1 latency in quiet was significantly shorter than the onset P1 latency in noise (Mean = ms, SD = 12.75, t = -5.96, df = 85, p < ). The ACC P1 latency was also significantly shorter in quiet than in noise (Mean = ms, SD = 9.67, t = -5.82, df = 85, p < ). 116

134 In Figure 28(b), N1 detectability shows a clear developmental delay in noise for the ACC (right column). The ACC N1 is not present in all subjects until late teenage years or about 18 years of age. In Figure 28(c), the N1-P2 peak-to-peak amplitude measured in noise increased with age significantly only for the onset response (p < ). The slope of the N1-P2 amplitude is about 0.6 µv/year. Age explains only about 20% of the variability in the amplitude of the onset response. A significant linear relationship was not found with the ACC N1-P2 amplitude. (Table B.5 in Appendix B). A paired t-test shows that the onset N1-P2 amplitude was significantly larger than the ACC N1-P2 amplitude (Mean = 6.74 µv, SD = 2.75, t =16.08, df = 42, p < ). The slope of onset N1-P2 amplitude with age was not significantly different between quiet and noise conditions at the α = 0.05 level. This suggests that the developmental pattern of N1-P2 amplitude may be same for the onset in quiet and noise listening conditions. When the onset N1-P2 amplitudes were compared for quiet and noise, the amplitude in quiet was significantly larger than in noise (Mean = 3.06 µv, SD = 2.74, t = 9.42, df = 80, p < ). The ACC N1-P2 amplitude was also significantly larger in quiet than in noise (Mean = 2.10 µv, SD = 1.66, t = 8.29, df = 42, p < ). 117

135 Figure 28. Comparisons of P1 latency and N1-P2 response between quiet and noise conditions in NH listeners Figures show means and standard errors for (c) P1 latency (b) N1 detectability, and (c) N1-P2 amplitudes in NH listeners in quiet (in gray) and noise (in colors) for the onset (in left columns) and the ACC (in right columns). 118

136 Finally, Figure 29 illustrates a delayed developmental pattern of correlation to adult waveforms in noise relative to that in quiet. This delay is more pronounced for the ACC. Adding background noise led to poorer correlations at ages 7-9 compared to those measured in quiet for the onset. These ages were when N1 started to emerge earlier in quiet than in noise for the onset P1-N1-P2, reaching a positive correlation coefficient first in quiet. For the ACC, correlation coefficients were poorer over a longer period and only reaching a maximum correlation in late teen years. Figure 29. Comparisons of correlations between quiet and noise conditions in NH listeners The left panel shows calculated correlation coefficients for the onset response in quiet and in noise. The right panel shows calculated correlation coefficients for the ACC in the two listening conditions. The gray lines with dots are data from the quiet listening condition. The colored lines with dots are data with background noise. 119

137 4.3. Changes of CAEPs with Age in CI users in Quiet The primary goal of this study was to investigate the impact of development on the onset P1-N1-P2 and the ACC in NH and CI listeners using long duration speech-like stimuli. Previous sections have shown results of NH listeners in quiet and with background noise. The following sections will focus on how the onset P1-N1-P2 and the ACC change with age in individuals who were born deaf but received a CI early in life. Before examining the CI children s data, Figures 30(a) and (b) present CAEPs measured in early-implanted, pre-lingually deafened adults, i.e., those who grew up with their CI, and post-lingually deafened adults, who received their CI after the age of 20. Onset and ACC responses obtained in quiet (the left panel) and with background noise (the right panel) are plotted for pre-lingually deafened (Figure 30(a)) and post-lingually deafened CI adults (Figure 30(b)). Individual data are shown in black and grand average waveforms are shown in blue and red for the quiet and noise listening conditions, respectively. Similar to the results of NH listeners, adding 10 db SNR background noise did increase peak latencies and decrease N1-P2 peak-to-peak amplitudes significantly for both onset and ACC responses at the α = 0.01 level. Table B.6 in Appendix B summarizes peak latencies and amplitudes and compares them for quiet and noise conditions. 120

138 Figure 30. Adult CI users CAEPs in quiet and noise conditions (a) plots cortical auditory evoked potentials obtained from pre-lingually deafened CI adults and (b) from post-lingually deafened adults. The left and center panels show responses measured in quiet and noise listening conditions, respectively. Individual data are shown in black and grand average waveforms are shown in blue and red for each listening condition. The right panel compares grand average waveforms between the two listening conditions for each of the CI groups. Figure 31 compares responses of pre-lingually deafened adults to those of NH listeners and post-lingually deafened adults. It is assumed that subjects from the three listening groups have a mature central auditory system since they are all over 20 years of age. Pre-lingually deafened CI adults appear to have comparable onset responses to both NH and post-lingually deafened adults. While the morphology of the ACC obtained from the two CI groups is similar to that of NH adults, its amplitude appears to be smaller than that from NH adults, especially with background noise. ANOVA tests and post hoc analyses confirmed these observations. The onset N1-P2 amplitude measured from the 121

139 pre-lingually deafened adults was not significantly different from NH adults responses for either quiet or noise listening conditions. However, for the ACC, the N1-P2 amplitudes measured from pre-lingually deafened adults were significantly smaller than NH adults for both quiet and noise conditions at the α = 0.01 level. Interestingly, the ACC amplitudes measured from post-lingually deafened adults were also smaller than NH adults in both quiet and noise conditions. There were no significant differences in the ACC amplitudes between the two CI adults groups. Table B.7 (a) and (b) in Appendix B summarizes results of ANOVA tests and post hoc analyses. Figure 31. Comparisons of CAEPs among the three adult listening groups The left and right panels compare grand average waveforms of the two CI groups to that of eleven NH adults for quiet and noise listening conditions, respectively. The gray lines indicate data from NH, the green lines indicate data from post-lingually deafened adults, and the blue (in quiet) and red (in noise) indicate data from pre-lingually deafened adults. 122

140 Developmental data from pre-lingually deafened CI listeners in quiet are shown in Figures The plots parallel those presented for NH subjects. Figure 32 shows morphologic changes of the P1-N1-P2 and the ACC with age. For onset P1-N1-P2 responses, P1 is the dominant component at 3 to 5 years, with longer latencies than any other age group. At 6 to 8 years, P1 is still present alone. The N1 of the ACC is shown as young as 8-years-old in the grand average waveform (the left panel). At 9 to 11 years, the N1and P2 responses become dominant in the onset response and the amplitude grows from ages At age years, the morphology of the onset response becomes adult-like. For ACC responses, P1 is the only component evident from 3 to 8 years of age. The N1 of the ACC is shown as young as 10 in the grand average waveform (the left panel). N1-P2 peaks are shown at 9 to 11 years, and amplitude grows with age. The morphology of the ACC becomes adult-like at 15 to 17 years, showing P2 amplitude larger than P1. 123

141 Figure 32. Waveforms of CAEPs in pre-lingually deafened CI users in quiet Grand average waveforms measured at the electrode Cz are shown for early-implanted, prelingually deafened CI listeners. The top panel shows individual waveforms and grand average waveforms for each age group. The bottom left panel shows grand average waveforms for each age group ranging from 3 to 18 years, and adults. The bottom right panel shows grand average waveforms for five groups of children, grouped in three year intervals, and adults. The three straight lines at 0, 400, and 800 ms indicate the onset, change, and offset of stimulus respectively. Dashed lines indicate when P1, N1, and P2 components shown in adults onset and ACC responses. 124

142 All peak latencies and N1-P2 amplitudes of CAEPs are summarized in Table B.8 in Appendix B. Previous studies show that early-implanted, pre-lingually deafened children exhibit similar P1 latency changes to those seen in NH children. Figure 33 illustrates the change of P1 latency with age for the ACC as well as for the onset in earlyimplanted, pre-lingually deafened children. This figure is parallel to Figure 22 in which data of NH are shown. Figure 33. Changes of P1 latency with age in CI users in quiet This graph shows changes of P1 latencies in both onset and ACC responses. Black dots indicate data from the onset response and red dots indicate data from the ACC. Error bars show standard errors. The P1 latency of the ACC was calculated based on the post stimulus change. 125

143 Linear regression analyses show that the P1 latency for both the onset (p < ) and the ACC (p < ) decreases significantly with age. The P1 latency decreases by about 3 ms/year for the onset response and about 2 ms/year for the ACC. Age alone predicts about 35% of P1 latency variations in the onset and 51% of P1 latency variations in the ACC. Details of these analyses are shown in Table B.9 in Appendix B. The slope of the onset P1 latency with age was compared with the slope of the ACC P1 latency with age, using a two sample t-test. The two slopes are not significantly different (t = 1.25, df = 80, p = 0.214). This indicates that the developmental pattern of P1 latency may be the same for the onset and ACC responses. P1 latencies of the onset and ACC responses were compared within subjects, using a paired t-test. The P1 latency of the onset response was significantly shorter than the Pl latency of the ACC response (Mean = ms, SD = 20, t = -9.36, df = 41, p < ). Like NH children, N1 is not present in young CI children for either onset or ACC responses; however, it emerges in older children, and its amplitude grows among teenagers (Figure 21 and 32). Figure 34 shows the course of development in the presence of N1 (in the left panel) and the peak-to-peak amplitude of the N1 and P2 (in the right panel). For the onset response, N1 emerges at around 8 years old, reaches about 50% at 9 years old, and becomes present in all subjects by 10 years old. For the ACC, the pattern of the N1 detectability is more variable up until 13 years of age. N1 is present in all subjects who are 14 and older. The right panel shows N1-P2 amplitude changes with age. Linear regression analyses show that the N1-P2 peak-to-peak amplitude increases with age significantly only for the onset at the α = 0.01 level, not for the ACC. The slope of the N1-P2 126

144 amplitude is about 0.6 µv/year and age explains about 26% of the variability in the amplitude of onset responses (Table B.9 in Appendix B). Additionally, the paired t-test shows that onset N1-P2 amplitude was significantly larger than the ACC N1-P2 amplitude within individuals (Mean = 3.88 µv, SD = 4.00, t =4.85, df = 25, p < ). Figure 34. Changes of N1 component with age in CI users in quiet The left panel shows detectability of N1 and the right panel shows the N1-P2 peak-to-peak amplitude with age for both onset and ACC responses. For both graphs, black dots indicate data from the onset and red dots indicate data from the ACC. Error bars show standard errors. 127

145 Figure 35 shows calculated correlation coefficients between grand average waveforms of the early-implanted CI children and NH adults. The identical method was used for calculating correlation coefficients for NH (shown in Figure 24) and CI children. All calculated correlation coefficients for the onset are shown in blue dots and for ACC in red dots. The pattern of the onset P1-N1-P2 development with age is noisy but similar to the normal hearing data. While the onset data from NH children plateau around 11 years old, the onset data from CI children reach high correlation coefficients at 11 years, and plateau around 14 years old. The pattern of the ACC is also noisy but continues to improve with age. However, its correlation coefficient is lower than that of onset in teenagers. Also, pre-lingually deafened CI adults (show as in PA in the x-axis in Figure 35) show higher correlation for the onset than the ACC with NH adults. Figure 35. Correlations between CI children and NH adults CAEPs in quiet Correlation coefficients are calculated between CI children grand average waveforms and NH adult s obtained in quiet. The blue line with dots represents calculated correlation coefficients for the onset and the red line with dots represents the ACC. PA stands for pre-lingually deafened CI adults. 128

146 4.4. Changes of CAEPs with Age in CI users in Noise This section will show how a challenging listening conditions impact the developmental patterns of the onset P1-N1-P2 and the ACC in CI users. The CAEPs measured from pre-lingually deafened CI users with background noise, i.e., speechshaped noise at 10 db SNR, are shown in Figures 36 and 37. Also, Figures directly compare CAEPs recorded in CI users between quiet and noise. Figure 36 shows morphologic changes of the P1-N1-P2 and the ACC with age. For onset responses, at 3 to 5 years, P1 is the dominant component with larger amplitude. At 6 to 8 years, P1 is still the only present peak, but with smaller amplitude. At 9 to 11 years, the N1 peak begins to emerge. N1 and P2 peaks become dominant in the onset response at 12 to 14 years. The morphology of the onset response becomes adult-like at 15 to 17 years. For ACC responses, P1 is the only component and is large in amplitude from 3 to 11 years. The N1-P2 peaks are evident at 12 to 14 years and the amplitude grows at 15 to 17 years. However, the N1-P2 peak-to-peak amplitude appears to be small in the 18 years olds and pre-lingually deafened adults grand average waveforms. Table B.10 in Appendix B summarizes peak latencies and N1-P2 amplitudes from pre-lingually deafened CI users in noise. 129

147 Figure 36. Waveforms of CAEPs in CI users in noise Grand average waveforms are shown for early-implanted, pre-lingually deafened CI users in background noise. The top panel shows individual waveforms and grand average waveforms for each age group. The bottom left panel shows a series of grand average waveforms from 3 to 19 years, and adults. The bottom right panel shows grand average waveforms for 5 groups of children, grouped in 3 year increments, and adults. Three straight lines indicate when the onset, change, and offset of stimulus occur. Dashed lines indicate when P1, N1, and P2 components are shown in adults responses. 130

148 Figure 37 shows correlation coefficients for both onset (shown in blue) and ACC (shown in red) responses, using data obtained with background noise. The onset response s correlation coefficients plateau around 16 years old. The ACC response s correlation coefficients are poor until 13 years old of age and improve at 14 years old age. After 14 years of age, they are stable, but maintain at a moderate positive correlation in late teenagers and pre-lingually deafened young adults. Figure 37. Correlations between CI users and NH adults' CAEPs in noise Correlation coefficients are calculated between CI children grand average waveforms and NH adult s obtained in speech-shaped noise at 10 db SNR. The blue line with dots represents calculated correlation coefficients for the onset and the red line with dots represents the ACC. PA stands for pre-lingually deafened CI adults. 131

149 Figures directly compare CAEPs recorded in quiet and noise conditions from CI users. Figure 38 shows CAEPs recorded in both quiet and noise. As shown in NH data, adding background noise clearly impacted the morphology, latency, and amplitude of the response. At 9-11 years, N1 and P2 peaks are identifiable for both the onset and ACC in quiet. However, in background noise, amplitudes decreased in the onset and are absent in the ACC. Interestingly, unlike NH adults, the ACC responses of the pre-lingually deafened CI adult group became much smaller when compared to those in quiet listening conditions. Figure 38. Comparisons of grand average waveforms between quiet and noise conditions in CI users Grand average waveforms measured in both quiet and noise conditions are shown for CI users. Gray lines indicate recordings in quiet and colored lines indicate recordings with background noise. The three straight lines at 0, 400, and 800 ms indicate the onset, change, and offset of stimulus respectively. Dashed lines indicate when P1, N1, and P2 components are shown in adults responses in a quiet listening condition. 132

150 Figure 39 further illustrates these changes, comparing (a) P1 latency, (b) N1 detectability, and (c) N1-P2 peak-to-peak amplitude between quiet and background noise for the onset and the ACC. Gray lines with dots represent data from quiet, and colored lines with dots represent data from noise. The left columns compare them for the onset, and the right columns for the ACC. First, the P1 latency in noise decreased significantly with age for both onset (p = ) and ACC (p < ) responses (Figure 39(a)). The slope of the P1 latency with age is about 3.0 ms/year for the onset and 2.7 ms/year for the ACC. Age explains about 28% of P1 latency variations in the onset and 44 % in the ACC (Table B.11 in Appendix B). The slopes of the onset P1 latency and the ACC P1 latency with age are not significantly different from each other at the α = 0.05 level. This suggests that the developmental pattern of P1 latency may be same for the onset and the ACC in noise. In addition, P1 latencies of the ACC were significantly longer than the P1 latency of the onset in noise (Mean = -24 ms, SD = 24.26, t = -6.26, df = 39, p < ). The slope of onset P1 latency with age was not significantly different between quiet and noise conditions at the α = 0.05 level. No significant difference was found in the slope of ACC P1 latency between quiet and noise at the α = 0.05 level. This may indicate that the developmental pattern of P1 latency may be the same in quiet and noise listening conditions. The onset P1 latency in quiet was significantly shorter than the onset P1 latency in noise (Mean = -15 ms, SD = 20.79, t = -4.93, df = 43, p < ). The ACC P1 latency was also significantly shorter in quiet than in noise (Mean = -11 ms, SD = 12.89, t = -5.46, df = 43, p < ). 133

151 In Figure 39(b), N1 detectability shows a clear developmental delay in noise for the ACC (right column). With some variability, the onset becomes present in all subjects by age 13. The ACC N1 is present in all subjects by age 16. In Figure 39(c), the N1-P2 peak-to-peak amplitude measured in noise increased with age significantly only for the onset response (p = ). The N1-P2 amplitude is about 0.5 µv/year. Age explains only about 37% of the variability in the amplitude of the onset response. Details of these analyses are shown in Table B.11 in Appendix B. A paired t-test shows that the onset N1-P2 amplitude was significantly larger than the ACC N1-P2 amplitude (Mean = 4.58 µv, SD = 3.03, t =6.59, df = 18, p < ). The slope of onset N1-P2 amplitude with age was not significantly different between quiet and noise conditions at the α = 0.05 level. This suggests that the developmental pattern of N1-P2 amplitude may be same for the onset in quiet and noise listening conditions. The onset amplitude in quiet did not show significant differences from that in noise (Mean = 1.1 µv, SD = 2.91, t = 1.84, df = 23, p = 0.079), but the ACC N1-P2 amplitude was significantly larger in quiet than in noise (Mean = 1.5 µv, SD = 1.27, t = 5.12, df = 18, p < ). 134

152 Figure 39. Comparisons of P1 latency and N1-P2 response between quiet and noise conditions in CI users Figures show means and standard errors for (c) P1 latency (b) N1 detectability, and (c) N1-P2 amplitude in early-implanted, pre-lingually deafened CI users in quiet (in gray) and noise (in colors) for the onset (in left columns) and the ACC (in right columns). 135

153 Finally, Figure 40 illustrates a delayed developmental pattern in correlations to adult waveforms in noise relative to those in quiet. Adding background noise led to poorer correlations at ages compared to those measured in quiet for the onset. These ages are when both amplitude and latency for the ACC are particularly affected by noise (Figure 38). The correlation coefficients for onset responses reach levels close to one for both quiet and noise; however, correlation coefficients of the ACC in noise remain relatively low, even in adults. Figure 40. Comparisons of correlations between quiet and noise conditions in CI users The left panel shows calculated correlation coefficients for the onset response in quiet and in noise. The right panel shows calculated correlation coefficients for the ACC in the two listening conditions. The gray lines with dots are data from the quiet listening condition. The colored lines with dots are data with background noise. PA stands for pre-lingually deafened CI adults. All correlation coefficients were calculated between pre-lingually deafened CI children and NH adults grand average waveforms obtained in quiet and noise. 136

154 This study was not designed to investigate individual changes within subjects; however, the two subjects in Figure 41 can be examples of observed changes within subjects for NH listeners and CI users. An NH subject was tested at 14 and 18 years old. At 18 years old, his onset and ACC responses in quiet were enhanced relative to those obtained at 14 years old in the same listening condition. In noise, the NH listener at 18 years old shows larger N1-P2 peak amplitudes in the ACC than those at 14 years old. A CI subject was tested at 12 and 14 years old. This subject shows morphologic changes in ACC responses in noise between these two recordings. At 14 years old, the P1 peak of the ACC was decreased, and the morphology became more adult-like in quiet listening conditions. With background noise, the CI subject had a prominent P1 only for the ACC at 12 years old. Two years later, the N1-P2 response was present in the ACC in noise listening conditions. 137

155 Figure 41. Examples of changes with age within individuals A top panel shows cortical auditory evoked potentials obtained from one NH listeners in quiet and noise conditions. The gray solid lines show responses obtained when the subject was 14 years old. The blue and red solid lines show responses obtained when the subject was 18 years old in quiet and noise listening conditions, respectively. A bottom panel shows cortical auditory evoked potentials obtained from one pre-lingually deafened CI users in quiet and noise conditions. The gray solid lines show responses obtained when the subject was 12 years old. The blue and red solid lines show responses obtained when the subject was 14 years old in quiet and noise listening conditions, respectively. 138

156 4.5. Summary 1. Using long duration speech-like sounds, onset and ACC responses were successfully obtained from NH listeners and early-implanted, pre-lingually deafened CI users, from 3 year olds to adults, in both quiet and noise listening conditions. 2. The morphology, latency, and amplitude of the onset and the ACC varied with age. The ACC followed similar developmental patterns to those of the onset response, but the ACC matured later than the onset response. Pl latency decreased significantly with age for both onset and ACC responses. These developmental patterns were the same between NH listeners and CI users in both quiet and noise conditions. 3. Adding background noise affected both onset and ACC responses: degraded morphologies, longer latencies, and smaller amplitudes were obtained in noise. These effects led to a delayed development of the onset and the ACC in noise in both NH and CI groups. 4. However, the ACC was affected by noise more than the onset response. This was apparent in the ACC of CI users. 5. These data outline developmental trends in NH listeners and CI users. They can provide an important basis for future studies of individual variations in development. 139

157 CHAPTER 5 DISCUSSION AND CONCLUSIONS Can a CI facilitate the development of the auditory brain in children who were born deaf but received a CI at a young age? Will the development of the central auditory system in these CI children be similar to that of NH children? The current experiment attempts to answer these questions in part by studying developmental effects on two auditory evoked potentials measured at the cortical level: the onset P1-N1-P2 and the ACC. The onset P1-N1-P2 response is related to detection of sounds. The ACC is related to discrimination of sounds. Previous studies have focused on the onset response alone in order to study changes with age. Results show that the P1 latency decreases with age over the first two decades of life in both NH listeners and CI users, especially those who were implanted before 3.5 years of age. This suggests similar developmental patterns in both groups (e.g., Ponton et al., 1996, 2000; Sharma et al., 2002a,b, 2007; Dorman et al., 2007). While the ACC has widely been used to estimate auditory discrimination abilities in various listening groups and ages, studies on developmental patterns of the ACC have not been reported. The current study adds to the literature on developmental effects on CAEPs, specifically the developmental changes of ACCs in both easy and difficult listening conditions. The hope is that this information will be useful in estimating behavioral discrimination abilities in NH and CI children using ACCs. The current study seeks to answer: (1) how onset and ACC responses change with age in NH listeners and early-implanted CI users in quiet conditions, and (2) how background noise impacts those developmental patterns in both NH and CI listeners. The two major findings of the study are that (1) the ACC matures late relative to the onset response in both NH and CI 140

158 groups. (2) Background noise impacts the development of ACC responses more than onset responses in both NH and CI groups. ACC responses with background noise in CI users did not reach NH adult levels even among late teenagers and young adults. All of these findings support the hypotheses described in the Chapter 1 Introduction Feasibility of Recording CAEPs using Long Duration of Complex Stimuli Previous CAEP development studies have used brief stimuli such as clicks (100 µsec/phase, for 100 ms) or syllables (90 ms) to elicit the onset P1-N1-P2 response alone (e.g., Ponton et al., 1996, 2000; Sharma et al., 1997, 2002a,b, 2007). In this study, we constructed an 800-ms-long stimulus using two vowels as a way to elicit both the onset and the ACC. The primary change appeared in F2 of the two vowels from about 1200 Hz to 2300 Hz, or vice versa. This stimulus was presented at a slow rate, about 1 stimulus every 3 seconds. Both the P1-N1-P2 complex and the ACC were successfully recorded from both NH and CI children. Our results of development of the onset P1-N1-P2 response are comparable to those reported in the literature that used a brief stimulus to elicit the onset alone. Figure 41 shows how onset P1 latency data from the current study fits the regression reported in the previous study. The solid line is the best fit line for NH listeners P1 latency in Sharma et al. (2002a). P1 latency data from the current study for both NH (in open circles) and CI (in red triangles) fit the line very well. 141

159 Figure 42. Comparisons of onset P1 latency between previous and current studies Onset P1 latency data from the current study were plotted on the regression line reported in Sharma et al. (2002a). Open circles indicate the P1 latency recorded from NH listeners and filled triangles from early-implanted, pre-lingually deafened CI users in the current study. In addition to P1 latency, morphologic changes with age in the onset response are also consistent with previous studies. All previous studies were done in quiet conditions. Our findings among NH listeners in quiet conditions are (1) P1 is the only evident peak in children from 3 to 5 years, (2) N1 is absent in young children, but emerges at ages 7 to 8 and is present in all subjects by 10 years old, (3) the response becomes adult like at age 11, and (4) the N1-P2 peak-to-peak amplitudes increase significantly with age. These results agree with the body of the development literature in NH listeners in quiet (e.g., Ponton et al., 1996b, 2000; Sharma et al., 1997; Cunningham et al., 2000; McArthur & Bishop, 2002; Gilley et al., 2005; Wunderlich & Cone-Wesson, 2006). Unfortunately, previous studies on developmental patterns in CI users are minimal and mostly focus on P1 latency development alone. Ponton et al. (1996b) found that the rate of P1 latency 142

160 development is similar to NH listeners, but that the age at which responses become adultlike is delayed for implanted children. Our CI onset response data agree with Ponton et al. (1996b). The P1 latency decreases at the same rate as that of NH listeners. The correlation between CI children and NH adults plateaus at around 14 years for the onset responses, while the same measure for NH children plateaus at around 11 years. In fact, the CI onset responses from the current study show systematic changes with age similar to NH listeners. Similar findings on the onset response for NH and CI groups are also found in the pilot study (Appendix A) where complex music-like stimuli were used. More discussion of results in CI users follows in the next section. In short, the developmental effects on the onset response in the current study support previous findings in the literature for both NH listeners and CI users. This is important because it indicates that it is feasible to use a long duration stimulus including an acoustic change at the halfway point to study developmental changes on the ACC, without affecting the onset response adversely. This is also encouraging for clinical and research applications. This stimulus can be constructed using any long duration speechand music-like stimuli close to the sounds that we hear every day. Findings show that we can also record both onset P1-N1-P2 complex and the ACC simultaneously, allowing estimation of both sound detection and discrimination Developmental Patterns of the ACC in NH listeners and CI users in Quiet The highlight of the findings of the current study is the developmental patterns of ACCs. First, the ACC P1 latency decreases as a function of age in both NH and CI listeners at the same rate (3 ms/year). The ACC P1 latency is significantly longer than the onset P1 latency in both NH and CI groups. Second, results show that the ACC 143

161 experiences the same morphologic changes, but with later development than those of the onset response. P1 is the only clear component in young children, N1 starts to emerge in older children, and N1-P2 amplitude grows with age in teenagers. However, the ages when N1 emerges and the response becomes adult-like occur later than the ages shown with the onset response. The N1-P2 peak-to-peak amplitude of the ACC is smaller than that of the onset. These findings are common among both NH and CI groups. In NH listeners, the N1 is seen at ages 6 to 8 for the onset, but not until ages 9 to 11 for the ACC. The correlation coefficient reaches a plateau, i.e., high positively correlated with adults, at 11 years for the onset, and at 14 years for the ACC. Therefore, the ACC has a longer developmental trajectory when compared to the onset response in NH listeners in quiet. The N1-P2 amplitude increases with age for both onset and the ACC about 0.5 µv/year. Similar developmental patterns of NH listeners are also shown in CI users with the onset and the ACC, i.e., the ACC matures later than the onset. However, there are a couple of differences in the developmental patterns of CI listeners compared to NH listeners. First, the onset response in CI listeners appears to develop slightly later than in NH children. They reach about the same level of development by teenagers. For onset responses, N1 is evident at ages 6-8 in NH listeners, while P1 is still the dominant peak in CI listeners. Comparing onset response correlation data in NH listeners and CI users may suggest a late maturation of the onset response in CI listeners by about 3 years when compared to NH listeners. The onset response correlation data in NH listeners reach high positive correlations (around 0.8 and 0.9) with NH adults at 11 years. CI users onset response correlation data are noisy compared to the NH data and did not reach high correlations until 14 years. This delay of onset maturation in CI users 144

162 may relate to the duration of deafness before implantation. Ponton et al. (1996b) also observed delayed development in the onset response. However, pre-lingually deafened adults do not have significant differences in morphology, latency, or amplitude of the onset response when compared to the NH listeners (Figure 35). This suggests that while the development of the onset response is delayed in CI children due to sound deprivation periods before cochlear implantation, it still reaches NH adult levels by 20 years old. Differences in ACCs between the NH listeners and CI users are interesting. The correlation of the ACC in CI listeners continues to improve even among late teenagers (Figure 35). The 17 year olds and pre-lingually deafened CI users had at best about 0.8 correlation coefficients with NH adults. The poor correlation of the ACC with NH listeners may relate to the smaller N1-P2 amplitudes, which would reflect perception abilities. The pre-lingually deafened CI users have comparable onset amplitudes to those of NH listeners, but the ACC amplitudes are significantly smaller in pre-lignaully deafened CI users. Post-lingually deafened adults, who received their CI after the age of 20, had significantly smaller ACC amplitudes than NH listeners (Figure 31). With the assumption that post-lingually deafened adults have similar auditory development experiences to NH listeners, the poor amplitudes of the ACC may relate to limitations of CI sound processing and poor frequency resolution, rather than development of the central auditory system. We also observed that the onset N1-P2 amplitudes responses increased significantly with age, while the ACC amplitudes did not. Previous studies have shown a relationship between ACC amplitudes and detection of the acoustic change (e.g., Martin & Boothroyd, 2000; Tremblay et al., 2004; He et al., 2012, 2013, 2014; Brown et al., 2015). 145

163 Why is there a similar developmental pattern, but slower time course of development of the ACC compared to the onset response? P1 is believed to be from the auditory cortex, known for the prominent myelinated structure of the temporal lobe (e.g., Bock et al., 2009). Numerous studies associate onset P1 latency decrease with increasing age with shorter travelling time of sounds along the central auditory system due to the development of myelination and synaptogenesis (e.g., Moore et al., 1995; Ponton et al., 1996a,c; Eggermont et al., 1997; Sharma et al., 1997, 2000a,b, 2005). While the generators of ACC responses have not been studied thoroughly, researchers generally believe that the change response originates from the auditory cortex. Jones (2003) hypothesized that a different set of neuron responses contribute to the change of the response, since most neurons can be in refractory stages. Ganapathy et al. (2013) suggested the sources are the same for both onset and ACC responses. Based on the results of the current study, the ACC may originate from the same generators as the onset response, because their developmental change with P1 latency and morphologic developmental patterns are the same. This is probably due to the development of myelination and synaptic pruning in the same areas. Longer P1 latencies, smaller N1-P2 amplitudes, and late maturation patterns are possibly explained by the hypothesis that the neurons are still in refractory when the ACC occurs. However, this does not rule out the possibility that the ACC originate from a different set of neurons which are not responding to the onset, and merely develop with a similar rate but delayed in time. The significant slope differences between the onset P1 latency and ACC P1 latency with increasing age in NH listeners may suggest developments of different groups of neurons. 146

164 5.3. Effects of a Challenging Condition on ACC Responses The current study also documents the developmental patterns of CAEPs in background noise listening conditions. First, P1 latency decreased with age for the onset and the ACC in both NH and CI listeners. The slopes of P1 latency in quiet and in noise are not significantly different, suggesting the same age effect on P1 latency regardless of listening conditions. However, P1 latency is significantly longer in noise than in quiet in both NH listeners and CI users, indicating that background noise delays the responses uniformly across all age groups. Both the onset and the ACC go through similar developmental changes in morphology as they do in quiet conditions. Both onset and ACC responses are affected by background noise. This leads to the delayed development compared to quiet conditions. The N1-P2 amplitude is significantly smaller in noise for both onset and ACC responses. The morphologic changes occur later in noise than in quiet for both onset and ACC responses (Figure 27 and Figure 38) for both NH and CI users. The different time of maturation for the onset and the ACC shown in quiet conditions is even more obvious in noise. Interestingly, the N1-P2 amplitudes significantly grow with age only for the onset, but not for the ACC in noise conditions. This may suggest that even though the morphology of the ACC becomes adult-like, characterized by dominant N1 and P2 peaks, the N1-P2 amplitudes may reflect perception abilities rather than development. In CI listeners, the ACC is greatly affected by noise. The ACC in CI listeners remains moderate correlations relative to the NH adults after 14 years of age in noise conditions. Both pre-lingually and post-lingually deafened CI adults have significantly smaller ACC N1-P2 amplitudes than NH adults in noise. In short, CAEPs have the same 147

165 rate of P1 latency decrease in noise, but overall the development appears to be delayed due to background noise. This may be a reflection of their perceptual difficulties in discriminating between sounds in background noise. Since the post-lingual adults also show significantly smaller N1-P2 amplitudes in the ACC in both quiet and noise conditions compared to NH listeners, the poor correlations of ACC between CI users and NH adults may not be related to age but more to perceptual abilities due to limitations of implant processing. The pilot study used two instruments similar in timbre to construct a perceptually more difficult condition. This corresponds to the noise conditions in the current study. For the easy-to-distinguish sound, the pilot study used a big pitch change from approximately 250 Hz to 1500 Hz. This corresponds to quiet conditions in the current study. The delayed developmental patterns, especially in the ACC, with more difficult-todistinguish sounds were also seen in the pilot study using the musical sounds. This is encouraging for clinic and research applications using different levels of stimuli or listening conditions Implications of the Findings We successfully obtained two obligatory CAEPs, both onset and change responses, using identical procedures from young children to adults in both NH and CI users. Findings of the current study support that a cochlear implant can facilitate the development of the central auditory system in children who were born deaf but implanted early. The developmental patterns of the CAEPs in CI children are similar to those of NH listeners in quiet conditions. Both onset and ACC responses reach comparable maturation levels to NH adults in quiet. However, ACC responses are affected by the noise more 148

166 than the onset response; they do change with age but their maturation may not reach NH adults level. This is primarily due to significantly small amplitudes. Since the change responses are thought to be related to discrimination abilities, the poor morphology, smaller amplitudes, and longer latencies recorded from CI users in noise listening conditions may relate to their perceptual difficulty in noise compared to NH listeners. The knowledge of developmental patterns of the ACC can help researchers to study the ACC in both NH and CI children. Since we know that ACCs can be recorded in almost all CI users regardless of age, we can study the relationship between the ACC and behavioral measures, not only in a group but also on an individual level. Knowing a typical development model of the ACC in NH and CI listeners in quiet conditions is important to understanding differences among individual implant users. Finally, knowledge of how the ACC changes with age after CI activation in noise conditions can provide a foundation from which to study development of the central auditory system in challenging conditions in CI users compared to NH listeners Limitations of the Study Limitations of the current study include a small number of subjects in the CI group, in particular, a small number of young adult CI subjects. It was difficult to recruit early-implanted young adult CI users in their twenties and thirties. They are small in number since they were among the first generation of cochlear implant recipients. They also have full-time jobs away from Iowa City, Iowa. In NH listeners, P1 latency results suggest that onset and ACC responses may have different developmental patterns in both quiet and noise conditions. However, in CI listeners, while P1 latency was longer in the ACC than the onset in both listening conditions, results did not show different 149

167 development patterns between the onset and the ACC. This needs to be tested again with a greater number of CI subjects. Also, it was assumed in the study that post-lingually deafened adults have the acoustic listening experiences necessary to develop typical speech and language before implantation. The post-lingually deafened adults were used like a control group in the study. However, they may not have the same maturation levels that NH adults have. Speech-shaped noise alone was used at one SNR level. CI users are known to do well with the background noise of speech-shaped noise. When a babble noise is used in the background, the results, especially in CI users, may be different: poorer morphology and amplitudes are expected with the babble noise compared to those obtained with speech-shaped noise. Also, different SNR levels may show different developmental patterns between NH and CI users. Finally, we used a particular set of cortical responses among many other evoked potentials and other types of cortical measurement to document the developmental patterns on these two particular evoked potentials. These particular responses are thought to reflect detection and discrimination but certainly do not fully characterize cortical sound processing. Studies with different stimuli and/or recording techniques could provide further insights into auditory processing development. The results from the current study cannot fully generalize typical cortical sound processing for NH and CI listeners Future Directions The results from the current study open up further questions. A future study needs to recruit NH adults who are between 50 and 80 years old in order to explore if there are 150

168 further changes in the CAEP with age, especially in ACCs in noise conditions. Also, a future study needs to continue to recruit CI children, teenagers, and young adults who were implanted early to strengthen the findings we found in the study. CI users who were implanted later than 3.5 years old should be studied to explore how these developmental patterns are affected by the late electrical stimulation of sounds. For the bilateral CI users, the current study was done using only one ear with a CI received early or performs better. In a future study, it would be interesting to see developmental changes using the two CIs separately (via direct input connection) within subjects. A future study could explore the difference in performance among children and adults who have bigger amplitudes of the ACC and smaller amplitudes of the ACC at the same age. Different individuals with various ACC responses may develop at different rates; examining how different development rates relate to perception may provide further insights into developmental processes. Finally, this study could be a basis for investigating the relationship of these evoked potentials and perceptual performance in individual implant users. Since the ACC was feasible in young children, a study can be designed to facilitate the hearing and hearing aid evaluations, hearing aid fitting, and programming of CI speech processors. 151

169 5.7. Conclusions A cochlear implantation helps a deaf child access all speech sounds and learn to listen and speak. It is assumed to facilitate the development of auditory cortex, which is a prerequisite to acquisition of aural communication. This study investigated the developmental changes of two types of cortical auditory evoked potentials, related to detection and discrimination of sounds, in both NH listeners and CI users. The results in quiet are very encouraging. Pre-lingually deafened CI children show similar developmental patterns and rates in both CAEPs to those of NH children. In noise, however, NH children had overall delayed maturation for both CAEPs compared to quiet conditions, while responses in CI children were affected greatly by noise, especially for the ACC, which is thought to be related to sound discrimination. This study shows how the ACC may complement the onset response to document the development of the central auditory system in both NH and CI children. The results suggest that the ACC can be used as a measurement for estimating perceptual performance of individuals in various age and listening groups. 152

170 APPENDIX A: A PILOT STUDY ACOUSTIC CHANGE COMPLEX IN CHILDREN USING MUSICAL STIMULI 153

171 A.1. Overview Previous studies have shown consistent results on how development affects the P1-N1-P2 onset response during the first two decades of life. These studies used short ( ms) syllables, clicks, tone bursts, etc. While previous studies have also shown that ACCs are obtainable from both NH and CI adults; however, developmental effects on ACCs in children have not yet been studied. This pilot study focuses on the feasibility of recording the ACC in both NH and CI children and the effects of development on the P1-N1-P2 and the ACC using musical sounds. In this study, 800-ms-long duration of musical stimuli were used. Musical stimuli are important, everyday sounds that are complex to perceive and can be easily modified to create any acoustic change in pitch, timbre, and duration. In this pilot study, we created two pairs of musical stimuli to elicit CAEPs: one pair was considered to have an easy to distinguish change in pitch and the other pair was considered to have a difficult to distinguish change in timbre. All stimuli were presented via sound field as it represented everyday listening conditions and allowed researchers to use identical equipment and recording procedures for both NH and CI listeners. It was hypothesized that it would be possible to obtain ACCs in both NH and CI children and that the developmental trends of ACCs would be similar to those of onset responses, regardless of listening groups. ACCs obtained using the easy contrast stimuli were hypothesized to have larger amplitudes than ACCs obtained using the difficult contrast stimuli in both groups. However, NH listeners were expected to have better ACCs than CI listeners. 154

172 A.2. Methods A.2.1. Participants Both children and adults with NH and CIs participated. Nineteen NH children 5 to 19 years of age and five NH young adults 23 to 26 years of age participated. Six prelingually deafened CI users 12 to 21 years of age, and six post-lingually deafened adult CI users 25 to 80 years of age participated. All CI children received their Nucleus CIs before 3.5 years of age. Post-lingually deafened adults received their Nucleus CIs between 20 and 68 years of age. All CI listeners received their CI at the University of Iowa Hospitals and Clinics and were reported as good users by their audiologists. All subjects speech processors were fitted within 6 months of the study and were working appropriately at the time of data collection. Table A.1 shows the demographic information of participants in the pilot study. Table A.1. 1 Demographic information of subjects 155

173 A.2.2. Stimuli A series of musical stimuli were created. Each stimulus was 800 ms in duration and was composed of two 400 ms segments. The two segments were equalized in overall RMS level and edited digitally to insure there was no discontinuity that could create an audible click or pop in the middle of the acoustic stimulus. All stimuli were created in pairs to allow measurement of both onset and change responses for each segment of the stimulus: one stimulus with segment a presented before segment b and another stimulus with segment b presented before segment a. Each stimulus contained either a pitch or timbre change. The stimulus with a pitch change was designed to be an easy-to-perceive contrast; the stimulus with a timbre change was designed to be a more difficult-to-perceive contrast. There was an additional control condition where there was no change in pitch or timbre. This same set of stimuli was used to test all subjects regardless of their age or hearing status. For the pitch change stimulus, a clarinet played a note (B3) corresponding to a fundamental frequency of 247 Hz changing to a note (Gb6) corresponding to a fundamental frequency of 1472 Hz and vice versa. For the timbre change stimulus, the fundamental frequency remained the same (B3 for 800 ms), but the instrument (i.e., the harmonic distribution) changed from clarinet to oboe and vice versa. Additionally, one stimulus was created using the same method, but with no change in pitch or timbre. This stimulus consisted of clarinet playing B3 for 800. Figure A.1 shows the spectrograms for two stimuli that consist of a change in either pitch or timbre: (a) clarinet playing B3 to clarinet playing Gb6 and (b) clarinet playing B3 to oboe playing B3. 156

174 All stimuli were presented at a slow stimulation rate approximately 1 stimulus every 2.8 to 3.8 seconds. The ISI was varied randomly within 1 second to minimize the effects of adaptation. Figure A.1. Spectrograms for stimuli (a) This stimulus includes a pitch change half way through (400 ms after the onset) from clarinet playing B3 to clarinet playing Gb6. (b) This stimulus includes a timbre change half way through from clarinet playing B3 to oboe playing B3. A.2.3. Procedures Identical stimuli, presentation, and recording equipment were used for all subjects. CAEPs were recorded using stimuli presented in the sound field at 70 dba from a loudspeaker located at 0º azimuth approximately 4 feet from the listener. Participants sat in a comfortable chair and were encouraged to read, play with an ipad, or watch captioned videos in order to stay awake during testing. The researcher monitored the subject s status using two video cameras mounted in the sound booth. Surface electrodes with six channels to record CAEPs were applied to the scalp on the side of the head opposite the test ear. Differential recording bipolar montages include three active electrodes (vertex Cz, the contralateral temporal lobe Tc, and high forehead Fz) and two reference electrodes for each active electrode (the contralateral mastoid Mc, and the Inion Oz). Additionally, two electrodes were placed above and 157

175 lateral to the eye in order to monitor and reject eye blinks, which may have contaminated the CAEP recordings. A ground electrode for all recording channels was placed contralateral to the test ear to reduce any artifact contamination in recordings, especially for CI users. All subjects (CI or NH) had the same recording systems, including the electrode montages. For NH listeners, one ear was selected randomly as the test ear. The other ear was plugged and muffed. For Bilateral CI users, the better CI ear, reported by the subject was selected. An Intelligent Hearing System Opti-amplifier (IHS 8000) was used to record the ongoing EEG activity: a gain of 10,000 and band-pass filter (between 1 and 30 Hz). Custom designed LabView software was used to average responses prior to off-line analysis using MATLAB. The recording time window was 1200 ms. Recording started 100 ms prior to the sound onset. Two repetitions of 100 sweeps were obtained for all subjects. A.2.4. Statistical Analysis General descriptive statistics were used to provide a profile of how development impacts CAEPs in terms of the peak latencies, amplitudes, and morphologies. These were compared across age groups and listening groups. To see the stimulus effect on the CAEPs, latencies, amplitudes, and morphologies of CAEPs obtained using pitch change stimuli were compared to those using a timbre change stimuli in both listening groups. 158

176 A.3. Results Figure A.2 shows responses from NH listeners under the control condition, where no change response was expected. This figure highlights the onset P1-N1-P2 responses. ACCs were not obtained using the control stimulus. The data show that the method used in this study to create a long duration stimulus using two segments did not produce distortions that elicit the ACC. The results from NH listeners show that P1 latency decreases with age: starting at approximately 100 ms at age 5 to 80 ms by age 7 and to about 65 ms in young adults. N1 and P2 components emerge at about 7 to 8 years of age and by years of age these appear to dominate the waveforms. The results further indicate that findings using long duration musical stimuli are similar to those presented in previous studies using short duration stimuli (e.g., Ponton et al., 1996a). Figure A.2. Changes of the P1-N1-P2 with age in NH listeners A control stimulus with no change was used: clarinet playing B3 for 400 ms followed by clarinet playing the same note for another 400 ms. 159

177 Figure A.3 displays grand average waveforms obtained from NH individuals of different ages for both stimuli. Results show that both the onset P1-N1-P2 and the ACC are obtainable in NH subjects of different ages for both stimuli. ACCs follow similar developmental patterns to onset P1-N1-P2 responses regardless of stimulus type. Maturation in ACC, however, appears to be more delayed than the onset P1-N1-P2 in both stimuli. For example, N1 appears to emerge in the onset response at about 8 or 9 years of age while the same peak is not evident in the ACC for another year or two. When we focus on how ACC develops with age using different stimuli (one with a pitch change and another with a timbre change), results show that developmental effects on the ACC appear to be affected by the stimulus type. The ACC obtained using the stimulus with a pitch change, considered to be an easy contrast in this study, appears to mature earlier than the ACC obtained using stimulus with a timbre change. At 10 years of age the N1 component of the ACC is recorded when the stimulus with a pitch change is used. At the same age, however, N1 is absent and P1 is dominant when the stimulus includes a change in timbre. At 14 years of age the ACC becomes adult-like when the easy contrast stimulus is used; however, the morphology of the ACC remains immature when the difficult contrast stimulus is used. In adults, there were no differences in morphology, latency, or amplitudes of ACCs when responses obtained using both stimuli were compared. 160

178 Figure A.3. Developmental effects on the ACC in NH listeners Grand average waveforms are shown for normal hearing listeners aged from 5 to 26. The left panel shows responses obtained using (a) stimuli with a pitch change, considered to be an easy contrast, and the right panel using (b) stimuli with a timbre change, considered to be a difficult contrast. 161

179 Figure A.4 shows grand average waveforms from NH listeners in three age groups. These three age groups were created by grouping ages together based on the similarity of each age group shown in Figure A.3, in order to summarize the data and clarify the developmental effects on the onset P1-N1-P2 and the ACC with age. Regardless of the stimulus, the onset P1-N1-P2 has similar developmental patterns. P1 dominates the responses in 7-8 year olds, N1 and P2 are shown by 9-10 years, and the P1-N1-P2 response becomes adult-like by years. Developmental patterns of the ACC, however, are stimulus dependent. At 7-8 years, P1 dominates both ACCs obtained using both stimuli (pitch and timbre change). However, N1 begins to emerge when the easy contrast stimulus is used; N1 grows and dominates the response by 9-10 years. When the stimuli with a timbre change are used, P1 still dominates in the 9-10 year old group. By years, the ACC has three robust components in both stimuli conditions, however, the amplitude of N1-P2 is smaller when the more difficult to contrast stimuli were used. Figure A.4. Summary of the developmental effects on CAEPs in NH listeners Grand average waveforms are combined for three age groups. The left panel shows responses using stimuli with (a) a pitch change and the left panel with (b) a timbre change. 162

180 Using the same stimuli, CAEPs were recorded from CI users. Figures A.5 shows responses at different ages obtained from pre-lingually deafened CI users and postlingually deafened CI users. The ages listed here are in numbers of years post stimulation, i.e., hearing age. This figure is parallel to Figure A.2 obtained from NH listeners. First, the onset P1-N1-P2 was recorded from all CI subjects. Results from CI users indicate that the general developmental trends apparent in the NH data seem also to be evident in CI users as well. Figure A.5 shows the CAEP responses for the control condition, where no change response was expected. The onset P1-N1-P2 responses from CI users were similar to those of NH listeners. By 13 years post-stimulation, the response becomes adult-like. Figure A.5. Changes of the P1-N1-P2 with age in CI users A hearing age calculated after initial stimulation. Grand average waveforms are shown from the post-lingually deafened adults (n=6). Waveforms form the pre-lingually deafened children are individual data for each hearing age. A control stimulus with no change was used: clarinet playing B3 for 400 ms followed by clarinet playing the same note for another 400 ms. 163

181 Figure A.6 displays responses obtained from pre-lingually deafened CI individuals at different hearing ages and grand average waveforms from post-lingually deafened CI adults using the both stimuli. This figure is comparable to Figure A.3 obtained from NH listeners. ACC was obtainable in all CI subjects. The topmost line combines data from 6 post-lingually deafened CI adults. Older pediatric CI listeners show onset and ACC responses comparable to post-lingually deafened CI users and NH listeners when the easy contrast stimulus was used. The overall developmental trends apparent in the NH data seem to be evident in CI users. As in NH listeners, ACC develops more slowly than onset responses. ACC responses recorded from CI users, in response to the timbre change stimulus, are much smaller in amplitude than those recorded using a pitch change. For example, at ages 17 and 18, the ACC response to a pitch change is apparent but the ACC response is not observed for a timbre change in the 18 year old and is greatly reduced in the 17 year old. Using different stimulus contrasts did not greatly impact ACC amplitude in NH listeners but had a greater impact in CI listeners. This likely reflects the perceptual difficulty this contrast posed for CI users. 164

182 Figure A.6. Developmental effects on the ACC in CI users Grand average waveforms are shown for post-lingually deafened CI adults (n=6). Individual waveforms are shown for pre-lingually deafened CI children hearing aged from 10 to 18 years. The left panel shows responses obtained using (a) stimuli with a pitch change, considered as an easy contrast, and the right panel using (b) stimuli with a timbre change, considered as a difficult contrast. 165

183 Figure A.7 summarizes the data to clarify the developmental effects with age in CI users. This figure is comparable to Figure A.4 obtained from NH listeners. Some of the ages are grouped together based on similarity of responses. The onset P1-N1-P2 has similar developmental patterns for both stimuli, as seen in NH listeners. P1 dominates the responses at hearing age years and the P1-N1-P2 response becomes adult-like by 13 years. However, developmental patterns of the ACC are stimulus dependent, as was shown in NH listeners (Figure A.7 (a)). When stimuli with a pitch change are used, P1 dominates ACCs at years, N1 appear at 13 years and N1-P2 amplitude becomes robust by years of CI use. However, when stimuli with a timbre change are used, similar patterns occur but with much smaller amplitudes (Figure A.7 (b)). Figure A.7. Summary of the developmental effects on CAEPs in CI users Grand average waveforms are combined for three age groups. The left panel shows responses using stimuli with (a) a pitch change and the left panel with (b) a timbre change. 166

184 Figure A.8 compares results from three adult subject groups: NH (n=5), postlingually deafened CI adults who grew up with acoustic hearing and received a CI after 20 years of age (n=6), and pre-lingually deafened CI adults who received a CI before 3.5 year of age (n=2). Both onset and change responses are identifiable in the three adult subject groups. When musical stimuli with a pitch change were used, pre-lingually deafened CI adults had both onset and change responses comparable to NH and postlingually deafened CI adults. When musical stimuli with a timbre change were used, both pre-lingually and post-lingually deafened CI adults had robust onset responses but much smaller ACCs compared to NH adults. In other words, the effect of stimuli types was greater in both post-lingually and pre-lingually deafened CI users in regards to the ACC. Figure A.8. Comparisons among three adult subject groups Grand average waveforms obtained from normal hearing (n=5), post-lingually deafened adults (n=6), and pre-lingually deafened adults (n=2) are shown. The left panel shows grand average waveforms obtained using with stimuli (a) a pitch change and the right panel using with stimuli (b) a timbre change. 167

185 A.4. Discussion Using complex musical stimuli, this experiment obtained the onset P1-N1-P2 response and the ACC from children and adults with NH and CIs. The results show that despite different stimuli and recording procedures, the pattern in changes of P1 latency in the P1-N1-P2 response obtained in this study are similar to those found in previous published data. For example, among NH listeners, the P1 latency of onset P1-N1-P2 waveform decreased with age, from approximately 100 ms at age 5 to 80 ms by age 7 and about 65 ms by young adulthood (Figure A.2). While more variability is seen in CI children than NH subjects, the CI user with the hearing age of 13 shows that onset P1- N1-P2 responses become adult-like, with a robust N1-P2 responses, by this age (Figure A.5). This finding is comparable to previous published data (e.g., Kurtzberg et al., 1984; Novak et al., 1989; Pasman et al., 1991; Ponton et al., 1996a,b, 2000, 2002; Sharma et al., 1997, 2002a,b, 2004, for a review, see Wunderlich & Cone-Wesson, 2006). This experiment shows that ACCs can be evoked in both NH and CI children using a pitch or timbre change in musical stimuli. The pattern of change in the general morphology of the ACC is similar to that described previously: the decrease in the ACC P1 latency with age is similar to that described previously for the onset response. The ACC s change with age, however, appears to occur over a longer period of time than the onset response. The ACC response appears to mature later than the onset P1-N1-P2 response; N1 component appears later in ACC responses than in onset P1-N1-P2 responses for both NH and CI users. ACCs are somewhat stimulus-dependent in both NH and CI listeners. The ACC recorded using the stimulus with a timbre change seems to mature more slowly than the 168

186 ACC response recorded using the stimulus with a pitch change. For example, at 10 years of age in NH listeners, the N1 component of the ACC is recorded with the pitch change stimulus (Figure A.3 (a)), but the P1 component of the ACC is still dominant with the timbre change stimulus (Figure A.3 (b)). ACCs recorded in CI users (Figure A.6) show similar developmental trends affected by the stimulus to those shown in the NH data. A primary difference in developmental patterns of ACCs between NH and CI participants is the amplitude of ACC, especially in the more difficult stimulus condition. ACCs recorded from CI users in response to the timbre change stimulus are much smaller in amplitude than those recorded using a pitch change; this is true across all age groups. This trend is not apparent in the data from NH listeners. When N1 and P2 peak-to-peak amplitudes are measured, NH subjects older than 16 years show amplitudes that are not affected by the stimulus type (Figure A.3). However, N1 and P2 amplitudes of ACCs recorded from CI users with years of CI use are significantly affected by the stimulus type (Figure A.6). The effect of stimulus type is greater in both pre-lingually and post-lingually deafened CI users (Figure A.8). CI users had much smaller ACC amplitudes in the difficult-to-distinguish stimulus condition. This likely reflects the perceptual difficulty this contrast poses for CI users. This may not be due to arrested development. However, this pilot study was limited in subjects, it is therefore challenging to fully understand the developmental effects on ACCs, especially in CI users, and the differences in ACCs between NH and CI listeners especially those with more challenging conditions. Based on this pilot study, the current experiment is designed to include more subjects from both NH listeners and CI users, to use speech stimuli, and to create difficult 169

187 conditions by recording CAEPs in background noise. Together with the pilot study using musical complex stimuli, results from the current study will provide more toward understanding the development of ACCs in both NH and CI children. 170

188 APPENDIX B ADDITIONAL TABLES AND FIGURES FROM THE CURRENT STUDY 171

189 Table B.1 shows means and standard deviations of responses measured from NH adults (n=11). It also shows the results of paired t-tests comparing responses in the two listening conditions. Table B.1. NH Adults CAEPs: Quiet vs. Noise Table B.2 shows means and standard deviations of peak latencies and N1-P2 peak-to-peak amplitudes for CAEPs recorded from all NH listeners in quiet conditions. Table B.2. Peak Latencies and N1-P2 Amplitudes: NH listeners in Quiet 172

190 Table B.3 shows results of linear regression analyses on Pl latency with age and the N1-P2 peak-to-peak amplitude with age for both onset and ACC responses. Data used for these analyses were collected from NH listeners in quiet listening conditions. Table B.3. Linear Regression on P1 latency and N1-P2 Amplitude with Age in Quiet In NH Listeners Table B.4 shows means and standard deviations of peak latencies and N1-P2 peak-to-peak amplitudes for CAEPs recorded from NH listeners with background noise, speech-shaped noise at 10 db SNR. Table B.4. Peak Latencies and N1-P2 Amplitudes: NH listeners in Noise 173

191 Table B.5 shows results of linear regression analyses on Pl latency with age and the N1-P2 peak-to-peak amplitude with age for both onset and ACC responses. Data used for these analyses were collected from NH listeners in noise listening conditions. Table B.5. Linear Regression of P1 latency and N1-P2 Amplitude with Age in Noise In NH Listeners Table B.6 shows means and standard deviations of responses measured from CI adults (n=15). It also shows the results of paired t-tests comparing responses in the two listening conditions. Table B.6. CI Adult s CAEPs: Quiet vs. Noise 174