Age-related changes in temporal resolution revisited: findings from cochlear implant users

Transcription

1 University of Iowa Iowa Research Online Theses and Dissertations Spring 2016 Age-related changes in temporal resolution revisited: findings from cochlear implant users Bruna Silveira Sobiesiak Mussoi University of Iowa Copyright 2016 Bruna Silveira Sobiesiak Mussoi This dissertation is available at Iowa Research Online: Recommended Citation Mussoi, Bruna Silveira Sobiesiak. "Age-related changes in temporal resolution revisited: findings from 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 AGE-RELATED CHANGES IN TEMPORAL RESOLUTION REVISITED: FINDINGS FROM COCHLEAR IMPLANT USERS by Bruna Silveira Sobiesiak Mussoi 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 Carolyn J. Brown

3 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 Bruna Silveira Sobiesiak Mussoi 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: Carolyn J. Brown, Thesis Supervisor Paul J. Abbas Melissa C. Duff Christopher W. Turner Yu-Hsiang Wu

4 To Marcelo M. B. Mussoi ii

5 Live as if you were to die tomorrow. Learn as if you were to live forever. Mahatma Gandhi iii

6 ACKNOWLEDGEMENTS Financial support for this project was provided by the American Academy of Audiology under a Student Investigator Research Grant Award, and by the National Institutes of Health, Institute on Deafness and Other Communication Disorders under award P50DC Bridget Zimmerman and Jake Oleson provided statistical support. A special thank you is owed to my husband and children, who have showered me not only with love and support but also with perspective. I would also like to thank Carolyn Brown and Paul Abbas for their guidance during this project. Finally, thanks are owed to my friends and colleagues for their encouragement. iv

7 ABSTRACT A decline in temporal resolution, or the ability of the auditory system to track fast changes in incoming sounds, is one factor thought to contribute to difficulties in speech perception that accompany the aging process. Aging effects on gap detection abilities, using behavioral or isolated electrophysiologic measures, have been studied previously. However, peripheral and central electrophysiological, and behavioral measures of temporal resolution have not been examined in the same subjects. Also, the relationship between age-related changes in temporal resolution and speech perception is still unclear, as is their interaction with cognition. By revisiting this question in CI users, it was possible to study aging effects on temporal resolution without the potential confound of age-related hearing loss. In addition, the device allows for manipulations of the temporal properties of a signal without concomitant changes in its spectrum, and for auditorynerve recordings. This study had two main goals: (1) to determine how aging affects temporal resolution at the auditory periphery, the cortex and perceptually; and (2) to explore the relationship between age-related changes in temporal resolution, general cognitive functioning and speech perception. Results showed that when the auditory system is stimulated with a cochlear implant, few effects of advancing age on temporal resolution are evident. It is possible that, by stimulating the auditory nerve with precise timing, cochlear implants can help users overcome temporal resolution deficits. Alternatively, and perhaps more likely, it is possible that previous studies that reported age effects on temporal resolution were v

8 largely influenced by differences in peripheral processing, which were minimized in this study by the use of a cochlear implant. Across the age groups, digit span was the only variable significantly correlated with speech perception in noise and perception of time-compressed speech. A longer memory span for digits was associated with better outcomes in both tests of speech perception. This finding is consistent with previous research, and underscores the notion that cognitive factors, not age, may be more important for speech perception. vi

9 PUBLIC ABSTRACT The aging process is accompanied by difficulties in speech perception, particularly in background noise. Temporal resolution, or the ability track fast changes in sounds, is thought to be a contributor of such difficulties. This study investigated whether there are age-related changes in temporal resolution in younger and older cochlear implant users. Cochlear implants are surgically implanted devices that allow listeners with severe-to-profound hearing loss to achieve various degrees of improved hearing. Another goal of this study was to examine the relationship between speech perception, temporal resolution and measures of cognition. Results showed that only one of the measures of temporal resolution was affected by advancing age. Also, a form of short-term memory, not age or temporal resolution, was significant associated with better speech perception in noise and of fast speech. As the population ages, these findings are important in the counseling and treatment of those with hearing-related complaints. vii

10 TABLE OF CONTENTS LIST OF TABLES... ix LIST OF FIGURES...x LIST OF ABBREVIATIONS... xi CHAPTER 1 INTRODUCTION...1 Background...1 Auditory Temporal Resolution...2 The Impact of Hearing Loss...3 Proposed Study...4 CHAPTER 2 REVIEW OF LITERATURE...7 Aging and Speech Perception...7 Aging and Temporal Resolution: Psychophysical Studies...11 Aging and Temporal Resolution: Electrophysiological Studies...17 Aging, Hearing and Cognition...29 CHAPTER 3 METHODS...34 Study Participants...34 Study Overview...35 General Procedures...35 CHAPTER 4 RESULTS...51 Demographics...51 Experiment 1: Age-Related Changes in Temporal Resolution...52 Experiment 2: Behavioral Measures and Relationships with Speech Perception...59 CHAPTER 5 DISCUSSION...77 REFERENCES...87 viii

11 LIST OF TABLES Table 1. Participant demographic and CI information...49 ix

12 LIST OF FIGURES Figure 1. Schematic of stimuli used in electrophysiologic testing...50 Figure 2. ECAP refractory recovery functions Younger CI users Figure 3. ECAP refractory recovery functions Older CI users Figure 4. ECAP refractory recovery functions Group comparisons...66 Figure 5. ECAP pulse train recovery functions Younger CI users...67 Figure 6. ECAP pulse train recovery functions Older CI users Figure 7. ECAP pulse train recovery functions Group comparisons...69 Figure 8. Cortical grand mean waveforms and individual responses 30 ms gap...70 Figure 9. Cortical latency-shifted grand mean waveforms Group comparison...71 Figure 10. Cortical onset N1 and P2 amplitudes and latencies Group comparisons...72 Figure 11. Cortical normalized ACC amplitudes and behavioral gap detection thresholds Group comparisons...73 Figure 12. Behavioral tests: Speech perception and Listening Effort Group comparisons...74 Figure 13. Behavioral tests: Cognition Group comparisons Figure 14. Significant relationships with speech perception measures...76 x

13 LIST OF ABBREVIATIONS AAA.... American Academy of Audiology ABR.. Auditory Brainstem Response ACC....Acoustic Change Complex ANOVA.. Analysis of Variance ASSR.. Auditory Steady State Response BKB.... Bamford-Kowal-Bench sentence test cabr...complex Auditory Brainstem Response CHABA...Committee on Hearing, Bioacoustics and Biomechanics CI Cochlear Implant C-level..Most Comfortable level CL....Current Level CNC...Consonant-vowel Nucleus-Consonant CUNY...City University of New York test EABR....Electrically Evoked Auditory Brainstem Response ECAP....Electrically Evoked Compound Action Potential EEG....Electroencephalographic HINT...Hearing in Noise Test MLR.Middle Latency Response MMN....Mismatch Negativity MMSE. Mini Mental State Examination MPI Masker-Probe Interval NIC.Nucleus Implant Communicator QuickSIN....Quick Speech In Noise test SNR. Signal-to-Noise Ratio SRT..Speech Reception Threshold xi

14 T-level..Minimum audible level VOT.. Voice-Onset Time xii

15 1 CHAPTER 1 INTRODUCTION Background Age-related hearing loss, or presbycusis, is the third most prevalent chronic condition in older adults, affecting one in four people 70 years and older (Dillon et al, 2010). In 2010, persons 65 years and older comprised 13% of the total population in the US, and their share is expected to reach 20% by 2030 when the last of the baby boomers reach the age of 65 (US Census Bureau, 2011). Many studies have suggested that, as a group, older adults have more difficulty in challenging listening situations such as understanding speech in fluctuating noise or in highly reverberant environments (Humes & Dubno, 2010) and their word understanding is often much poorer than what hearing thresholds would suggest. Over two decades ago, a seminal report on speech understanding and aging was published by the Committee on Hearing, Bioacoustics and Biomechanics (CHABA, 1988). Among the topics identified by the working group as needing further attention was the impact of peripheral and central processing on speech understanding in younger and older adults with similar amounts of hearing loss. This group also pointed to the importance of determining how age-related changes in cognition affect speech understanding. In the years that followed the CHABA report, many studies attempted to address the effects of advancing age on speech perception but rarely was attention given to assessing the relative contribution of neural processing at peripheral and central levels in the human auditory system. More recently, the American Academy of Audiology (AAA) Task Force on Central Presbycusis published a report concluding that Peripheral-auditory, central auditory, and cognitive factors are intertwined and difficult to disentangle using behavioral methods from older adults (Humes et al, 2012). They recommended use of

16 2 objective measures in addition to behavioral measures of age-related changes in the auditory system that focused on use of non-speech stimuli. It is this gap in the knowledge base that the current study sought to fill. More specifically, this study explored the effect of advancing age on measures of temporal resolution. Behavioral measures of temporal resolution were compared with electrophysiologic measures of temporal processing obtained from both peripheral and central portions of the auditory system. In addition, correlations between those behavioral and electrophysiologic measures of temporal resolution and measures of speech perception and cognition were explored. Auditory Temporal Resolution The term temporal resolution refers to how accurately changes in the temporal properties of an acoustic signal are perceived by a listener. Perception of temporal cues in an ongoing signal is important because much of the information needed to understand speech is conveyed not in the steady parts of the signal, but in the temporal variations themselves (Moore, 2003). Investigators have long reasoned that age-related changes in temporal resolution may be one factor that contributes to the speech perception difficulties experienced by the elderly. Temporal resolution has been studied using a range of different experimental paradigms including gap detection, modulation detection, and duration discrimination (Moore, 2003). Most employ psychophysical methods, but evoked potential techniques can also be used to determine how temporal cues are coded in the auditory system. All of these approaches are complicated by the fact that changing an acoustic signal in the time domain produces changes in the frequency domain. Additionally, older adults typically present with varying degrees of hearing loss. Both factors present challenges to investigators who have attempted to explore the relationship between aging and temporal resolution in the auditory system.

17 3 The Impact of Hearing Loss Presbycusis is a term used to describe hearing loss associated with the aging process. Noise and chemical exposure, as well as lifestyle and genetic predisposition to hearing loss can all have a cumulative and adverse effect on hearing. A major challenge in exploring how hearing mechanisms change with advancing age is that it is difficult to separate the effects of age from the effects of hearing loss. It should also be noted that while many of these factors primarily affect the peripheral auditory structures, it is possible that advancing age and/or reduced peripheral input could lead to degeneration of higher order structures within the auditory system. Researchers studying age-related changes in temporal resolution have attempted to control for the effects of peripheral hearing loss by (1) selecting older adults with normal hearing as subjects, (2) matching hearing thresholds between groups of younger and older adults, (3) masking high frequencies in younger adults in order to simulate the threshold shift characteristic of presbycusis, and (4) selecting subjects with a variety of degrees of hearing loss in both age groups in order to eliminate correlations of a specific hearing loss with age, then accounting for threshold effects statistically. None of these methods is ideal. The most commonly used method of controlling for hearing loss in studies of aging is the first method described above recruiting older adults with normal hearing (the golden ears ). However, it can be argued that older adults with normal hearing are not representative of the general population. Also, the fact that these elderly adults have normal hearing suggests that they may not be undergoing the typical aging process. In addition, the definition of normal hearing itself is based on studies conducted several decades ago, which tested mostly male subjects and did not explicitly rule out the effects of noise exposure. In fact, data from a recent national survey showed that median high frequency hearing thresholds are significantly lower (i.e., better) than those reported in the 1960s and used as the basis for current international (ISO-1999, 1990) and national (ANSI S ) hearing threshold standards (Hoffman et al., 2010, 2012).

18 4 Furthermore, it is possible that even a small amount of hearing loss can have effects on the auditory system that are not quantifiable by pure-tone audiometry, as evidenced by findings of reduced performance with filtered speech in low-to-mid frequency regions of normal to near-normal hearing (Lèger et al, 2012). This can also happen with the threshold matching approach, as matching is usually done in a restricted frequency range. Masking of high frequency thresholds in the younger cohort to match thresholds to the older group seems to be even less appropriate, as threshold elevation is not the only effect of hearing loss. Finally, using statistical methods to account for varying degrees of hearing loss may require the use of large numbers of subjects in order to obtain sufficient statistical power. Proposed Study This study examined age-related changes in temporal resolution in cochlear implant (CI) users. Cochlear implants are surgically implanted devices that bypass the outer and middle ears as well as the cochlear hair cells and stimulate the auditory nerve directly. As surgical techniques, devices and speech processing strategies have evolved in the past two decades, increasing numbers of older adults are receiving a cochlear implant. These individuals often are able to obtain levels of performance comparable to their younger counterparts on measures of speech perception (Kelsall et al, 1995; Orabi et al, 2006; Williamson et al, 2009) and their scores on these tests appear to remain stable over time, both in quiet and in noise (Dillon et al, 2013). We have opted to study the impact of aging on temporal resolution in cochlear implant users for several reasons. First, all traditional cochlear implant users, regardless of their age, have similar hearing prior to surgery. That is, they all have severe to profound sensorineural hearing loss. It seems reasonable to assume that the peripheral auditory structures will be similarly affected by hearing loss for older and younger listeners, particularly if these two groups present with similar durations of profound

19 5 hearing loss and etiologies of deafness. This assumption is supported by findings of similar recovery from forward masking measured using the electrically evoked compound action potential (ECAP) in younger and older adults with cochlear implants (Lee et al., 2012). By focusing on cochlear implant users rather than individuals with different degrees of acoustic hearing, it should be possible to assess the impact of advancing age on auditory temporal resolution without confounds of degree of hearing loss. A second reason to study the effect of age on temporal resolution in cochlear implant users is that the device is equipped with two-way telemetry systems which allow for stimulation of the auditory nerve and recording of the neural response in a noninvasive way. Another advantage of working with cochlear implant users is that for this group of listeners it is possible to control the implanted electronics directly. The place in the cochlea that is stimulated is determined by which electrode is activated (Chatterjee et al., 1998); therefore spectral splatter is likely not an issue. Finally, because spectral information provided by the device is limited, success with a cochlear implant is likely to depend more heavily on the effective use of temporal cues. Research Questions and Hypotheses The goals of this study were to explore age-related changes in temporal resolution in cochlear implant users at both peripheral and central levels of the auditory system, to compare those results to behavioral measures of temporal resolution and relate that information to cognition and speech perception. More specifically, the following questions were addressed: (1) Do older listeners with cochlear implants exhibit poorer temporal resolution abilities than their younger counterparts? (2) Can temporal resolution measures and/or cognitive factors explain the variability in performance exhibited by cochlear implant users in terms of speech perception in noise?

20 6 To answer the first question, peripheral and central electrophysiological measures of temporal resolution were obtained and compared with behavioral measures of temporal resolution in two groups of younger and older CI users. To answer the second question, the electrophysiological and psychophysical temporal resolution measures were combined with scores on tests of cognitive functioning and correlated with performance on tasks requiring perception of speech in background noise and perception of fast speech. This study will test the following hypotheses: (1) The group of younger CI users will show better temporal resolution than their older counterparts in the psychophysical and electrophysiological measures, as evidenced by smaller gap detection thresholds, faster ECAP recovery and larger ACC amplitudes. (2) Temporal resolution and cognitive processing measures will be significantly correlated with speech perception, in that listeners with better temporal resolution and higher cognitive functioning will exhibit better speech perception scores. Results of the current study will offer a novel perspective on how age-related changes in temporal resolution affect hearing in humans. The results of this work will broaden our understanding of how advancing age impacts the way temporal cues are coded at different levels of the auditory system and may help us understand the extent to which changes in temporal resolution associated with aging are related to speech perception.

21 7 CHAPTER 2 REVIEW OF LITERATURE Aging and Speech Perception Older adults generally have more difficulty understanding speech than their younger counterparts. This is especially true in challenging listening situations such as in the presence of background noise or when the environment is highly reverberant (CHABA, 1988). Along with the aging process, presbycusis, a characteristic high-frequency hearing loss, also poses a significant problem for older adults. Presbycusis can act as a confounder in studies of the effects of aging on the auditory system, and investigators have used a range of different experimental methods to circumvent this problem. The results have been varied. Several studies have not found age effects on speech recognition beyond those posed by hearing loss. Souza & Turner (1994), for example, carefully matched hearing thresholds up to 4000 Hz, and used a high-pass masker to prevent listeners from using speech information in the unmatched high frequencies. They found no effects of age on word recognition in different types of background noise, and both groups of younger and older hearing-impaired listeners performed more poorly than a group of younger subjects with normal hearing. Likewise, Takahashi & Bacon (1992) showed that differences in the perception of sentences in modulated background noise were mostly related to hearing thresholds, and not age. Hearing loss has also been found to be the strongest predictor of speech perception in quiet and in low levels of background noise in a large group of elderly listeners, among several tests of auditory function and cognition (Humes et al, 1994). This was also the case for SRTs recorded in high-level noise (van Rooij & Plomp, 1990), although cognitive factors accounted for an additional third of variance in speech perception.

22 8 In an attempt to explain these findings, researchers have argued that aging effects should be more salient with complex stimuli, such as sentences rather than words, and fluctuating background noise as opposed to steady-state (Pichora-Fuller & Souza, 2003). Several studies support this theory. For example, Dubno et al (1984) used an adaptive procedure to find speech reception thresholds (SRTs) and showed both age and hearing loss effects for spondees and sentences presented at different levels in multitalker babble. Using similar procedures, Frisina & Frisina (1997), reported age effects when testing younger and older adults that were matched on the basis of hearing thresholds and SRTs in quiet. Effects of age beyond those of hearing loss were also were also found for the perception of time-compressed and reverberated sentences (Gordon-Salant & Fitzgibbons, 1993, 1999). In all of these studies, undistorted speech recognition in quiet was only affected by hearing loss. Another reason proposed for the lack of age effects is that surveyed subjects may not be old enough. Humes & Christopherson (1991) tested a group of young old adults (aged years) and a group of old old adults (aged years) and compared their performance to that of group of young subjects with normal hearing and simulated hearing loss. Although they concluded that hearing loss was the main factor underlying the speech difficulties of their older subjects, they also found that the group of old old adults performed more poorly than the group of young old adults, despite having similar audiometric thresholds. These studies seem to show that while hearing loss can act as a confounder and complicate interpretation of the results, age related changes in speech perception can occur beyond those that may result from presbycusis per se. These effects are likely more evident when complex tasks are used, and may be related to changes in cognition that accompany the aging process.

23 9 Electric Hearing The effects of advancing age on speech perception have also been assessed in cochlear implant users. Reports of performance with early versions of cochlear implants, which processed speech in crude ways, showed no differences in speech recognition in quiet between younger and older CI users with matched (Kelsall et al., 1995; Labadie et al., 2000) and unmatched duration of deafness (Shin et al, 2000). These findings are corroborated by a large study that looked at CNC word scores in 627 cochlear implant users years old and showed no correlation of those scores with age at implantation (Leung et al, 2005). The rate of improvement of speech perception from pre-operative to 3, 12, and 24 months post-implantation is also unaffected by age (Budenz et al, 2011; Lenarz et al, 2012 [for speech perception in quiet]). Two studies did report that age was negatively correlated with speech perception in quiet (Chatelin et al, 2004 [for CNC words only]; Carlson et al, 2010 [for AzBio sentences only]), however, these results should be interpreted with caution, as both studies also did further speech perception testing in quiet and in noise, and found no evidence of declining performance with advancing age. Researchers have also explored the speech perception abilities of CI users in more challenging listening situations. In general, no difference in performance between younger and older CI recipients has been found for BKB sentences presented in +10dB SNR pink noise (Orabi et al., 2006), HINT sentences in +10dB SNR speech-shaped noise (Poissant et al, 2008), and CUNY sentences in +10dB SNR (unspecified noise) (Budenz et al, 2011). Interestingly, when comparing two groups of younger and older adults that had been matched on pre-operative sentence scores (with several subjects showing floor performance) and duration of deafness, Friedland et al (2010) reported significantly better 12-month post-implantation performance in the younger group for speech perception in quiet (CNC words, HINT sentences), but no difference for HINT sentences in speechshaped noise (SNR not provided). However, when a less favorable SNR of -10dB was

24 10 used, older CI users had consistently poorer sentence recognition in speech-shaped noise than their younger counterparts, up to 24 months post-implantation (Lenarz et al, 2012). A limitation of all the CI studies outlined above is that they were retrospective in nature, limiting study design and data analyses due to missing data points. Several of those studies lacked important details in the characterization of the subject groups and stimuli used. In fact, speech perception was not the main focus of many of those studies, but just one of several outcomes examined. Also, the choice of fixed SNRs may not have been optimal for investigating age effects. Finally, cochlear implant users typically show a wide range of performance regardless of age, which could have prevented researchers from finding significant effects of older age. An alternative approach that has been used to investigate age effects in electric hearing is to simulate the output of cochlear implants using noise-vocoded strategies in listeners with acoustic hearing. Vocoding involves sampling acoustic signals in short time intervals, filtering into pre-determined frequency bands and extracting the envelope from each band. This is used to modulate broadband noise that is filtered into the same frequency bands. Studies have shown that increasing the number of frequency bands improves speech perception (Shannon et al, 1995), especially for more complex stimuli [up to 7-10 bands] (Friesen et al., 2001). Using these techniques, Sheldon et al (2008a) found that younger and elderly listeners were equally able to understand open-set vocoded words with differing number of frequency bands when a gating paradigm was used, providing repetition of the words. However, when not allowed the opportunity of repetition (and feedback on responses), older adults needed the temporal envelope of a larger number of frequency bands to obtain the same speech perception as their younger counterparts. A similar conclusion of age effects on the number of bands required for speech perception was reached in their companion paper, independently of significant benefits of priming and context (Sheldon et al, 2008b). However, the older age group had higher high frequency thresholds and

25 11 those effects did not seem to have been ruled out by the authors. In contrast, Souza & Boike (2006) recruited participants with a variety of degrees of hearing loss and found that their ability to understand vocoded consonants with different numbers of channels was related to their age, not hearing loss. Older and younger adults were equally able to integrate temporal envelope cues from an increasing number of frequency bands. These results are corroborated by those of Schvartz et al (2008), who also found that when a frequency-place mismatch was simulated in normal hearing listeners, middle-aged and older participants were more negatively affected than their younger counterparts. In sum, there is mixed evidence from cochlear implant users to support aging effects in speech perception skills. This conclusion may be related to study design and choice of stimuli used. In more controlled studies using vocoded stimuli, results suggest that older adults may not be able to use the temporal information provided by cochlear implants to understand speech as well as their younger counterparts. Aging and Temporal Resolution: Psychophysical Studies Historically, age-related decrements in temporal resolution abilities have been postulated as a possible underlying mechanism for the speech perception difficulties observed in the elderly (CHABA, 1988). Different psychophysical methods have been used to assess temporal resolution. Gap detection is a widely used experimental paradigm in assessments of age-related changes in temporal resolution. It consists of placing a silent gap between two noise or sinusoidal markers, and varying the size of the gap until the smallest detectable gap is found (i.e., the gap threshold). A problem with the gap detection task is that introduction of a gap into a signal that is not broadband will result in spectral splatter particularly if short rise/fall times are used. That means listeners can use spectral rather than [or in addition to] temporal cues to detect the presence of the gap (Moore, 2003). A second approach that can be used to study temporal resolution is to assess the ability of the listener to detect changes in the duration of a sound, when

26 12 discriminating smaller duration differences is taken as a sign of better temporal resolution. Duration discrimination is the technique that has been used most infrequently to study the effects of aging on temporal resolution, likely because the underlying mechanisms involved in this task are still poorly understood (Moore, 2003). A third method used in investigations of temporal resolution is to assess how easily an individual is able to perceive small intensity fluctuations in an ongoing stimulus. This is usually accomplished by finding the smallest modulation depth that can be detected in an amplitude-modulated stimulus. Across all paradigms, there are some studies that show age-related declines in temporal resolution as measured behaviorally (e.g. Snell, 1997; Schneider and Hamstra [for short marker durations only] (1999), Strouse et al, 1998; Gordon-Salant & Fitzgibbons, 1999; He et al, 2008; Humes et al, 2010), while others do not (e.g. Moore et al, 1992; Takahashi and Bacon, 1992). These studies are subject to the hearing loss confounder, in addition to the difficulty in assessing temporal resolution while keeping spectral changes at a minimum. The present study will focus on gap detection, a widely used measure of temporal resolution. The following section will review studies that used acoustic stimuli to explore the effects of older age on gap detection abilities. Next, studies that used electrical stimulation to investigate gap detection abilities of cochlear implant users will be reviewed. Acoustic Hearing In an early study of age-related changes in gap detection, Moore et al (1992) used 500 ms sinusoids in low-passed white noise of various frequencies and levels to investigate the gap detection abilities of elderly subjects with hearing loss and with normal hearing up to 2 khz. They found that the distributions of gap detection thresholds were largely overlapping between those two groups of elderly listeners and a group of

27 13 younger listeners with normal hearing from a previous study. This led the authors to conclude that although the poorer performance in both groups of elderly subjects suggested the presence of age effects, those were likely due to a few outliers. In this study, the authors disregarded the presence of hearing loss above 2 khz in the group of elderly subjects with normal hearing because the test signals were restricted to the lower frequency range. However, there is evidence that high frequency hearing loss can also disrupt perception of lower frequencies (e.g., Leger et al, 2012). In contrast, following careful matching of thresholds (up to 4 khz) between younger and older listeners, Snell (1997) found significant age effects for all gap detection conditions tested. These effects were independent of thresholds at the unmatched frequencies of 6 and 8 khz. Stimuli consisted of 150 ms low-pass noise bursts with 1 or 6 khz cutoff frequencies presented at two intensities, with two levels of modulation and three background noise conditions. When the author examined the distribution of gap detection scores for both groups, there was also overlap in the distributions with a larger range of gap detection abilities in the older age group. However, due to the fact that some older adults have exceptionally poor gap detection abilities and some younger subjects have exceptionally good gap detection thresholds, the conclusion was that there is a shift in the distribution of gap detection thresholds with age [differently than Moore et al 1992]. In order to address these discrepancies, Schneider and Hamstra (1999) measured the effects of marker duration on gap detection thresholds of younger and older adults. They found age effects for markers up to 200 ms (agreeing with Snell, 1997) but not for longer-duration markers such as 500 ms (similar to the findings of Moore et al, 1992). However, it should be noted that, even though there was no correlation between gap detection thresholds and audiometric thresholds at 2 khz, older listeners had more hearing loss at higher frequencies, which was not taken into account.

28 14 Humes and colleagues attempted to address issues in previous studies by conducting a large-scale study of temporal resolution across three different sensory modalities (hearing, vision and touch). Gap detection was among several measures of auditory temporal resolution, and was examined with 400 ms narrow bands of noise centered at 1 and 3.5 Hz, for subjects with audiometric thresholds up to the moderate-tosevere hearing loss range. In order to minimize the effects of presbycusis, the authors introduced broadband noise with the gapped stimuli, and used very high presentation levels (91 db SPL). In a preliminary report (Humes et al, 2010), no effects of age on gap detection thresholds were evident for 339 young, middle-aged and older adults, once hearing loss was factored out. Investigators have also proposed that age effects on temporal resolution may be more evident in challenging tasks. For example, He et al (1999) investigated the effects of gap location within the noise markers, as well as location uncertainty (fixed versus random gap location) and marker duration, for low-pass filtered noise bursts. Because of the large variability in gap detection thresholds reported in previous studies, the authors used a constant stimuli procedure to obtain the complete psychometric functions. Hearing thresholds were closely matched across the audiometric frequency range. However, only six subjects in each age group (young, old) were tested in each experiment. Results showed age effects in the random condition only when the gaps were located at the edges of the noise markers. For the two marker durations (100 and 400 ms), there were no age effects on the slope and threshold of the psychometric functions, which differs from the findings of Schneider & Hamstra (1999). Using similar procedures, Harris et al (2010) investigated the effects of gap location within 500 ms wide band noise bursts and location uncertainty, as well as their effects on perceived workload. The authors also explored the relationship between those challenging gap detection tasks and cognition. Results showed age effects when the gap was fixed at the beginning of the noise interval (differently from He et al, 2009), but not

29 15 when it was located in the middle or end of the noise. Like He et al (2009), when gap location was randomly varied, age effects were seen for gaps at both the beginning and end of the noise. Interestingly, measures of processing speed and gap detection thresholds in the fixed gap location conditions explained a large part of the variability in the random-location gap detection and after controlling for these factors, age was no longer a significant predictor of gap detection in the random conditions. Regarding perceived workload of gap detection, there was an age-related increase in mental demand only [out of six subscales on the NASA Task Load Index questionnaire (Hart & Staveland, 1988)]. Lister et al (2002) took another approach to making the task more challenging. They used leading and trailing markers surrounding the gap of different frequencies (i.e., across-channel gap detection). The leading narrow-band noise was fixed at 2 khz and the trailing marker was varied between 0.5 and 3 khz, for a total duration of 500 ms (including the gap). They found that older adults had poorer gap detection thresholds than younger and middle-aged adults, and these effects were unrelated to the differences in detection thresholds for the narrow bands of noise between the groups. It should be noted that since the authors kept the total duration of the stimuli constant, listeners could have used marker duration cues to perceive the gaps. In a follow-up study, Lister & Tarver (2004) investigated across-channel gap detection using synthetic speech sounds of varying duration as the markers. Despite efforts to match the two age groups in hearing thresholds, the older age group had significantly higher high frequency audiometric thresholds. In fact, the significant age effects observed for gap detection across the conditions tested did not hold once hearing sensitivity was taken into account. In sum, aging seems to affect the gap detection abilities of at least some listeners, beyond what would be expected as a result of hearing loss. Yet, several investigators have pointed out that many elderly listeners exhibit gap detection thresholds that fall within the range of their younger counterparts.

30 16 Electric Hearing A few studies have investigated the gap detection abilities of cochlear implant users, but most did not focus on age effects. For example, Shannon (1989) used biphasic pulses between 0.5 and 2 khz with Nucleus 22 users, and 1 khz sinusoidal and pulsatile stimuli with Ineraid users. The author reported that gap detection thresholds in cochlear implant users were in the normal hearing range (when high-level stimuli are used), regardless of stimulation type (sinusoids, pulse trains) and electrode stimulated (basal, apical). Busby & Clark (1999) tested early deafened Nucleus 22 CI users with a wide range of age at the time of implantation (3.5 to 20 years). They used 500 and 1000 ms pulse trains presented at a single electrode at pulse rates varying from 200 to 1000 Hz, to find gap detection thresholds. Presentation rate and pulse train duration affected the gap detection abilities of a few subjects only. In general, the authors found gap detection thresholds of up to 12.8 ms for most of their early-deafened CI users, although three users had higher thresholds, in the 17 to 32.1 ms range. Using more complex stimuli, Grose & Buss (2007) explored the gap detection abilities of five MED-EL COMBI 40+ users. The markers had different degrees of modulation, achieved by varying modulation bandwidth and degree of amplitude compression. Gap detection thresholds were in the normal hearing range for unmodulated markers, and for modulated markers with high degree of amplitude compression (regardless of modulation bandwidth). With no envelope compression (i.e., more modulated envelopes), gap detection thresholds were poorer, especially for the markers with narrow bands of modulation. Finally, Sagi et al (2009) investigated gap detection abilities in CI users and normal hearing subjects using vowel-like markers of 500 ms (total duration was 1 s for the no-gap condition, and 1.09 s for the longest gap condition). Different from most studies, Sagi et al (2009) employed a more difficult 7- alternative identification task for the gaps. While all subjects were equally able to discriminate between the no-gap and the smallest gap condition of 15 ms, overall d

31 17 scores were significantly poorer in the CI group. Some CI users performed in the normal hearing range, but most CI users performed no better than the lower range of normal hearing subjects. Cumulative d scores were positively correlated with speech perception (CNC words, Iowa Consonant Identification Task) in the CI group. Like in acoustic hearing, gap detection thresholds for CI users have been shown to increase as the markers surrounding the gap are more dissimilar [in level or presentation rate] (Chatterjee et al, 1998). Also, like in acoustic hearing, gap detection thresholds in cochlear implant users are poorer for noise markers than for sinusoidal markers (Moore & Glasberg, 1988). The conclusion reached by most of these investigations is that cochlear implant listeners have similar gap detection abilities to their normal hearing counterparts. This is remarkable given that CI users have severe-to-profound hearing loss, and that hearing loss can potentially affect temporal resolution (e.g., Takahashi & Bacon, 1992; Humes et al., 2010). However, those findings are not unanimous (e.g., Sagi et al, 2009). Aging and Temporal Resolution: Electrophysiological Studies Electrophysiologic techniques can also be used to assess the effects of aging in the auditory system. Evoked potentials offer the advantage of allowing researchers to explore how auditory signals are processed at different levels of the auditory pathways. Evoked potentials such as the ones used in this study are also more resistant to effects of cognitive factors such as memory and attention, which are known to decline with age and can negatively impact behavioral measures of auditory perception. Peripheral Evoked Potential Studies Auditory evoked potentials of peripheral origin have a relatively short latency. As a result, they are typically recorded using brief stimuli. Although this limits stimulus complexity, early potentials are less affected by attention and development than evoked

32 18 potentials of more central origin. Additionally, waveform morphology tends to be more consistent across listeners (Miller et al, 2008). The compound action potential (CAP) is a recording of a large group of auditory neurons firing in synchrony. It consists of a negative peak (N1) occurring 0.2 to 0.4 ms after stimulus onset, followed by a smaller positive peak (P1). The Auditory Brainstem Response (ABR) is composed of a series of 5-7 peaks that occur as a result of synchronized neural response within the first 10 ms following stimulation. Each peak is generated at progressively more central structures of the auditory system, from the auditory nerve to the inferior colliculus. Most of the published studies on age-related changes in temporal resolution in humans have used acoustic stimulation to record the ABR. For example, Poth et al (2001) investigated the effects of different gap durations on wave V amplitude in groups of younger and older adults with normal hearing. Stimuli consisted of two 50 ms broadband noise bursts separated by a gap. Results revealed that three of the eight older subjects did not show a wave V to the second noise burst in the shortest gap conditions (4 and 8 ms), while most of the young subjects did (no statistical analyses were performed). However, when wave V was present, it was similar in amplitude and latency between the two groups. Walton, Orlando & Burkard (1999) used ABR wave V latency as an indicator of recovery from forward-masking in younger and older adults with normal hearing. The probe consisted of 1, 4 and 8 khz tonebursts fixed at 40 db SL and presented at different time intervals (0-64 ms) after a 30 ms-long tonal masker of the same frequency as the toneburst. Results showed that wave V latencies were prolonged for the shortest maskerprobe intervals in both groups at all frequencies. However, the older age group had significantly greater latency shifts at the shortest time intervals (up to 16 ms) for the 4 and 8 khz tonebursts. Temporal processing can also be assessed by evaluating the effects of increasing the stimulus repetition rate on auditory evoked potentials. Burkard & Sims (2001)

33 19 investigated the effects of high click rates on waves I and V of the ABR in younger and older adults with normal hearing and with high frequency hearing loss. Wave I of the ABR is generated primarily at the auditory nerve. Wave V reflects the synchronous response of auditory neurons at the level of the synapse between the lateral lemniscal tract and the [contralateral] inferior colliculus (Møller, 2000). They used high-level clicks (115 db pspl) and presentation rates ranging between 11 and 75 Hz ( regular ABR) and 100 and 500 Hz ( maximum length sequence MLS ABR, or ABR to clicks with pseudorandom interpulse intervals). Although wave I was smaller and absolute latencies tended to be longer in older adults, the I-V interval did not differ as a function of age, and no differential effects of presentation rate between the age groups were found in terms of amplitudes or peak latencies. The ABR can also be elicited using speech-like sounds, in which case it is called complex ABR (or cabr). Vander Werff & Burns (2011) investigated age effects on the cabr in groups of younger and older adults with normal/near normal hearing. Using clicks at 80 db nhl and a synthesized /da/ stimulus presented at 82 db SPL, many differences in latencies and amplitudes of the click-abr and FFR component of the cabr (called S-ABR by the authors) were found. However, after the analysis was redone with high-frequency pure-tone sensitivity as a covariate, few age-related differences remained. These included a smaller wave V-V amplitude of the click-abr in the older group (although all click-abr values were within clinically used norms), as well as a smaller amplitude A peak of the cabr (which is similar to V of the click-abr). Also, peak O of the cabr, which is thought to reflect a response to stimulus offset, had longer latency in the older group. The authors suggest this could indicate differences in encoding of transient stimuli by older adults. In contrast, Anderson et al (2012) found age effects on the cabr in listeners with clinically normal hearing up to 8 khz. Using the same stimuli as Vander Werff & Burns (2011), they found longer peak latencies in response to the syllable onset and formant

34 20 transition (but not to the steady-state part of the syllable). Also, older adults had less consistent responses (after controlling for their higher baseline EEG activity), less phase locking and smaller amplitude responses. However, as the authors point out, there were significant differences between the groups in hearing thresholds at 4 and 8 khz, and these differences were correlated with age. Yet, in a previous study with two groups of older adults matched in age and high frequency hearing loss but differing in speech perception in noise abilities (in the HINT test), those authors had reported larger F0 magnitudes and RMS amplitudes in the C-ABR of the better-performing group (Anderson et al, 2011). Because the groups were matched for high frequency hearing loss, it is possible that these effects are related to different forms of the aging process (i.e., the better-performing group could be aging successfully ). However, this hypothesis was not directly tested. It is possible to use acoustic stimuli to record compound action potentials from individuals with residual acoustic hearing, but it is not easy often requiring a transtympanic electrode. In cochlear implant users, electrically evoked auditory-nerve recordings (ECAP) can be obtained with relative ease. This is possible because most current generation cochlear implants are equipped with two-way telemetry systems that allow the intracochlear electrodes to be used both as a stimulating and as a recording electrode. While contamination of these responses by electrical artifact can be problematic, techniques have been developed to minimize the effect of the artifact on neural recordings. The subtraction technique is the most commonly used. It consists of the sequential presentation of two current pulses often referred to as a masker and a probe pulse, exploiting the refractory properties of the auditory nerve (Brown et al, 1990). The ease of recording these potentials resulted in a reduction in the number of studies focusing on recording the electrically evoked auditory brainstem response (EABR) (Brown, 2003). Advantages of ECAPs when compared to EABR include no need for additional recording equipment, larger amplitude responses and less contamination from muscle artifact, which allows subjects to move during recordings and decreases data

35 21 collection time (Miller et al, 2008). In general, electrically evoked auditory potentials have been shown to be comparable to their acoustic counterparts, except for small differences in absolute and interpeak latencies (Firszt et al, 2002). These differences have been attributed to the fact that electrical stimulation drives the auditory nerve directly, bypassing the hair cell-neuron synapse and eliminating delays introduced by the cochlear traveling wave. To date, a single study has reported on the effect that advancing age has on temporal resolution in cochlear implant users using peripheral evoked potentials. Lee et al. (2012) measured ECAP recovery functions from seven older and five younger cochlear implant users. They used a two-pulse forward masking paradigm, varied the interval between the two pulses and measured the response of the auditory nerve to the second pulse in the sequence. In addition, they measured psychophysical recovery from forward masking using a 300 ms pulse train masker and a single biphasic probe pulse, separated by different inter-stimulus intervals (i.e. gaps). Results indicated no difference between the two age groups on the peripheral electrophysiological measure, while a slower recovery from forward masking was found behaviorally for the older subjects. The two measures were not correlated. Correlations with speech perception in quiet (CNC words and HINT/Q) and in noise (HINT/N at +8 db SNR) were only significant for the psychophysical recovery time constants and CNC word scores. Besides having small sample sizes, a potential limitation of this study is their choice of age groups. The younger subjects had an average of 46.8 years, with the oldest subject being 57 years old. Thus, these subjects might be considered more middle-aged than young. Collectively, the reviewed studies suggest that there may be age-related changes in temporal resolution as measured by auditory brainstem responses. No such evidence has been found for auditory-nerve responses in the limited number of studies available.

36 22 Middle-Latency Evoked Potential Studies Middle-latency auditory evoked potentials consist of a series of positive and negative peaks recorded within 8-40 ms following stimulus onset. These responses are thought to be generated in the thalamus, primary auditory cortex or association cortex (Picton et al, 1974). The Auditory Steady State Response (ASSR) is an evoked potential originating from phase-locked activity at several generators. The ASSR to amplitude-modulated stimuli was originally designed to assess hearing acuity, but it can also be considered as a measure of temporal resolution since it reflects the auditory system s time locking abilities (to the modulation frequency but note that neurons are tuned to the carrier frequency). Different modulation frequencies produce responses coming from different levels of the auditory system. The most commonly used modulation frequency in ASSR testing is 40Hz, which is thought to generate activity primarily in the thalamo-cortical regions of the auditory nervous system. Boettcher et al (2001) recorded ASSRs from three different groups of subjects, all of whom had normal hearing: young, young-old and older-old adults (who were only years old). The stimuli consisted of sinusoids modulated at 40 Hz (0-100% modulation depth) with carrier frequencies of 520 and 4000 Hz. No differences in amplitude or phase of the responses were found between the age groups as a function of modulation depth or carrier frequency. In addition, in a second experiment older adults with high frequency hearing loss and young adults with simulated hearing loss were tested. Results showed that, for the low frequency carrier, ASSR amplitudes were smaller in these two groups as compared to young adults (despite huge variability). This underscores the importance of controlling for off-frequency hearing loss, as subjects had normal hearing in the test frequency region. Leigh-Paffenroth & Fowler (2006) investigated ASSR responses at two different carrier frequencies (500 and 2000 Hz) in young and older adults that were matched in

37 23 hearing thresholds in the low-mid frequency range. By using different modulation rates (20, 40 and 90 Hz), this study attempted to examine temporal resolution at different levels of the auditory system, in that lower modulation rates would stimulate more central levels. The number of phase-locked ASSR responses was used as the outcome measure, instead of the amplitude and phase of those responses, which most researchers report. Results showed a significantly larger number of responses in the young age group for the 500 Hz carrier frequency, but not for 2000 Hz. Younger subjects also had more responses than their aged counterparts to the 90 Hz modulation rate, but not to 20 and 40 Hz. These age effects were further complicated by significant correlations in a larger subject group (with six subjects in addition to those of the first experiment), between hearing sensitivity and the number of phase-locked ASSR responses at the 500 Hz carrier frequency - the only frequency to show age effects in the matched threshold group. No published studies were found using middle latency responses to investigate age effects in cochlear implant users. Results from the two [acoustic] studies reviewed above suggest little to no age effects independently of hearing loss. Long Latency Evoked Potential Studies Long latency auditory evoked potentials are recorded ms after the onset of an acoustic stimulus. They follow the MLR and are thought to be generated at cortical or pre-cortical levels within the auditory system. The P1- N1- P2 complex is the most widely studied obligatory auditory evoked potential. It consists of a series of three peaks typically recorded between approximately 50 and 250 ms following stimulus onset. In adults, the N1-P2 peaks are the most robust. N1 and P2 are clearly distinct events, reflecting contributions from various sources, and each likely having their own significance in central auditory processing (Hyde, 1997). N1 is thought to originate at the primary auditory cortex. Less is known about P1 responses. A disadvantage of long latency potentials is that they can be affected by sleep and attention. But they also have

38 24 distinct advantages over more peripherally generated responses. For example, they can be evoked using longer duration stimuli such as speech-like sounds. Also, because these potentials are generated at the level of the cortex, they are more likely to be correlated with speech perception and cognitive abilities. A number of studies have used cortical evoked potentials to assess age related changes in temporal resolution. In a series of studies published between 2002 and 2004, Tremblay and colleagues explored the effects of age on cortical responses evoked with syllables differing in voice-onset time (VOT) (a timing cue that is analogous to a temporal gap). In the first study, a synthesized 7-step /ba/-/pa/ VOT continuum with 180 ms duration was presented to groups of younger and older adults with normal hearing (Tremblay et al., 2002). Older subjects had more difficulty discriminating the 10 ms differences in VOT behaviorally, and electrophysiological results showed longer N1 latencies for the longest VOTs (i.e., longest gaps between the onset of the consonant and the voiced vowel), and longer P2 latencies across the continuum for older adults. No effects of age on amplitudes were observed. The authors did not report correlations between the behavioral and electrophysiological results. In a follow-up study, Tremblay et al. (2003) used similar stimuli and conditions to explore the effects of age and age-related hearing loss in groups of younger subjects with normal hearing, older subjects with normal hearing, and older subjects with age-related hearing loss (this seems to have been an assumption; subjects had high frequency hearing loss, but apparently other factors such as noise exposure were not ruled out). The behavioral, N1 and P2 latency differences between the two age groups with normal hearing were the same as in their previous study. Contrary to their hypothesis, however, the presence of hearing loss in the second group of older adults did not produce additional N1 and P2 latency delays or amplitude reductions; in fact N1 amplitude was found to be larger in older subjects with hearing loss (for longer VOTs only). When comparing the two older groups, besides the N1 amplitude difference (in the opposite

39 25 direction than predicted), the only difference was in the behavioral discrimination task, in which the group with hearing loss performed more poorly, as expected. That is, the perceptual differences (and differences in hearing thresholds) observed between the two groups of older adults were only reflected electrophysiologically by an increase in N1 amplitude for longer VOTs (i.e., voiceless stimuli). Therefore, it is not possible to rule out the possibility of their previous age-related findings being related to subtle effects of hearing loss that are not apparent on standard audiometric testing. Alternatively, it is possible that audibility could have impacted the findings from subjects with hearing loss, as the authors did not seem to have controlled for this variable. In a subsequent study Tremblay et al. (2004) investigated rate and stimulus complexity effects in younger and older adults with normal hearing. They used the syllable /pa/ and 1kHz tonebursts of same duration and overall level, presented at interstimulus intervals of 510, 910 and 1510 ms. For speech stimuli, age-related effects with rate changes were found for N1, which was delayed for the older group at the medium and fast rates, and for P2, which was delayed at all rates. For tones, N1 latency was delayed in the older group at the faster rate only. There were no age-related differences in amplitude as a function of rate for either stimulus. In short, although there were speechspecific N1 and P2 delays in older adults, increasing rate affected N1 latencies for both speech (to a greater extent) and tones in older adults. The authors suggest that this could be related to age-related refractory differences that would be more evident with complex stimuli such as speech. When a continuous sound of relatively long duration changes in its spectral and/or temporal characteristics, a second set of P1-N1-P2 responses is generated. This has been called the Acoustic Change Complex (ACC), and it is different than a simple onset response to the second part of the stimulus (Ostroff et al, 1998). The ACC is generally smaller in amplitude than the onset response, and it is only recorded if there is a change in the stimulus. An ACC has been recorded for changes in spectrum and amplitude

40 26 separately and combined (Martin & Boothroyd, 1999), with more salient changes in the stimuli producing larger amplitude responses (Martin & Boothroyd, 1999; Brown et al., 2008). Therefore, the ACC is thought to reflect discriminability of the change in the stimulus. Lister et al. (2011) used the acoustic change complex (ACC) to investigate the effect of advancing age on gap detection. They tested 24 older adults with up to mild hearing loss at the test frequencies of 1 and 2 khz (but with a wide range of high frequency thresholds). Across- and within-channel gap detection was assessed using psychophysical and electrophysiological methods that employed narrow-band noise markers. Spectral splatter was controlled for in the behavioral testing by introducing a very short 1-ms gap in the standard stimulus. Marker duration was randomly varied between 250 and 350 ms to prevent listeners from using total duration cues to detect the gaps. Also, detection thresholds were obtained for the noise markers to ensure audibility. When compared to responses of younger adults from a previous study, older subjects showed longer P2 latencies for both onset and change responses, as well as larger P1 amplitudes for the change response. P1 amplitudes tended increase with gap duration in older adults but not in their younger counterparts. A gap detection paradigm was also used by Harris et al (2012) to evoke cortical responses in younger and older adults with normal hearing up to 4 khz. A continuous broadband noise stimulus with interruptions every 2 to 2.2 s was used, producing an ACC response only, with P1 reduced or absent. Results showed significantly reduced N1, P2 and N2 amplitudes and longer P2 latencies in older adults compared with younger listeners. This was the case for gap durations that were individually determined at a specific point of each subject s psychophysical gap detection curve (gap durations were longer in the older age group), and for 15 ms gaps. Finally, Palmer & Musiek (2014) measured behavioral and change N1-P2 thresholds to gaps in broadband noise in younger and older adults with mostly normal

41 27 hearing. Their end goal was to assess the utility of the electrophysiological measure in predicting behavioral gap detection. Cortical stimuli were presented every 4s. Gap duration was adaptively varied and smaller steps were used close to threshold. Older subjects had larger behavioral and electrophysiological thresholds, but there was no correlation between the two measures in the combined sample of 22 listeners. Also, puretone averages were correlated with the two gap measures, suggesting that results were not independent of hearing loss. Interestingly, the behavioral and electrophysiological thresholds were within 2 ms of each other for most subjects (in fact, no difference was found within the groups between the two measures). Event-related cortical potentials are usually obtained in an oddball paradigm, when an acoustically deviant (rare) stimulus is presented among frequently occurring (standard) stimuli. The mismatch negativity (MMN) and P300 are the most investigated event-related auditory potentials. The first is generated by automatic (i.e. unattended) processing of the deviant stimuli, and it is derived by finding the difference waveform between the responses to standard and deviant stimuli. The P300 is generated by controlled (i.e. attended) processing of the deviant stimuli. Cortical N1-P2 peaks and the MMN were used to investigate gap detection in a group of older adults (Bertoli et al, 2002). 1 khz tones in low-pass masking noise (cutoff frequency 3 khz) served as the standard stimuli, while deviant stimuli were of same total duration but contained gaps ranging from 6 to 24 ms, in 3-ms steps. Similar gap stimuli were used for behavioral gap detection, and results from this set of older adults with normal hearing (only thresholds up to 3 khz are reported) were compared to results from a previous study with younger subjects. Although psychophysical gap detection was not different between the two age groups, the minimum gap that elicited a group MMN response was different: 9 ms for individuals in the younger group and 15 ms for individuals in the older age group. The authors compared mean group MMN responses at 15 ms gap and concluded that MMN amplitude was significantly smaller and latency was

42 28 longer for the older subjects. However, no-responses were included in the group means. In the 15 ms gap condition, only one older subject (of ten) had an MMN present, while that was the case for nine (of ten) young subjects. Four older and one younger subject (of ten) did not have an MMN in response to any of the gaps. Regarding the N1-P2 analysis to the standard stimuli (averaged across the 6, 9 and 15 ms gap conditions), P2 amplitudes were smaller and latencies were longer for older adults. Ostroff et al (2003) and Alain et al (2004) reported on different aspects of the same study, in which they assessed age effects on tone duration with an oddball paradigm. Tones varied between 8 and 18 ms, and were either continuous or contained a gap of 3-13 ms. In the oddball paradigm, blocks containing 85% probability of no gap ( standard stimuli) and 15% probability of stimuli with a gap ( deviant stimuli) were presented to three groups of young, middle-aged and older adults. Subjects had thresholds of up 30 db at the test frequency of 2 khz, and thresholds were significantly poorer in each group of increasing age. Ostroff et al (2003) focused on the analysis of automatic N1-P2 responses to the standard stimuli. Results revealed increased P2 amplitudes with increasing tone duration in young and middle aged subjects but not in the older age group (the authors accepted a significance level of p<0.06). These differences were unrelated to hearing thresholds. Alain et al (2004) reported on the analyses of the gap stimuli (deviant) closest to asymptotic behavioral detection (i.e, at perceptually equivalent conditions across subjects). Results showed longer P3b latencies in the older age group when compared to the other two groups. Also, no MMN was detected in this gap condition in the two older age groups. When all the gap durations were averaged within the groups, longer N2b and P3b and MMN waves were observed in the older age group, as well as smaller amplitude MMN. However, it is unclear whether no-response conditions in the two older age groups were eliminated from this analysis. The authors interpret these findings as evidence of age-related decline in automatic processing of sounds (due to the lack/small amplitude of MMN responses in older subjects) that can be

43 29 compensated for by focused attention (because P300 responses showed fewer age effects). Neuromagetic evoked responses have also been used to assess the effect of advancing age on gap detection thresholds. Ross et al (2010) used 1 khz tones of different durations (34-76 ms) that were either continuous or contained a gap to record neuromagnetic middle and long latency evoked potentials. Recordings that were analogous to middle-latency auditory potentials showed longer latencies and larger amplitude responses in the older age group (mostly in the left hemisphere). Recordings that were analogous to the cortical P1-N1-P2 showed that P1 latencies and amplitudes increased with age. Difference waves between the no-gap and gap conditions (not interpreted by the authors as the MMN, likely because the probability of each stimulus occurring was 50%) had smaller amplitudes in the older age groups. Older adults had significantly higher thresholds in the test frequency range, but because of the high stimulus presentation levels used (80dB SPL), the authors did not explicitly rule out the effects of hearing loss. In short, most studies of age-related loss of temporal resolution using cortical auditory evoked potentials report some latency and/or amplitude changes in older adults. However, these effects differ across studies, likely as a result of methodological differences. No studies to date investigated age-related changes in temporal resolution in cochlear implant users through cortical evoked potential measures. Aging, Hearing and Cognition Cognition refers to the mental abilities and processes involved in storing, making and manipulating knowledge (Bayles & Tomoeda, 2007), and it is known to decline with age. One of the prevailing theories of cognitive aging proposes that a general slowing of the nervous system happens with age, impairing cognitive functioning (Salthouse, 1996).

44 30 Therefore, when investigating age-related changes in auditory temporal processing, a concomitant analysis of cognitive factors seems warranted. The need to explore the contributions of peripheral, central and cognitive processes on age-related changes in speech perception has long been established (CHABA, 1988). Still, it was not until the last decade that studies linking hearing and cognition received more attention. The urge to investigate more realistic listening situations, to fully understand the effects of aging and to explore reasons for the variability in benefit from current hearing technology were among the factors that motivated those advances (Arlinger et al, 2009). The interaction between age-related changes in the auditory system (peripheral and central) and cognitive decline in the elderly remains unclear. In a recent review of post-chaba report aging studies, the AAA Task Force on Central Presbycusis found that evidence from most temporal processing studies showed effects of age unconfounded by hearing loss (but with cognition largely unassessed) (Humes et al., 2012). However, the group concluded that across all behavioral studies it was not possible to establish the existence of age-related changes in the central auditory system without confounds of peripheral hearing loss and cognitive factors. Likewise, a review of studies of speech perception in noise and cognition [not limited to the elderly cohort] concluded that there seems to be a link between the two, although hearing loss seems to be the primary predictor of performance in noise (Akeroyd, 2008). As Houtgast & Festen (2008) point out, although variables typically used by researchers to predict speech perception in noise do not fully explain the variance in speech perception in noise (albeit explaining around 70% of the variance), it is possible that further advances in cognitive testing may help fill that gap. Epidemiological studies also suggest that there may be a link between sensory loss and cognition. Results from a large sample of older adults in the Baltimore Longitudinal Study of Aging revealed an increased risk of cognitive decline and dementia

45 31 in older adults with hearing loss, with the risk being proportional to the degree of hearing loss (Lin et al, 2011). This is in agreement with earlier cross-sectional studies that had documented increased risk of dementia with progressively greater degree of hearing loss (e.g. Uhlmann et al, 1989), and greater decline in memory performance with change in hearing thresholds over time [to a lesser extent than change in vision] (Valentijn et al, 2005). Proposed mechanisms for this relationship are (1) the existence of a common factor mediating hearing and cognitive decline, (2) that hearing impairment leads to increased effort in communication, which in turn decreases cognitive resources available for other mental tasks, (3) that sensory deprivation leads to atrophy of neural structures and permanent cognitive decline, and (4) that an impoverished sensory input leads to poorer [not permanent] cognitive performance. On the other hand, many older adults seem to use increased cognitive efforts and preserved linguistic knowledge to their advantage, to overcome different forms of signal degradation (Wingfield et al, 2005). Older adults seem to be able to take advantage of contextual cues in speech better than their younger counterparts (Pichora-Fuller et al, 2007), with increasing advantage for older adults with larger amounts of hearing loss (Frisina & Frisina, 1997) (although Dubno et al, 2000 showed that when provided with equivalent audibility, younger and older adults benefit equally from context). This compensatory brain activity has been documented in neuroimaging studies. PET imaging showed that high-performing older adults activate additional areas of the brain during memory tasks, when compared to younger adults (Cabeza et al, 2002). Consistent with this hypothesis, when presented with words in background noise (at +20 and -5 db SNR), fmri results revealed increased activation of working memory and attentional areas of the brain in older adults, especially in the more difficult noise condition (in which those subjects performed more poorly than their younger counterparts) (Wong et al, 2009). Understanding the interplay between hearing and cognition also has potential implications for auditory rehabilitation in the elderly, especially because conventional

46 32 hearing technology (hearing aids, assistive listening devices) cannot improve temporal resolution. For example, previous music training in older adults has been found to result in improved speech perception in background noise, auditory working memory and temporal acuity [through backward masking] (Parbery-Clark et al, 2011). Also, auditory training focused on consonant-vowel transitions has been shown to reduce age-related delays in the formant transition area of the speech-evoked ABR in noise, and to decrease interpeak variability in those responses (Anderson et al, 2013). More specific to the current study, Kishon-Rabin et al (2013) reported that auditory training in detecting gaps improved gap detection thresholds of both younger and older adults, decreasing the difference between the two age groups. Finally, cochlear implants require remarkable ability of their users to adapt to a completely novel way of hearing. Thus, it seems reasonable to hypothesize that a listener s cognitive status would impact success with the device. In fact, the finding that cochlear implant users generally improve in performance within the first few months of device use would seem to support the role of cognition. In one study, verbal learning was found to be the most significant predictor of speech perception scores at 6 months postimplantation (although verbal learning itself was predicted by duration of hearing loss and working memory), followed by baseline speech perception scores and lip-reading abilities, which together accounted for 82% of the variability in speech perception (Heydebrand et al, 2007). However, in a subsequent study with a larger sample of 114 adult CI users, cognitive measures that were originally significant predictors of speech perception were also found to be significantly correlated with age at implantation (Holden et al, 2013). Once the extreme age groups were removed, cognition (and age at implantation) ceased to predict speech perception. Collectively, these studies suggest that cognitive factors are important when considering age-related changes in the auditory system. However, the exact role of cognition without confounds of hearing loss (or its covariance with hearing loss) is still

47 33 not completely understood. It is possible, as Wingfield et al (2005) pointed out, that the boundaries between peripheral, central and cognitive mechanisms contributing to speech perception are not clearly distinct. Assessing all three in the same subjects should significantly advance our understanding of these mechanisms and their interactions.

48 34 CHAPTER 3 METHODS Study Participants Two groups of 10 postlingually deafened Nucleus cochlear implant users were recruited to participate in this study, for a total of 20 subjects. The younger group consisted of individuals who were between 18 and 40 years of age. The older group was between 65 and 80 years. Study participants were only invited if they used a Nucleus CI24RE TM, CI512 TM or CI422 TM cochlear implant. More recent generation CI systems were required because the neural telemetry systems used to record the ECAP are better in these implants than in previous generation devices. All subjects were native English speakers and had at least one year experience with their cochlear implant at the time they participated in this study. Exclusion criteria included (1) etiologies of hearing loss known to potentially reduce performance with cochlear implants such as meningitis, (2) prelingual severeprofound deafness, (3) less than full array insertion in the cochlea, and (4) clinical map with more than three electrodes deactivated. In addition, subjects were excluded if they did not have measurable electrophysiological responses (ECAP and ACC). Finally, subjects who did not meet a passing score of 27 out of a maximum possible of 30 points on the Mini-Mental State Evaluation (MMSE) (Folstein et al, 1975) were excluded. Additionally, attempts were made at recruiting subjects with similar duration of profound deafness, as it has been found to be a significant predictor of speech perception with cochlear implants by a number of investigators (e.g., Blamey et al, 1996, Rubinstein et al, 1999; Leung et al, 2005). Individual demographic and CI information can be found on Table 1. This study was conducted in accordance with the norms established by the Institutional Review Board of the University of Iowa.

49 35 Study Overview In order to investigate age-related changes in temporal resolution in cochlear implant users, the following measures were obtained from each subject: 1. Cognitive testing (MMSE screening, subtests from the WAIS-III and IV) 2. Peripheral Electrophysiological Measures a. AutoNRT across the electrode array b. Behavioral T- and C- levels (for pulses presented at 80Hz and for pulse trains presented at 1000Hz) c. ECAP growth d. ECAP recovery from refractoriness (two-pulse paradigm) e. ECAP recovery from forward masking (pulse train paradigm) 3. Central Electrophysiological Measures (ACC to gaps in pulse trains) 4. Psychophysical gap detection threshold 5. Speech perception in noise testing (QuickSIN) 6. Speech perception of time-compressed sentences (30% and 20% timecompressed SPIN) 7. Listening effort Details specific to each task are included in the following sections. General Procedures For all the auditory tests (except for speech perception and cognitive testing), a single research processor was used. The microphone of the speech processor was bypassed with direct audio input, and the intracochlear electrodes were stimulated

50 36 directly using the Nucleus Implant Communicator (NIC) routines. Bypassing the speech processor of the cochlear implant in this way ensured that for all temporal resolution measures the same electrode was stimulated similarly across participants. Also, with this approach there is no concern about residual acoustic hearing a subject may have. A single, mid-array electrode was used for all tests (electrode 11, 12 or 13). For bilateral CI users, only one ear was tested. Stimuli were presented at the levels that were reported as loud but comfortable, and were kept constant throughout testing. Data were obtained from most subjects in a single session lasting approximately 5.5h and included breaks, as needed, to ensure cheerful compliance. Temporal Resolution: Peripheral Evoked Potentials Cochlear implants are equipped with two-way telemetry systems that allow recording of the neural response to electrical stimulation. The electrically evoked compound action potential (ECAP) is a population response to electrical stimulation of the auditory nerve. It consists of a negative peak (N1) occurring 0.2 to 0.4 ms after stimulus onset and it is followed by a smaller positive peak (P2). An ECAP is characterized by the difference in amplitude (i.e., voltage) between these two peaks. Initially, ECAP thresholds were obtained across the electrode array using the manufacturer s automated algorithm (AutoNRT). This step ensured that ECAP threshold at the test electrode did not differ significantly from thresholds obtained at the other electrodes. Once this was done, the stimulus train used for measuring the ECAP was presented to the subject and the lowest level that could be detected (T-level) as well as the maximum level that was still comfortable (C-level) for the listener were obtained at the test electrode (typically, electrode 12). This was done using the Stimulation Only mode of the Cochlear Custom Sound EP program. The process was repeated using pulse trains created in the cortical evoked potential module on the Custom Sound EP program. A visual scale helped subjects identify T- and C-levels. The scale is the same that is used

51 37 to program their speech processor and contained 10 points. Number 1 was labeled Just noticeable and number 10 was labeled Too loud. T-level was defined as the stimulus level that was ranked as 1. C-level was defined as the stimulation level ranked as 7 and was labeled Loud but comfortable. A modified ascending method of adjustment was used to define these two loudness levels. The Nucleus CI codes intensity on a scale that ranges from 1 to 255 CL (Current Level) units. Initially a step size of 5CL was used, which was reduced to 2CL after the first ascending run. Loudness judgments were obtained at a relatively low stimulation rate of 80Hz, as this is the rate used for ECAP growth and ECAP two-pulse recovery functions. In addition, T- and C-levels were obtained for pulse trains presented at 1000Hz, which was the rate used for ECAP pulse train recovery, psychophysical gap detection and ACC measures. Temporal integration is known to affect loudness perception such that faster stimulation rates produce lower T- and C-levels (e.g. Potts et al, 2007; Botros & Psarros, 2010). In order to equate presentation levels across the experiments in this study, T- and C-levels were measured for both presentation rates. Next, Custom Sound EP software was used to record ECAP thresholds and growth functions on the test electrode. Stimulation and recording parameters such as amplifier gain and sampling delay were customized for each subject using a method outlined originally by Abbas et al (1999). These settings were used for all of the ECAP experiments, except in circumstances where those settings produced sub-optimal recordings (e.g., sampling delay was at times increased to reduce artifact in the pulse train paradigm when short MPIs were used). Two different measures of temporal resolution at the level of the auditory nerve were obtained using the ECAP. The first method is illustrated schematically in the top panel of Figure 1 and requires presentation of two pulses (a masker and a probe ) separated by a brief gap. The masker and probe were biphasic, 25 µs/phase pulses with 7 µs interphase gap. The time interval between masker and probe was systematically

52 38 varied. When very short (up to 500 µs) masker-probe intervals (MPIs) are used, the auditory nerve fibers that respond to the masker are in refractory state when the probe is presented, resulting in little or no response to the probe (Brown & Abbas, 1990). As the time interval between the masker offset and probe onset is increased, progressively more nerve fibers will be out of their refractory state and able to respond to the probe (as described in Brown, Abbas & Gantz, 1990). A recovery function consists of the ECAP amplitude in response to the probe as a function of the masker-probe time interval (MPI). The following MPIs were used: 0.3, 0.5, 1, 1.2, 1.5, 2, 3, and 5 ms. The top panel of Figure 1 shows a schematic of the stimuli used to record the two-pulse [refractory] recovery functions. High masker and probe levels such as those used in this study tend to produce larger ECAP amplitudes and possibly fast recovery. Stimulus presentation rate was 80Hz, the software default for ECAP recordings and a rate that is widely used in clinical and research settings. Two repetitions of 100 sweeps were recorded. This two-pulse recovery from refractoriness is similar to the one used by Lee et al. (2012), who did not show differences in the ECAP recovery function between the two age groups. Although the two-pulse paradigm can provide insight into the readiness of auditory nerve fibers in tracking fast changes in the incoming signal, its time course is very short and may therefore not represent how longer-duration phonemes and syllables are processed peripherally. In addition, while studies with subjects using early cochlear implants showed correlations between the slope of ECAP refractory recovery and speech perception (e.g., Brown et al, 1990), recent studies with newer generation devices have not found such relationship (e.g., Lee et al, 2012). The second paradigm was also used to assess temporal processing at the level of the auditory nerve, but this second procedure allowed us to track recovery over a longer time period. The stimuli used were trains of biphasic current pulses identical to those used for the central electrophysiological and psychophysical measures of temporal

53 39 resolution described below. This design allowed for direct comparison of temporal resolution at different levels of the auditory system. For this procedure, the masker consisted of a train of biphasic current pulses that was 400 ms in duration. Amplitude of each pulse in the train was held constant. The time interval between the masker offset and the onset of a single probe pulse was then systematically varied. The following masker-probe intervals were used: 1, 2, 5, 10, 20, 50, 100, 150 and 250 ms. The middle panel in Figure 1 shows the stimuli used for the pulse train ECAP recovery measures. Masker rate was 1000Hz, which is typical of clinically used settings. The repetition rate was 1.5 Hz. A recovery function showing the ECAP amplitude in response to the probe as a function of the masker-probe interval was plotted, and slopes were estimated using exponential functions for each subject. In this paradigm the response to the probe is likely affected by, among other factors, the degree of adaptation in the auditory nerve, or the decrease in neural response probability over time with repeated stimulation. In both ECAP experiments, stimuli were presented and responses were recorded using the manufacturer s research software (Custom Sound EP). Recordings were analyzed offline using a custom script in Matlab TM. In both ECAP experiments, the response amplitudes were normalized, which allows for comparison across subjects and helps control for long term adaptation effects that may have been encountered. ECAP Data Analysis When recording and analyzing ECAPs, it is important to consider the stimulusrelated artifact, which is usually very close in time to the ECAP and can obscure the neural response. The subtraction technique (Brown et al, 1990) is currently implemented in the CI manufacturer s software, and it is generally effective in uncovering the neural response. It involves the sequential recording of four frames using different masker and probe pulse settings, which take advantage of forward masking and are combined to

54 40 cancel out the probe and masker artifacts. However, when longer masker-probe intervals are used such as in the present experiments, a modified subtraction technique proposed by Miller et al (2000) may be more effective. It involves recording the masker and probe frames in an additional template condition, using high masker levels and a very short masker-probe interval. This template of the probe artifact is combined with each recording to uncover the neural response. Thus, we also recorded a template condition in both ECAP recovery paradigms, to subsequently aid in analysis of the neural responses. ECAP responses to the two-pulse stimulus were analyzed for each subject using a 300 µs MPI template (which was originally one of the test MPIs), a 400 µs MPI template, and the standard subtraction technique. Across all subjects, the 300 µs MPI template yielded the best responses, and was therefore used in all subsequent analyses. The ECAP amplitudes were then normalized to the ECAP amplitude to the probe alone frame in the template condition (for each subject), to allow for comparisons between subjects. ECAP responses to the probe pulse following the pulse train masker were analyzed for each subject using a 400 µs MPI template and the standard subtraction technique. Across all subjects, the standard subtraction technique yielded better morphology responses, and was therefore used for all subsequent analyses. For the modified subtraction technique to be effective, it is necessary that the masker be of sufficiently high level (and better yet, higher than that of the probe pulse) in order to drive the auditory nerve fibers into refractory state. However, because the presentation levels used in the pulse train stimulus were as loud as the subjects could comfortably tolerate, and the template probe needed to be presented at the same level, we were not able to increase the masker level beyond that of the probe level without risking subject discomfort. Thus, the 400 µs MPI template we collected may not have been optimal. It should be noted that because of the computations involved in the standard subtraction technique, the present ECAP pulse train recovery functions have a decaying exponential form rather than the expected rising exponential form. The ECAP amplitudes were

55 41 subsequently normalized to the neural response amplitude of the 1 ms MPI condition, in order to allow comparisons across subjects. Temporal Resolution: Central Evoked Potentials Temporal resolution was assessed centrally using the acoustic change complex (ACC), which consists of obligatory P1-N1-P2 cortical responses recorded with scalp electrodes, in response to temporal gaps. When a listener is presented with a relatively long stimulus that includes a change in the center such as the presence of a gap, two separate P1-N1-P2 responses are generated. The first is a response to the onset of the initial stimulus and reflects detection of that sound. A second P1-N1-P2 potential is generated in response to the change in the ongoing stimulus in this case the gap. This response, the ACC, has an amplitude that is roughly proportional to how perceptually different the two stimuli are (e.g., Martin & Boothroyd, 1999; Brown et al., 2008). In this study, listeners were presented with two 400 ms constant amplitude pulse trains identical to those used for the ECAP pulse train paradigm, separated by a gap (analogous to the MPI used for peripheral measures of temporal resolution). A schematic of the stimuli is shown in the bottom panel of Figure 1. Cortical responses were recorded for the following gap conditions: 5, 10, 15, 20 and 30 ms. One set of stimuli was presented every 3s, at the same presentation level as the stimuli used in the ECAP pulse train paradigm. Two repetitions of 100 artifact-free epochs were recorded for each gap condition. Presentation order was counterbalanced across subjects. During recordings, subjects were seated in a comfortable chair and were encouraged to either read or watch a captioned movie. Frequent breaks were offered to the study participants in order to help maintain their alertness and comfort. Responses were recorded with six disposable scalp electrodes. Two channels of differential recordings were obtained: vertex (+) to contralateral mastoid (-), and vertex (+) to the inion (-). A ground electrode was placed off-center on the forehead and two

56 42 electrodes were placed laterally, above and below the contralateral eye to monitor eyeblinks. A custom script using the NIC routines delivered the stimuli to the implant and triggered the recording software. Responses were amplified with a gain of 10,000 and digitally filtered between 1 and 30 Hz using an Intelligent Hearing Systems differential amplifier (IHS 8000). The analog electroencephalographic (EEG) activity was digitized with a National Instruments Data Acquisition board sound card (DAQCard-6062E) at a rate of 10,000 Hz per channel. The ongoing EEG activity was displayed and averaged for offline analysis using custom LabVIEW TM version software (National Instruments Corp., 2009). Online artifact rejection eliminated sweeps containing large voltage excursions. Amplitudes and latencies of the cortical onset and change responses were analyzed offline using a custom script in Matlab TM. Amplitudes of the ACC N1 and P2 peaks were computed relative to a 300 ms pre-stimulus baseline. Those amplitudes were normalized to that of the onset responses for each condition, to account for long-term adaptation effects as well as changes in subject state. Temporal Resolution: Psychophysics Gap detection was assessed behaviorally with the same stimuli used for central electrophysiologic recordings, i.e., two 400 ms constant amplitude pulse trains separated by a gap of varying duration. Presentation levels were the same as those used for cortical recordings. Studies have shown that gap detection thresholds are lowest at moderate-tohigh presentation levels both in acoustic and electric hearing (e.g., Chatterjee et al, 1998; Shannon et al, 1989). Studies have also shown that gap detection thresholds of cochlear implant users vary across the electrode array in a non-uniform way across listeners (Garadat & Pfingst, 2011). However, for most subjects virtually no across-site difference is observed at high stimulus levels. Therefore, testing a single electrode at their loud but comfortable level seemed appropriate.

57 43 The same intracochlear electrode used for electrophysiologic recordings was tested psychophysically. Stimuli were delivered directly to each subject s implant, bypassing the speech processor. A three-interval, three-alternative forced choice task with feedback was used. Two reference intervals contained the two 400 ms pulse trains presented sequentially (without a gap) while the comparison interval contained an adaptively varied gap. The stepping rule was two-down, one-up, converging on the 70.7% point on the psychometric function (Levitt, 1971). Thus, after a subject correctly identified the interval containing the gap in two consecutive set of intervals, gap size decreased; however the step size was increased after a single incorrect response. Initial gap duration was 50 ms. Initial step size was 40 ms, and it decreased after each reversal by a factor of 1.5. This adaptive paradigm terminated after eight reversals. The last six reversals were used to calculate threshold. Gap detection thresholds were measured four times and the results of the best three runs were averaged together. The duration of the markers surrounding the gap were randomly varied between 350 and 450 ms to avoid the use of total duration cues when determining the presence of a gap. The interval between each stimuli and also between each trial was 300 ms. Subjects were allowed to respond at their own pace. They were instructed to select the box in a computer monitor corresponding to the interval that sounded different from the other two (i.e., the interval containing the gap). Subjects were explicitly told to ignore total duration variations. Speech Perception: QuickSIN Speech perception in noise was assessed with the Quick Speech Perception in Noise test (QuickSIN) (Killion et al., 2004). Stimuli were delivered to each subject s own speech processor through direct audio input (bypassing the microphone), at a presentation level that was perceived by subjects as loud but ok (as suggested in the test manual). Once again a visual analog scale was used to help subjects determine this level.

58 44 The QuickSIN is a non-adaptive test that computes the signal-to-noise ratio (SNR) necessary for 50% correct speech recognition in a background of four-talker babble noise. It contains 12 lists of 6 sentences that are each presented at a different SNR ranging from 25 to 0 db in 5dB steps. Scoring is based on key words repeated correctly (5 per sentence). All sentences in a list are presented to the listener and the total number of words repeated correctly is computed and subtracted from 25.5 to yield the listener s SNR loss relative to normal hearing listeners. The QuickSIN test has been shown to more effectively separate groups of normal hearing listeners from those with hearing loss, in comparison with other commonly used sentence tests (Wilson et al., 2007). It is also not widely used in routine testing of CI users, which avoids practice effects. The presence of noise was deemed to reflect a more realistic listening situation in which older adults have difficulty, and that is more likely to be affected by age-related changes in temporal resolution and by increased cognitive load. Using a sentence rather than word test is also more representative of everyday speech, and may be more prone to the effects of age due to increased cognitive load required to remember and repeat whole sentences rather than single words. Finally, the different levels of background noise were likely to allow for the wide range of performance that is characteristic of cochlear implant users, avoiding floor and ceiling effects. Seven randomly chosen lists were administered to each subject and scores were averaged across the lists, yielding an SNR loss that is accurate within 1.0 db at the 95% confidence level (Killion et al., 2004). Speech Perception: Time-Compressed SPIN Sentences While speech perception in noise is a typical outcome measure of auditory performance, given that we investigated temporal resolution, it was of interest to examine the perception of time-compressed speech. In addition, this paradigm has been frequently used in studies of age-related changes in temporal resolution. The SPIN-R test (Kalikow

59 45 et al, 1977) was used to assess the effects of time compression. As with the speech in noise test, stimuli were delivered to each subject s own speech processor through direct audio input (bypassing the microphone), at a presentation level that was perceived by subjects as loud but ok. A visual analog scale was used to help subjects determine this level. The SPIN-R test consists of eight lists of 50 sentences each. Half of the sentences in each list have context that facilitates predictability of the final word; the other half of the sentences consist of low-predictability final words. Percent correct scores are calculated from the number of final words correctly repeated. For the speech perception task, each subject was presented with one list of sentences per time compression condition, excluding list number 7, which was used in the baseline recall condition. Performance on high and low predictability sentences was averaged together for the analyses in this study. Note that for the listening effort dual task recall task outlined below (i.e., the 20% time compression condition), in order to equate the number of words between the baseline and recall tasks, the last two sentences of the list used in that condition were eliminated. Two conditions of time compression were used in this study: 20% and 30%. They were chosen based on results from pilot data with cochlear implant users, in an effort to find conditions that would avoid floor and ceiling effects. To achieve the time compression, each of the SPIN-R lists was processed using the IZotope Radius algorithm available on Adobe Audition CS 5.5 software. This algorithm shortens or extends audio files without compromising pitch quality. No distortions were evident in any of the lists after time compression was applied. Listening Effort Two measures of listening effort were used in this study. The first consisted of a dual task paradigm, and followed a similar procedure to the one used by Sarampalis et al.

60 46 (2009), with the addition of a baseline condition. In the baseline condition, subjects were presented with one sentence of the SPIN-R list number 7 at a time in a computer monitor, in large black font against white background. They were instructed to repeat each sentence aloud and remember the last word. After every 8 sentences, a prompt appeared in the monitor asking subjects to recall as many of the last 8 final words as they could remember, in any order. Six blocks of eight sentences were presented, for a total of 48 sentences (the last two sentences in the list were eliminated). The baseline recall was scored as the percent correct of words correctly recalled. In the dual task paradigm, subjects listened to one list of 48 sentences that had been time-compressed by 20%, via direct audio input to their speech processor (as outlined above). They were asked to repeat each sentence and remember the final word. After every 8 sentences, subjects were prompted to recall the as many of the last 8 final words as they could remember, in any order, similarly to the baseline condition. Of note, for the dual task recall condition, words recalled were considered correct if they matched what the subject had understood (even when this was different than the word presented). Also, no-responses were eliminated from the number of possible items to be recalled. The dual task effect was calculated as the difference between the percent correct of words recalled in the dual task and in the baseline condition. The second measure of listening effort was a subjective rating. Immediately after completing the dual task, subjects received a form asking How hard did you have to work to achieve your level of speech understanding? A 20-point visual analog scale was provided, with anchors at the extreme points labeled 0 Not at all, and 100 Very, very hard. Subjects marked on the scale what their perceived level of effort was. Cognition Measures Any study focusing on the effects of aging needs to consider the role that agerelated changes in cognition might play in determining the outcome on behavioral

61 47 assessment procedures such as those used in this study. The main focus of this study was on temporal resolution. Therefore, it seemed important to assess the subjects processing speed abilities. In addition, working memory has been shown to most consistently relate to speech perception in noise (Akeroyd, 2008), and was therefore also tested. Working memory refers to the transient information that is held and the active processes involved in the conscious person (Bayles & Tomoeda, 2007). The two core tests of speed of processing from the WAIS-IV test were used in this study. The Symbol Search test consists of two target symbols that should be searched for within a list of five symbols. Subjects had to complete as many of the 60 sets of stimuli as possible within a 120 s time frame. Raw scores were computed by subtracting the number of incorrectly completed items from the number of correct responses. The same time limit applied to the Coding test, in which subjects were presented with a key of numbers 1-9 and their corresponding symbols. The subject s task was to draw the symbol corresponding to each of 135 numbers, using the provided key. Raw scores were based on the number of items correctly completed. Before the testing began, subjects were provided with the standard instructions that accompany the tests, and practiced with demonstration items. All subjects demonstrated a clear understanding of the tests. The Digit Span Test is a widely used test of working memory. Subjects were presented with sequences of increasing numbers of digits, which they had to repeat in the same order. In the first part of the test, digits were repeated in the same order as heard, for lists of up to nine digits. In the second part, digits had to be repeated in reverse order (i.e., backward), for lists of up to eight digits. To avoid differences across presentation trials that are inherent to the live voice presentation mode [especially given that the subjects have hearing loss], a pre-recorded audio-visual version of the digit span test was used. Digit span for each response mode was determined as the longest string of digits that subjects correctly repeated in two consecutive presentations. The total score was computed by averaging each subject s digit span in the forward and backward modes.

62 48 Statistical Analyses To address the first research question, the normalized ECAP recovery amplitudes obtained at different masker-probe intervals (i.e., time) were log-transformed (ECAP amplitudes to the two-pulse condition were inverse-transformed first) and analyzed using a linear mixed model analysis. In this model, age group and time are fixed factors. Time is treated as a continuous variable, which allows for computation of the individual recovery slopes and subsequent comparison between the groups. Individual intercepts and slopes are considered as random factors. This analysis was carried out separately for the ECAP recovery functions following presentation of the pulse train masker and single pulse masker. Normalized ACC amplitudes were analyzed with a two-way ANOVA (factors were age group and gap condition) with repeated measures on one factor (gap condition). Cortical onset response amplitudes and latencies, psychophysical gap detection thresholds, speech perception scores, listening effort measures, and results from cognitive tests were compared between the two age groups using t-tests. When the data did not pass normality tests, a Mann-Whitney test was used in these group comparisons instead. Significance levels were set at α = To address the second research question, two separate stepwise multiple regression analyses were used to compare each of the speech perception measures (QuickSIN and 20% time-compressed SPIN) with the three electrophysiological temporal resolution measures, gap detection thresholds, and the three cognitive tests. Significance levels for inclusion and exclusion from the model were set at α = 0.10.

63 49 Table 1. Participant demographic and CI information ID Age Gender Yrs education Ear Device (Nucleus) Yrs CI use Yrs prof deafness Electrodes deactivated Y1 32 M 13 L 24RE Y2 40 M 12 L , 2 Y3 27 F 14.5 R , 2 Y4 19 F 12 L , 2 Y5 18 M 13 L 24RE Y6 30 F 12 R 24RE none Y7 18 F 12 L Y8 31 F 18 R , 4, 14 Y9 34 F 12 L 24RE none Y10 29 F 13 R O1 73 M 12 R none O2 82 M 12 R 24RE O3 68 F 14 R none O4 80 F 12 L none O5 77 M 13 L 24RE none O6 70 M 12 L 24RE none O7 74 M 14 R , 2, 3 O8 74 F 18 L , 2, 3 O9 68 M 21 R none O10 82 F 12 L none

64 Figure 1. Schematic of stimuli used in electrophysiologic testing. Top panel: two-pulse paradigm, used for ECAP recovery from refractoriness (Peripheral). Middle panel: pulse-train paradigm, used for ECAP recovery from forward masking (Peripheral). Bottom panel: gaps in pulse trains, used for ACC recordings (Central). 50