NORTHWESTERN UNIVERSITY

Perceptual Learning on Auditory Spectral and Spectro-Temporal Modulation Tasks

A DISSERTATION

SUBMITTED TO THE GRADUATE SCHOOL IN PARTIAL FULFILLMENT OF THE REQUIREMENTS

For the degree

DOCTOR OF PHILOSOPHY

Field of Communication Sciences and Disorders

By Andrew Todd Sabin

EVANSTON, ILLINOIS December 2011

  2   © Copyright by Andrew Sabin, 2011 All rights reserved

  3   ABSTRACT Perceptual Learning on Auditory Spectral and Spectro-Temporal Modulation Tasks Andrew Todd Sabin Sensory systems must extract meaningful information from complex patterns of activity distributed across the sensory periphery. In the auditory system, these patterns are distributed along the cochlea and represent peaks and valleys of sound energy that are spread across frequency (spectral modulation) and vary over time (spectro-temporal modulation). Processing of these cochlear patterns at more central stages of the auditory system has been proposed to depend upon a bank of modulation filters, each tuned to a unique combination of a spectral and a temporal modulation frequency. However, while there is considerable physiological evidence for such combined spectro-temporal tuning, direct behavioral evidence is lacking. Here I examined the central processing of spectral and spectro-temporal modulation behaviorally using a perceptual-learning paradigm. Human listeners were trained ~1 hour/day for 7 days to either detect the presence, or discriminate the depth, of a single spectral, temporal, or spectro-temporal modulation, and examined how that training affected performance on the trained (learning) and untrained (generalization) modulations. In young adults with normal hearing (YNH), the rate and magnitude of learning differed across trained modulations. Critically, this learning was specific to the trained modulation; it did not generalize to untrained spectral modulation frequencies, or to untrained spectro-temporal modulations, even when the trained and untrained stimuli had one modulation frequency in common. The same training regimen led to different outcomes in older listeners with hearing impairment (OHI), including a more protracted influence of the initial testing session and generalization to an untrained spectral modulation frequency. The dependencies of this learning upon the particular trained modulation in YNH listeners indicate

  4   that training modified an aspect of processing that was selective for the trained modulation. These results therefore provide behavioral evidence that is consistent with the proposal that the auditory system analyzes cochlear activity through filters tuned to combined spectro-temporal modulation. The population differences in turn imply that the modified aspect was less selective in OHI than YNH listeners. On the practical level, these results suggest that training could aid real-world skills that depend upon the accurate perception of spectro-temporal modulation, such as sound localization and speech and music perception.

  5   ACKNOWLEDGMENTS This document would simply not exist if it had not been for the critical support that I received from numerous individuals. First and foremost, I thank my adviser Beverly Wright for her support throughout the process of completing this degree. Her guidance, patience, and encouragement have improved the scientific quality of my work, as well as my ability to communicate it. David Eddins also provided critical help in the design and interpretation of each of the reported experiments. I also thank the other members of my dissertation committee, Nina Kraus and Sumitrajit Dhar for their time and effort. The past and present member of the psychoacoustics lab provided tremendous support and general camaraderie that was essential to this process. These lab members include Matt Fitzgerald, Julia Mossbridge, Jeanette Ortiz, Yuxuan Zhang, Julia Huyck, Nicole Marrone, Karen Banai, and David Little. Several audiology students worked on these experiments as part of their capstone projects. Cara Depalma helped with data collection on the experiments described in Chapter 2. Cynthia Clark and Nikki Taft helped with data collection for the work described in Chapter 4. I am also grateful to the National Institute on Deafness and Other Communication Disorders who provided financial support for this work (grant F31DC009549). Collaborations with other researchers have enriched my scientific development over the past seven years at Northwestern. Such researchers include: Bryan Pardo, Zafar Rafii, Sumit Dhar, Nicole Marrone, Lauren Hardies, Pamela Souza, Holly Wiles, David Little, Chun Liang Chan, Ewan Macpherson, John Middlebrooks, Guillaume Andeol, and Joshua Cooper. Completing this degree was a great challenge for reasons that I could not have predicted at the beginning. The love and support that I received from my family and friends was immeasurably valuable during this process. In particular I’d like to recognize Robert Sabin, Janet

  6   Sabin, Bradley Sabin, Donna Sabin, Arlene Bizar, Mary Donoghue, Brendan Kirwin, and Alex Hall. Their support kept me (relatively) sane throughout this process. There are also several individuals who are no longer with us, but who are dearly important to me and I wish they could be here to celebrate: Irwin Bizar, Maurice Sabin, June Sabin, Conor Redig, and Katie Helmstadter.

  7   TABLE OF CONTENTS PRELIMINARY PAGES Abstract…………………………………………………………………… Acknowledgments………………………………………………………… Dedication ………………………………………………………………... Table of Contents…………………………………………………………. List of Figures …………………………………………………………….. List of Tables …………………………………………………………….. CHAPTER 1. Introduction ……………………………………………………….. Previous Work on Central Encoding of Spectral and Spectro-Temporal Modulation …………………….…………………..…………………….. Perceptual Learning of Spectral and Spectro-Temporal Modulation …… Perceptual Learning in Clinical Populations ……..….………………….. The Current Work………………………………….……………………..

3 5 6 7 9 10 11 11 17 19 20

CHAPTER 2. Perceptual learning of auditory spectral modulation detection………………………………...……………………………………….. Abstract…………………………………………………………………. Introduction…………………………………………………………….. Method………………………………………………………………….. Overview……………………………………………………….. Conditions……………………..……………………………….. Task and Procedure…………………………………………….. Stimulus Synthesis………………………….………………….. Stimulus Presentation………………………………………….. Listeners……………………………………….……………….. Results………………………………………………………………….. Performance on the Trained Conditions……………………….. Performance on the Untrained Conditions……………….…….. Discussion…………………………………………..………………….. Basis for Improvements…………………………..…………….. Stimulus Dependence of Learning……………….…………….. Practical Implications………………………………….……….. Comparison to Learning on Visual Contrast Detection…….…..

22 22 23 25 25 26 27 28 28 29 29 29 33 33 34 35 37 38

CHAPTER 3. Auditory perceptual learning specificity to combined spectro-temporal modulation……………….…………………………………… Abstract………………………………………………………………….. Introduction…………………………………………………………….... Method………………………….……………………………………….. Overview………………………………………………………… Conditions………………………….……………………………. Tasks and Procedures…………………………………………...…

40 40 41 43 43 43 44

  8   Stimulus Synthesis and Presentation ……………………………. Listeners………………………………..……………………….. Results…………………………….…………………………………….. Learning on the Trained Conditions…………………………….. Generalization to Untrained Conditions. ..……………………… Depth Discrimination. …………….…………………….. Detection of Spectro-Temporal Modulation. ……….….. Discussion……………………………………………………………….. Combined Spectro-Temporal Processing. …………………..….. Detection vs Discrimination. …………………………………… Summary and Conclusion. …..………………………………….. CHAPTER 4. Different patterns of auditory perceptual learning on spectral modulation detection between older hearing-impaired and younger normal-hearing adults ………………..….……….……………….. Abstract…………………………………………………………………. Introduction……………………………………………………………... Method………………………….……………………………………….. Listeners.………………………………………………………… Overview.………………………….……………………………. Tasks and Procedures…………………………………………...… Stimulus Synthesis and Presentation ……………………………. Results…………………………….…………………………………….. Performance on the Trained Condition.……………………….... Performance on Untrained Conditions......……………………… Discussion……………………………………………………………….. Differences in Learning Between OHI and YNH listeners…..….. A Potential Account for These Learning Differences…………… Potential Contributions of Age and Hearing Loss……………..… Practical Implications..…..…………………………………...…..

45 46 47 47 49 49 50 50 50 53 55

57 57 58 59 59 60 61 62 64 64 66 68 68 70 72 73

CHAPTER 5. Implications and Future Directions……………...……………….. Overall Result Summary………………………..……………………….. Implications About Encoding of Spectral and Spectro-Temporal Modulation in the Central Auditory System…………………………….. Processing Advantages of Spectral and Spectro-Temporal Modulation Filterbanks………………………………………………….. Comparison to Vision………………………………………………….... Practical Implications …………………………………………………... Future experiments………..……………………………………………..

75 75

Figures…………………………………………………………………………….. Table……………………………………………………….……………………… References………………………………………………………………………… Curriculum Vitae………………………………………………………………….

88 111 112 123

76 79 80 82 84

  9   LIST OF FIGURES Figure 1.1

Vowel Spectral Modulations

Figure 1.2

Musical Instrument Timbre Spectral Modulations

Figure 1.3

Vowel Spectro-Temporal Modulations

Figure 1.4

Musical Instrument Timbre Spectro-Temporal Modulations

Figure 2.1

Spectral modulation detection

Figure 2.2

Learning Curves

Figure 2.3

Learning Curve Slopes

Figure 2.4

Within-Session Performance

Figure 2.5

Untrained Spectral Modulation Detection Conditions

Figure 3.1

Spectro-Temporal Modulation Filterbank

Figure 3.2

Performance on the Trained Depth-Discrimination Conditions

Figure 3.3

Performance on the Untrained Modulation Depth-Discrimination Conditions

Figure 3.4

Performance on the Untrained Spectro-Temporal Modulation Detection Condition

Figure 4.1

Tasks

Figure 4.2

Learning Curves

Figure 4.3

Within Session Performance

Figure 4.4

Performance on Untrained Conditions

  10   LIST OF TABLES Table 4.1

Listener Audiograms

  11   CHAPTER 1. Introduction A primary function of any sensory system is to extract meaningful information from the complex patterns of activity distributed across the sensory periphery that are evoked by natural stimuli. In the auditory system, the pattern of excitation distributed across the basilar membrane of the cochlea represents the peaks and valleys of sound level spread across audio frequency, known as spectral modulation. For most natural sounds, these spectral modulations vary across time and are therefore more completely characterized as spectro-temporal modulations. Everyday hearing depends upon the accurate perception of these modulations because they provide fundamental cues for distinguishing between, grouping, and identifying natural sounds (e.g., Bregman, 1990, Woolley et al., 2005). For instance, differences in spectral modulation provide some of the most important cues that are used to distinguish vowels (e.g., Peterson and Barney, 1952; Fig. 1.1), musical instrument timbres (e.g., Grey, 1977; Fig. 1.2), and sound locations in the vertical plane (for review see Middlebrooks and Green, 1991). Spectro-temporal modulations provide critical information about the transitions between speech sounds (e.g., Lindblom and Studdert-Kennedy, 1967; Fig. 1.3) and also contribute to musical timbre identification (e.g., McAdams et al., 1995; Fig. 1.4). The initial encoding of these modulations in the cochlea has been the focus of a large number of investigations (for review see von Helmholtz, 1863, Von Békésy, 1960, Dallos, 1992). However, much less is known about how the central auditory system transforms this cochlear activity to extract meaningful information from it and to ultimately provide a vivid and rich representation of the auditory world. Determining the characteristics of these transformations is one of the primary motivations of this dissertation. Previous Work on Central Encoding of Spectral and Spectro-Temporal Modulation.

  12   The importance of spectral modulation perception to everyday hearing is supported by observations of degraded performance on a variety of real world tasks after distortions to naturally occurring spectral modulations, and by correlations in performance between real world tasks and tasks that isolate spectral modulation perception. In normal hearing listeners, artificial reductions in spectral modulation depth imposed using techniques such as spectral flattening or smearing lead to progressively more errors in speech sound identification (Ainsworth and Millar, 1972, Zahorian and Jagharghi, 1993, Sakayori et al., 2002), musical instrument identification (McAdams et al., 1999), and vertical sound localization (Sabin et al., 2005). Similarly, in hearing-impaired listeners, internal reductions in the representation of spectral modulation depth (Bacon and Brandt, 1982, Sidwell and Summerfield, 1985, Leek et al., 1987, Summers and Leek, 1994), attributed to widened peripheral auditory filters (Glasberg and Moore, 1986), contribute to difficulties in speech perception (Bacon and Brandt, 1982, Sidwell and Summerfield, 1985, Van Tasell et al., 1987, Henry and Turner, 2003, Henry et al., 2005, Won et al., 2007). Further supporting this link, in cochlear implant users, performance on tasks that isolate spectral modulation perception is correlated to speech perception in quiet (Henry et al., 2005, Litvak et al., 2007) and in noise (Won et al., 2007), as well as to several aspects of music perception (Won et al., 2010). Despite the importance of spectral modulation perception to everyday hearing, little is known about how this modulation is processed by the central auditory system. There is, however, a considerable literature about how analogous activity is processed in the visual system. The auditory and visual systems face a similar challenge. For both systems, critical information about the identity of a stimulus is encoded in the pattern activity distributed across the sensory receptor (cochlea or retina). Therefore, the central auditory and visual systems must

  13   employ transformations that extract information from these patterns. In the visual system, psychophysical (e.g., Graham, 1972, De Valois, 1977) and physiological (e.g., Maffei and Fiorentini, 1973, De Valois et al., 1982) work (primarily from the 1970’s) has clearly demonstrated that the visual system transforms activity from the retina via filters tuned to spatial frequency (see DeValois and DeValois, 1988 for review). Visual filters tuned to low spatial frequencies encode patterns of light that change slowly over space (coarse patterns) while those tuned to high frequencies encode patterns of light that change rapidly over space (fine patterns). The outputs of these visual filters are then sent to higher areas of processing that facilitate the identification of natural objects, such as faces (Smith et al., 2005, Sowden and Schyns, 2006). A growing body of recent evidence has led to the proposal that the central auditory system might use analogous processing to the visual spatial frequency filters. That is, the central auditory system might analyze cochlear activity via a bank of spectral modulation frequency filters, each tuned to a unique modulation frequency (frequency in cycles per octave: cyc/oct) (Schreiner and Calhoun, 1994, Shamma et al., 1995, Saoji and Eddins, 2007). The filters are analogous to the visual spatial frequency filters (Shamma, 2001). These auditory filters might also be tuned to particular ranges of audio (i.e., carrier) frequency (Shamma et al., 1995). Four separate lines of behavioral investigation support the possibility that spectral modulation frequency filters exist in the auditory system. First, as the spectral modulation frequencies of two sounds become more similar, they are more likely to mask each other, even after accounting for masking that would occur in the cochlea (Saoji and Eddins, 2007). This phenomenon may be due to both sounds engaging the same spectral modulation frequency filter. Second, the ability to detect spectral modulation after extended exposure to a target modulation frequency is more impaired at the target frequency than at other frequencies (Eddins and Harwell, 2002). This

  14   observation might be due to adapted processing in the filter that was tuned to the target modulation frequency. Third, the ability to detect complex spectral modulation (comprised of multiple spectral modulation frequencies) is better predicted by a listener’s sensitivity to the component modulation frequencies rather than by the overall depth of the complex modulation (Eddins and Saoji, 2004). This could occur if a decision maker was receiving input from multiple spectral modulation frequency filters, rather than a single mechanism with broad tuning to spectral modulation frequency. Fourth, sound localization performance in the vertical plane (an ability known to depend upon spectral modulation perception; Middlebrooks and Green, 1991) becomes degraded by imposing an interfering spectral modulation, but only for a narrow range of spectral modulation frequencies (~ 0.5-2 cycles/octave) (Macpherson and Middlebrooks, 2003). Similarly, removing spectral modulation in the same range of modulation frequencies can impair vertical sound localization (Qian and Eddins, 2008). These impairments in performance could occur if the underlying sound localization mechanism was extracting information from one or more spectral modulation frequency filters tuned to that narrow range. Further, physiological observations also provide some evidence for spectral modulation frequency filters. Single neurons with bandpass tuning to a specific spectral modulation frequency have been observed in ferret and cat primary auditory cortex (Schreiner and Mendelson, 1990, Schreiner and Sutter, 1992, Shamma et al., 1993, Schreiner and Calhoun, 1994, Shamma et al., 1995, Versnel et al., 1995, Kowalski et al., 1996, Calhoun and Schreiner, 1998, Versnel and Shamma, 1998, Klein et al., 2000, Keeling et al., 2008). There is also some indication from fMRI that portions of human primary and secondary auditory cortex are tuned to spectral modulation frequency (Langers et al., 2003, Schonwiesner and Zatorre, 2009). The individual neurons and/or regions of auditory

  15   cortex that are tuned to spectral modulation frequency could be the biological basis of a spectral modulation frequency filterbank. An extension of the spectral modulation filterbank proposal suggests that central analysis of cochlear activity depends upon a spectro-temporal modulation filterbank (Chi et al., 1999, Chi et al., 2005), where spectral modulation filters are one part of this larger filterbank. As mentioned above, the spectral modulations of many natural sounds change over time (Figs. 1.3 and 1.4). Therefore a transformation that ultimately represents cochlear input in terms of the component spectro-temporal modulations might facilitate better processing of natural sounds. It has been proposed that this type of transformation occurs in the central auditory system through a spectro-temporal modulation frequency filterbank. A schematic of the proposed spectrotemporal modulation filterbank is shown in Figure 3.1. Each spectrogram in Figure 3.1 depicts the spectro-temporal modulation to which a particular filter is tuned. Each filter has a preferred spectral modulation frequency (vertical spacing of bars), temporal modulation frequency (horizontal spacing of bars), and direction (up or down: direction of bar tilt). Note that in this proposal, the spectral modulation frequency filters discussed above simply reflect one “slice” of this larger filterbank (Fig. 3.1; middle column) where the temporal modulation frequency is zero. Similarly, the spectro-temporal modulation filterbank also contains filters selective only for temporal modulation frequency (Fig. 3.1; bottom row). However, evidence for the proposed combined spectro-temporal tuning comes primarily from physiological observations. Such tuning has been observed in individual neurons in inferior colliculus (e.g., Versnel et al., 2009), thalamus (e.g., Miller et al., 2002), and auditory cortex (e.g., Kowalski et al., 1996, deCharms et al., 1998, Klein et al., 2000, Depireux et al., 2001), as well as in population responses of human auditory cortex as identified with fMRI (Langers et al., 2003, Schonwiesner and Zatorre, 2009).

  16   Indirect behavioral support for this proposal comes from the success of a model involving this filterbank in capturing several psychophysical phenomena (Chi et al., 1999, Carlyon and Shamma, 2003, Chi et al., 2005, Elhilali and Shamma, 2008), and in predicting speech intelligibility (Elhilali et al., 2003a, Elhilali et al., 2003b). However, to date there is no direct behavioral evidence supporting the possibility of combined spectro-temporal tuning in the central auditory system. Though not the focus of this dissertation, there is considerable agreement that filters tuned to temporal modulation frequency also underlie auditory perception. Behavioral evidence for these filters comes from temporal-modulation-frequency-specific effects of masking (Bacon and Grantham, 1989, Houtgast, 1989, Yost and Sheft, 1989, Yost et al., 1989), adaptation (Tansley and Suffield, 1983, Richards et al., 1997, Wojtczak and Viemeister, 2003), expectation (Wright and Dai, 1998), and training (Fitzgerald and Wright, 2005, 2011), and physiological evidence comes from observations of individual neurons tuned to specific temporal modulation frequencies at cortical (Schreiner and Urbas, 1988, Eggermont, 1994, Langner et al., 1997, Giraud et al., 2000) and subcortical levels (Moller, 1974, Schreiner and Langner, 1988) of the central auditory system. In summary, a review of the research into modulation processing gives some indication that the central auditory system acts as a modulation frequency filterbank that transforms input from the cochlea via filters tuned to unique combinations of spectral modulation frequency, temporal modulation frequency, and audio frequency. The possibility that these filterbanks exist is supported primarily from physiological observations. Here I use a behavioral perceptual learning paradigm to explore this proposal. Perceptual Learning of Spectral and Spectro-Temporal Modulation.

  17   In this dissertation, I used perceptual learning to explore the central processing of spectral and spectro-temporal modulation. Until recent decades, it was thought that the adult auditory system could not be changed by experience. However research over the past 20 years has shown that in adults, performance on many basic auditory tasks can improve with practice – a process called perceptual learning (e.g., Wright and Zhang, 2009a, b). Investigations of perceptual learning have both theoretical and practical implications. On the theoretical level, perceptual learning experiments can reveal properties of the aspect of processing that was modified by training such as the selectivity of that aspect for stimulus features as well as how that aspect affects performance on other tasks (e.g., Wright, 2001). These inferences rely on the widely used assumption that generalization of learning from a trained condition to an untrained condition occurs if and only if the neural circuitry modified during training also influences performance on the untrained condition (Ahissar and Hochstein, 2004, Wright and Zhang, 2009b). Comparisons between patterns of generalization and those of physiological responses can be used to place constraints on the anatomical site(s) of modification (Julez, 1971). On the practical level, perceptual learning experiments indicate that perception can improve. Such improvements could potentially help remediate impaired sensory processing (e.g., Merzenich et al., 1996, Tallal et al., 1996, Moore et al., 2001), or to enhance normal sensory processing to expert levels (e.g., Sowden et al., 2000, Kraus and Chandrasekaran, 2010). Investigations from the visual system suggest that perceptual learning could be used to identify the transformations of peripheral input that occur in central sensory systems. Multi-day training to detect visual sinusoidal gratings (the visual analog of auditory sinusoidal spectral modulation detection) led to improvements that were tuned to the trained spatial frequency (Sowden et al., 2002, Huang et al., 2008, Huang et al., 2009). Further, the bandwidth of that

  18   tuning matched other estimates of the bandwidth of spatial frequency filters derived from either behavioral or physiological measurements (DeValois and DeValois, 1988). Therefore, if filters tuned to spectral or spectro-temporal modulation frequency underlie auditory processing, their existence could be revealed by examining the generalization of perceptual learning to untrained stimuli. Prior to the current work, little was known about how practice influences the perception of spectral and spectro-temporal modulation. The previous work gave some indication that sensitivity to spectral modulation shape can improve with practice. However, there has been no investigation of the generalization of learning to untrained stimulus features, thereby precluding the opportunity of observing tuning to the trained modulation frequency. In the only two previous investigations of the influence of training on spectral modulation perception in humans (Kidd et al., 1986, Drennan and Watson, 2001), listeners were asked to distinguish a reference stimulus comprised of multiple simultaneous pure-tone components from a signal that had an increase in the intensity of one of the components, but was otherwise identical to the reference except for a randomly selected higher or lower overall level (profile analysis; for review see Green, 1987). Performance on the trained stimulus improved gradually over multiple days, but it is not possible to gain a deeper understanding of the nature of the influence of this training because neither the influence of the characteristics of the trained stimulus on improvement nor the generalization to untrained stimuli were evaluated. Though I am not aware of any investigation of training on spectro-temporal modulation perception, that influence has been studied on temporal modulation alone (Fitzgerald and Wright, 2005, 2011). In these studies, both detection of temporal modulation and discrimination of temporal modulation frequency improve

  19   with practice, and those improvements are largely, but not entirely, specific to the features of the trained modulation and the trained task (detection vs. discrimination). Here, I examined how any learning induced by training with a single spectral or spectrotemporal modulation generalizes to untrained modulation frequencies. A lack of generalization to untrained modulation frequencies was interpreted to indicate that perception can be influenced in a manner that is dependent on modulation frequency. Such a result was interpreted to be consistent with the spectral and/or spectro-temporal modulation frequency filterbank proposal. Perceptual Learning in Clinical Populations. Finally, I was also interested in how training affected performance in a population known to have difficulties perceiving spectral modulation: older listeners with hearing impairment (OHI). As mentioned above, OHI listeners have widened peripheral auditory filters (Glasberg and Moore, 1986). These widened filters reduce the internal representation of modulation depth, and contribute, in part, to difficulties in speech perception (Bacon and Brandt, 1982, Sidwell and Summerfield, 1985, Van Tasell et al., 1987, Henry and Turner, 2003, Henry et al., 2005, Won et al., 2007). These OHI listeners are a population that could receive considerable benefit from auditory training on spectral modulation. Therefore I explored how practice can influence spectral modulation perception in this population. I also recognized that by examining the influence of the same training regimen in young adults with normal hearing (YNH) and older listeners with impaired hearing (OHI), I could determine whether the influence of training is population dependent. While previous investigations have shown that training can lead to improved perception in OHI listeners, especially on linguistic tasks (e.g., Walden et al., 1981, Robinson and Summerfield, 1996, Stecker et al., 2006, Sweetow and Sabes, 2006, Burk and Humes, 2007, 2008), comparisons

  20   across YNH and OHI populations given the same training regimen are lacking. Similarly, little is known about how OHI listeners learn on basic auditory tasks. On the theoretical level, the comparison of learning between populations can provide some insight into whether perceptual learning is mediated by similar or different factors between populations. On the practical level, it is important to examine whether knowledge about perceptual learning derived from YNH listeners can be applied to OHI listeners. The possibility that the influence of spectral modulation detection training might differ between OHI and YNH listeners receives some support from reports that the influence of the same training regimen can differ between clinical and non-clinical populations on visual tasks. Most of this visual work comes from training individuals with amblyopia, (Polat et al., 2004, Fronius et al., 2006, Huang et al., 2008, Astle et al., 2010), a disorder in which individuals are unable to perceive fine spatial details despite having a physically normal eye (Attebo et al., 1998).

This training often involves repeated practice using a single spatial frequency. In

comparison to individuals with normal vision, the learning of amblyotes shows a longer time course (Li et al., 2008) and a larger magnitude (Polat et al., 2004, Chen et al., 2008), and this learning generalizes to a broader range of untrained spatial frequencies (Huang et al., 2008, Huang et al., 2009, Astle et al., 2010). If similar population differences exist in the auditory system, it is possible that spectral modulation detection training could differ between OHI and YNH listeners. The Current Work. In this dissertation, I used perceptual learning to explore the processing that underlies perception of spectral and spectro-temporal modulation. I also compared the influence of spectral

  21   modulation training between YNH and OHI listeners. The experiments were primarily designed to address the following five questions: 1. Can detection of spectral modulation detection improve with practice? (Chapter 2) 2. If so, are improvements mediated by processing that is tuned to spectral modulation frequency and/or audio frequency? (Chapter 2) 3. Can the discrimination of spectral-temporal modulation depth improve with practice? (Chapter 3) 4. If so, are improvements mediated by processing that is tuned to unique combinations of spectral and temporal modulation frequency? (Chapter 3) 5. Does the aspect of processing modified by training on spectral modulation detection differ between YNH and OHI listeners? (Chapter 4) I addressed these questions with a series of experiments that all used the same design. In a pre-test, I tested the performance of all listeners on a variety of modulation perception conditions. I then trained a subset of listeners. Training consisted of seven days of practice for ~1 hr/day on a single modulation perception condition. I then tested all listeners in a post-test that was identical to the pre-test. Improvements in the trained listeners were always compared to those of a separate group of controls who did not participate in the training, but were otherwise identical to the trained listeners. Therefore, the influence of the training phase itself on the trained and untrained conditions was isolated by comparing the improvements between the trained listeners and controls.

  22   CHAPTER 2. Perceptual learning of auditory spectral modulation detection Abstract Normal sensory perception requires the ability to detect and identify patterns of activity distributed across the receptor surface. In the visual system, the ability to perceive these patterns across the retina improves with training. This learning differs in magnitude for different trained stimuli and does not generalize to untrained spatial frequencies or retinal locations. Here we asked whether training to detect patterns of activity across the cochlea yields learning with similar characteristics. Differences in learning between the visual and auditory systems would be inconsistent with the suggestion that the ability to detect these patterns is limited by similar constraints in these two systems. We trained three groups of normal-hearing listeners to detect spectral envelopes with a sinusoidal shape (spectral modulation) at either 0.5, 1, or 2 cycles/octave and compared the performance of each group to that of a separate group that received no training. On average, as the trained spectral modulation frequency increased, the magnitude of training-induced improvement and the time to reach asymptotic performance decreased, while the tendency for performance to worsen within a training session increased. The training-induced improvements did not generalize to untrained spectral modulation frequencies or untrained carrier spectra. Thus, for both visual-spatial and auditory-spectral modulation detection, learning depended upon and was specific to analogous features of the trained stimulus. Such similarities in learning could arise if, as has been suggested, similar constraints limit the ability to detect patterns across the receptor surface between the auditory and visual systems.

  23   Introduction One of the primary functions of a sensory system is to detect and identify patterns of activity distributed across the receptor surface. In the visual system, the activity pattern across the retina reflects the distribution of light in space and provides a primary cue for visual object recognition. In the auditory system, the analogous information is conveyed through the activity pattern across the cochlea, reflecting the peaks and valleys of sound level spread across audio frequency (the spectral envelope). The ability to detect and discriminate visual patterns improves with practice, indicating that the perception of these patterns is malleable (e.g., Mayer, 1983, Sowden et al., 2000, Sowden et al., 2002, Adini et al., 2004, Polat et al., 2004, Yu et al., 2004, Wenger and Rasche, 2006, Zhou et al., 2006, Huang et al., 2008, Huang et al., 2009), but it is not known whether the characteristics of this visual learning are mirrored in the auditory system. Differences in learning between these two systems would imply that different constraints underlie improvements in the ability to detect these patterns, while similarities would be consistent with the idea (Shamma, 2001) that the underlying constraints are comparable. Here, to enable comparison to visual learning, we investigated the extent to which training-induced improvements on an auditory spectral-modulation detection task are influenced by and specific to basic characteristics of the trained stimulus. In the visual system, training-induced improvements in the ability to detect luminancedefined spatial patterns differ in magnitude for different trained stimuli and fail to generalize to most untrained stimulus features. In the majority of training experiments of visual contrast detection, participants were asked to distinguish a uniform-contrast image from a sinusoidal grating. The amplitude of that sinusoid was varied adaptively to determine the minimum contrast required to detect the grating. Performance on this contrast-detection task gradually improved

  24   across multiple days of practice (Mayer, 1983, Sowden et al., 2002, Polat et al., 2004, Wenger and Rasche, 2006, Zhou et al., 2006, Huang et al., 2008). However, the magnitude of the improvement on the trained condition varied across different trained stimuli. Huang et al. (2008), noted large improvements on contrast detection for a spatial frequency of ~27 cycles/degree, but only smaller, if any, improvements for a spatial frequency of ~10 cycles/degree. Further, the learning was specific to a subset of the features of the stimulus used during training. Sowden et al. (2002) reported that improvements were narrowly tuned to the trained spatial frequency (see also Huang et al., 2008), retinal location, and eye, but generalized broadly to untrained orientations. In the auditory system, while there is some indication that sensitivity to spectral envelope shape can improve with practice, there has been no investigation of the influence of the trained stimulus on learning or of the generalization of learning to untrained stimulus features. We are aware of only two previous human auditory investigations of the influence of training on spectral envelope shape perception (Kidd et al., 1986, Drennan and Watson, 2001). In both, listeners were asked to distinguish a reference stimulus comprised of multiple simultaneous pure-tone components from a signal that had an increase in the intensity of one of the components, but was otherwise identical to the reference except for a randomly selected higher or lower overall level (profile analysis, Green, 1987). Performance on the trained stimulus improved gradually over multiple days, but comparison to the visual results described above is not possible because neither the influence of the characteristics of the trained stimulus nor generalization were tested. Further precluding this comparison, the type of trained stimuli differed considerably between the visual (sinusoidal grating) and auditory (a complex with a single peak) experiments.

  25   Here we examined how practice affected the ability to detect the presence of each of three different auditory spectral shapes and how improvements in that ability generalized to untrained stimuli. To do so, we trained listeners to detect the presence of auditory sinusoidal spectral modulation (Eddins and Bero, 2007), a task that parallels the one used in the visual contrast-detection training experiments (Sowden et al., 2002). In this task, listeners distinguished a reference noise with a flat spectral envelope (Fig. 2.1, left) from a signal with a spectral envelope that had a sinusoidal shape on a logarithmic frequency axis (Fig. 2.1, right).

The

frequency of the sinusoid, the spectral modulation frequency, was measured in cycles/octave (cyc/oct). To investigate whether the particular properties of the stimulus used during training affect learning on this spectral-modulation detection task, we trained three separate groups of listeners, each with a different stimulus, using a multiple-day training regimen. To determine the pattern of generalization to untrained stimuli on this task, before and after training we tested listeners on stimuli that differed from the trained stimulus in carrier spectrum (i.e., cochlear location) and spectral modulation frequency. The learning patterns were qualitatively similar to those previously observed with visual stimuli suggesting that similar factors may limit improvement in the ability to detect patterns of activity distributed across the receptor surface for these two senses. Method Overview. Three separate groups of trained listeners and two separate groups of controls participated in this study. All trained listeners participated in an initial screening, a pre-training session, a training phase, and a post-training session. During the screening, pure tone detection thresholds were measured at octave frequencies from 250-8000 Hz. In the pre-training session, performance was evaluated on the trained, and two or three untrained, spectral modulation

  26   detection conditions. For all trained listeners the training phase consisted of seven daily practice sessions (each approximately 1 hr in length) in which thresholds were measured repeatedly on a single spectral modulation detection condition. That condition differed across the three trained groups. The post-training session followed the training phase and was identical to the pretraining session. The pre-training session and first day of training were conducted on consecutive days, as were the final day of training and the post-training session. The controls participated in all of the same stages, except for the training phase. Thus, any difference between the trained and control groups can be attributed to the training phase. The pre- and post-tests were separated by an average of 15.6 days for the trained listeners and 14.9 days for the controls. The order of the conditions in the pre- and post-training sessions was randomized across listeners, but held constant between the pre- and post-training sessions for each individual listener. Conditions. The trained condition differed across the three trained groups.

The

conditions tested in the pre- and post-training sessions were the same for two of the trained groups, but differed for the third, and thus two different control groups were employed. Two of the trained groups practiced detecting either 0.5 (n = 8) or 1 (n = 12) cyc/oct spectral modulation spanning 200-1600 Hz. During the pre- and post-training sessions, these two groups were tested on their ability to detect 0.5, 1, and 2 cyc/oct spectral modulation spanning 200-1600 Hz as well as 1 cyc/oct spectral modulation spanning 1600-12800 Hz.

One group of controls (Control

Group 1: n = 12) was tested on the same conditions as the 0.5- and 1-cyc/oct trained listeners. The third trained group practiced detecting 2 cyc/oct (n = 7) spectral modulation spanning 4003200 Hz and were tested on their ability to detect 1, 2, and 4 cyc/oct spectral modulation spanning 400-3200 Hz during the pre- and post-training sessions. Another group of controls (Control Group 2: n = 8) was tested on the same conditions as the 2-cyc/oct trained listeners.

  27   Subsets of these listeners were also tested on modulation-masking and speech-identification-innoise conditions before and after training, however for the purposes of this paper we limit our analyses to the spectral modulation detection performance. Task and Procedure. In the spectral-modulation detection task, listeners had to distinguish a signal, spectrally modulated stimulus (Fig. 2.1, right) from a reference, flatspectrum stimulus (Fig. 2.1, left). Stimuli were presented using a three-alternative, forcedchoice method. On a given trial, three intervals, two containing the reference stimulus and one containing the signal were presented in random order. Listeners indicated which of the three intervals contained the signal stimulus by using a computer mouse to click on a visual display. After every trial, visual feedback was provided indicating whether the response was correct or incorrect. The modulation depth (peak to valley difference in dB) was adjusted adaptively across trials to estimate the spectral modulation detection threshold. Modulation-depth adjustment followed a 3-down/1-up rule and therefore converged on the 79.4% correct point on the psychometric function (Levitt, 1971). The modulation depths at which the direction of change reversed from decreasing to increasing or vice versa are referred to as reversals. The depth was initially 20 dB and was adjusted in steps of 2 dB until the third reversal; subsequent steps were 0.4 dB. In each block of 60 trials, the first three reversals were discarded, and the modulation depths at the largest remaining even number of reversals were averaged and taken as the spectral modulation detection threshold. Blocks that contained fewer than 7 reversals (5 % in total) or single trials that were longer than 20 sec (from the first observation interval through the response, 2 % in total) were excluded from analysis.

  28   During the pre- and post-tests, listeners completed four threshold estimates (240 trials) for each of the tested modulation conditions. During each session of the training phase, listeners completed twelve threshold estimates (720 trials) for the single trained condition. Stimulus Synthesis. The protocol for stimulus generation was adapted from a previous study on spectral modulation detection (Eddins and Bero, 2007). An 8192-point buffer was first filled with a sinusoid computed on a log2 frequency axis with the appropriate spectral modulation frequency (0.5, 1, 2, or 4 cyc/oct) and modulation depth (expressed in dB). The sinusoid was first multiplied by an equivalently sized buffer filled with randomly numbers drawn from a Gaussian distribution, and then multiplied by the magnitude response of a Butterworth filter (-32 dB/octave) with cutoff frequencies that were determined by the condition (200-1600 Hz, 160012800 Hz, or 400-3200 Hz). The resulting magnitude response was combined with a random phase spectrum and the real inverse Fourier transform was computed. Once in the time domain, the sound was shaped by a 100-ms amplitude envelope with 10-ms raised cosine on/off ramps. Finally the stimuli were scaled to have the same RMS amplitude. Two steps were taken to help prevent the listeners from basing detection on the use of local level cues (comparing the intensity at a single frequency across intervals). First, the phase of the sinusoid that determined the spectral modulation frequency and depth was randomly selected from a uniform distribution spanning 0-2π, causing the spectral peaks and valleys to be located randomly in frequency. Second, the presentation level on each observation interval was roved +/- 8 dB around a spectrum level of 35 dB SPL. This synthesis procedure was repeated before each stimulus presentation. Stimulus Presentation. All stimuli were presented using custom software written in MATLAB and played through a 16-bit digital-to-analog converter (Tucker-Davis-Technologies

  29   DD1) followed by an anti-aliasing filter with a 16-kHz cutoff frequency (TDT FT6-2), a programmable attenuator (TDT PA4), a sound mixer (TDT SM3), and a headphone driver (TDT HB6). The sounds were presented through the left earpiece of Sennheiser HD265 circumaural headphones. Listeners were tested in a sound-attenuated room. Listeners. Forty-seven participants (29 female) between 18 and 40 years of age served as listeners. Listeners had normal hearing sensitivity (< 20 dB HL) in the test ear at the standard audiometric frequencies as measured in the screening session and no previous experience with psychoacoustic tasks. All listeners gave informed consent and were financially compensated for their participation. All procedures were approved by the Institutional Review Board at Northwestern University. Listeners whose pre-training thresholds were greater than two standard deviations above the mean of all listeners on a particular condition were removed from analysis of that condition (3.5% of the entire dataset). Results Performance on the Trained Conditions. Spectral modulation detection thresholds improved with training, but the influence of practice depended upon the trained stimulus. The listeners who were trained using the lowest spectral modulation frequency (0.5-cyc/oct spectral modulation spanning a 200-1600 Hz carrier) improved gradually across multiple sessions. The thresholds of these listeners decreased by 4.9 dB, from 14.4 dB (the highest average pre-training threshold of the three trained groups) to 9.5 dB (Fig. 2.2A, circles). This improvement was confirmed by both a significant negative slope of a single line fitted to the population of withinlistener daily mean thresholds over the log10 of the session number (slope = -4.7, p < 0.0001), and a significant one-way analysis of variance (ANOVA) using session number as a repeated measure (F8,56 = 6.4, p < 0.0001). Both statistics were calculated across all sessions, including

  30   the pre- and post-training tests. The controls for this trained group (Control Group 1) also improved between the pre- and post-training tests (t11 = 2.4, p = 0.04) (Fig. 2.2A, diamonds). However, the magnitude of improvement of the trained listeners was larger than that of the controls, as determined by an analysis of covariance (ANCOVA) computed on the post-training thresholds after accounting for the influence of pre-training threshold1 (F1,17 = 5.1, p = 0.037). As can be seen in the average learning curve, the largest improvement in the trained listeners occurred between the pre-training test and the first training session. Nevertheless, the acrosssession improvement was still significant when the pre-training test was removed from the analyses (slope = -3.36; p = 0.028; ANOVA p = 0.008). The listeners who were trained to detect the intermediate spectral modulation frequency (1-cyc/oct spectral modulation spanning a 200-1600 Hz carrier) also improved with practice, but with most of the learning occurring early in training. The thresholds of these listeners decreased from 9.8 dB (the second highest average pre-training threshold) to 7.1 dB, an improvement of 2.7 dB (Fig. 2.2B, triangles).

As a group, these listeners improved significantly when

performance was evaluated across all of the sessions including the pre- and post-training tests (slope = -2.1; p = 0.01; ANOVA; p = 0.001). They also improved more than the controls for this trained group (Control Group 1) (ANCOVA; F1,20 = 4.3, p = 0.05) who themselves did not

1

The ANCOVA should be interpreted with caution because a test of the heterogeneity of regression was nearly significant (p = 0.06), suggesting that the relationship between the pre- and post-training thresholds differed between groups, which would violate an ANCOVA assumption. Nevertheless, either the ANCOVA or the heterogeneity of regression indicate that the 0.5 cyc/oct-trained listeners distinguished themselves from controls, suggesting an influence of the training phase. Also consistent with this conclusion, the 0.5-cyc/oct-trained listeners had significantly lower thresholds than the controls at the post-training test but the thresholds of the two groups did not differ at the pre-training test. Note that, for all other ANCOVAs reported, the heterogeneity of regression was not significant.

 

improve between the pre- and post-training tests (t10

31   = 0.69, p = 0.50).However, when the pre-

training test was omitted from the analyses the trained listeners showed no improvement across sessions (slope = -0.76; p = 0.55; ANOVA; p = 0.38), indicating that most of their improvement occurred between the pre-training test and the first training session. The performance of the controls on this condition is particularly interesting in this context. The controls participated in the same pre-training test that seems to have induced the learning in the trained listeners, but did not improve between the pre- and post-training tests. Thus, it appears that by the post-training test the controls lost any improvements resulting from exposure to the pre-training test, while the practice sessions served to maintain those improvements in the trained listeners. The listeners who were trained to detect the highest spectral modulation frequency (2cyc/oct spectral modulation spanning a 400-3200 Hz carrier) did not improve. These listeners began with the lowest average pre-training threshold (6.9 dB) and ended at nearly the same value (6.6 dB) (Fig. 2.2C, squares). They showed no significant learning either across all sessions (slope = -0.99; p = 0.21; ANOVA; p = 0.47) or when the pre-training test was omitted from the analyses (slope -1.7; p = 0.15; ANOVA; p = 0.44). The controls for this trained group (Control Group 2) also showed no improvement between the pre- and post-training tests (t7 = 1.67, p = 0.14). Analyses of the individual learning-curve slopes support the same conclusions as those reached through the analyses of the average learning curves. For each trained stimulus and individual listener, we computed the slope of a regression line fitted to each threshold estimate over the log10 of the session number. Each slope is a point in Fig. 2.3. For the 0.5-cyc/octtrained listeners, the proportion of slopes that were significantly different from zero and negative (filled circles) was the same (0.75) when all sessions were included in the analysis (Fig. 2.3A,

  32   left column) as when the pre-training test was omitted (Fig. 2.3A, right column). In both cases, the population of slopes was significantly less than zero (pre-training test included: p = 0.002; excluded: p = 0.01). For the 1-cyc/oct trained listeners, the proportion of significantly negative slopes was greater when the pre-training test was included (0.58) compared to excluded (0.17) (Fig. 2.3B) and the population of slopes was significantly less than zero only when the pretraining test was included (included: p = 0.02; excluded: p = 0.30). For the 2-cyc/oct-trained listeners, the same proportion of slopes were significantly negative both with and without the pre-training test, but in neither case was the population of slopes less than zero (all p > 0.25, Fig. 2.3C). Finally, another difference in the influence of training across the three modulation frequencies was that performance within session did not change consistently for the lowest trained frequency but worsened for the other two frequencies. We investigated whether performance changed systematically within training sessions by computing, for each trained listener on each session, the mean of the first three and the last three threshold estimates. The group averages are plotted in Figure 2.4A-C. We evaluated within-session performance using a 2 time (first vs. last) by 7 session ANOVA with time as a repeated measure. There was no consistent within-session change in performance for the 0.5-cyc/oct trained listeners (time: F1,49 = 0.86, p = 0.36; time x session: F6,49 = 1.78; p = 0.12). In contrast, the listeners who practiced either of the two higher modulation frequencies showed a consistent within-session worsening (1 cyc/oct: time: F1,77 = 13.79, p < 0.001; time x session: F6,77 = 0.07, p = 0.99) (2 cyc/oct: time: F1,42 = 11.7, p = 0.001; time x session: F6,42 = 0.5, p = 0.8). Therefore, performance deteriorated within sessions for those frequencies for which there was no change in performance across training sessions.

  33   Performance on the Untrained Conditions. No trained group learned significantly more than controls on any untrained spectral modulation detection conditions. We evaluated whether training led to improvements on untrained conditions (generalization) using the same criteria we used to determine the effect of training on the trained condition (a significant ANCOVA between the trained listeners and controls, using pre-training performance as a covariate). The data displayed in Figure 2.5 show individual and group-mean thresholds at the post-training test, after adjusting the values based on their relationship to the pre-training thresholds. In each panel, the dashed line indicates the average pre-training threshold, and the horizontal box represents the 95% confidence interval of the mean of the controls’ post-training thresholds. The controls improved between the pre- and post-training tests only on the two conditions with the highest pre-training thresholds (0.5 cyc/oct, 200-1600 Hz (p = 0.04) and 1cyc/oct, 1600-12800 Hz (p < 0.001) all other p > 0.14). The trained groups did not distinguish themselves from the controls on any untrained spectral modulation frequency (Fig. 2.5, middle columns; all p > 0.18) and/or carrier spectrum (Fig. 2.5, right column; all p > 0.50). This lack of training-induced improvement on untrained conditions occurred for the two groups that learned more than controls on their trained condition (0.5 cyc/oct- and 1 cyc/oct-trained listeners) as well as for the group that did not (2 cyc/oct-trained listeners) (Fig. 2.5, left column). Thus, it appears that the effect of training was specific to the trained spectral modulation detection condition. Discussion The present data demonstrate that the influence of training spectral modulation detection in normal-hearing adults for multiple days is dependent upon the characteristics of the trained stimulus. On average, both the magnitude of training-induced improvement and the time to reach asymptotic performance decreased as the trained spectral modulation frequency increased from

  34   0.5 to 2 cyc/oct. Pre-training thresholds also decreased as spectral modulation frequency increased. Within the training sessions, performance consistently worsened for 1 and 2 cyc/oct, but did not change for 0.5 cyc/oct. Finally, in no case did the training-induced improvements generalize to untrained spectral modulation frequencies or untrained carrier spectra. Basis for improvements. It appears that the present improvements on the trained conditions actually reflect an improved sensitivity to spectral modulation despite several potential alternatives. We considered (and ultimately rejected) two alternative accounts of the present learning. First we asked whether the observed learning might have arisen solely from an improvement in the ability to ignore the randomization of stimulus features. In the current procedure, the spectral modulation phase and presentation level were randomized to minimize the use of local (audio-frequency specific) intensity cues. Thus, one possibility is that these randomizations initially distracted listeners from the target spectral modulation detection task, but that the ability to ignore these randomizations increased with training. However, if that were the case, there would have been similar learning on all trained conditions and complete generalization to all untrained conditions, because each of the conditions used the same randomizations. Instead, listeners only improved on a subset of the trained and tested conditions, making this alternative unlikely. Second, we asked whether the observed learning might simply have resulted from improved memory of the reference spectrum. Previously observed training-induced improvements in profile analysis (Kidd et al., 1986, Drennan and Watson, 2001) have been attributed to memorization of the reference stimulus, because listeners still improve with practice on a novel reference spectrum after reaching asymptotic performance with a trained reference (Kidd et al., 1986). However, if the present learning were due to the memorization of the

  35   reference (a flat-spectrum bandpass noise), it would have generalized to all conditions that employed the same reference. Instead there was no such generalization. The lack of support for these alternative accounts, leads us to the idea that the present learning resulted from the enhancement of sensitivity to spectral modulation. Stimulus Dependence of Learning. The differences in the learning patterns across the three trained stimuli and lack of generalization among them suggests that different factors may limit performance for different spectral modulation frequencies. Though the trained stimuli differed in carrier spectrum (the spectrum of the trained stimulus for the 2-cyc/oct group was 1 octave higher than that for the 0.5- and 1- cyc/oct groups), there are at least two reasons to think that the differences in learning were instead due to the spectral modulation frequency. First, the rate and magnitude of learning as well as the pattern of within-session performance differed between the 0.5- and 1-cyc/oct trained groups even though the carrier spectrum was the same (200-1600 Hz) for both trained stimuli (Figs. 2.2-2.4). In addition, the controls improved more at 0.5 than at 1 cyc/oct with that same carrier spectrum. These differences show that the spectral modulation frequency itself can influence learning separately from the carrier spectrum. Second, there was no consistent relationship between carrier spectrum and improvement. In a direct comparison of improvement with different carriers but the same spectral modulation frequency (the only one possible in this data set), the magnitude of improvement at 1 cyc/oct in controls was greater with the higher (1600-12800 Hz) than the lower (200-1600 Hz) carrier spectrum. This result indicates that the carrier spectrum can affect the magnitude of pre-test induced learning on spectral modulation detection. However, for training-induced learning, the pattern was reversed. The improvement was greater with the lower (200-1600 Hz at 0.5 or 1 cyc/oct) rather than the higher (400-3200 Hz at 2 cyc/oct) carrier spectra. The opposite influence of

  36   increasing the frequency range of the carrier spectrum for pre-test and training-induced learning suggests that the carrier spectrum is unlikely to have been the dominant feature that determined the learning pattern. Instead, it appears that the different training-induced learning outcomes with the present three trained stimuli were determined primarily by the spectral modulation frequency. The stimulus dependence of learning observed here is consistent with evidence that different factors limit spectral modulation detection at different spectral modulation frequencies. Eddins and Bero (2007) quantified the modulation depth in the excitation pattern (an approximation of the peripheral representation, Moore and Glasberg, 1987) at the spectral modulation detection threshold for a range of spectral modulation frequencies. They reasoned that if spectral modulation detection performance for all stimuli were limited by the modulation depth in the excitation pattern (the peak to valley difference), then the threshold level of depth in that pattern would be identical across all detectable spectral modulation frequencies, but this was not the case. For spectral modulation frequencies greater than 2 cyc/oct, listeners were highly sensitive to spectral modulation, requiring about 1 dB of modulation depth in the excitation pattern for detection. However, as the spectral modulation frequency decreased below 2 cyc/oct, detection required increasingly greater modulation depth in the excitation pattern, reaching about 7 dB at 0.25 cyc/oct. A similar analysis reported by Summers and Leek (1994) revealed the same pattern. These analyses suggest that a factor beyond the depth of modulation in the excitation pattern affects the detection of spectral modulation. One possibility is that this modulation is detected by a mechanism that compares the output levels across audio frequency channels to find the peaks and valleys in the excitation pattern, and that the mechanism’s capacity to make these comparisons decreases as the channel separation between the peaks and valleys increases (i.e., spectral modulation frequency decreases). If we extend this idea to the current investigation, we

  37   can account for the results by assuming that the capacity to compare across nearby channels is already at asymptotic performance before training, but that the capacity to compare across more distant channels can improve with practice. Interpreted in this context, the current results further suggest that the mechanism modified by training was selective for both spectral modulation frequency and carrier spectrum. If the mechanism were not selective for these features, improvements would have generalized broadly. It is of interest to note that neurons tuned to combinations of spectral modulation frequency and carrier spectrum have been documented in ferret and cat auditory cortex (Schreiner and Calhoun, 1994, Shamma et al., 1995, Kowalski et al., 1996, Calhoun and Schreiner, 1998, Klein et al., 2000, Keeling et al., 2008). Selectivity for spectral modulation frequency using stimuli with the same carrier has also been observed in humans. For such stimuli, the ability to detect a target spectral modulation in the presence of an interfering modulation decreases as the modulation frequencies of the target and interferer become more similar (Saoji and Eddins, 2007). There is also evidence that portions of human auditory cortex identified with fMRI are tuned to specific ranges of spectral modulation frequency (Langers et al., 2003, Schonwiesner and Zatorre, 2009). Thus the current training may have resulted from increased sensitivity to modulation in units such as these (or other units with similar selectivity), or from optimizing the weights of such units on a more central decision maker (for discussions of these two views of perceptual learning see (Dosher and Lu, 1998) and (Ahissar and Hochstein, 2004)). Practical Implications. The present results demonstrate that the detection of spectral modulation at low frequencies (< 1 cyc/oct) can be improved with training, and thus suggest that training might lead to improvement on real-world tasks for which performance is limited by the

  38   ability to detect modulation at these frequencies. The ability to detect low spectral modulation frequencies appears to be important for several real-world tasks. For example, in individuals with cochlear implants, the ability to detect low (< 0.5 cyc/oct) spectral modulation frequencies is positively correlated with the ability to identify speech sounds (Litvak et al., 2007, Saoji et al., 2009). This result suggests that for cochlear implant users, improved spectral modulation detection at these frequencies might lead to improvements in speech perception. In addition, in normal-hearing listeners, vertical sound localization appears to depend upon detection of low spectral modulation frequencies < 1 cyc/oct; (Macpherson and Middlebrooks, 2003, Qian and Eddins, 2008), suggesting that improved sensitivity to these modulations might aid performance on this task. However, the specificity of training-induced learning to the trained spectral modulation frequency and carrier spectrum implies that spectral modulation detection training will only be effective if listeners train at the point along these dimensions that is most crucial for the target real-world task. Comparison to Learning on Visual Contrast Detection. The influence of practice on auditory spectral modulation detection documented here shares several qualitative similarities to that previously reported for the analogous visual task (contrast detection of a sinusoidal grating). In both cases the influence of training depended upon characteristics of the trained stimulus, such that improvements were largest for modulation frequencies where naïve performance was poorest (Huang et al., 2008, Huang et al., 2009). The modulation transfer functions representing naïve performance are “bowl-shaped” as a function of spatial (DeValois and DeValois, 1988) or spectral modulation (Summers and Leek, 1994, Eddins and Bero, 2007) frequency, with performance being best for middle frequencies and worse at either extreme.

Training at

frequencies near the edges of the transfer function led to large improvements in performance

  39   while training at frequencies in the flat portion led to little, if any, improvement in the visual (Huang et al., 2008) as well as the auditory (here) systems. Further, contrast detection improvements in the visual system were specific to the trained spatial frequency and retinal location (Sowden et al., 2002, Huang et al., 2008), and the current auditory improvements were specific to the analogous features: spectral modulation frequency and cochlear location (i.e., carrier spectrum). Finally, in both cases, the specificity of training-induced improvements resembled the selectivity of neurons in the primary sensory cortex associated with the trained modality (visual (e.g., Tootell et al., 1981) or auditory (e.g., Kowalski et al., 1996)). Thus it appears that improvements in the ability to detect the presence of patterns of activity distributed across the sensory epithelium might be mediated by similar mechanisms in the auditory and visual systems.

  40   CHAPTER 3. Auditory perceptual learning specificity to combined spectro-temporal modulation Abstract Natural sounds are characterized by complex patterns of sound energy distributed across both frequency (spectral modulation) and time (temporal modulation). Perception of these patterns has been proposed to depend upon a bank of modulation filters, each tuned to a unique combination of a spectral and a temporal modulation frequency. There is considerable physiological evidence for such combined spectro-temporal tuning. However direct behavioral evidence is lacking. Here we examined the processing of spectro-temporal modulation behaviorally using a perceptual-learning paradigm. We trained human listeners ~1 hr/day for 7 days to discriminate the depth of either spectral (0.5 cyc/oct; 0 Hz), temporal (0 cyc/oct; 32 Hz), or upward spectro-temporal (0.5 cyc/oct; 32 Hz) modulation. Each trained group learned more on their respective trained condition than controls who received no training. Critically, this depth-discrimination learning did not generalize to the trained stimuli of the other groups or to downward spectro-temporal (0.5 cyc/oct; -32 Hz) modulation. Learning on discrimination also led to worsening on modulation detection, but only when the same spectro-temporal modulation was used for both tasks. Thus, these influences of training were specific to the trained combination of spectral and temporal modulation frequencies, even when the trained and untrained stimuli had one modulation frequency in common. This specificity indicates that training modified circuitry that had combined spectro-temporal tuning, and therefore that circuits with such tuning can influence perception. These results are consistent with the possibility that the auditory system analyzes sounds through filters tuned to combined spectro-temporal modulation.

  41   Introduction Most natural sounds have peaks and valleys of energy that vary across both frequency and time. Accurate perception of these spectro-temporal modulations is necessary for fundamental auditory skills such as the discrimination and grouping of auditory objects (Bregman, 1990, Woolley et al., 2005) and the perception of speech (Elhilali et al., 2003a). It has been proposed that the perception of spectro-temporal modulation is mediated by a bank of filters, with each filter tuned to a particular combination of a spectral and a temporal modulation frequency (Chi et al., 1999, Chi et al., 2005). At present there is considerable physiological, but little behavioral, evidence for such spectro-temporal tuning. A schematic of the proposed spectro-temporal modulation filterbank is shown in Figure 3.1. Each spectrogram depicts the spectro-temporal modulation to which a particular filter is tuned. Each filter has a preferred spectral modulation frequency (vertical spacing of bars), temporal modulation frequency (horizontal spacing of bars), and direction (up or down: direction of bar tilt). Note that the filters that are selective for isolated spectral (middle column) or temporal (bottom row) modulation frequencies simply constitute one “slice” of this spectrotemporal modulation filter bank where the preferred modulation frequency in the other dimension is zero. Models involving such filterbanks have successfully described performance on a variety of psychophysical tasks (Carlyon and Shamma, 2003, Elhilali et al., 2003b). These demonstrations indicate that combined spectro-temporal processing could underlie perception, but do not provide direct behavioral evidence that is does. Current evidence for combined spectro-temporal tuning comes only from physiological observations. Such tuning has been observed in individual neurons in inferior colliculus (e.g., Versnel et al., 2009), thalamus (e.g., Miller et al., 2002), and auditory cortex (e.g., Kowalski et al., 1996, Depireux et al., 2001), as

  42   well as in population responses identified with fMRI in human auditory cortex (Langers et al., 2003, Schonwiesner and Zatorre, 2009). Investigations focusing on a single modulation dimension have also revealed tuning along the tested dimension (temporal: e.g., Schreiner and Langner, 1988) (spectral: e.g., Shamma et al., 1995). Behavioral support for modulation tuning comes from modulation-frequency-specific effects of masking (temporal: e.g., Bacon and Grantham, 1989) (spectral: Saoji and Eddins, 2007), adaptation (temporal: e.g, Wojtczak and Viemeister, 2003) (spectral: D.A. Eddins, unpublished observations), expectation (temporal: Wright and Dai, 1998), and training (temporal: e.g., Fitzgerald and Wright, 2005, Fitzgerald and Wright, 2011) (spectral: A.T. Sabin, unpublished observations), on modulation perception. However, in all of these cases only one dimension of modulation was manipulated, thereby preventing potential observations of combined spectro-temporal tuning. As a potential means to observe behavioral evidence of combined spectro-temporal tuning in humans, we examined the influence of training on the discrimination of modulation depth.

We reasoned that if training with a particular combination of spectral and temporal

modulation frequencies modified processing that was tuned to that combination, then traininginduced improvements would not generalize to any untrained combinations. Such specificity could arise from a change in a modulation filter itself, or in the weight assigned to that filter by a decision maker. To test this possibility, we trained three groups of listeners to discriminate the depth of either spectral-only, temporal-only, or spectro-temporal modulation, and examined generalization of depth-discrimination learning to the trained conditions of the other groups and to downward spectro-temporal modulation. To assess specificity to the depth-discrimination task, we also examined generalization to the detection of upward spectro-temporal modulation. The three training regimens each led to depth-discrimination learning that was largely restricted to

  43   the trained combination of modulation frequencies, and had different influences on the ability to detect spectro-temporal modulation ranging from improvement to worsening. This result provides some behavioral support for the proposal that filters tuned to combinations of spectral and temporal modulation underlie modulation perception. Method Overview. We tested three separate groups of trained listeners and one group of controls. All trained listeners participated in an initial screening, a pre-test, a training phase, and a posttest. During the screening, we measured pure-tone detection thresholds at octave frequencies from 250 to 8000 Hz. In the pre-test, performance was evaluated on five modulation perception conditions, including the trained one. For the trained listeners, the training phase consisted of seven daily practice sessions (each approximately 1 hr in length) in which thresholds were measured repeatedly on a single modulation depth-discrimination condition. The trained condition differed across the three trained groups. The post-test followed the training phase and was identical to the pre-test. The controls participated in all of the same stages, except for the training phase. Thus, any difference between the trained and control groups can be attributed to the training phase. Conditions. The pre- and post-tests consisted of four modulation depth-discrimination conditions and one modulation detection condition. Each modulation is characterized by both a spectral modulation frequency (in cycles/octave) and a temporal modulation frequency (in Hz). The four depth-discrimination conditions were spectral modulation alone (0.5 cyc/oct, 0 Hz), temporal modulation alone (0 cyc/oct, 32 Hz), upward spectro-temporal modulation (0.5 cyc/oct, 32 Hz), and downward spectro-temporal modulation (0.5 cyc/oct, -32 Hz). The detection condition was upward spectro-temporal modulation (0.5 cyc/oct, 32 Hz). Each of the three

  44   trained groups practiced a different depth-discrimination condition from the pre- and post-tests, either spectral modulation alone (n = 8), temporal modulation alone (n = 8), or upward spectrotemporal modulation (n = 8). Tests of vowel and consonant identification in noise were also included in the pre- and post-tests, but those data are not reported here. Tasks and Procedure. For the modulation depth-discrimination conditions listeners had to distinguish a modulated noise stimulus with a 50 dB depth (standard) from a one with a shallower depth (signal). Stimuli were presented using a three-interval, forced-choice method. On each trial, the standard was presented in two intervals, and the signal in one, with the signal interval determined randomly. Listeners indicated which of the three intervals contained the signal by using a computer mouse to click on a visual display. After every trial, visual feedback was provided indicating whether the response was correct or incorrect. The signal modulation depth (peak to valley difference in dB) was adjusted adaptively across trials to estimate the modulation depth-discrimination threshold.

Modulation-depth

adjustment followed a 3-down/1-up rule and therefore converged on the 79.4% correct point on the psychometric function (Levitt, 1971). The modulation depths at which the direction of change reversed from decreasing to increasing or vice versa are referred to as reversals. The depth was initially 0 dB and was adjusted in steps of 6 dB until the third reversal; subsequent steps were 2 dB. In each block of 60 trials, the first three reversals were discarded, and the modulation depths at the largest remaining even number of reversals were averaged. The difference between this average and the standard modulation depth is reported as the depthdiscrimination threshold. Blocks were excluded from analysis if they contained fewer than 7 total reversals (2 % in total) or any trials that took longer than 20 sec (from the first observation interval through the response, 3.4 % in total).

  45   For the modulation detection condition listeners had to distinguish an unmodulated noise stimulus (standard) from a modulated one (signal), and the modulation depth of the signal was varied adaptively to estimate the detection threshold. The procedure was the same as for depthdiscrimination except that the initial signal modulation depth was 20 dB. The average of the signal modulation depths at the included reversals is reported as the detection threshold. During the pre- and post-tests, listeners completed five threshold estimates (300 trials) for each modulation condition. The order of the conditions was randomized across listeners, but held constant between the pre- and post-tests for each individual listener. During the training phase, trained listeners completed twelve threshold estimates (720 trials) for the single trained condition on each session. The pre-test and first day of training were conducted on consecutive days, as were the final day of training and the post-test. The training sessions occurred on most weekdays. The pre- and post-tests were separated by an average of 20.3 days for the trained listeners and 22.3 days for the controls. Stimulus Synthesis and Presentation. The protocol for stimulus generation was adapted from previous reports (e.g., Kowalski et al., 1996), and primarily used code from the neural systems laboratory toolbox. The spectro-temporal envelope (S) of each stimulus was a twodimensional sinusoid defined across both time (t) and frequency (x; number of octaves above the fundamental frequency) as follows:

S(t,x) = 1+ A sin (2πωt + 2πΩx + Φ)

where A is the linear modulation depth, ω is the temporal modulation (in Hz), Ω is the spectral modulation (in cycles/octave), and Φ is the starting phase of the sinusoid. The fundamental

  46   frequency was 100 Hz and the components were spaced in steps of 2.5 Hz up to 16314 Hz. The starting phase was randomized. To generate the modulated noise, the spectro-temporal envelope was multiplied by random numbers drawn from a Gaussian distribution. The modulated noise was passed through a Butterworth filter with cutoff frequencies of 400 and 3200 Hz and a slope of -32 dB/Octave. The duration of each stimulus was 125 ms including 10-ms raised cosine on/off ramps. Due to computational constraints, the stimuli were synthesized prior to testing. To introduce some randomization, we pre-computed 25 instances of each combination of spectral modulation frequency, temporal modulation frequency, and modulation depth. On a given presentation one of those 25 stimuli was randomly selected and played to the listener. The presentation level on each observation interval was roved +/- 8 dB around a spectrum level of 35 dB SPL. The level and starting-phase randomizations were employed to reduce the availability of local intensity cues in the spectral-modulation condition, but were used in all conditions for uniformity. All stimuli were presented using custom software written in MATLAB and played through a 16-bit digital-to-analog converter (Tucker-Davis-Technologies DD1) followed by an anti-aliasing filter with a 16-kHz cutoff frequency (TDT FT6-2), a programmable attenuator (TDT PA4), a sound mixer (TDT SM3), and a headphone driver (TDT HB6). The sounds were presented through the left earpiece of Sennheiser HD265 circumaural headphones. Listeners were tested in a sound-attenuated room. Listeners. Thirty-two participants (26 female) between 18 and 36 years of age served as listeners. All had normal hearing sensitivity in the left ear (< 20 dB HL from 250-8000 Hz, re: ANSI, 1996) and no previous experience with psychoacoustic tasks. All listeners gave informed consent and were financially compensated for their participation. All procedures were approved

  47   by the Institutional Review Board at Northwestern University. The data from listeners whose pre-test thresholds were greater than two standard deviations above the mean of all listeners on a particular condition were removed from the analyses of that condition (3.1% of the entire dataset). Results Learning on the Trained Conditions. Multiple sessions of depth-discrimination training aided performance on spectral- (Fig. 3.2A; triangles), temporal- (Fig. 3.2B; diamonds), and spectro-temporal- (Fig. 3.2C; squares) modulation. We evaluated the influence of each training regimen by computing, separately for each trained condition, a group (trained vs. control) x session (pre vs. post) ANOVA, using time as a repeated measure. In each case there was a significant group by session interaction (all p < 0.03). The trained listeners improved between the pre- and post-tests (filled symbols, paired t-test, all p < 0.004) while the controls did not (open symbols, all p > 0.26). Each trained group improved over the course of the experiment, though the temporalmodulation trained listeners reached asymptote before the other two trained groups. For each trained group, this improvement was confirmed by a significant negative slope of a single line fitted to the population of within-listener daily mean thresholds over the log10 of the session number (all p < 0.003). Further, for each individual we computed the slope of a regression line fitted to each threshold estimate over the log10 of the session number (values displayed in Fig 3.2 panels D-F, left side). The population of slopes for each trained group was significantly negative (all p < 0.007), again indicating improvement in each group. However, when the same set of analyses were repeated with the pre-test removed (Fig 3.2, panels D-F, right side), the spectral and spectro-temporal modulation trained listeners still showed improvement (all p < 0.04), while

  48   the temporal-modulation trained listeners did not (all p > 0.42). These analyses suggest that participating in the pre-test brought the temporal-modulation trained listeners to asymptotic performance. Interestingly, the controls, who also participated in the same pre-test, showed no improvement, indicating that by the post-test the controls lost any improvements resulting from exposure to the pre-test, while the practice sessions served to maintain those improvements in the temporal-modulation trained listeners. The relative contributions of within- and between-session improvements that occurred during the training phase differed among the trained groups. To examine when on a given day training-phase improvements emerged, we compared within- to between-session improvements for the two groups that showed improvement during this phase (the spectral and spectro-temporal modulation trained listeners). For each listener we computed the means of the first three and of the last three threshold estimates for each training session, and used those values to calculate performance changes within sessions (first 3 minus last 3 estimates) as well as between sessions (last 3 estimates of current session minus first three of the subsequent session). The relative magnitudes of these changes were evaluated, separately for each trained group, using a change (within-session vs between-session) x session (all training days) ANOVA with both change and session as repeated measures. For spectral-modulation trained listeners, there was a main effect of change (F1,42 = 11.3; p = 0.012), which arose because the listeners tended to get worse within sessions (T55 = -2.1; p = 0.04) and to improve between sessions (T55 = 3.7; p < 0.001). There was no main effect of session and no change x session interaction for this group (all p > 0.031). Thus the improvements of the spectral-modulation trained listeners emerged during the time between sessions. In contrast, for the spectro-temporal modulation trained listeners, there were no

  49   significant main effects or interactions (all p > 0.25), indicating that for these listeners there was no marked distinction between within and between-session improvements. Generalization to Untrained Conditions. Depth Discrimination. While practice led to improved depth discrimination on each trained condition, in no case did that learning generalize to an untrained depth-discrimination condition. We evaluated generalization, separately for each untrained condition, with the same criterion that we used to assess learning on the trained condition -- a significant group (trained vs. control) x session (pre vs. post) interaction of an ANOVA, using session as a repeated measure. By this criterion, there was no evidence of generalization. None of trained groups generalized their improvement to the trained conditions of the other groups (all p > 0.12; SM Fig 3.3A, TM Fig 3.3B, STM+ Fig 3.3C), or to a depthdiscrimination condition on which none of the groups trained -- downward spectro-temporal modulation (all p > 0.13; STM- Fig 3.3D). Though none of the trained groups improved more than controls on any of untrained depth-discrimination conditions, the trained groups did show some improvement that was uniform across these conditions. We further examined performance on the untrained conditions separately for each group (3 trained and 1 control) by computing a session (pre vs. post) x condition (all untrained depth-discrimination conditions) ANOVA with session as a repeated measure. For each trained group, there was a main effect of session (all p < 0.001) and no session x condition interaction (all p > 0.55). In contrast, for the controls neither the main effect of session nor the session x condition interaction was significant (all p > 0.19). Thus, the trained listeners improved somewhat on the untrained depth-discrimination conditions, but not by amount that was distinguishable from the controls.

  50   Detection of Spectro-Temporal Modulation. Finally, the influence of practicing modulation depth-discrimination on the ability to detect upward spectro-temporal modulation (Fig. 3.4) differed across the three trained groups. On this detection condition, in comparison to controls, the spectral modulation trained listeners learned (F1,14 = 9.8; p = 0.007), the temporal modulation trained listeners showed no change (F1,14 = 0.50, p = 0.49), and the spectro-temporal modulation trained listeners actually got worse (F1,13 = 4.8; p 0.048). The decrements in performance on detection in the spectro-temporal modulation trained listeners were correlated with improvements in depth-discrimination for the same stimulus (their trained condition) (Fig. 3.4E; r=-0.80, p =0.03). There were no such tradeoffs between detection and discrimination of this stimulus in the remaining three groups (Fig. 3.4B-D; all p > 0.13), or between the detection of this stimulus and the discrimination of the respective trained stimuli of the other two trained groups (all p > 0.47). Discussion Combined Spectro-Temporal Processing. The present results provide behavioral evidence that stimuli can be processed by a mechanism that is selective for particular combinations of spectral and temporal modulation frequencies. We trained listeners to discriminate the depth of either spectral, temporal, or upward spectro-temporal modulation, and tested the generalization of their learning to the trained conditions of the other groups as well as to the depth discrimination of downward spectro-temporal modulation, and the detection of upward spectro-temporal modulation. Three influences of training were dependent upon the spectro-temporal characteristics of the trained stimulus. First, and most importantly, there was no generalization to untrained depth-discrimination conditions. Learning on isolated spectral (0.5 cyc/oct; 0 Hz) or temporal modulation (0 cyc/oct; 32 Hz) did not generalize to either direction of

  51   spectro-temporal modulation (0.5 cyc/oct; +/-32 Hz), even though the spectro-temporal conditions shared one modulation frequency with each of the isolated modulations (Fig. 3.3 CD). There was also no generalization from spectro-temporal modulation to the isolated modulations or to the opposite direction of spectro-temporal modulation (Fig. 3.3 A-B and 3.4 D). Second, there was a unique relationship between detection and discrimination when both tasks were assessed using stimuli that had the same combination of spectral and temporal modulation frequencies. Learning on depth discrimination of spectro-temporal modulation led to worsening on (negative generalization to) the detection of that spectro-temporal modulation, and the individual magnitudes of these changes were significantly correlated. The other two trained groups did not show this pattern, indicating that this relationship was specific to the combined spectro-temporal modulation. Instead, learning on spectral modulation depth discrimination led to improvements, rather than worsening, on the detection of spectro-temporal modulation. This generalization implies a connection between these two conditions, but one that differs from the connection between the two spectro-temporal conditions. Learning on temporal modulation depth discrimination did not influence performance on the detection condition, implying separate processing of these two conditions. Third, the time courses of improvement on modulation depth discrimination differed for different combinations of spectral and temporal modulation frequencies. Asymptotic performance was reached by the first training day for temporal modulation, but later in the training phase for spectral and spectro-temporal modulation. Further, the more gradual improvements occurred primarily between sessions for spectral modulation, but not for spectrotemporal modulation.

  52   It is unlikely that these outcomes would have occurred if training had modified an aspect of processing in which the spectral and temporal components were represented separately. Consider, for example, the lack of generalization to untrained depth-discrimination conditions. If training on spectro-temporal modulation had modified separate processing of spectral and/or temporal modulation, learning would have generalized to spectral-alone and/or temporal-alone modulation. Similarly, if training on these isolated modulations had modified such separate processing, learning would have generalized, at least in part, to spectro-temporal modulation. The present behavioral support for combined spectro-temporal processing is reminiscent of the combined spectro-temporal tuning observed in numerous physiological investigations. As mentioned in the introduction, individual neurons that are tuned to particular combinations of spectral and temporal modulation frequencies have been observed at cortical (e.g., Kowalski et al., 1996) and subcortical levels (e.g., Versnel et al., 2009) in animals, and similar tuning has been observed in regions of human primary and secondary auditory cortex (Langers et al., 2003, Schonwiesner and Zatorre, 2009). Thus there are many candidate neural populations that could have been involved in the learning. However, the specificity of learning to the trained direction of spectro-temporal modulation helps narrow the possibilities. Neurons that preferentially respond to one direction of spectro-temporal modulation are more common in higher than lower areas of auditory cortex (Loftus and Sutter, 2001). Therefore, it is more likely that the traininginduced modification involved higher rather than lower areas of auditory cortex. The training could have led to a refinement in these modulation-tuned neurons themselves, or in the weighting assigned to inputs from these neurons by a decision maker (for more general discussions of these alternatives see: Dosher and Lu, 1998, Ahissar and Hochstein, 2004).

  53   Despite the specificity of learning to the trained condition, each trained group showed a modest improvement across all of the untrained depth-discrimination conditions (Fig. 3.3). Thus, the trained groups appear to have learned some general aspect of modulation depth discrimination. One possibility is that this uniform improvement came from an increased ability to ignore the randomization of presentation level. Since the same level randomization was used on all conditions, an increased ability to ignore this randomization would lead to improvement on all untrained conditions. Another possibility is that the listening strategy for the modulation itself changed over the course of training. For example, listeners might have initially chosen the interval that was most correlated to an internal template of an unmodulated noise, and then later chosen the interval that was least correlated to a template of the trained standard. Refinements of the unmodulated template with practice would lead to modest uniform improvement across all depth-discrimination conditions, while refinements of the template of the trained standard would lead to improvements that were specific to the trained condition. It is also worth noting that the magnitude of the improvement between the pre- and post-tests on the untrained conditions was comparable to that between the pre-test and the first day of training on the trained conditions. This indicates that the pre-test itself was sufficient to generate some learning on untrained conditions. It is possible that further training simply maintained these improvements in the trained groups, and that these improvements declined in the controls due to the lack of training. Detection vs Discrimination. The current results also replicate and extend an existing literature indicating that improvements in modulation discrimination can come at the cost of worsening on detection. Training on depth discrimination of spectro-temporal modulation led to learning on that condition, but worsening on the detection of that same modulation (Fig. 3.4E). A related tradeoff has been observed in auditory temporal modulation perception where listeners

  54   who practiced discriminating modulation rate improved on that task, but got worse at detecting the same stimulus (Fitzgerald and Wright, 2005) (though detection training did not influence discrimination; Fitzgerald and Wright, 2011). Similarly, in the visual system, adaptation to a sinusoidal grating led to better orientation discrimination but poorer detection (Regan and Beverley, 1985). The present data extend this previous work by demonstrating that the discrimination/detection tradeoff can be specific to the trained stimulus. This tradeoff did not occur when depth-discrimination training and detection testing used different modulations (Fig. 3.4C-D). Training to discriminate spectral modulation actually led to improvements in detection of spectro-temporal modulation. There are several potential listening strategies, which if adopted over the course of training, could have led to a discrimination/detection tradeoff that is specific to the trained modulation. For example, trained listeners could have learned to focus on the processing of the trained modulation at deep modulation depths, and to ignore that of shallow depths. Adoption of this listening strategy would improve depth discrimination because this task requires listeners to distinguish deep modulation depths (signal) from deeper ones (standard), but worse detection because this task depends upon the ability to distinguish shallow modulation depths (signal) from no modulation at all (standard). Alternatively, training could have led listeners to adopt a strategy to choose the interval that is least correlated to an internal template of the trained standard (as suggested above). This strategy would aid depth discrimination because the signal interval in this task (the shallower modulation depth) has the weakest correlation to that template, but hurt detection because the signal interval in this task (the modulated stimulus) has the strongest correlation to that template.

  55   One possibility is that this uniform improvement came from an increased ability to ignore the randomization of presentation level. Since the same level randomization was used on all conditions, an increased ability to ignore this randomization would lead to improvement on all untrained conditions. Another possibility is that the listening strategy for the modulation itself changed over the course of training. For example, listeners might have initially chosen the interval that was most correlated to an internal template of an unmodulated noise, and then later chosen the interval that was least correlated to a template of the trained standard. There are several potential listening strategies, which if adopted over the course of training, could have led to a discrimination/detection tradeoff that is specific to the trained modulation. In one such strategy, listeners would increasingly learn to focus on the processing of deep modulation depths (of the trained modulation), and ignore that of shallow ones. Adoption of this listening strategy would lead to improved depth discrimination because this task requires listeners to distinguish deep modulation depths from deeper ones, but worse detection because this task requires the listener to distinguish shallow modulation depths from no modulation at all. Alternatively, training could have led listeners to adopt a strategy to choose the interval that is least correlated to an increasingly refined internal template of the trained standard (as suggested two paragraphs earlier). If so, such a strategy would be helpful for depth-discrimination where the signal interval (the shallower modulation depth) has the weakest correlation to that template, but harmful for detection where the signal (the modulated interval) has the strongest correlation to that template. Summary and Conclusion. A fundamental function of any sensory system is to extract information from the patterns of activity distributed across the sensory periphery. The auditory system must analyze the complex patterns of time-varying cochlear activity that are evoked by

  56   natural sounds such as speech. The current data demonstrate that perceptual learning can be specific to particular combinations of spectral and temporal modulation frequency, and thus provide some of the first direct behavioral evidence that a mechanism with such combined tuning can influence perception. These data are consistent with the proposal that auditory perception is mediated a bank of modulation filters that are tuned to particular combinations of spectral and temporal modulation frequency, and further suggest that this proposal should allow the contributions of each filter to be influenced separately by previous experience. They also suggest that experience-induced physiological modifications that are restricted to neurons tuned to particular combined spectro-temporal modulations could be observed. Finally, these results imply that training on spectro-temporal modulation might improve performance on real-world perceptual tasks, but only if the training involves the particular combination of spectral and temporal modulation frequencies that limits performance of those tasks.

  57   CHAPTER 4. Different patterns of auditory perceptual learning on spectral modulation detection between older hearing-impaired and younger normal-hearing adults Abstract Young adults with normal hearing (YNH) can improve their sensitivity to basic acoustic features with practice. However, it is not known to what extent the influence of the same training regimen differs between YNH listeners and older listeners with hearing impairment (OHI) -- the largest population seeking treatment in audiology clinics. Here we trained OHI listeners on a basic auditory task (spectral modulation detection) using a training regimen previously administered to YNH listeners (~1 hr/session for 7 sessions on a single condition). The YNH listeners who received training learned more than matched controls who received none, but that learning did not generalize to any untrained spectral modulation frequency. In contrast, the OHI trained listeners and controls learned by comparable amounts on the trained condition, even though the trained listeners both improved over the training phase and generalized their learning to an untrained spectral modulation frequency. These population differences suggest that learning consolidated more slowly, and that training modified an aspect of processing that had broader tuning to spectral modulation frequency, in OHI than YNH listeners. More generally these results demonstrate that conclusions about perceptual learning that come from examination of one population do not necessarily apply to another.

  58   Introduction Listeners can improve their sensitivities to basic acoustic features with practice. Such perceptual learning has been documented on a number of auditory tasks for young, college-aged, listeners with normal hearing (YNH) (for review see Wright and Zhang, 2009a, b). However, far less is known about the influence of auditory training in the largest population seeking treatment in audiology clinics – older listeners with hearing impairment (OHI). While perceptual learning has been observed in OHI listeners especially on speech-perception tasks (e.g., Walden et al., 1981, Sweetow and Sabes, 2006, Burk and Humes, 2007, 2008), how the learning in this population compares to that of YNH listeners given the same training has received little attention. Thus, it is not known whether perceptual learning itself differs between OHI and YNH listeners. Any differences in learning between these populations would indicate that the processing mediating perceptual improvement differs in some way between them, and that conclusions about perceptual learning that arise from studying one population do not necessarily apply to the other. Therefore we trained OHI listeners on a basic auditory task (spectral modulation detection) and compared their learning patterns to those we previously reported for YNH listeners given the same training (Chapter 2). The clearest evidence that a sensory disorder itself can influence perceptual learning comes from recent reports in which the effects of the same multi-session training regimen were compared between young adults with and without amblyopia. Amblyopia is characterized by an inability to perceive fine spatial details despite a physically normal eye (Attebo et al., 1998). In these investigations, participants practiced detecting sinusoidal gratings at a particular spatial frequency or discriminating between spatial frequencies. In comparison to individuals with normal vision, learning in amblyotes had a larger magnitude (Polat et al., 2004, Chen et al.,

  59   2008), a longer time course (Li et al., 2008), and generalized to a broader range of untrained spatial frequencies (Huang et al., 2008, Huang et al., 2009, Astle et al., 2010). Here we compared learning between OHI and YNH listeners on auditory spectral modulation detection, the auditory analog of the detection task that the amblyotes practiced. Listeners had to distinguish a noise with a flat spectral envelope from one with a spectral envelope that had a sinusoidal shape on a logarithmic frequency axis (Eddins and Bero, 2007). We selected this task both because it requires the accurate perception of sinusoidal patterns of activity distributed across the sensory periphery, as did the visual detection of a sinusoidal grating practiced by the amblyotes, and because the learning patterns on this auditory task are similar to those on the corresponding visual task in young adults with normal sensory processing (auditory: Chapter 2)(visual: Sowden et al., 2002). In the current investigation, we trained a group of OHI listeners on a single spectral modulation detection condition, 720 trials/session for 7 daily sessions. Before and after this training phase, we evaluated performance on the trained condition, on the detection of two untrained spectral modulation frequencies, and on a ripple reversal task (another measure of spectral modulation perception; Henry et al., 2005). A separate group of OHI controls only participated in the pre- and post-tests. We compared the results of the OHI listeners to those we previously reported for YNH listeners who practiced detecting spectral modulation using the same training regimen (Chapter 2). There were qualitative differences in both learning and generalization between these two populations, indicating that the processing underlying perceptual improvement can differ between OHI and YNH listeners. Method Listeners. Sixteen participants (8 female) between 56 and 82 years of age were recruited from the Northwestern University Audiology Clinic. All listeners had binaural sensorineural

  60   hearing loss ranging from moderate to profound (see Table 1). The hearing loss was symmetric between the two ears in all but one listener (listener C7). All listeners reported no previous experience with psychoacoustic tests, gave informed consent, and were financially compensated for their participation. All procedures were approved by the Institutional Review Board at Northwestern University. Overview. The participants were divided into a group of trained listeners (n = 8) and a separate group of controls (n = 8). All trained listeners participated in an initial screening, a pretest, a training phase, and a post-test. During the screening, pure-tone air- and bone-conduction thresholds were measured at frequencies from 250-8000 Hz, and the uncomfortable listening level was determined. In the pre-test, performance was evaluated on three spectral modulation detection conditions (1, 2 and 4 cyc/oct) as well as on a ripple reversal task. For some listeners, vowel and consonant identification in noise was also measured during this session, but those data are not reported here. For all trained listeners the training phase consisted of seven daily practice sessions (each approximately 1 hr in length) in which thresholds were measured repeatedly on a single spectral modulation detection condition (2 cyc/oct). The post-test followed the training phase and was identical to the pre-test. The order of the conditions in the pre- and post-tests was randomized across listeners, but held constant between the pre- and post-tests for each individual listener. The pre-test and first day of training were conducted on consecutive days, as were the final day of training and the post-test. The controls participated in all of the same stages, except for the training phase. Thus, any difference between the trained listeners and controls can be attributed to the training phase. Pre- and post-tests were separated by an average of 17.0 days for the trained listeners and 18.8 days for the controls.

  61   Tasks and procedures. Thresholds were estimated using an adaptive three-alternative, forced-choice method. On a given trial, three intervals were presented in random order. One interval contained a signal stimulus and two contained a reference stimulus. In the spectral modulation detection task, the signal was a spectrally modulated stimulus and the reference had a flat spectrum (Fig 4.1A). In the ripple reversal task, the signal was spectrally modulated and the reference was identical to the signal except that the locations of the peaks and valleys were interchanged (Fig 4.1B). Listeners indicated which of the three intervals contained the signal stimulus by using a computer mouse to click on a visual display. After every trial throughout the experiment, visual feedback was provided indicating whether the response was correct or incorrect. Spectral modulation detection thresholds were estimated by adaptively adjusting the modulation depth of the signal (peak to valley difference in dB) using a 3-down/1-up rule. This procedure converged on the 79.4% correct point on the psychometric function (Levitt, 1971). The modulation depths at which the direction of change reversed from decreasing to increasing or vice versa are referred to as reversals. The depth was initially 20 dB and was adjusted in steps of 2 dB until the third reversal; subsequent steps were 0.4 dB. In each block of 60 trials, the first three reversals were discarded, and the modulation depths at the largest remaining even number of reversals (the useable reversals) were averaged and taken as the spectral modulation detection threshold. This procedure followed that used by Sabin et al. (Chapter 2). Ripple reversal thresholds were estimated by adaptively adjusting the spectral modulation frequency using a 2-down/1-up rule. This procedure converged on the 70.7% correct point on the psychometric function (Levitt, 1971). The spectral modulation frequency was always the same for both the signal and the reference. It was initially 1.414 cyc/oct and was adjusted in one-half

  62   octave steps. All other aspects of the adaptive procedure the same as for spectral modulation detection, except that the threshold was computed by taking the geometric, rather than the arithmetic, mean of the useable reversals. This procedure was based on that used by Henry et al. (2005). For both tasks, blocks that contained fewer than 7 reversals (5.3% of all blocks) or single trials that were longer than 20 sec (from the first observation interval through the response, 1.1% of all blocks) were excluded from analysis. We also removed from analysis blocks for which the threshold estimate was more than two standard deviations higher than the mean of all the estimates from all the listeners on that condition (4.4% of all blocks). The uncomfortable listening level was determined using a modified version of the contour test of loudness perception (Cox et al., 1997). On each trial, the listener was presented with a flat spectrum noise spanning 400-3200 Hz -- the reference used in the spectral modulation detection tasks. The listener had to select which of seven loudness categories, ranging from “inaudible” to “uncomfortably loud,” best described that presentation. On the first trial, the noise was presented at 50 dB SPL. On each subsequent trial, the presentation level increased by a random amount ranging from 2 to 5 dB until the listener selected the “uncomfortably loud” category. Each listener repeated this procedure three times. The average of the three sound pressure levels that were characterized as uncomfortably loud was taken as the uncomfortable listening level (UCL). Stimulus synthesis and presentation. For spectral modulation detection, the protocol for stimulus generation was adapted from a previous study on spectral modulation detection (Eddins and Bero, 2007). An 8192-point buffer was first filled with a sinusoid computed on a log2 frequency axis with the appropriate spectral modulation frequency (1, 2, or 4 cyc/oct) and

  63   modulation depth (expressed in dB). The phase of the sinusoid was randomly drawn from a uniform distribution spanning 0-2π. The sinusoid was first multiplied by an equivalently sized buffer filled with randomly drawn numbers from a Gaussian distribution, and then multiplied by the magnitude response of a Butterworth filter (-32 dB/octave) with cutoff frequencies at 400 and 3200 Hz. The resulting magnitude response was combined with a random phase spectrum and the real inverse Fourier transform was computed. Once in the time domain, the sound was shaped by a 100-ms amplitude envelope with 10-ms raised cosine on/off ramps. All stimuli were scaled to have the same RMS amplitude. For the ripple reversal task, the protocol for stimulus generation was also adapted from a previous study (Henry et al., 2005). Each stimulus was comprised of 200 sinusoids that had random phases and were evenly spaced along a log2 frequency axis from 100 to 5000 Hz. The amplitudes of the individual components were shaped by a full-wave rectified sinusoid that was defined across a log2 frequency axis (spectral frequency in cyc/oct) and had a modulation depth of 30 dB. The phase of that sinusoid was chosen randomly for the signal stimulus, and the opposite phase was chosen for the references. To approximate the long-term speech spectrum, an overall spectral tilt of -6 dB/oct was applied to the frequencies above 750 Hz. Each stimulus was 500 ms in duration including 150-ms raised cosine on/off ramps, and was scaled to have the same RMS amplitude. To help prevent the listeners from basing their decisions on the use of local intensity cues (comparing the intensity at a single audio frequency across intervals), we randomized the presentation level of each stimulus in both tasks. The presentation levels were randomly drawn from a uniform distribution spanning 15 dB, where the top of that range was 5 dB below the individual’s UCL. Each stimulus was synthesized before each trial.

  64   All stimuli were presented using custom software written in MATLAB. For some listeners, the stimuli were presented through a 16-bit digital-to-analog converter (Tucker-DavisTechnologies DD1) followed by an anti-aliasing filter with a 16-kHz cutoff frequency (TDT FT6-2), a programmable attenuator (TDT PA4), a sound mixer (TDT SM3), and a headphone driver (TDT HB6). For others, the stimuli were presented through a two-channel USB sound card (Edirol UA-25) and a headphone amplifier (Crown D75). The stimuli were always presented through both earpieces of Sennheiser HD265 circumaural headphones. Listeners were tested in a sound-attenuated room. Results Performance on the trained condition. The performance of the trained listeners on the trained spectral modulation detection condition improved over the course of multiple sessions, but the controls also improved on that condition by a comparable amount (Fig. 4.2A). The magnitudes of improvement between the pre- and post-tests for the trained listeners (squares; T7 = 2.9, p = 0.024) and controls (circles; T7 2.6, p = 0.036) did not differ according to either the interaction term of a 2 group (trained vs. control) x 2 session (pre vs. post) analysis of variance (ANOVA) using session as a repeated measure (F1,14 =0.1; p = 0.8) or the effect of group in an analysis of covariance (ANCOVA) using pre-test performance as a covariate (F1,13 = 0.8, p =0.39). It is the case that the improvements were more consistent across the individual trained listeners (Fig. 4.2B, lines) than the controls (circles). The post-test thresholds (adjusted for pretest performance; Cohen, 1977) of three of the eight controls were more than two standard deviations higher than the mean post-test threshold of the trained listeners, while none of the thresholds of the individual trained listeners were this extreme. Nevertheless, removing the data of the three aberrant controls did not change any of the statistical conclusions arising from the

  65   between-group comparisons of pre- to post-test improvement on either the trained or untrained conditions. Neither age nor severity of hearing loss (pure tone average) was predictive of pretest performance (all r2 < 0.09, all p > 0.25) or of the amount of improvement (all r2 < 0.009, all p > 0.72) on the trained condition. Despite the similar improvement by trained listeners and controls, the trained listeners actually showed a clear learning curve over the training phase. On the group level, this training phase improvement is indicated by a significant negative slope of a single line fitted to the pretest-adjusted daily mean thresholds of each trained listener vs. the log10 of the session number, computed across all sessions except the pre-test (r = -0.36, p = 0.004). This gradual improvement is further confirmed on the individual level where nearly all (7 of 8) trained listeners improved over these sessions, as indicated by a significant negative slope (all p < 0.048) of a regression line fitted to each adjusted threshold estimate over the log10 of the session number. To help determine when these improvements emerged, we examined how performance changed within training sessions by computing, for each trained listener on each training session, the means of the first three and of the last three threshold estimates (Fig. 4.3). We evaluated within-session performance using a 2 time (first vs. last) by 7 session (all training days) ANOVA using both time and session as repeated measures. While there was no main effect of time (F1,42 =0.4, p = 0.53), there was a main effect of session (F6,42 = 5.8, p < 0.0001), and a time x session interaction (F6,42 = 4.0, p = 0.003). This interaction arose because performance improved within the first training session (T7 = 4.7, p = 0.002), but not within any other sessions (all p > 0.14). However, even when the entire first day of training is excluded from analysis, the slope of the remaining learning curve is still significantly negative (r =-0.31, p = 0.001). Therefore, the learning of the

  66   trained listeners was comprised of both a fast within-session improvement during the first training session and a more gradual improvement over the other sessions. The apparently contradictory result of training-phase learning despite comparable improvement by trained listeners and controls is at least partially clarified by evidence that the influence of the pre-test emerged over multiple days. To examine the time course of the influence of the pre-test, we compared the performance of the controls at the post-test to that of the trained listeners on the first training day, because at these points, the only prior experience for both groups came from the pre-test. Three performance differences between the two groups suggest that the influence of the pre-test took more than 1 day to fully emerge. First, the post-test thresholds of the subset of controls who learned were lower than those of the trained listeners on the first training day (F1,10 = 13.2, p = 0.005). Second, while the trained listeners showed no improvement between the pre-test and the beginning (first three estimates) of the first day of training (T7 = -0.09, p = 0.92), the controls improved between the pre- and post-tests (see above). Third, the trained listeners improved during the first day of training (r = -0.37, p < 0.001), but the controls showed no improvement during the post-test. The controls did not improve during the testing of the trained condition itself (r = -0.08, p = 0.67). They also did not improve over the course of the entire post-test, either when thresholds were expressed as raw values (r = -0.13, p = 0.21) or when they were z-score normalized on a condition-by-condition basis to the post-test performance of all controls on that condition (r = -0.14, p = 0.17). Therefore, at least a portion of the training-phase learning shown by the trained listeners may simply reflect this implied multiday influence of the pre-test. Performance on untrained conditions. While the contribution of the multiple-session practice to improvements on the trained condition is unclear, analyses of the untrained conditions

  67   indicate that this training phase did affect performance. For the detection of the untrained lower spectral modulation frequency (1 cyc/oct; Fig 4.4A, middle column), the trained listeners improved significantly more than controls (ANOVA: F1,12 = 4.9, p = 0.047; ANCOVA: F1,11 = 6.0, p = 0.03). This between-group difference indicates that the training phase itself led to some improvement. However, this improvement was limited to (specific to) the lower untrained spectral modulation frequency. The improvements of the trained listeners and controls were not distinguishable from each other for the detection of the higher untrained spectral modulation frequency (4 cyc/oct, Fig. 4.4A, right column) or for the ripple reversal task (Fig 4.4B) (ANOVA: all p > 0.43 ; ANCOVA all p > 0.41). Finally, participation in the pre-test itself led to improvements that were restricted to a subset of conditions, suggesting that these improvements did not arise solely from procedural learning. Of the four conditions, exposure to the pre-test induced learning in only two. Trained listeners and controls improved significantly and equivalently between the pre- and post-tests on the trained condition (2 cyc/oct; see above) and on the detection of the higher untrained spectral modulation frequency (4 cyc/oct; Fig. 4.4A, right column, ANOVA main effect of session: F1,12 = 6.0, p = 0.03). The similar improvements shown by the two groups for these two conditions suggests that this learning arose from participation in the pre-test, because prior to the post-test this was the only experience shared by the two groups. In contrast, there were no such improvements on the other two conditions. For the detection of the lower untrained spectral modulation frequency (1 cyc/oct), the controls did not improve (T6 = 0.05, p = 0.96), even though improvements on this condition were possible, as demonstrated by the trained listeners. Neither group improved on the ripple reversal task (T-test: all p > 0.2 ; ANOVA: F1,13 = 3.1, p = 0.1). Thus, the influence of the pre-test was restricted to detection of the higher spectral

  68   modulation frequencies (2 and 4 cyc/oct). This specificity implies that the improvements arising from participation in the pre-test do not simply reflect the learning of procedural aspects that were common to all of the conditions or even of those that were common only to the detection task. Discussion The goal of this investigation was to assess the extent to which auditory perceptual learning differs between older listeners with hearing impairment (OHI) and younger, collegeaged, listeners with normal hearing (YNH). Toward this end, we trained OHI listeners on a spectral modulation detection task so that we could compare their results to those we previously reported for YNH listeners who participated in the same training regimen. Here we identify four differences in perceptual learning between OHI and YNH listeners, present a potential account for these differences, consider the potential distinct influences of age and hearing loss on these differences, and discuss the practical implications of these results. Differences in learning between OHI and YNH listeners. We compared the learning patterns on spectral modulation detection of the current OHI listeners to those we previously reported for YNH listeners who participated in the same training regimen (Chapter 2). Like the OHI listeners, the YNH listeners practiced a single spectral modulation detection condition ~ 1 hour/day for seven days and were tested on multiple conditions before and after the training phase. Different groups of YNH listeners practiced either the same condition that was trained here (2 cyc/oct spectral modulation spanning 400-3200 Hz) or one of two conditions with a lower spectral-modulation and carrier frequency (0.5 cyc/oct, 200-1600 Hz, or 1 cyc/oct 2001600 Hz). Below, we describe four differences between the OHI and YNH listeners in the influence of spectral modulation detection training.

  69   First, when given multiple-session training on the same condition (2 cyc/oct), the OHI listeners improved over the training phase (Fig. 4.2A), while the YNH listeners did not. Thus, it appears that the YNH, but not the OHI, listeners were already at optimal performance on this condition prior to training. However, YNH listeners did show training-phase improvement on a different condition (0.5 cyc/oct). This improvement was similar in time course and magnitude to that of the OHI listeners, demonstrating that gradual learning on spectral modulation detection is not unique to OHI listeners. Second, for those cases in which the trained listeners improved gradually (OHI trained at 2 cyc/oct and YNH trained at 0.5 cyc/oct), how that learning compared to the improvement of controls differed between the YNH and OHI populations. The YNH trained listeners learned more on their trained condition than did matched controls who received no training, indicating that the training itself led to learning. In contrast, even though the trained OHI listeners improved over the course of the training phase, the OHI controls improved by a similar amount, suggesting that the learning in this population was driven primarily by the pre-test. Thus, the training-phase learning in these two populations appears to have arisen from different components of the same experimental protocol. Third, the improvement of the controls appeared to emerge over a longer time course in OHI than in YNH listeners. The controls did not participate in the training phase, so any improvements between the pre- and post-tests in this group can be attributed to an influence of the pre-test. We inferred the time course of this influence by comparing the performance of the controls at the post-test (~18 days after the pre-test) to that of the trained group on the first training session (1 day after the pre-test), because in both cases, the only previous experience came from the pre-test. By these analyses, for the YNH listeners, the pre-test either led to modest

  70   improvements that emerged in full by the following day and were maintained (0.5 cyc/oct) or lost (1 cyc/oct) by the post-test, or did not lead to improvement at all (2 cyc/oct) (Chapter 2). On the other hand, for the OHI listeners, participation in the pre-test led to marked improvement on the trained condition (2 cyc/oct) that seemed to take multiple sessions to fully emerge (see results). Thus, the two populations also differed in how exposure to a pre-test influenced their performance. Fourth, the influence of practice generalized to untrained spectral modulation frequencies in the OHI, but not the YNH, listeners. Across all three groups of YNH trained listeners there were no cases in which trained listeners learned more than controls on an untrained spectral modulation frequency; training-induced learning was specific to the trained condition. In contrast, the OHI trained listeners learned more than controls on an untrained lower spectral modulation frequency (1 cyc/oct), despite not having done so on their trained condition (2 cyc/oct). Thus, the influence of training generalized more broadly in OHI than in YNH listeners. A potential account for these learning differences. One possible unifying account for these population differences is that learning consolidated more slowly, and that training modified an aspect of processing that had broader tuning to spectral modulation frequency, in OHI than YNH listeners. It is generally held that after training has ceased there is an extended period of consolidation during which the new and fragile learning becomes more stable (for review see McGaugh, 2000, Walker and Stickgold, 2004). This process of consolidation is also thought to underlie improvements that emerge well after the end of a practice session (e.g., Korman et al., 2003, Roth et al., 2005). In this context, the evidence that the influence of the pre-test took multiple days to fully emerge in OHI listeners, but only a single day in YNH listeners, suggests that consolidation proceeded over a slower time course in OHI than in YNH listeners. A

  71   difference in consolidation rate could account for the population differences in how the learning in the trained listeners compared to that in the controls. In YNH listeners, the gradual learning on the trained condition (in listeners trained at 0.5 cyc/oct) could have arisen from multiple short (~1 day) consolidation periods initiated by each training session, leading to a gradual learning curve and ultimately to greater learning than controls. However, because the OHI trained listeners showed comparable learning to controls, their gradual learning curve on the trained condition could, at an extreme, simply reflect the multiple-session consolidation initiated by the pre-test, rather than an influence of the training phase. A more moderate possibility is that the learning curve of the OHI trained listeners reflects the influences of both the pre-test and the training phase. Evidence that the training itself did have some effect on this population comes from the observations that the trained listeners generalized to an untrained condition (1 cyc/oct, Fig 4.4A) and showed more consistent learning than controls (Fig 4.2B). Even so, these influences of the training could have consolidated over a slower than normal time course. Similar arguments hold if, instead of consolidating the same modifications at different rates, training in the two populations modified different aspects of processing that normally consolidate at different rates. Different consolidation times course have been reported for learning different aspects of the same condition in YNH listeners (Ortiz and Wright, 2010). It also appears from these data that the aspect of processing that was modified by the training phase had broader tuning to spectral modulation frequency in OHI than in YNH listeners. When the influence of practice on one condition affects performance on another (generalization), it is assumed that training modified processing that is engaged by both conditions (e.g., Ahissar and Hochstein, 2004, Wright et al., 2010). The observed population difference in the breadth of generalization therefore implies that whatever was modified by

  72   training had broader tuning in OHI than YNH listeners. One possibility is that the traininginduced modification involved the neural circuitry underlying the filtering of stimuli into their component spectral modulation frequencies (e.g., Chi et al., 2005). Behavioral support for filters tuned to particular spectral modulation frequencies in YNH listeners comes from spectralmodulation-frequency-specific effects of masking (Saoji and Eddins, 2007) and adaptation (Eddins and Harwell, 2002), while physiological support comes from observations of individual neurons in auditory cortex with bandpass tuning to a particular spectral modulation frequency (Schreiner and Calhoun, 1994, e.g., Shamma et al., 1995, Versnel et al., 1995, Kowalski et al., 1996). The training could have modified these filters directly or modified a more central decision maker that receives input from these filters (for discussions of these views see Dosher and Lu, 1998, Ahissar and Hochstein, 2004). Either way, this possibility requires that these filters had broader tuning in OHI than YNH listeners, at least at the post-test. Potential contributions of age and hearing loss. Both age and hearing status differed between the YNH and OHI populations, and therefore both factors could have contributed to the observed population differences. While the present data do not allow us to separate the influences of these factors, comparisons to the limited literature regarding their isolated influences on perceptual learning raise the possibility that age primarily affected the trained condition, while hearing loss affected generalization. For the trained condition, the performance pattern differences between the populations described here were quite similar to those between groups of older and younger listeners with normal hearing (ONH and YNH) who received multisession training on auditory temporal interval discrimination: ONH listeners who started poorly improved over the training phase but matched controls improved as much and YNH listeners given the same training learned more than controls (Marrone et al., 2010). Thus, age alone could

  73   have led to the current population differences on the trained condition. Age has also been shown to have an effect on perceptual learning even when the starting performance is similar across groups (adolescents vs young adults; (Huyck and Wright, 2011); younger vs older adults; (Marrone et al., 2010)). However, age may not affect perceptual learning in all cases. Younger and older participants with different starting performance had qualitatively similar learning on two basic visual tasks (Ball and Sekuler, 1986, Andersen et al., 2010), though we note that this learning has not been compared between groups differing in age but not starting performance on the same condition. For the untrained conditions, the current population differences in the breadth of generalization resemble those described in the introduction between young adults individuals with and without amblyopia -- a different sensory disorder. After training to detect a sinusoidal grating at a single spatial frequency (the visual analog of the current trained task) or to discriminate that frequency, young adults with amblyopia generalized to a broader range of untrained spatial frequencies than did those with normal vision (Huang et al., 2008, Huang et al., 2009, Astle et al., 2010). This pattern matches the present broader generalization across spectral modulation frequency in OHI than YNH listeners. Thus hearing loss alone could have led to the current population differences in generalization. Practical implications. Overall, this investigation adds to others (e.g., Peelle and Wingfield, 2005, Huyck and Wright, 2011) indicating that conclusions about perceptual learning are not easily applied across populations. Given the same training regimen, different populations can learn differently. It appears then that the development of clinical training regimens should focus primarily on the target population. For OHI listeners, the long time course of the pre-testinduced improvement observed here indicates that brief periods of practice on some conditions can have an extended influence in this population. This time course could be harnessed in a

  74   clinical training regimen by providing training sessions that are widely spaced in time. Doing so could decrease the overall time and effort spent by both the clinician and the patient. Further, the improvement of the OHI listeners on an untrained condition following multi-session practice on a single trained condition demonstrates that generalization, the goal of most clinical training regimens, can occur in this population. Finally, this work more specifically indicates that the detection of spectral modulation in OHI listeners can improve with practice. Therefore it is possible that, in this population, training could aid real world skills that are limited by spectral modulation detection such as sound localization in the vertical plane (Macpherson and Middlebrooks, 2003, Qian and Eddins, 2008) or speech perception (Litvak et al., 2007).

  75   CHAPTER 5. Implications and Future Directions Overall Result Summary. In three separate lines of experiments I have examined how practice influences the perception of spectral (Chapter 2) and spectro-temporal modulation (Chapter 3) in younger, college-aged, listeners with normal hearing (YNH) and have compared the influence of spectral modulation detection training between YNH listeners and older listeners with impaired hearing (OHI) (Chapter 4). The main results of these investigations are briefly summarized here. (1) In YNH listeners, the influence of spectral modulation detection training was both dependent upon and specific to the trained spectral modulation frequency. On average, as the trained spectral modulation frequency increased, the magnitude of traininginduced improvement and the time to reach asymptotic performance decreased, while the tendency for performance to worsen within a training session increased. The training-induced improvements did not generalize to untrained spectral modulation frequencies or to an untrained carrier spectrum. (Chapter 2) (2) In YNH listeners, training to discriminate the depth of spectral, temporal, or spectrotemporal modulation led to improvements that were specific to the combined spectrotemporal modulation of the trained stimulus. Each group of trained YNH listeners learned more on their respective trained condition than controls that received no training. Critically, this depth-discrimination learning did not generalize to the untrained stimuli even when the trained and untrained stimuli had one modulation frequency (spectral or temporal) in common. Learning on modulation depth discrimination also led to worsening on modulation detection, but only when the same combined spectro-temporal modulation was used for both tasks. (Chapter 3)

  76   (3) Spectral modulation detection training had a qualitatively different influence in OHI as compared to YNH listeners. The OHI trained listeners and controls learned by comparable amounts on the trained condition, even though the trained listeners both improved over the training phase and generalized their learning to an untrained spectral modulation frequency (Chapter 4). This learning differed from that of YNH listeners who, for some trained stimuli, learned more than matched controls but did not generalize that improvement to any untrained spectral modulation frequency. Implications About Encoding of Spectral and Spectro-Temporal Modulation in the Central Auditory System. One of the primary motivations of this dissertation was to explore the transformations that the auditory system uses to extract information from the complex time-varying patterns of cochlear activity that are evoked by natural stimuli. Central to the design and interpretation of the experiments described in Chapters 2 and 4 was the proposal (Schreiner and Calhoun, 1994, Shamma et al., 1995, Saoji and Eddins, 2007) that central analysis of cochlear activity depends upon modulation filters tuned to unique spectral modulation frequencies. In Chapter 3, I also considered the wider proposal (Chi et al., 1999, Chi et al., 2005) that such filters are a portion of a larger spectro-temporal modulation frequency filterbank where each filter is tuned to a unique combination of spectral modulation frequency and temporal modulation frequency. Overall, the results of the reported experiments provide behavioral evidence that is consistent with these proposals. This work also extends these proposals to indicate how such processing is influenced by training.

  77   The specificities of perceptual learning in the YNH listeners observed in Chapter 2 is consistent with the idea that the central auditory system acts as a spectral modulation frequency filterbank. In these listeners, there were no cases in which training with a single spectral modulation frequency led to improvements on an untrained frequency. This observation indicates that perception can be influenced in a manner that is dependent upon spectral modulation frequency. This specificity could have been mediated by a training-induced modification involving a filter tuned to the trained spectral modulation frequency. For example, training could have directly increased the sensitivity of the filter, or it could have increased the weight that this filter exerts on a more central process (such as a decision maker) that pools information from multiple filters (for more general discussions of these alternatives see: Dosher and Lu, 1998, Ahissar and Hochstein, 2004). The lack of generalization to an untrained carrier spectrum suggests that if spectral modulation frequency filters exist, they might also be tuned to particular ranges of audio frequency. The observed improvements are inconsistent with the possibility that learning was mediated by a modification involving processing with broad tuning to spectral modulation frequency, because a modification of this type would generalize to untrained modulation frequencies. The specificity of depth discrimination learning to the trained combination of spectral and temporal modulation frequencies is consistent with the proposal that filters tuned to combined spectro-temporal modulation underlie perception. Training improved the ability of YNH listeners to discriminate the depth of spectral, temporal, or spectro-temporal modulation, but those improvements did not generalize to stimuli containing only one of the trained modulation frequencies (spectral or temporal). This result indicates that stimuli can be processed in a manner that is specific to a particular combination of spectral and temporal modulation frequencies.

  78   Using the same logic as above, this specificity could have arisen from a modification involving a filter tuned to the trained combination of spectral and temporal modulation frequencies, and is therefore consistent with the possibility that such filters exist. In contrast, this result is not consistent with the idea that training modified two independent filters tuned to either spectral or temporal modulation frequency. This result does not imply that such single-dimension filters do not underlie perception, but instead it implies that if they do, they did not underlie the traininginduced improvements observed here. In addition to providing behavioral support for the spectral and spectro-temporal modulation filterbank proposals, the current work extends these proposals by describing how this processing can be influenced by experience. The specificity of learning to the trained modulation indicates that, with experience, the processing in each filter can change independently. Further, the different amounts of learning across trained stimuli indicate that each filter is differently susceptible to training. The current work also extends the modulation filterbank proposal to suggest that the processing of spectral modulation might differ between YNH and OHI listeners. In particular, this work suggests that spectral modulation tuning might be broader in OHI than in YNH listeners. While in YNH listeners, training never led to improvement on untrained spectral modulation frequencies, in OHI listeners training at a higher spectral modulation frequency (2 cyc/oct) led to improvement at a lower one (1 cyc/oct). Interpreted in the context of the spectral modulation frequency filterbank, this result could indicate that in OHI listeners there is a filter that is sensitive to 2 cyc/oct (because it was excited by training) that also influences performance at 1 cyc/oct. Using similar logic, in YNH listeners, training engages filters that are narrowly tuned to the trained spectral modulation frequency. Thus, this difference in generalization could

  79   arise if the tuning of spectral modulation frequency filters in OHI listeners is broader than that in YNH listeners. If so, then this broader tuning should be observable on other tests of modulation filter bandwidth such as adaptation (Tansley and Suffield, 1983, Richards et al., 1997, Wojtczak and Viemeister, 2003) and modulation masking (Bacon and Grantham, 1989, Houtgast, 1989, Yost and Sheft, 1989, Yost et al., 1989). Individual differences in spectral or spectro-temporal modulation tuning could provide a new way to explain the large individual differences in performance on real world tasks that are not accounted for by the audiogram (Neher et al., 2011). Processing Advantages of Spectral and Spectro-Temporal Modulation Filterbanks. If the auditory system transforms cochlear input via filters tuned to spectral and/or spectro-temporal modulation, such a transformation likely facilitates perception of natural sounds. Supporting this possibility, research from computer science indicates that representing natural sounds in terms of their component spectral and/or spectro-temporal modulation frequencies provides an especially useful and compact way to capture meaningful information contained within them. Systems that are designed to automatically classify speech and music sounds perform particularly well when they use an input representation that resembles the output of a spectral modulation frequency filterbank. This representation, known as mel-frequency cepstral coefficients (MFCCs) (Davis and Mermelstein, 1980), is the cosine transform of the log power spectrum of a sound after that spectrum has been warped to the mel-frequency scale (Stevens et al., 1937). In other words, MFCCs represent the spectrum of the audio frequency spectrum after it has been warped to a cochlea-like scale. The coarse changes in the spectral envelope that can be used to identify sounds can be represented by a few MFCCs. In automatic speech recognition (Davis and Mermelstein, 1980) and musical instrument identification (Eronen, 2001) classifiers that are given MFCCs perform better than those that are given

  80   representations of the audio frequency power spectrum. MFCCs are a highly efficient representation. Audio information retrieval systems perform well with only 10-15 coefficients (~ filters) (Davis and Mermelstein, 1980, Eronen, 2001) and can potentially use far fewer (Pardo, personal communication). Thus an internal representation based on spectral modulation frequency in the central auditory system might be particularly efficient, and it has been proposed that creating such efficient representations is a primary goal of information processing in sensory systems (Field, 1994). Though the use of combined spectro-temporal modulations is less common in audio information retrieval systems, a growing body of evidence indicates that these modulations might be particularly helpful for those systems. Supporting the value of this representation, an independent components analysis performed on a library of speech recordings resulted in a variety of localized sinusoidal spectro-temporal modulations (Adballah and Olumbey, 2001). This analysis indicates that a sparse representation based on spectro-temporal modulation frequency could efficiently capture much of the independent acoustical information in speech via a relatively small number of channels. Thus, it appears that transforming cochlear input using a spectral or spectro-temporal modulation filterbank would facilitate perception of natural sounds. Comparison to Vision The similarities between the auditory and visual systems in terms of the perceptual learning on analogous tasks add to the proposal (e.g., Shamma, 2001) that the central nervous system employs similar transformations to extract information from the patterns of activity distributed across the retina and the cochlea. As mentioned in Chapters 1 and 2, it is widely accepted that these transformations in the visual system involve spatial frequency filters (for review see DeValois and DeValois, 1988). Filters tuned to low spatial frequencies encode

  81   patterns of light that change slowly over space (coarse details) while those tuned to high frequencies encode patterns of light that rapidly slowly over space (fine details). Further in the visual system, spatial frequency analysis occurs independently at different spatial regions (retinal locations). Therefore, the central visual system transforms retinal activity via filters tuned to spatial frequency and retinal location. Analogous processing in the auditory system would be dependent upon spectral modulation frequency and audio frequency range (cochlear location). The spectral modulation detection learning experiments reported here in normal-hearing (chapter 2) and hearing-impaired (chapter 4) populations provide additional information that supports the possibility of analogous processing between the auditory and visual systems. The visual task that is most similar to spectral modulation detection is the detection of a sinusoidal grating. Training on this visual task leads to improvements that are tuned to spatial frequency and retinal location (Sowden et al., 2002), matching the selectivity of the visual filters described above. Similarly, in YNH listeners, the specificity of spectral modulation detection learning to the trained spectral modulation frequency and audio frequency is consistent with the tuning of the proposed auditory filters. Further, the influence of training in populations with sensory disorders is also qualitatively similar between the visual and auditory systems. Contrast detection training in individuals with amblyopia, a visual disorder in which perception of fine spatial details is impaired despite a normally functioning eye (Attebo et al., 1998), yields learning that generalizes more broadly to untrained spatial frequencies than does comparable training in individuals with normal vision (Huang et al., 2008, Astle et al., 2010). This population difference is qualitatively similar to the generalization to untrained spectral modulation frequencies that occurred in OHI but not in YNH listeners (Chapter 4). Together, these similarities in the influence of practice on analogous tasks between the visual and auditory systems provide further

  82   evidence that analogous central transformations of peripheral activity could be occurring between these two systems. Practical Implications. The experiments described in this dissertation indicate that the processing underlying the perception of some spectral and spectral-temporal modulations can improve with practice, and therefore suggest that real-world tasks that are limited by perception of these modulations can improve with training. In practical terms, an ideal training regimen would lead to broad generalization such that improvements acquired through practicing with a limited set of stimuli would generalize to the wide variety of stimuli encountered in everyday life. Instead, the improvements arising from the training in YNH listeners were specific to the characteristics of the trained stimulus (Chapters 2 and 3). These specificities suggest that if applied training regimens based on this work were created for YNH listeners, such regimens should target the specific modulations that are most important to the real world task of interest. For example, if the goal is to improve vertical sound localization, training should focus on the lower spectral modulation frequencies that have been shown to be crtical for this skill (0.5 - 2 cyc/oct; Macpherson and Middlebrooks, 2003, Qian and Eddins, 2008). This specificity of the learning observed in YNH listeners might be due to the fact that only a single modulation was used during training. Broader generalization could be achieved through high variability training (e.g., Bradlow et al., 1997, Bradlow et al., 1999). In such training, the tested stimulus changes rapidly during the course of the training session and, at least on speech perception tasks, ultimately leads to broader generalization. However, trial-to-trial randomization of the trained stimulus can also stop learning entirely (Adini et al., 2004, Yu et al., 2004, Kuai et al., 2005, Otto et al., 2006).

  83   The possibility of using spectral modulation detection training as a clinical treatment for OHI listeners is more promising. Unlike YNH listeners, in OHI listeners, the influence of training generalized to a lower untrained spectral modulation frequency (Chapter 4). This result indicates that, in comparison to YNH listeners, OHI listeners generalize to a broader range of stimuli. This observation raises the possibility that, in this population, focused training can lead to improvement on a wide variety of real world stimuli. Also promising is the observation that in OHI listeners, brief periods of training can lead to substantial improvements. Participation in the pre-test led to substantial improvements on the two highest tested spectral modulation frequencies. Those improvements appear to take multiple days to reach their final state (Chapter 4). This observation raises the possibility that a clinical training regimen could use either a single training session or multiple sessions spaced widely in time. Reducing the number of training sessions would save time and effort for both the patient and clinician, and spacing sessions widely in time would allow an evaluation of the necessity of each training session. Though not tested in the current experiments, there is some indication that individuals with cochlear implants could receive the most benefit from improvements in the perception of spectral and/or spectro-temporal modulation. This population is the one in which spectral modulation perception is most highly correlated to performance on real world tasks. Specifically, in this population, performance on tasks that isolate spectral modulation perception are correlated to speech identification in quiet (Henry et al., 2005, Litvak et al., 2007) and in noise (Won et al., 2007), as well as to several aspects of music perception (Won et al., 2010). Particularly relevant to the current work, in listeners with cochlear implants, perception of speech in quiet is highly correlated to the detection of low spectral modulation frequencies (0.5

  84   cyc/oct: Litvak et al., 2007). This is particularly fortunate because, at least in YNH listeners, training has the largest effect at low spectral modulation frequencies (Chapter 2). Finally, independent of learning, electronic systems involving audio processing might be improved if the underlying signal processing is designed to exploit or simulate the spectral or spectro-temporal modulation frequency filterbanks that are implied by the current work. For example there is some indication that speech-in-noise performance in OHI listeners can be improved by artificially boosting the spectral (Eddins and Liu, 2006, 2008, 2009, Liu and Eddins, 2010) or temporal modulation frequencies that are critical for speech perception and reducing those that are not (Drullman et al., 1994). The current work further suggests that it might be best to boost these modulations frequencies in terms of their combined spectrotemporal modulations, thereby providing maximal excitement of the central filters that are tuned to the target sound and reducing excitement in other filters. Incorporation of this type of signal processing into hearing aids could potentially improve a patient’s ability to listen in difficult environments (Edwards, 2007). Further, audio information retrieval systems often are designed to perform tasks that are relatively easy for humans such as speech recognition in noise. These systems might be able to perform as well as humans if the input features mimic those that would be output by an auditory spectral or spectro-temporal modulation filterbank. Supporting this possibility, it appears that systems designed to extract a signal from noise perform particularly well when they are designed to extract a target sound via filters tuned to combined spectrotemporal modulations (Raj and Smaragdis, 2005, Elhilali and Shamma, 2008). Future experiments. The experiments reported in this paper are an initial exploration into the effects of practice on the perception of spectral and spectro-temporal modulation. There are numerous

  85   directions that future work could take. The following is a list of nine unresolved issues raised by the current work, each followed by a brief discussion of experiments that could shed light on those issues. (1) What, specifically is the bandwidth of spectral and spectro-temporal modulation perceptual learning? The reported experiments only examined performance on untrained modulations that had large differences from the trained one. Future experiments could examine generalization at a more fine-grained level to get more reliable estimates of the bandwidth of learning. This bandwidth could be compared to other estimates of the bandwidth of modulation tuning. This comparison could be used to evaluate whether all measures of modulation tuning reflect the same processing. (2) Are there other behavioral measures of combined spectro-temporal tuning in humans? Direct measures of combined spectro-temporal tuning depend upon the ability to rule out the possibility that separate spectral and temporal modulation processing underlies the observed behavior. For example, an experiment involving changes in performance based on expectation can separate these possibilities. If listeners are primed to expect a particular spectro-temporal modulation, they should show better performance on the expected modulation, but worse performance on modulations differing from the expected one in either the spectral or temporal modulation frequencies or both. (3) How do improvements in spectral and spectro-temporal modulation relate to performance on real world tasks? It is unclear how the observed improvements relate to real world skills. This issue could be evaluated if performance on a variety of real

  86   world tasks that are thought to depend upon this modulation perception (e.g., vertical sound localization, speech perception in noise, musical instrument identification) were included in pre- and post-tests. The converse is true as well. Tests of modulation perception could be included in pre- and post-tests around direct training on these real world tasks. (4) How would the specificity of learning differ if the training regimen used multiple trained stimuli? As mentioned previously, such high variability training on speech perception tasks yields broader generalization (Bradlow et al., 1997, Bradlow et al., 1999). Therefore broader generalization might occur if the tested spectral or spectrotemporal modulation frequency was randomly varied on a trial-by-trial basis. However, as mentioned previously, such randomization can also stop learning entirely. (5) Are population differences in perceptual learning between OHI and YNH listeners due to age or hearing loss? Both age and hearing loss varied between the two populations trained on spectral modulation detection and therefore their relative influences can not be determined. To isolate the influences of hearing loss and age, separate groups of younger listeners with hearing loss and older listeners with normal hearing should be trained. (6) Are population differences in perceptual learning between OHI and YNH listeners due to presentation level? The OHI listeners were tested at a higher presentation level than the YNH listeners. Therefore the presentations level itself, rather than age or hearing loss, could have led to the observed differences in learning. The influence of

  87   presentation level could be isolated by training YNH listeners at a sound pressure level that was comparable to the one used for OHI listeners. (7) How did the OHI controls learn as much as the OHI trained listeners even when OHI trained listeners showed a gradual learning curve? A proposed interpretation of this result (in Chapter 4) was that the pre-test induced improvement in OHI controls took multiple days to consolidate. If so, then separate control groups tested at increasing temporal distances from the pre-test should show progressively more learning. Further, consolidation also leads to increased resistance to interference. Therefore the time course over which learning is susceptible to interference should also be longer in OHI than YNH listeners. (8) Is spectral modulation frequency tuning broader in OHI than YNH listeners? The broader generalization to untrained spectral modulation frequencies in OHI than YNH listeners provides some indication that tuning to spectral modulation frequency might be broader in OHI than YNH listeners. If so, it is possible that broader tuning in OHI listeners would be observable in other behavioral measures such as adaptation and modulation masking. (9) What is the influence of spectral and spectro-temporal modulation training in listeners with cochlear implants? As mentioned in the previous section, in cochlear implant listeners, improved spectral modulation perception could be particularly valuable for everyday perception. Future work could determine the basic parameters of such learning in this population and could examine whether that learning generalizes to real-world tasks.

  88  

Figure 1.1 Vowel Spectral Modulations. Spectral modulations for twelve words spoken by an adult male. Each word begins with /b/ and ends with /t/, but differs in the medial vowel. Sound files were passed through an audio frequency filterbank designed to mimic the audio frequency selectivity of individuals with normal hearing. Each graph is therefore a rough approximation of the pattern of excitation along the cochlea evoked collapsed over time, where frequency (~ cochlear place) is on the abscissa and magnitude (~ excitation) is on the ordinate. Note that each vowel evokes a unique and complex pattern of excitation distributed across cochlear place. The differences between these patterns can be used by the central auditory system to identify the speech sound.

  89  

Figure 1.2. Musical Instrument Timbre Spectral Modulations. As in Figure 1.1, except plotted for twelve different orchestral instruments each playing the same note (B3; F0 = 247 Hz) at the same overall level. As with the vowels in Figure 1, differences in spectral modulation provide a reliable cue that can be used to identify musical instrument timbre.

  90  

Figure 1.3. Vowel Spectro-Temporal Modulations. As in Figure 1.1, except that spectral modulations are plotted over time. In each graph, time is on the abscissa, frequency (~ cochlear place) is on the ordinate, and magnitude (~ excitation) is indicated by the color. Note that, for these words, spectral modulations change over time, and are therefore better characterized as spectro-temporal modulations. These spectro-temporal modulations differ across words, and therefore can be reliably used to identify speech sounds.

  91  

Figure 1.4 Musical Instrument Timbre Spectro-Temporal Modulations. As in Figure 1.3, except plotted for the same orchestral instruments as in Figure 1.2. Though not as dramatic as differences between vowels (Figure 1.3), the spectro-temporal modulations differ across musical instruments, and provide a reliable cue for instrument identity.

  92  

Figure 2.1 Spectral modulation detection. Schematic diagrams of the stimuli in the spectralmodulation detection task. Listeners had to distinguish a noise with a sinusoidal spectral shape over a logarithmic frequency axis (signal; left) from noises with a flat spectrum (references; middle and right) in each three-interval forced-choice trial. The signal interval was chosen at random. The modulation depth was varied adaptively to determine the spectral modulation detection threshold.

  93  

Figure 2.2 Learning Curves. The group average spectral modulation detection thresholds (79% correct performance) as a function of testing session for listeners trained at 0.5 (A, circles; n=8), 1 (B, triangles; n=12), and 2 (C, squares; n=7) cyc/oct as well as for controls who received no training (diamonds; n=8-12). The schematics near the top left corner of each panel illustrate the trained stimulus for that panel, with audio frequency on the abscissa and magnitude on the ordinate (axes not shown; see Figure 1). Error bars indicate one standard error of the mean. Asterisks indicate a significant difference (p < 0.05) between trained listeners and controls based on analyses of covariance using pre-training performance as a covariate. The 0.5- and 1-cyc/oct trained listeners learned significantly more than controls between the pre- and post-tests, while the 2-cyc/oct trained listeners did not.

  94  

Figure 2.3 Learning Curve Slopes. The slopes of regression lines fitted to the daily mean thresholds over the log10 of the session number for each listener in the 0.5- (A), 1- (B), and 2- (C) cyc/oct trained groups. Filled symbols indicate that the slope for that listener was significant and negative. The boxplots to the left of the individual points reflect the distribution of points, where the box is comprised of lines at the upper quartile, median, and lower quartile values and the whiskers extend from each end of the box to the maximum and minimum values (excluding outliers). Slopes were computed either across all sessions (left in each panel) or across all sessions excluding the pre-test (right in each panel). Asterisks indicate that the population of slopes was significantly (p < 0.05) less than zero according to a one-sample t-test. When all sessions were considered, the population of slopes was significantly negative for the 0.5- and 1cyc/oct trained groups, but not the 2-cyc/oct trained group. Only the population of slopes for the 0.5-cyc/oct trained group was significantly negative when the pre-training session was excluded.

  95  

Figure 2.4 Within-Session Performance. The group-average spectral modulation detection thresholds from the beginning and the end of each training session. The averages of the first three (left open symbols) and last three (right open symbols) threshold estimates from the same training session are connected by a line. The average pre- and post-training thresholds are also shown (filled symbols). The results are plotted separately for the 0.5- (A, circles), 1- (B, triangles), and 2- (C, squares) cyc/oct trained listeners. Error bars indicate one standard error of the mean. The 0.5-cyc/oct trained listeners showed no consistent within-session performance change, while the 1- and 2- cyc/oct-trained listeners consistently got worse within each session.

  96   Figure 2.5 Untrained Spectral Modulation Detection Conditions. Post-training spectral modulation detection thresholds for each tested condition for the 0.5- (circles; top row), 1(triangles; middle row), and 2- (squares; bottom row) cyc/oct trained listeners as well as for the controls (diamonds). Results are shown for the group averages (filled symbols) and the individual listeners (unfilled symbols). The thresholds were adjusted using pre-training performance as a covariate. The horizontal boxes represent the 95% confidence interval of the post-training performance of the controls, and the dashed lines represent the mean pre-training performance across groups. Error bars indicate one standard error of the mean.

Asterisks

indicate a significant (p < 0.05) difference between trained listeners and controls based on an analysis of covariance using pre-training performance as a covariate. Performance was evaluated for the trained condition (left column), untrained spectral modulation frequencies (middle columns), and an untrained carrier spectrum (right column). The top right panel is for both an untrained spectral modulation frequency and an untrained carrier spectrum. There were no untrained conditions on which trained listeners distinguished themselves significantly from controls.

  97  

  98  

Figure 3.1. Spectro-Temporal Modulation Filterbank. A schematic of hypothetical filters in a spectro-temporal modulation filterbank. Each spectrogram depicts the particular spectral modulation frequency (vertical spacing of bars), temporal modulation frequency (horizontal spacing of bars), and direction (up or down: direction of bar tilt) to which that filter is tuned. These filters are tuned to spectral-alone (middle column), temporal-alone (bottom row), or spectro-temporal (downward: left two columns; upward: right two columns) modulation.

  99   Figure 3.2. Performance on the Trained Depth-Discrimination Conditions. (A-C) The group average modulation depth-discrimination thresholds (79% correct performance) as a function of testing session for listeners trained on spectral (A, triangles; n=8), temporal (B, diamonds; n=8), or spectro-temporal (C, squares; n=8) modulation. Results are also shown for controls who received no training (circles; n=8). Spectrograms of each trained stimulus are depicted at the top of each column.

Error bars indicate +/- one standard error of the mean. Asterisks indicate a

significant interaction (p < 0.05) of a group (trained vs. control) x time (pre- vs. post-test) ANOVA using time as a repeated measure. (D-F) The slopes of individual regression lines fitted to all threshold estimates vs. the log10 of the session number for each listener in the spectral- (D), temporal- (E), and spectro-temporal- (F) modulation trained groups. Filled symbols indicate that the slope for that listener was significant and negative. The boxplots to the left of the individual points reflect the distribution of points, where the box is comprised of lines at the upper quartile, median, and lower quartile values and the whiskers extend to the maximum and minimum values (excluding outliers). Slopes were computed either across all sessions (left in each panel) or across all sessions excluding the pre-test (right in each panel). Asterisks indicate that the population of slopes was significantly less than zero (p < 0.05) according to a one-sample t-test. Training led to improvement for all three trained modulations, but with different time courses.

  100  

  101  

Figure 3.3 Performance on the Untrained Modulation Depth-Discrimination Conditions. The difference in threshold between the pre-test and post-test (pre minus post) for each group (bars) and listener (symbols) on the depth-discrimination conditions. Results are shown for spectral(A) and temporal- (B) modulation, as well as for upward (C) and downward (D) spectrotemporal-modulation. Spectrograms of the tested stimuli are displayed near the top right corner of each panel. Error bars indicate +/- one standard error of the mean. Asterisks indicate a significant interaction (p < 0.05) of a group (trained vs. control) x time (pre- vs. post-test) repeated measures ANOVA. The learning on the trained conditions did not generalize to any untrained depth-discrimination conditions.

  102  

  103   Figure 3.4 Performance on the Untrained Spectro-Temporal Modulation Detection Condition. (A) As in Figure 3.3, but for the detection (rather than discrimination) of upward spectrotemporal modulation. (B-E) The difference in threshold between the pre-test and post-test (pre minus post) for discrimination (abscissa) and detection (ordinate) of upward spectro-temporal modulation. Results are plotted separately for listeners trained to discriminate the depth of spectral (B), temporal (C), or spectro-temporal (D) modulation as well as for controls (E). Discrimination training led to worsening on detection but only when both tasks used the same modulation.

  104  

  105  

Figure 4.1. Tasks. Schematic diagrams of the stimuli used in the two tested tasks. (A) In the spectral modulation detection task, listeners had to distinguish a noise with a sinusoidal spectral shape over a logarithmic frequency axis (signal; solid line) from one with a flat spectrum (reference; dotted line). The modulation depth was varied adaptively to determine the spectral modulation detection threshold. (B) In the ripple reversal task, listeners had to distinguish a stimulus with a full wave rectified sinusoidal shape on a logarithmic frequency axis (signal; solid line) from one in which the peaks and valleys were interchanged with those of the signal stimulus (reference; dotted line). The spectral modulation frequency was varied adaptively to determine the ripple reversal threshold. Note that one spectral modulation cycle contains two peaks, because the sinusoid was full wave rectified.

  106  

Figure 4.2. Learning Curves. (A) The group average spectral modulation detection threshold on the trained condition (2 cyc/oct) as a function of testing session for the trained listeners (squares; n = 8) and controls (circles; n = 8). Values are adjusted using individual differences in pre-test performance as a covariate (Cohen, 1977). The schematic illustrates the trained stimulus, with audio frequency on the abscissa and magnitude on the ordinate (axes not shown; see Figure 1). Error bars indicate one standard error of the mean. (B) Individual performance for the trained listeners (lines) and controls (circles). Trained listeners improved gradually over the training phase, but controls also improved by a comparable amount.

  107  

  108  

Figure 4.3. Within Session Performance. The group-average spectral modulation detection thresholds toward the beginning and end of each training session. The averages of the first three (left square) and last three (right square) adjusted threshold estimates from the same training session are connected by a line. Values are adjusted using individual differences in pre-test performance as a covariate (Cohen, 1977). Error bars indicate one standard error of the mean. Trained listeners showed a rapid within-session improvement during the first training session and a gradual improvement across the remaining sessions.

  109  

Figure 4.4 Performance on Untrained Conditions. (A) Adjusted post-test spectral modulation detection thresholds for each tested condition for the trained listeners (squares) and controls (circles). Results are shown as group averages (filled symbols) and for each individual listener (unfilled symbols). Values are adjusted using individual differences in pre-test performance as a covariate (Cohen, 1977). Performance was evaluated for the trained (left column) and untrained spectral modulation frequencies (middle and right columns). A schematic of the tested stimulus is displayed above each column. The dashed lines represent the mean pre-training performance across groups, and the horizontal boxes represent the 95% confidence interval of the post-test performance of the controls. Error bars indicate one standard error of the mean. Asterisks indicate a significant (p < 0.05) difference between trained listeners and controls based on an analysis of covariance using pre-test performance as a covariate. (B) As in A, but for the ripple reversal task. Trained listeners learned more than controls on the untrained 1-cyc/oct condition, while both groups showed comparable improvements on the trained 2- and the untrained 4cyc/oct conditions, and neither group improved on the untrained task.

  110  

  111  

Listener T1 T2 T3 T4 T5 T6 T7 T8 C1 C2 C3 C4 C5 C6 C7 C8

0.25 40/40 40/35 20/25 10/10 25/30 20/25 45/45 30/30 25/30 35/40 35/40 15/15 25/25 35/30 45/45 35/30

0.5 45/40 40/40 30/40 15/20 35/35 25/30 45/50 40/35 35/35 35/40 20/40 25/20 25/35 40/30 50/40 45/45

1 40/35 35/40 45/45 25/20 40/45 30/35 45/45 45/45 45/35 45/50 30/40 30/20 50/50 50/50 60/35 50/50

Frequency  (kHz) 2 3 65/65 65/65 35/45 35/35 55/50 55/55 40/35 55/60 35/40 60/50 55/60 60/60 45/50 45/50 50/50 55/50 55/45 55/50 55/50 55/55 50/55 50/55 65/65 75/70 50/55 55/55 70/60 70/55 65/40 70/40 70/70 75/70

4 65/60 30/40 50/55 70/75 60/55 65/65 50/50 65/60 60/60 65/60 50/55 70/80 55/60 70/55 85/50 80/75

6 75/85 45/45 50/55 80/85 70/60 70/70 55/45 70/60 80/70 75/70 55/55 75/80 60/60 70/60 85/50 75/80

8 90/95 60/55 55/55 90/85 80/65 75/75 55/45 80/65 80/75 80/70 55/60 75/80 60/75 80/70 85/50 75/80

Age 72 67 74 74 70 76 72 60 76 77 78 77 72 82 56 81

Sex M F F M F F F F M M M M F M F M

Table 1. Listener Audiograms. Audiometric thresholds in dB HL are shown for the left/right ears of the trained listeners (T1-T8) and controls (C1-C8) at frequencies ranging from 0.25 to 8 kHz. Also shown are age (in years) and sex.

  112   References Adballah S, Olumbey M (2001) If the independent components of natural images are edges, what are the independent components of natural sounds. In: International Conference on Independent Component Analysis and Blind Signal Separation, pp 534–539 San Diego, CA. Adini Y, Wilkonsky A, Haspel R, Tsodyks M, Sagi D (2004) Perceptual learning in contrast discrimination: the effect of contrast uncertainty. J Vis 4:993-1005. Ahissar M, Hochstein S (2004) The reverse hierarchy theory of visual perceptual learning. Trends Cogn Sci 8:457-464. Ainsworth WA, Millar JB (1972) The effect of relative formant amplitude on the perceived identity of synthetic vowels. Language and speech 15:328-341. Andersen GJ, Ni R, Bower JD, Watanabe T (2010) Perceptual learning, aging, and improved visual performance in early stages of visual processing. J Vis 10:4. Astle AT, Webb BS, McGraw PV (2010) Spatial frequency discrimination learning in normal and developmentally impaired human vision. Vision Res 50:2445-2454. Attebo K, Mitchell P, Cumming R, Smith W, Jolly N, Sparkes R (1998) Prevalence and causes of amblyopia in an adult population. Ophthalmology 105:154-159. Bacon SP, Brandt JF (1982) Auditory processing of vowels by normal-hearing and hearingimpaired listeners. Journal of speech and hearing research 25:339-347. Bacon SP, Grantham DW (1989) Modulation masking: effects of modulation frequency, depth, and phase. The Journal of the Acoustical Society of America 85:2575-2580. Ball K, Sekuler R (1986) Improving visual perception in older observers. J Gerontol 41:176-182. Bradlow AR, Akahane-Yamada R, Pisoni DB, Tohkura Y (1999) Training Japanese listeners to identify English /r/ and /l/: long-term retention of learning in perception and production. Percept Psychophys 61:977-985. Bradlow AR, Pisoni DB, Akahane-Yamada R, Tohkura Y (1997) Training Japanese listeners to identify English /r/ and /l/: IV. Some effects of perceptual learning on speech production. The Journal of the Acoustical Society of America 101:2299-2310. Bregman A (1990) Auditory Scene Analysis: The Perceptual Organization of Sound. Cambridge, MA: The MIT Press. Burk MH, Humes LE (2007) Effects of training on speech recognition performance in noise using lexically hard words. J Speech Lang Hear Res 50:25-40.

  113   Burk MH, Humes LE (2008) Effects of long-term training on aided speech-recognition performance in noise in older adults. J Speech Lang Hear Res 51:759-771. Calhoun BM, Schreiner CE (1998) Spectral envelope coding in cat primary auditory cortex: linear and non-linear effects of stimulus characteristics. Eur J Neurosci 10:926-940. Carlyon RP, Shamma S (2003) An account of monaural phase sensitivity. The Journal of the Acoustical Society of America 114:333-348. Chen PL, Chen JT, Fu JJ, Chien KH, Lu DW (2008) A pilot study of anisometropic amblyopia improved in adults and children by perceptual learning: an alternative treatment to patching. Ophthalmic Physiol Opt 28:422-428. Chi T, Gao Y, Guyton MC, Ru P, Shamma S (1999) Spectro-temporal modulation transfer functions and speech intelligibility. The Journal of the Acoustical Society of America 106:2719-2732. Chi T, Ru P, Shamma SA (2005) Multiresolution spectrotemporal analysis of complex sounds. The Journal of the Acoustical Society of America 118:887-906. Cohen J (1977) Statistical power analysis for the behavioral sciences. Hillsdale, New Jersey: Lawrence Erlbaum Associates. Cox RM, Alexander GC, Taylor IM, Gray GA (1997) The contour test of loudness perception. Ear Hear 18:388-400. Dallos P (1992) The active cochlea. J Neurosci 12:4575-4585. Davis S, Mermelstein P (1980) Comparison of parametric representations for monosyllabic word recognition in continuously spoken sentences. IEEE Transactions on Acoustics, Speech and Signal Processing 28:357-366. De Valois KK (1977) Spatial frequency adaptation can enhance contrast sensitivity. Vision Res 17:1057-1065. De Valois RL, Albrecht DG, Thorell LG (1982) Spatial frequency selectivity of cells in macaque visual cortex. Vision Res 22:545-559. deCharms RC, Blake DT, Merzenich MM (1998) Optimizing sound features for cortical neurons. Science 280:1439-1443. Depireux DA, Simon JZ, Klein DJ, Shamma SA (2001) Spectro-temporal response field characterization with dynamic ripples in ferret primary auditory cortex. Journal of neurophysiology 85:1220-1234. DeValois R, DeValois K (1988) Spatial Vision. New York: Oxford Univ. Press.

  114   Dosher BA, Lu ZL (1998) Perceptual learning reflects external noise filtering and internal noise reduction through channel reweighting. Proc Natl Acad Sci U S A 95:13988-13993. Drennan WR, Watson CS (2001) Sources of variation in profile analysis. I. Individual differences and extended training. The Journal of the Acoustical Society of America 110:2491-2497. Drullman R, Festen JM, Plomp R (1994) Effect of reducing slow temporal modulations on speech reception. The Journal of the Acoustical Society of America 95:2670-2680. Eddins D, Harwell R (2002) Spatial Frequency Channels in Audition? In: Midwinter Meeting of the Association for Research in Otolaryngology. Eddins D, Liu C (2006) Enhancement of spectral modulation frequency in acoustics and identification of vowels. J Acoust Soc Am 119:3338(A). Eddins D, Liu C (2008) Improvement of speech recognition thresholds by spectral modulation enhancement. J Acoust Soc Am 123. Eddins D, Liu C (2009) Consonant identification in noise with and without spectro-temporal enhancement by normal hearing and hearing impaired listeners. Assoc Res Otolaryngol Abstr 892. Eddins D, Saoji A (2004) Phase sensitivity as a test of spectral modulation frequency channels. Assoc Res Otolaryngol Abstr 892. Eddins DA, Bero EM (2007) Spectral modulation detection as a function of modulation frequency, carrier bandwidth, and carrier frequency region. The Journal of the Acoustical Society of America 121:363-372. Edwards B (2007) The future of hearing aid technology. Trends Amplif 11:31-45. Eggermont JJ (1994) Temporal modulation transfer functions for AM and FM stimuli in cat auditory cortex. Effects of carrier type, modulating waveform and intensity. Hear Res 74:51-66. Elhilali M, Chi T, Shamma S (2003a) Intelligibility and the spectrotemporal representation of speech in the auditory cortex. Speech Communication 41:331-348. Elhilali M, Chi T, Shamma S (2003b) A spectro-temporal modulation index (STMI) for assessment of speech intelligibility. Speech Communication 41:331-348. Elhilali M, Shamma SA (2008) A cocktail party with a cortical twist: how cortical mechanisms contribute to sound segregation. The Journal of the Acoustical Society of America 124:3751-3771.

  115   Eronen A (2001) Comparison of features for musical instrument recognition. In: IEEE Workshop on the Applications of Signal Processing to Audio and Acoustics, pp 19-22 New Platz, NY. Field D (1994) What is the goal of sensory coding? Neural Computation 6:559-601. Fitzgerald MB, Wright BA (2005) A perceptual learning investigation of the pitch elicited by amplitude-modulated noise. The Journal of the Acoustical Society of America 118:37943803. Fitzgerald MB, Wright BA (2011) Perceptual learning and generalization resulting from training on an auditory amplitude-modulation detection task. The Journal of the Acoustical Society of America 129:898-906. Fronius M, Cirina L, Kuhli C, Cordey A, Ohrloff C (2006) Training the adult amblyopic eye with "perceptual learning" after vision loss in the non-amblyopic eye. Strabismus 14:7579. Giraud AL, Lorenzi C, Ashburner J, Wable J, Johnsrude I, Frackowiak R, Kleinschmidt A (2000) Representation of the temporal envelope of sounds in the human brain. J Neurophysiol 84:1588-1598. Glasberg BR, Moore BC (1986) Auditory filter shapes in subjects with unilateral and bilateral cochlear impairments. The Journal of the Acoustical Society of America 79:1020-1033. Graham N (1972) Spatial frequency channels in the human visual system: effects of luminance and pattern drift rate. Vision Res 12:53-68. Green DM (1987) Profile Analysis : Auditory Intensity Discrimination. Oxford, UK: Oxford University Press. Grey JM (1977) Multidimensional perceptual scaling of musical timbres. J Acoust Soc Am 61:1270-1277. Henry BA, Turner CW (2003) The resolution of complex spectral patterns by cochlear implant and normal-hearing listeners. The Journal of the Acoustical Society of America 113:2861-2873. Henry BA, Turner CW, Behrens A (2005) Spectral peak resolution and speech recognition in quiet: normal hearing, hearing impaired, and cochlear implant listeners. The Journal of the Acoustical Society of America 118:1111-1121. Houtgast T (1989) Frequency selectivity in amplitude-modulation detection. The Journal of the Acoustical Society of America 85:1676-1680. Huang CB, Lu ZL, Zhou Y (2009) Mechanisms underlying perceptual learning of contrast detection in adults with anisometropic amblyopia. J Vis 9:24 21-14.

  116   Huang CB, Zhou Y, Lu ZL (2008) Broad bandwidth of perceptual learning in the visual system of adults with anisometropic amblyopia. Proc Natl Acad Sci U S A 105:4068-4073. Huyck JJ, Wright BA (2011) Late maturation of auditory perceptual learning. Dev Sci 14:614621. Julez B (1971) Foundations of the Cyclopean Perception: University of Chicago Press. Keeling MD, Calhoun BM, Kruger K, Polley DB, Schreiner CE (2008) Spectral integration plasticity in cat auditory cortex induced by perceptual training. Exp Brain Res 184:493509. Kidd G, Jr., Mason CR, Green DM (1986) Auditory profile analysis of irregular sound spectra. The Journal of the Acoustical Society of America 79:1045-1053. Klein DJ, Depireux DA, Simon JZ, Shamma SA (2000) Robust spectrotemporal reverse correlation for the auditory system: optimizing stimulus design. J Comput Neurosci 9:85111. Korman M, Raz N, Flash T, Karni A (2003) Multiple shifts in the representation of a motor sequence during the acquisition of skilled performance. Proc Natl Acad Sci U S A 100:12492-12497. Kowalski N, Depireux DA, Shamma SA (1996) Analysis of dynamic spectra in ferret primary auditory cortex. I. Characteristics of single-unit responses to moving ripple spectra. Journal of neurophysiology 76:3503-3523. Kraus N, Chandrasekaran B (2010) Music training for the development of auditory skills. Nat Rev Neurosci 11:599-605. Kuai SG, Zhang JY, Klein SA, Levi DM, Yu C (2005) The essential role of stimulus temporal patterning in enabling perceptual learning. Nat Neurosci 8:1497-1499. Langers DR, Backes WH, van Dijk P (2003) Spectrotemporal features of the auditory cortex: the activation in response to dynamic ripples. NeuroImage 20:265-275. Langner G, Sams M, Heil P, Schulze H (1997) Frequency and periodicity are represented in orthogonal maps in the human auditory cortex: evidence from magnetoencephalography. J Comp Physiol A 181:665-676. Leek MR, Dorman MF, Summerfield Q (1987) Minimum spectral contrast for vowel identification by normal-hearing and hearing-impaired listeners. The Journal of the Acoustical Society of America 81:148-154. Levitt H (1971) Transformed up-down methods in psychoacoustics. The Journal of the Acoustical Society of America 49:Suppl 2:467+.

  117   Li RW, Klein SA, Levi DM (2008) Prolonged perceptual learning of positional acuity in adult amblyopia: perceptual template retuning dynamics. J Neurosci 28:14223-14229. Lindblom BE, Studdert-Kennedy M (1967) On the role of formant transitions in vowel recognition. The Journal of the Acoustical Society of America 42:830-843. Litvak LM, Spahr AJ, Saoji AA, Fridman GY (2007) Relationship between perception of spectral ripple and speech recognition in cochlear implant and vocoder listeners. The Journal of the Acoustical Society of America 122:982-991. Liu C, Eddins D (2010) Improvement of consonant identification in noise by spectro-temporal enhancement. J Acoust Soc Am 127:1904(A). Loftus WC, Sutter ML (2001) Spectrotemporal organization of excitatory and inhibitory receptive fields of cat posterior auditory field neurons. J Neurophysiol 86:475-491. Macpherson EA, Middlebrooks JC (2003) Vertical-plane sound localization probed with ripplespectrum noise. The Journal of the Acoustical Society of America 114:430-445. Maffei L, Fiorentini A (1973) The visual cortex as a spatial frequency analyser. Vision Res 13:1255-1267. Marrone N, Hu R, Wright BA (2010) Age-related changes in auditory perceptual learning. Association for Research in Otolaryngology Abstracts 33. Mayer MJ (1983) Practice improves adults' sensitivity to diagonals. Vision Res 23:547-550. McAdams S, Beauchamp JW, Meneguzzi S (1999) Discrimination of musical instrument sounds resynthesized with simplified spectrotemporal parameters. The Journal of the Acoustical Society of America 105:882-897. McAdams S, Winsberg S, Donnadieu S, De Soete G, Krimphoff J (1995) Perceptual scaling of synthesized musical timbres: common dimensions, specificities, and latent subject classes. Psychol Res 58:177-192. McGaugh JL (2000) Memory--a century of consolidation. Science 287:248-251. Merzenich MM, Jenkins WM, Johnston P, Schreiner C, Miller SL, Tallal P (1996) Temporal processing deficits of language-learning impaired children ameliorated by training. Science 271:77-81. Middlebrooks JC, Green DM (1991) Sound localization by human listeners. Annual review of psychology 42:135-159. Miller LM, Escabi MA, Read HL, Schreiner CE (2002) Spectrotemporal receptive fields in the lemniscal auditory thalamus and cortex. Journal of neurophysiology 87:516-527.

  118   Moller AR (1974) Responses of units in the cochlear nucleus to sinusoidally amplitudemodulated tones. Exp Neurol 45:105-117. Moore BC, Glasberg BR (1987) Formulae describing frequency selectivity as a function of frequency and level, and their use in calculating excitation patterns. Hear Res 28:209225. Moore DR, Hogan SC, Kacelnik O, Parsons CH, Rose MM, King AJ (2001) Auditory learning as a cause and treatment of central dysfunction. Audiol Neurootol 6:216-220. Neher T, Laugesen S, Jensen N, Kragelund L (2011) Can basic auditory and cognitive measures predict hearing-impaired listeners’ localization and spatial speech recognition abilities? The Journal of the Acoustical Society of America 130:1542-1558. Ortiz JA, Wright BA (2010) Differential rates of consolidation of conceptual and stimulus learning following training on an auditory skill. Exp Brain Res 201:441-451. Otto TU, Herzog MH, Fahle M, Zhaoping L (2006) Perceptual learning with spatial uncertainties. Vision Res 46:3223-3233. Pardo B Unpublished Data . Peelle JE, Wingfield A (2005) Dissociations in perceptual learning revealed by adult age differences in adaptation to time-compressed speech. J Exp Psychol Hum Percept Perform 31:1315-1330. Peterson GE, Barney HL (1952) Control methods used in a study of the vowels. J Acoust Soc Am 24:175-184. Polat U, Ma-Naim T, Belkin M, Sagi D (2004) Improving vision in adult amblyopia by perceptual learning. Proc Natl Acad Sci U S A 101:6692-6697. Qian J, Eddins DA (2008) The role of spectral modulation cues in virtual sound localization. The Journal of the Acoustical Society of America 123:302-314. Raj B, Smaragdis P (2005) Latent variable decomposition of spectrograms for single channel speaker separation. In: IEEE Workshop on Applications of Signal Processing to Audio and Acoustics, pp 17-20. Regan D, Beverley KI (1985) Postadaptation orientation discrimination. J Opt Soc Am A 2:147155. Richards VM, Buss E, Tian L (1997) Effects of modulator phase for comodulation masking release and modulation detection interference. The Journal of the Acoustical Society of America 102:468-476.

  119   Robinson K, Summerfield AQ (1996) Adult auditory learning and training. Ear Hear 17:51S65S. Roth DA, Kishon-Rabin L, Hildesheimer M, Karni A (2005) A latent consolidation phase in auditory identification learning: time in the awake state is sufficient. Learn Mem 12:159164. Sabin AT, Macpherson EA, Middlebrooks JC (2005) Vertical-Plane Localization of Sounds with Distorted Spectral Cues. Abstracts of the Association for Research in Otolaryngology 1389. Sakayori S, Kitama T, Chimoto S, Qin L, Sato Y (2002) Critical spectral regions for vowel identification. Neuroscience research 43:155-162. Saoji AA, Eddins DA (2007) Spectral modulation masking patterns reveal tuning to spectral envelope frequency. The Journal of the Acoustical Society of America 122:1004-1013. Saoji AA, Eddins DA (2007) Spectral modulation masking patterns reveal tuning to spectral envelope frequency. The Journal of the Acoustical Society of America 122:1004-1013. Saoji AA, Litvak L, Spahr AJ, Eddins DA (2009) Spectral modulation detection and vowel and consonant identifications in cochlear implant listeners. The Journal of the Acoustical Society of America 126:955-958. Schonwiesner M, Zatorre RJ (2009) Spectro-temporal modulation transfer function of single voxels in the human auditory cortex measured with high-resolution fMRI. Proc Natl Acad Sci U S A 106:14611-14616. Schreiner C, Calhoun B (1994) Spectral envelope coding in cat primary auditory cortex: properties of ripple transfer functions. Audit Neurosci 1:39–61. Schreiner CE, Langner G (1988) Periodicity coding in the inferior colliculus of the cat. II. Topographical organization. J Neurophysiol 60:1823-1840. Schreiner CE, Mendelson JR (1990) Functional topography of cat primary auditory cortex: distribution of integrated excitation. Journal of neurophysiology 64:1442-1459. Schreiner CE, Sutter ML (1992) Topography of excitatory bandwidth in cat primary auditory cortex: single-neuron versus multiple-neuron recordings. Journal of neurophysiology 68:1487-1502. Schreiner CE, Urbas JV (1988) Representation of amplitude modulation in the auditory cortex of the cat. II. Comparison between cortical fields. Hear Res 32:49-63. Shamma S (2001) On the role of space and time in auditory processing. Trends Cogn Sci 5:340348.

  120   Shamma S, Versnel H, Kowalski N (1995) Ripple analysis in ferret primary auditory cortex. I. Response characteristics of single units to sinusoidal rippled spectra. Audit Neurosci 1. Shamma SA, Fleshman JW, Wiser PR, Versnel H (1993) Organization of response areas in ferret primary auditory cortex. J Neurophysiol 69:367-383. Sidwell A, Summerfield Q (1985) The effect of enhanced spectral contrast on the internal representation of vowel-shaped noise. The Journal of the Acoustical Society of America 78:495-506. Smith ML, Cottrell GW, Gosselin F, Schyns PG (2005) Transmitting and decoding facial expressions. Psychol Sci 16:184-189. Sowden PT, Davies IR, Roling P (2000) Perceptual learning of the detection of features in X-ray images: a functional role for improvements in adults' visual sensitivity? J Exp Psychol Hum Percept Perform 26:379-390. Sowden PT, Rose D, Davies IR (2002) Perceptual learning of luminance contrast detection: specific for spatial frequency and retinal location but not orientation. Vision Res 42:1249-1258. Sowden PT, Schyns PG (2006) Channel surfing in the visual brain. Trends Cogn Sci 10:538-545. Stecker GC, Bowman GA, Yund EW, Herron TJ, Roup CM, Woods DL (2006) Perceptual training improves syllable identification in new and experienced hearing aid users. J Rehabil Res Dev 43:537-552. Stevens S, Volkman J, Newman E (1937) A scale for the measurement of the psychological magnitude pitch. J Acoust Soc Am 8:185-190. Summers V, Leek MR (1994) The internal representation of spectral contrast in hearingimpaired listeners. The Journal of the Acoustical Society of America 95:3518-3528. Sweetow RW, Sabes JH (2006) The need for and development of an adaptive Listening and Communication Enhancement (LACE) Program. J Am Acad Audiol 17:538-558. Tallal P, Miller SL, Bedi G, Byma G, Wang X, Nagarajan SS, Schreiner C, Jenkins WM, Merzenich MM (1996) Language comprehension in language-learning impaired children improved with acoustically modified speech. Science 271:81-84. Tansley BW, Suffield JB (1983) Time course of adaptation and recovery of channels selectively sensitive to frequency and amplitude modulation. The Journal of the Acoustical Society of America 74:765-775. Tootell RB, Silverman MS, De Valois RL (1981) Spatial frequency columns in primary visual cortex. Science 214:813-815.

  121   Van Tasell DJ, Fabry DA, Thibodeau LM (1987) Vowel identification and vowel masking patterns of hearing-impaired subjects. The Journal of the Acoustical Society of America 81:1586-1597. Versnel H, Kowalski N, Shamma SA (1995) Ripple analysis in ferret primary auditory cortex. III. Topographic distribution of ripple response parameters. Audit Neurosci 1:271-285. Versnel H, Shamma SA (1998) Spectral-ripple representation of steady-state vowels in primary auditory cortex. The Journal of the Acoustical Society of America 103:2502-2514. Versnel H, Zwiers MP, van Opstal AJ (2009) Spectrotemporal response properties of inferior colliculus neurons in alert monkey. J Neurosci 29:9725-9739. Von Békésy G (1960) Experiments in hearing. Oxford, England: Mcgraw Hill. von Helmholtz H (1863) On the Sensations of Tone as a Physiological Basis for the Theory of Music. Oxford, England: Dover Publications. Walden BE, Erdman SA, Montgomery AA, Schwartz DM, Prosek RA (1981) Some effects of training on speech recognition by hearing-impaired adults. Journal of speech and hearing research 24:207-216. Walker MP, Stickgold R (2004) Sleep-dependent learning and memory consolidation. Neuron 44:121-133. Wenger MJ, Rasche C (2006) Perceptual learning in contrast detection: presence and cost of shifts in response criteria. Psychon Bull Rev 13:656-661. Wojtczak M, Viemeister NF (2003) Suprathreshold effects of adaptation produced by amplitude modulation. The Journal of the Acoustical Society of America 114:991-997. Won JH, Drennan WR, Kang RS, Rubinstein JT (2010) Psychoacoustic abilities associated with music perception in cochlear implant users. Ear Hear 31:796-805. Won JH, Drennan WR, Rubinstein JT (2007) Spectral-ripple resolution correlates with speech reception in noise in cochlear implant users. J Assoc Res Otolaryngol 8:384-392. Woolley SM, Fremouw TE, Hsu A, Theunissen FE (2005) Tuning for spectro-temporal modulations as a mechanism for auditory discrimination of natural sounds. Nat Neurosci 8:1371-1379. Wright BA (2001) Why and how we study human learning on basic auditory tasks. Audiol Neurootol 6:207-210. Wright BA, Dai H (1998) Detection of sinusoidal amplitude modulation at unexpected rates. The Journal of the Acoustical Society of America 104:2991-2996.

  122   Wright BA, Wilson RM, Sabin AT (2010) Generalization lags behind learning on an auditory perceptual task. J Neurosci 30:11635-11639. Wright BA, Zhang Y (2009a) Insights into human auditory processing gained from perceptual learning. In: The cognitive neurosciences vol. IV (Gazzaniga, M. S., ed), pp 353-366 Cambridge, MA: MIT. Wright BA, Zhang Y (2009b) A review of the generalization of auditory learning. Philos Trans R Soc Lond B Biol Sci 364:301-311. Yost WA, Sheft S (1989) Across-critical-band processing of amplitude-modulated tones. The Journal of the Acoustical Society of America 85:848-857. Yost WA, Sheft S, Opie J (1989) Modulation interference in detection and discrimination of amplitude modulation. The Journal of the Acoustical Society of America 86:2138-2147. Yu C, Klein SA, Levi DM (2004) Perceptual learning in contrast discrimination and the (minimal) role of context. J Vis 4:169-182. Zahorian SA, Jagharghi AJ (1993) Spectral-shape features versus formants as acoustic correlates for vowels. The Journal of the Acoustical Society of America 94:1966-1982. Zhou Y, Huang C, Xu P, Tao L, Qiu Z, Li X, Lu ZL (2006) Perceptual learning improves contrast sensitivity and visual acuity in adults with anisometropic amblyopia. Vision Res 46:739-750.

  123   Curriculum Vitae Andrew Todd Sabin Frances Searle Building 2240 Campus Drive, 2-226 Evanston, IL 60208 USA [email protected] Telephone: 847.491.2462 Fax: 847.491.4975 Education 2000-2004 Bachelor of Science with Highest Honors Biopsychology and Cognitive Science University of Michigan, Ann Arbor, MI, USA Adviser: John Middlebrooks 2004-2011 Ph.D. Communication Sciences and Disorders Northwestern University, Evanston, IL, USA Adviser: Beverly A. Wright 2010 Certificate Management Program for Scientists and Engineers Kellogg School of Management Northwestern University, Evanston, IL, USA Awards 2011 - American Auditory Society Graduate Student Poster Session grant (funded by National Institutes of Health) 2010 - Fellow, American Speech, Hearing and Language Association Lessons for Success Conference (funded by National Institutes of Health) 2008 - Present Fellow. Ruth Kirschstein National Research Service Award. National Institutes of Health. 2011 - Ph.D. Certificate. Northwestern Institute on Complex Systems: Language and Music. 2004 - Graduate Student Fellowship. Northwestern University. Department of Communication Sciences and Disorders. 2004 - Honorable Mention: National Science Foundation Graduate Research Fellowship. 2004 - Pillsbury prize for the most distinguished Undergraduate thesis in Psychology at the University of Michigan Publications Peer-Reviewed Journal Articles

  124   Sabin AT, Eddins D, Wright BA. (in Revision) Perceptual learning of auditory spectral modulation detection. In Revision for: Exp BrainRes Sabin AT, Eddins D, Wright BA. (in Review) Auditory perceptual learning specificity to combined spectro-temporal modulation. In review at: J Neurosci Sabin AT, Clark C, Dhar S, Eddins D, Wright BA. (in Preparation) Different patterns of auditory perceptual learning on spectral modulation detection between older hearing-impaired and younger normal-hearing adults. For submission to: Learn Mem. Macpherson EA, Sabin AT (in Revision) Vertical-plane sound localization at low and high sound levels modeled with distorted spectral cues. Reviewed at: Hear Res. Sabin AT, Rafii Z, Pardo B. (2011) Regression-Based Rapid Mapping of User Descriptors to Audio Processing Parameters. J Audio Eng Soc 59(6):419-430. Banai K, Sabin AT, Wright BA. (2011) Separable developmental trajectories for the abilities to detect auditory amplitude and frequency modulation. Hear Res. Sabin AT, Marrone N, Hardies L, Dhar S (2011) Weighting Function-Based Mapping of Descriptors to Frequency-Gain Curves in Listeners With Hearing Loss. Ear Hear 32(3):399-409. Wright BA, Sabin AT, Zhang Y, Marrone N, Fitzgerald M (2010) Enhancing perceptual learning by combining practice with periods of additional sensory stimulation. J Neurosci 30(38): 12868-12877. Wright BA, Wilson, R, Sabin AT (2010) Generalization lags behind learning on an auditory perceptual task. J Neurosci 30(35): 11635-11639. Wright BA, Sabin AT (2007) Perceptual learning: how much daily training is enough? Exp Brain Res 180: 727-736. Macpherson EA, Sabin AT (2007) Binaural weighting of monaural spectral cues for sound localization. J Acoust Soc Am 121: 3677-3688. Sabin AT, Macpherson EA, Middlebrooks JC (2005) Human sound localization at near-threshold levels. Hear Res 199: 124-134 Peer-Reviewed Conference Papers Sabin AT, Chan CL (2011) myMicSound: An online sound-based microphone recommendation system. ACM Recommender Systems. Chicago, IL, October 23-27. Best V, Brungart DS, Carlile S, Jin F, Macpherson EA, Martin RL, McAnally KI, Sabin AT, Simpson BD (2009) A Meta-Analysis of Localization Errors Made In The Anechoic Free Field. IWPASH. Kyoto, Japan, November 11-13. Sabin AT, Pardo B (2009) A method for rapid personalization of audio equalization parameters. ACM Multimedia. Beijing, China, October 19-24. Sabin AT, Pardo B (2009) Facilitating Creativity With A Novel Audio Equalizer. ACM Creativity and Cognition. Berkeley, CA, October 27-30. Sabin AT, Pardo B (2008) Rapid learning of subjective preference in equalization. Proceedings of the Audio Engineering Society 125th Convention. San Francisco, CA , October 2-5. Fox B, Sabin AT, Zopf A, Pardo B (2007) Modeling Perceptual Similarity of Audio Signals for Blind Source Separation Evaluation In: Intl. Conf. on Independent Component Analysis and Signal Separation. Springer, London, UK

  125   Peer-Reviewed Conference Abstracts Sabin AT, Eddins DA, Wright BA (2011) Spectro-temporal specificity of learning on modulation depth discrimination. 161st Meeting of the Acoustical Society of America Lee J, Souza P, Sabin AT, Kwon B, Brennan M, Poling G, Petersen C (2011) Dynamic range compression effects on modulation detection interference. 161st Meeting of the Acoustical Society of America Sabin AT, Wiles H, Souza P (2011) Temporal-based non-linear hearing aid prescription using a genetic algorithm. American Auditory Society Meeting Annual Meeting. Sabin AT, Eddins DA, Wright BA (2010) The influence of practice on the discrimination of spectro-temporal modulation depth. 158TH Meeting of the Acoustical Society of America. Andeol G, Sabin AT (2010) Sound localization and listener's spectral sensitivity to shape. 10th French Congress of Acoustics. Sabin AT, Eddins DA, Wright BA (2009) Spectral modulation detection training: Different outcomes for different trained stimuli. 157TH Meeting of the Acoustical Society of America: 125, 2526. Sabin AT, Marrone N, Dhar S (2009) A weighting-function-based approach to subjectively modify the frequency response of a hearing aid. 157TH Meeting of the Acoustical Society of America: 125, 2724. Sabin AT, Clark CA, Eddins DA, Dhar S, Wright BA (2009) Spectral modulation detection training in older adults with hearing loss. 157TH Meeting of the Acoustical Society of America: 125, 2633. Wright BA, Banai K, Sabin AT, Zhang Y (2009) Distinct phases of auditory learning identified by differences in vulnerability to intervening events. Abstracts of the Association for Research in Otolaryngology: 142. Sabin AT, DePalma CL, Eddins DA, Wright BA (2008) Training induced improvements in the ability to detect spectral modulation. Abstracts of the Association for Research In Otolaryngology: 915. Wright BA, Sabin AT, Wilson RM (2008) Disruption of the Consolidation of Learning on an Auditory Temporal-Interval Discrimination Task. Abstracts of the Association for Research In Otolaryngology: 1294. Wright BA, Sabin AT (2007) Different daily training requirements for learning versus generalization on duration discrimination. Abstracts of the Association for Research In Otolaryngology: 398. Wright BA, Sabin AT, Wilson RM (2007) Enhancing perceptual learning by practicing an irrelevant task. Abstracts of the Cognitive Neuroscience Society Annual Meeting: E95. Banai K, Sabin AT, Kraus N, Wright BA (2007) The Development of Sensitivity to Amplitude and Frequency Modulation Follow Distinct Time Courses. Abstracts of the Association for Research In Otolaryngology: 928. Sabin AT, Wright BA (2006) Contribution of Passive Stimulus Exposures to Learning on Auditory Frequency Discrimination. Abstracts Association for Research In Otolaryngology: 217. Wright BA, Sabin AT, Ortiz JA, Stewart CC, Fitzgerald MB (2006) Different Influences of Varying the Number of Daily Training Trials on Learning on Frequency- and Temporal-

  126   Interval Discrimination. Abstracts of the Association for Research in Otolayngology: 214. Banai K, Sabin AT, Kraus N, Wright BA (2006) The development Of sensitivity to dynamic auditroy stimuli in school-age children. Abstracts of the 15th Israel Society for Neuroscience Meeting: 10. Sabin AT, Macpherson EA, Middlebrooks JC (2005) Vertical-Plane Localization of Sounds with Distorted Spectral Cues. Abstracts of the Association for Research in Otolaryngology: 1389. Sabin AT, Macpherson EA, Middlebrooks JC (2004) Localization of Sound at Near-Threshold Intensities. Abstracts of the Association for Research In Otolaryngology: 1096. Macpherson EA, Sabin AT, Middlebrooks JC (2004) Binaural Weighting of Monaural Spectral Cues for Sound Localization. Abstracts of the Association for Research In Otolaryngology: 1081. Patent (pending) SYSTEMS, METHODS, AND APPARATUS FOR Equalization Preference Learning. Inventors: Andrew Sabin, Bryan Pardo. Assignee: Northwestern University. Teaching Rush University: Adjust Lecturer (Fall 2011): Anatomy and physiology of hearing and speech Northwestern University: Teaching Assistant (2004-2007) Introduction to Psychoacoustics (3 quarters) Biological Foundations of Speech and Music (3 Quarters) Introduction to Amplification (1 quarter) Computational Modeling in Communication Sciences and Disorders (1 quarter) University of Michigan: Instructor (2003) Music Technology Camp (High School Students) Advising 2007 Cara DePalma Au.D. Capstone Project 2008 Cynthia Clark, Au.D. Capstone Project 2009 Nichele Taft, Au.D, Capstone Project 2010 Lauren Hardies, Au.D. Capstone Project 2011 Holly Wiles, Au.D. Capstone Project Ad-Hoc Reviewer Journal of the Acoustical Society of America Computer Music Journal

  127   International Society for Music Information Retrieval Conference PLoS ONE Public Display Sabin AT (2004) An Acoustical Analysis of the Motown Reverberation Chamber: the Motown Historical Museum. Detroit, MI Popular Press Work mentioned in New Scientist, Wired, US News and World Report, Audiology Today, Sound on Sound, Radio World, Computer Music Magazine

NORTHWESTERN UNIVERSITY Perceptual ...

A DISSERTATION. SUBMITTED TO THE GRADUATE SCHOOL. IN PARTIAL FULFILLMENT OF THE REQUIREMENTS. For the degree. DOCTOR OF PHILOSOPHY. Field of Communication Sciences and Disorders ... These results therefore provide behavioral evidence that is consistent with the proposal that the auditory ...

2MB Sizes 1 Downloads 313 Views

Recommend Documents

NORTHWESTERN UNIVERSITY Perceptual Learning ...
Processing Advantages of Spectral and Spectro-Temporal. Modulation Filterbanks ..... All stimuli were presented using custom software written in. MATLAB and ...

Northwestern University
the risk management system of the rural banks in Ilocos Norte that help mitigate ... To test the difference between the responses of the rural bank managers and ...

Northwestern University
Two sets of questionnaires were used. The first was administered to the employees who performed the ICS-treated tasks before the implementation of the ICS.

Northwestern University
network development along road influenced areas in Laoag City. The study utilized ... minimal establishment/expansion of business in the road influences areas ...

Convention Paper - Computer Science - Northwestern University
Listeners were seated in a quiet room with a computer ... channel, there were 75 data points, where the within- .... plotted in the rightmost box plot of Figure 2.

Convention Paper - Computer Science - Northwestern University
represented as points on a grid. On each trial, the listener makes two paired preference judgments: one in which the two settings differ in high frequency gain, ...

Dynamic Adverse Selection - Economics - Northwestern University
Apr 14, 2013 - capturing our main idea that illiquidity may separate high and low quality assets in markets ... that she might later have to sell it, the owner of an asset had an incentive to learn its quality. ..... The proof is in an online appendi

Dynamic Adverse Selection - Economics - Northwestern University
Apr 14, 2013 - Of course, in reality adverse selection and search frictions may coexist in a market, and it is indeed ..... The proof is in an online appendix. Note that for .... Figure 1: Illustration of problem (P) and partial equilibrium. Figure 1

Northwestern University, Department of Communication ...
Compute Gain Control Signal. Apply Gain ... Gain. 0 dB. 70 dB uniform. The Unmodified Channel-Specific. Threshold. UCL. Notch- .... NAL-NL2 fitting software.

Convention Paper - Computer Science - Northwestern University
Machine ratings generated by computing the similarity of a given curve to the weighting function are ..... [11] Richards, V.M. and S. Zhu, "Relative estimates of.

NORTHWESTERN UNIVERSITY Tribological Contact ...
deformation. Surface heating causes the distortion of contacting bodies and temperature rises, which are responsible for interfacial degradation. The coupled thermomechanical contact analysis is critical to understanding the origin of heat-induced fa

pdf-12103\northwestern-university-law-review-volume-9-by ...
pdf-12103\northwestern-university-law-review-volume-9-by-anonymous.pdf. pdf-12103\northwestern-university-law-review-volume-9-by-anonymous.pdf. Open.

When shock waves hit traffic - Northwestern University Transportation ...
Jun 25, 1994 - That can cause the third driver in line to slow down even more, and so on, in an amplifying wave of deceleration. If enough drivers overreact, those at ... against traffic congestion, he says, planners need to make tactical decisions s

man-53\northwestern-state-university-transcript-request.pdf ...
man-53\northwestern-state-university-transcript-request.pdf. man-53\northwestern-state-university-transcript-request.pdf. Open. Extract. Open with. Sign In.

Download - Northwestern University School of Education and Social ...
Nov 11, 2008 - This article was downloaded by: [Adler, Jonathan M.] ... and conditions of use: http://www.informaworld.com/terms-and-conditions-of-access.pdf.

please scroll down for article - Northwestern University School of ...
Nov 11, 2008 - Spence (1982) echoed this sentiment, writing that ... retain what he [sic] learned during the analysis''. (p. 270). ..... But I now have better tools to work with*One of the ... else, but I can still do good work and enjoy life. In sum

Behavioral Theories of the Business Cycle - Northwestern University
Abstract. We explore the business cycle implications of expectation shocks and ... The fundamental shock in our model is investment-specific technical change.

Perceptual Reasoning for Perceptual Computing
Department of Electrical Engineering, University of Southern California, Los. Angeles, CA 90089-2564 USA (e-mail: [email protected]; dongruiw@ usc.edu). Digital Object ... tain a meaningful uncertainty model for a word, data about the word must be

PERCEPTUAL CoMPUTINg - CS UTEP
“Perceptual Computing Programs (PCP)” and “IJA Demo.” In the PCP folder, the reader will find separate folders for Chapters 2–10. Each of these folders is.

Perceptual Reward Functions - GitHub
expected discounted cumulative reward an agent will receive after ... Domains. Task Descriptors. Figure 3: Task Descriptors. From left to right: Breakout TG, ...

PERCEPTUAL CoMPUTINg
Map word-data with its inherent uncertainties into an IT2 FS that captures .... 3.3.2 Establishing End-Point Statistics For the Data. 81 .... 7.2 Encoder for the IJA.

Similarity-Based Perceptual Reasoning for Perceptual ...
Dongrui Wu, Student Member, IEEE, and Jerry M. Mendel, Life Fellow, IEEE. Abstract—Perceptual reasoning (PR) is ... systems — fuzzy logic systems — because in a fuzzy logic system the output is almost always a ...... in information/intelligent