Cognitive and Auditory Factors Underlying Auditory Spatial Attention in Younger and Older Adults

Transcription

1 Cognitive and Auditory Factors Underlying Auditory Spatial Attention in Younger and Older Adults by Gurjit Singh A thesis submitted in conformity with the requirements for the degree of Doctor of Philosophy Department of Psychology University of Toronto Copyright by Gurjit Singh 2010

2 Cognitive and Auditory Factors Underlying Auditory Spatial Attention in Younger and Older Adults Abstract Gurjit Singh Doctor of Philosophy Department of Psychology University of Toronto 2010 Listening to speech with competing speech in the background is challenging and becomes harder with age. Three experiments examined the auditory and cognitive aspects of auditory spatial attention in conditions in which the location of the target was uncertain. In all experiments, word identification was measured for target sentences presented with two competitor sentences. On each trial, the three sentences were presented with one from each of three spatially separated loudspeakers. A priori cues specified the location and identity callsign of the target. In Experiments I and II, sentences were also presented in conditions of simulated spatial separation achieved with the precedence effect. Participants were younger and older adults with normal hearing sensitivity below 4 khz. For both age groups, the contributions of richer acoustic cues (those present when there was real spatial separation, but absent when there was simulated spatial separation) were most pronounced when the target occurred at unlikely spatial listening locations, suggesting that both age groups benefit similarly from richer acoustical cues. In Experiment II, the effect of time between the callsign cue and target word on word identification was investigated. Four timing conditions were tested: the original sentences (which contained about 300 ms of filler speech between the callsign cue and the onset of the target words), or modified sentences with silent pauses of 0, 150, or 300 ms replacing the filler speech. For targets ii

3 presented from unlikely locations, word identification was better for all listeners when there was more time between the callsign cue and key words, suggesting that time is needed to switch spatial attention. In Experiment III, the effects of single and multiple switches of attention were investigated. The key finding was that, whereas both age groups performed similarly in conditions requiring a single switch of attention, the performance of older, but not younger listeners, was reduced when multiple switches of spatial attention were required. This finding suggests that difficulties disengaging attention may contribute to the listening difficulties of older adults. In conclusion, cognitive and auditory factors contributing to auditory spatial attention appear to operate similarly for all listeners in relatively simple situations, and age-related differences are observed in more complex situations. iii

4 Acknowledgments What an amazing six years it has been! First, I would like to thank my supervisor Dr. Kathy Pichora-Fuller who has been the ideal mentor. I also owe deep gratitude to Dr. Bruce Schneider. To you both, I wish to express my heartfelt gratitude for your generosity of time and spirit and for creating for your students the model environment in which to conduct research. It has been a pleasure working with and learning from both of you. There are also a number of fantastic colleagues without whom this dissertation would not have been possible. Many thanks to Susan Alison, Meital Avivi, Jessica Banh, Boaz Ben-David, Jane Carey, Alham Chelehmalzadeh, Marco Colleta, Stephan de la Rossa, Christine De Luca, Sabrena Deonarain, Kate Dupuis, Payam Ezzation, Huiwen Goy, Antje Heinrich, Salima Jiwani, Lulu Li, Mitra Mehra, Ewen MacDonald, Danielle Minghella, Alan Resendes, Signy Sheldon, James Qi, and Bob Quelch. I would also like to thank my examination committee members, Drs. Larry Humes, Kathy Pichora-Fuller, Bruce Schneider, Eyal Reingold, and Claude Alain. Thank you for the opportunity to cross swords, your generosity of time, and for the many helpful comments. Of course, this research would not have been possible without the contributions from the participants. Thank you for your time and for your many insightful observations. Finally, it would have been next to impossible to complete this work without Jessica s encouragement, help, guidance, support, and love. iv

5 Table of Contents Acknowledgments...iv Table of Contents...v List of Tables...x List of Figures...xi List of Appendices...xiii Chapter 1 Introduction Purpose Objectives Significance Document overview... 4 Chapter 2 The listening difficulties of older adults The information-processing perspective Methodological issues Presbycusis Age-related changes in audition Audibility Temporal processing Aging and cognition Speed of processing Working memory Inhibitory deficits Attention Selective attention v

6 2.6.2 Divided attention Auditory spatial attention Summary of literature review Current research Experiment I Experiment II Experiment III Concluding remark Chapter 3 The Effect of Age on Auditory Spatial Attention in Conditions of Real and Simulated Spatial Separation Introduction Spatial separation Real and simulated spatial separation Aging Current study: Dissertation study I Methods Participants Stimuli Equipment Design Procedures Results Age Location certainty Callsign cue Real vs. simulated spatial separation vi

7 3.2.5 Spatial listening expectations The role of natural spatial cues Discussion Performance of younger listeners The effect of age on performance Auditory perception of spatial separation Selective auditory spatial attention Interaction of auditory and attentional factors Chapter 4 The Time Course of Auditory Spatial Attention Introduction Methods Participants Stimuli Equipment Procedures Design Results Overall results Spatial listening expectations: Overall results Spatial listening expectations: Effect of pause duration Spatial listening expectations: Effect of filled and unfilled gaps Discussion The time course of auditory spatial attention The influence of age on the time course of attention Cognitive and acoustic mechanisms underlying auditory spatial attention...63 vii

8 4.3.4 Findings compared to the literature Chapter 5 The Effect of Single vs. Multiple Switches of Spatial Attention Introduction Current research Methods Participants Stimuli Equipment Procedure Design Results Overall results Spatial listening expectations Error analysis Discussion Alternate explanations of the age x switch complexity interaction Chapter 6 Aggregate Data Analysis Results Overall data Spatial listening expectations Individual data Discussion Effect of target location uncertainty Effect of spatial listening expectations Between-experiment performance viii

9 6.2.4 Individual data Alternative explanations Chapter 7 Conclusions Experiment I Experiment II Experiment III Summary of contributions Future directions References Appendices Data from Experiments I, II, and III ix

10 List of Tables Table 1. Individual and group left- (L) and right-ear (R) audiometric thresholds (in db HL) for the eight younger (Y) adult participants in Experiment I...27 Table 2. Individual and group left- (L) and right-ear (R) audiometric thresholds (in db HL) for the eight older (O) adult participants in Experiment I...28 Table 3. Individual and group left- (L) and right-ear (R) audiometric thresholds (in db HL) for the eight younger (Y) adult participants from Experiment II...48 Table 4. Individual and group left- (L) and right-ear (R) audiometric thresholds (in db HL) for the eight older (O) adult participants from Experiment II...49 Table 5. Mean percent correct word-identification scores and standard errors of the mean (SEM) for RSS and SSS presentation conditions, presented for the four location certainty (1.0, 0.8, 0.6, and 0.33) and four sentence (original and 0, 150, and 300 ms) conditions for younger and older listeners in Experiment II...55 Table 6. Mean percent correct word-identification scores and standard errors of the mean (SEM) for RSS and SSS presentation conditions, presented for the four sentence (original and 0, 150, and 300 ms) conditions at the likely and unlikely listening location for younger and older adults in Experiment II. Data are collapsed across location certainties 0.8 and Table 7. Individual and group left- (L) and right-ear (R) audiometric thresholds (in db HL) for the eight younger (Y) adult participants from Experiment III...68 Table 8. Individual and group left- (L) and right-ear (R) audiometric thresholds (in db HL) for the eight older (O) adult participants from Experiment III...68 x

11 List of Figures Figure 1. Schematic of loudspeaker configuration for real (left) and simulated spatial separation (right). The circle indicates the position of the participant and the squares indicate the position of the loudspeakers...31 Figure 2. Mean percent correct word-identification scores and standard errors of the mean for younger (top, unfilled symbols) and older (bottom, filled symbols) adults for the four location certainties. Scores on the y-axis were calculated as the percentage of trials where participants correctly identified both the color and number associated with the target callsign. Solid lines indicate RSS; dashed lines indicate SSS; circles indicate callsign cue before conditions; and triangles indicate callsign cue after conditions...34 Figure 3 Mean percent correct word-identification scores and standard errors of the mean for RSS (solid lines) and SSS (dashed lines) presentation conditions, depicted for the likely and unlikely spatial locations. Unfilled and filled circles represent data collected on younger and older adults respectively. Triangles represent the results of Kidd et al. (2005) for younger adults when performance was averaged over conditions in which the likely location was left, center, or right. Data are collapsed across location probabilities 0.8 and Figure 4. Mean percent correct word-identification scores and standard errors of the means from the younger adults listening with RSS in the current study (solid lines) and in the study of Kidd et al. (2005; dashed lines) for the four location certainties. Circles indicate callsign cue before conditions, and triangles indicate callsign cue after conditions...38 Figure 5. Mean percent correct word-identification scores for younger (left panel) and older (right panel) adults depicted across four location certainty conditions. The data presented are from Singh et al. (2008) (white bars), the current study (black bars), and from Kidd et al. (2005) (grey bars), with error bars representing the standard errors of the mean...53 Figure 6. Mean percent correct word-identification difference scores for RSS presentation conditions, depicted for the likely and unlikely listening locations for younger and older adults. Differences in word-identification scores between four pairs of sentence conditions, 150 and 0 ms, 300 and 150 ms, 300 and 0 ms, and original and 300-ms, are shown in white, grey, striped, and black bars, respectively...59 Figure 7. Mean percent correct word-identification scores and standard errors of the mean for younger (open circles) and older (filled circles) adults for the four location certainty conditions. Solid lines indicate conditions requiring simple switches and dashed lines indicate conditions requiring complex switches...73 Figure 8. Mean percent correct word-identification scores and standard errors of the mean for younger and older adults in conditions requiring simple and complex switches for targets presented from the likely (white bars) and unlikely (black bars) listening location...75 Figure 9. Cumulative proportion of color and number errors by source location for younger and older adults in conditions requiring simple (top) and complex (bottom) switches. xi

12 White bars indicate the incorrect responses were reports of distractors presented from the center loudspeaker, grey bars indicate the incorrect responses were reports of distractors presented from the incorrect side loudspeaker, and black bars indicate the incorrect responses were reports of tokens that were not presented Figure 10. Mean percent correct word-identification scores and standard errors of the mean for younger and older adults in conditions requiring simple and complex switches of attention for targets presented from the unlikely listening location. Data are presented by testing session...81 Figure 11. Mean percent correct word-identification scores for younger (Y) and older (O) adults in Experiments (Expt) I, II, and III. Data are presented as a function of target location certainty...85 Figure 12. Mean percent correct word-identification scores for younger (top) and older (bottom) adults in Experiments (Expt) I, II, and III. Data are presented as a function of target location certainty...86 Figure 13. Mean percent correct word-identification scores for Experiments I, II, and III. Data are presented as a function of target location certainty and collapsed across younger and older adults...87 Figure 14. Mean percent correct word-identification scores for younger and older adults for Experiments I, II, and III. Data are presented as a function of spatial listening expectation Figure 15. Mean percent correct word-identification scores and standard errors of the mean for targets presented from the likely (white bars) and unlikely (black bars) listening location in Experiments I, II, and III. Data are collapsed across younger and older adults and across location certainties 0.8 and Figure 16. Mean percent correct identification score for each younger participant for targets presented from the likely and unlikely listening location for Experiments I, II, and III...90 Figure 17. Mean percent correct identification score for each older participant for targets presented from the likely and unlikely listening location for Experiments I, II, and III...91 xii

13 List of Appendices Appendix A. Experiment I mean individual and group mean word-identification scores and standard deviations (SD) for younger (Y) and older (O) adults for the four target location certainty conditions (1.0, 0.8, 0.6, and 0.33). The data are categorized by the RSS and SSS presentation conditions when the callsign cue was provided before or after stimulus presentation Appendix B. Experiment I mean individual and group mean word-identification scores and standard deviations (SD) for younger and older adults for the likely and unlikely spatial positions. The data are categorized for the RSS and SSS presentation conditions Appendix C. Experiment II mean individual and group mean word-identification scores and standard deviations (SD) for younger (Y) and older (O) adults for the four target location certainty conditions (1.0, 0.8, 0.6, and 0.33). Sentence condition = 0 ms. The data are categorized by the RSS and SSS presentation conditions Appendix D. Experiment II mean individual and group mean word-identification scores and standard deviations (SD) for younger (Y) and older (O) adults for the four target location certainty conditions (1.0, 0.8, 0.6, and 0.33). Sentence condition = 150 ms. The data are categorized by the RSS and SSS presentation conditions Appendix E. Experiment II mean individual and group mean word-identification scores and standard deviations (SD) for younger (Y) and older (O) adults for the four target location certainty conditions (1.0, 0.8, 0.6, and 0.33). Sentence condition = 300 ms. The data are categorized by the RSS and SSS presentation conditions Appendix F. Experiment II mean individual and group mean word-identification scores and standard deviations (SD) for younger (Y) and older (O) adults for the four target location certainty conditions (1.0, 0.8, 0.6, and 0.33). Sentence condition = original. The data are categorized by the RSS and SSS presentation conditions Appendix G. Experiment II mean individual and group mean word-identification scores and standard deviations (SD) for younger (Y) and older (O) adults for the likely and unlikely spatial position. Sentence condition = 0 ms. The data are categorized by the RSS and SSS presentation conditions Appendix H. Experiment II mean individual and group mean word-identification scores and standard deviations (SD) for younger (Y) and older (O) adults for the likely and unlikely spatial position. Sentence condition = 150 ms. The data are categorized by the RSS and SSS presentation conditions Appendix I. Experiment II mean individual and group mean word-identification scores and standard deviations (SD) for younger (Y) and older (O) adults for the likely and unlikely spatial position. Sentence condition = 300 ms. The data are categorized by the RSS and SSS presentation conditions xiii

14 Appendix J. Experiment II mean individual and group mean word-identification scores and standard deviations (SD) for younger (Y) and older (O) adults for the likely and unlikely spatial position. Sentence condition = original. The data are categorized by the RSS and SSS presentation conditions Appendix K. Experiment III mean individual and group mean word-identification scores and standard deviations (SD) for younger (Y) and older (O) adults for the four target location certainty conditions (1.0, 0.8, 0.6, and 0.33). The data are categorized for the simple and complex conditions Appendix L. Experiment III mean individual and group mean word-identification scores and standard deviations (SD) for younger (Y) and older (O) adults for the likely and unlikely spatial positions. The data are categorized for the simple and complex conditions xiv

15 1 Chapter 1 Introduction Forgive me when you see me draw back when I would have gladly mingled with you. My misfortune is doubly painful to me because I am bound to be misunderstood; for me there can be no relaxation with my fellow men, no refined conversations, no mutual exchange of ideas. I must live almost alone, like one who has been banished; I can mix with society only as much as true necessity demands. If I approach near to people a hot terror seizes upon me, and I fear being exposed to the danger that my condition might be noticed. 1.1 Purpose Ludwig van Beethoven, October 6, 1802, The Heiligenstadt Testament Historically, most research attempting to understand the mechanisms underpinning hearing difficulties has attempted to identify anatomical sites-of-lesion, with greater emphasis placed on the role of the peripheral auditory system. This approach has greatly advanced our understanding of hearing for it has yielded methods of assessment that have diagnostic utility, as well as fostered rehabilitation approaches that are clinically meaningful. Although earlier research in hearing was rooted in creating a better understanding of the ear it was recognized that factors beyond the peripheral sense organ itself had a non-trivial role in the understanding of spoken language (e.g., Cherry, 1953; Davis & Silverman, 1960). These factors were broadly categorized under the banner of psychology and included terms such as memory, perception, motivation, and attention. Research has explored the critical role of the auditory periphery in hearing; however, throughout the 1980s and 1990s, important contributions to the body of knowledge on auditory perception developed from embracing research perspectives that also investigated hearing phenomena in which regions beyond the cochlea play a critical role. Two examples are the pioneering investigations exploring central auditory processing (for a review see ASHA, 1993) and those exploring how higher cognitive centers contribute to the processing of incoming auditory signals as separate perceptual representations (i.e., auditory scene analysis; Bregman, 1990). The current research builds on this tradition and endeavors to move toward more realistic research designs and methods that could enable a better understanding of how both cognitive and auditory mechanisms contribute to listening abilities in everyday challenging

16 2 environments such as listening situations where the location of a target talker in a multi-talker environment is uncertain (for a review see Arlinger, Lunner, Lyxell, & Pichora-Fuller, 2009). The current research is motivated by the fact that many older adults, even those with clinically normal audiometric thresholds, often report difficulty understanding speech in background noise (e.g., CHABA, 1988; Dubno, Dirks, & Morgan, 1984; Humes & Dubno, 2010; Pichora-Fuller, Schneider, & Daneman, 1995; Schneider, Pichora-Fuller, & Daneman, 2010; Willott, 1991). The quotation noted at the beginning of this chapter provides insight into the effects of hearing-impairment on the experiences of the listener. For the listener experiencing difficulty understanding spoken language, there are reports of reduced feelings of well-being, increased fatigue and anxiety, a reduction in social contact resulting in social isolation, increased rates of depression, and a reduction in reported quality of life (Bess, Lichtenstein, Logan, Burger, & Nelson, 1989; Erdman, Crowley, & Gillespie, 1984; Hickson, Worrall, Bennett, & Yiu, 1995; Scherer & Frisina, 1998; Seniors Research Group, 1999; Weinstein & Ventry, 1982). In general, hearing difficulty in older adults has been attributed to changes in the cochlea (e.g., outer hair cell damage or changes in endocochlear potentials) and/or losses of neural synchrony (Mills, Schmeidt, Schulte, & Dubno, 2006; Schuknecht, 1964), or a genetic predisposition (Ohlemiller, 2004; Friedman et al., 2008). However, given that auditory perception arises from a complex interaction of cognitive and auditory processing, and that adult age-related differences have been documented on a wide variety of cognitive variables (for reviews see Craik & Salthouse, 1992; 2000), it stands to reason that part of the difficulties in spoken language comprehension reported by older adults may also be due to age-related changes in cognitive processing. Hence, the current research explores the contribution of cognitive factors to the difficulties older listeners report regarding understanding language spoken in background noise. 1.2 Objectives Behavioural experiments in which participants heard three sentences spoken simultaneously with varying degrees of certainty about the location of a target sentence were used to address the following research objectives:

17 3 1. To test the hypothesis that there are age-related differences in word-identification performance on a task involving auditory spatial attention. 2. To test the hypothesis that time influences word-identification performance on a task involving auditory spatial attention, and to explore if there are possible age-related differences depending on the availability of time. 3. To test the hypothesis that multiple, compared with single, switches of auditory spatial attention are more detrimental to word-identification performance, and to determine if age moderates the possible effect of conducting multiple switches of auditory spatial attention. 1.3 Significance By better understanding how cognitive factors interact with auditory factors, it becomes possible to develop and refine rehabilitation strategies (e.g., hearing aid processing algorithms or behavioural training programs) that are better able to address a person s reduced listening abilities. This is important for several reasons. First, even in the absence of elevated audiometric thresholds, older adults experience at least some difficulty understanding language spoken when there is competing speech. Because these difficulties cannot be solved by amplification, more appropriate audiological rehabilitation interventions based on the listening needs and abilities of the older adult population need to be developed. Second, many people who are hard-of-hearing may potentially benefit from research exploring the link between audition, cognition, and aging. Although the prevalence of hearing loss is approximately 8-10% in the general population, it is closer to 30% in adults over the age of 65 (Cossette & Duclos, 2002; Kochkin, 2005). The prevalence further increases to 35-50% when taking into consideration difficulties in understanding language spoken in complex listening situations, such as when communicating in the presence of background noise or in environments containing high levels of reverberation (CHABA, 1988; Cruickshanks et al., 1998). Third, there are currently a limited number of available treatment options to counter the effects of hearing impairment in complex listening situations. Often these options include the provision of one or two hearing aids, and although hearing aid users can experience considerable benefit in listening environments containing relatively little background noise, less success is reported in environments containing high levels of background noise (Kochkin, 2010; Plomp, 1978). Finally, hearing aid manufacturers have

18 4 already started to incorporate factors related to cognition into the design of commercially available hearing aid processing algorithms. These algorithms have only started to reflect cognitive phenomena previously thought to be beyond the signal-processing capabilities of hearing aids; however, evidence suggests this trend for research on cognition to influence the engineering of communication devices such as hearing aids is likely to continue (Edwards, 2007; Lunner, Rönnberg, & Rudner, 2009; Pichora-Fuller & Singh, 2006) and would benefit from research linking auditory and cognitive processing. 1.4 Document overview Chapter 2 will include a review of the extant research literature relevant to the goals of the current research. This review explores the general problem area (i.e., the word-identification difficulties often reported by older listeners in challenging listening environments) and reviews auditory mechanisms (i.e., audibility, binaural processing, and temporal auditory processing) as well as cognitive mechanisms (i.e., attention, working memory, speed of processing, inhibitory processes, and auditory spatial attention) implicated in the poorer listening performance exhibited by older adults compared with younger adults. Chapter 2 concludes with a discussion of the need to research age-related hearing difficulties in testing environments that replicate features of everyday listening environments which negatively affect speech intelligibility performance. In particular, attention is drawn to the need to better understand how uncertainty about the location of a target talker in the presence of competitor speech influences wordidentification performance. Chapter 3 will describe the rationale, methods, results and discussion of a study (Experiment I) that investigates the capabilities of younger and older adults to allocate auditory spatial attention. The goal of Experiment I was to investigate how auditory and cognitive factors contribute to possible age-dependent differences in the ability to allocate auditory spatial attention. In this experiment, listeners were presented with three concurrent sentences in a multitalker, multi-spatial, listening environment in conditions where the spatial location of each of the sentences was either simulated or not. The sentences have the format Ready [callsign], go to [color] [number] now. In conditions of real spatial separation (RSS), the target sentence was typically presented from a central location and competing sentences were most often presented from left and right locations. In conditions of simulated spatial separation (SSS), different

19 5 apparent spatial locations of the target and competitors were induced using the precedence effect. The identity of the target was cued by a callsign presented to the listener either prior to or following each target sentence. The probability that the target would be presented at the three locations was specified at the beginning of each block. Younger and older adults with normal hearing sensitivity below 4 khz completed all 16 conditions (2 spatial separation method x 2 callsign conditions x 4 probability conditions). Overall, younger adults performed better than older adults. For both age groups, performance improved with target location certainty, with a priori cueing of the target callsign, and when differences in the location of the target and masker sentences were induced by real rather than simulated spatial separation. For both age groups, the contributions of natural spatial cues were most pronounced when the target occurred at the unlikely spatial listening locations. The results from Experiment I suggest that both age groups benefit similarly from richer acoustical cues and a priori information in a listening environment where a listener was uncertain about the location of the target. The failure to observe age-related differences in word-identification performance for targets presented from unlikely compared with likely listening locations led to the development of a second experiment which is described in Chapter 4. One limitation of Experiment I was that word-identification performance for targets presented at the unlikely listening location was tested in a fashion that did not systematically manipulate the time course of attention switching. Thus, the relatively long time delay between the cue and target may have been sufficient to accommodate the possibly slower attention switching capacities of older compared with younger adults. To address this limitation, the hypothesis that the availability of time may contribute to word-identification performance was investigated in Experiment II. Specifically, the time course of auditory spatial attention switching was investigated in younger and older adult listeners with normal hearing sensitivity below 4 khz. Word-identification performance was measured for target sentences presented with two spatially separated competitor sentences. A priori cues specified the location and identity callsign of the target. Four timing conditions were tested: the original sentences (which contained about 300 ms of filler speech between the callsign cue and the onset of the key words), or modified sentences with silent pauses of 0, 150, or 300 ms replacing the filler speech. For targets presented from the likely location, word-identification performance was better for the original than for the modified sentences. For targets presented from unlikely locations, word-identification performance was better when there was more time

20 6 between the callsign cue and key words. All listeners benefited similarly from the availability of more compared with less time and the presentation of continuous compared with interrupted speech. These results suggest that stream continuity facilitates the processes underlying the allocation and maintenance of auditory spatial attention when a target sentence is presented from an expected location, but that time is needed for the reallocation of auditory spatial attention when a target sentence is presented from an unexpected location. The failure to observe an age x target location interaction in Experiment II, even when controlling for the time available to switch attention, led to the development of a third experiment designed to explore the influence of relatively simple and relatively complex types of switching of auditory spatial attention, with the rationale being that perhaps age-related differences in word-identification performance are more readily observed in more challenging listening situations where multiple switches of attention are required. Chapter 5 describes Experiment III, which explored the effects of single and multiple switches of auditory attention in a multi-talker, multi-spatial listening situation in a group of younger and older listeners with normal audiometric thresholds below 4 khz. In all conditions, a target sentence was presented from one spatial location and competing sentences were presented from two different locations. Cues were provided pre-trial about the identity of the target callsign and the probability of the target callsign being presented at each of the three locations. There were four different probability conditions that varied in the likelihood of the target being presented at the left, center, and right locations ( , , , ). When the callsign cue was presented from an unlikely listening location, for half the trials the listener s task was to report key words from the sentence containing the callsign cue; for the remaining trials, the listener s task was to report key words from the sentence presented from the opposite unlikely listening location. Sixteen adults (8 younger and 8 older) participated. The key finding was that, whereas both age groups performed similarly in conditions requiring a single switch of attention, the performance of older but not younger listeners was reduced when multiple switches of spatial attention were required. The findings from Chapter 5 suggest that difficulties disengaging attention may explain, at least in part, why older adults with good audiograms in the speech range report difficulty communicating in multi-talker listening situations. Chapter 6 describes an analysis of the cumulative data from Experiments I, II, and III that were collected in similar baseline conditions. In this chapter, the overall patterns across the series

21 7 of experiments are considered, as well as similarities and differences in performance between the three experiments and across the younger and older adult groups in the baseline conditions. The baseline conditions were common to all three experiments and consisted of stimuli being presented using RSS and when the probability of the target being presented from the center location was 1.0, 0.8, 0.6, and Overall, a small group difference was observed, whereby younger adults outperformed the older adults. Furthermore, age-related differences were greatest in listening conditions with the most uncertainty regarding the location of the target. When the data were categorized by spatial listening expectations, similar to the pattern observed in each of the experiments, age-related differences in word-identification performance were not observed when targets were presented from an unlikely compared with the likely listening location. This finding suggests that in the baseline conditions both groups of listeners benefit similarly from spatial listening expectations, demonstrate similarly preserved abilities to locate an auditory target presented from one of two possible unlikely listening locations, and to switch spatial attention to that target. Across the three experiments, the patterns of performance in the baseline conditions included on each of the studies were similar, although minor between-experiment differences were observed and are discussed. Chapter 6 concludes with a consideration of the data observed for individual participants. The document will conclude with chapter 7 which will provide a summary discussion of the findings and contributions of the current research, as well as a discussion of future research directions.

22 8 Chapter 2 The listening difficulties of older adults The listening difficulties of older adults in background noise can be explored from a number of theories and perspectives. The following sections of Chapter 2 will provide a theoretical framework through which the topic may be considered (Section 1), a discussion of one of the methodological issues relevant when investigating the influence of age on hearing abilities (Section 2), background information on the topic of age-related hearing impairment (Section 3), and a description of different auditory (Section 4) and cognitive processes (Section 5) that contribute to the ability to listen in the presence of background noise. Based on this literature review, current gaps in our understanding are identified (Section 6), and a summary of three experiments designed to address these gaps is provided (Section 7). 2.1 The information-processing perspective From an information-processing perspective, auditory perception arises from a combination of peripheral and central auditory processes, involving not only the afferent and efferent auditory pathways, but also interactions with other sensory modalities (e.g., Sumby & Pollack, 1954) and cognitive functions (Pichora-Fuller & Singh, 2006; Schneider & Pichora- Fuller, 2000). One assumption of the information-processing perspective is that a listener has a limited pool of cognitive resources available to him/her (Kahneman, 1973; Sternberg, 1969) and that as listening becomes effortful more resources are consumed thereby leaving fewer resources available for other cognitive and perceptual operations (e.g., Pichora-Fuller, 1997). The following sections present a review of the background literature on topics in audition and cognition that influence auditory perception and which may possibly shed light on why older adults with audiometric thresholds in the normal range experience difficulty understanding speech in complex listening environments. 2.2 Methodological issues Before continuing, it is important to describe one of the methodological issues present when investigating age-related difficulties in spoken language comprehension. One of the complicating factors in conducting research on aging and auditory perception is distinguishing between the effects of age and those resulting from a loss of audibility of the sound signal due to

23 9 changes in hearing sensitivity. This distinction is important because rehabilitation and treatment approaches may differ considerably depending whether spoken language comprehension difficulties arise as a result of elevated audiometric thresholds or as a result of other age-related changes in processing of the supra-threshold speech signal. In general, four approaches are used to distinguish between the effects of age and a loss of audibility. The first approach is to match younger and older participant groups with respect to their audiometric thresholds. Typically, this is accomplished by recruiting younger adults with elevated audiometric thresholds or recruiting older adults with relatively good audiometric thresholds compared to their peers (e.g., Humes, 1996). A second approach is to use simulations, such as by presenting a masking noise to approximate the typical high-frequency hearing loss observed in older adults (e.g., Humes & Christopherson, 1991). In contrast to the matching or simulation approaches used in group designs, a third approach is correlational. This method examines individual differences by examining correlations between age, auditory measures and word-identification performance (e.g., Helfer & Wilber, 1990). A fourth approach involves investigating participants who range in both the degree of audiometric threshold elevation and age, and utilizing regression analyses to separate out effects of age from hearing loss (e.g., Souza & Boike, 2006). The relative merits of each approach differ, but each approach is considered a valid means of distinguishing between the effects of age and audibility (Pichora-Fuller & Souza, 2003). 2.3 Presbycusis Presbycusis, or hearing-impairment associated with various types of auditory system dysfunction, peripheral or central, that accompany aging and cannot be accounted for by extraordinary ototraumatic, genetic, or pathological conditions (Willott, 1999, p. 136), is the umbrella term used to describe hearing loss associated with aging. For older adults, presbycusis is one of the most prevalent chronic conditions (Cruickshanks et al., 1998). Presbycusis is typically characterized by elevated, bilateral, high-frequency audiometric thresholds and impaired auditory temporal processing. Functionally, these deficits manifest as difficulty in understanding spoken language, particularly if the speech signal is degraded (e.g., by filtering or time compression) or presented in complex listening situations (e.g., in background noise or reverberant environments), poorer sound localization abilities, and diminished central auditory processing capabilities (for reviews see Divenyi & Simon, 1999; Gates & Mills, 2005; Pichora- Fuller & Souza, 2003). In addition to those with elevated audiometric thresholds, older adults

24 10 with clinically normal audiograms in the speech range often report difficulty understanding speech in noisy, reverberant, or multi-talker environments (e.g., CHABA, 1988; Dubno et al., 1984; Duquesnoy, 1983; Frisina & Frisina, 1997; Gordon-Salant & Fitzgibbons, 1995; Helfer, 1992; Pichora-Fuller et al., 1995; Tun & Wingfield, 1999). Hearing difficulties in everyday listening situations associated with auditory aging, regardless of the degree of pure-tone threshold elevation, arise from an interaction of age-related differences in auditory and cognitive processing (for reviews see Pichora- Fuller, 1997; 2003; Schneider, Daneman, & Pichora-Fuller, 2002). In typical everyday multi-talker environments, listeners rely on auditory processing to localize sources originating from different locations and on cognitive processing to attend to the target(s) of interest and to ignore distracting sounds from other sources (Bregman, 1990; Greenberg & Larkin, 1968; Moore, 1989; Yost, 2006). A number of auditory mechanisms have been implicated to account for the poorer performance of older compared with younger adults. These mechanisms include differences in hearing sensitivity, binaural processing, and temporal processing (for a review see Moore, 1996). In addition to auditory factors, disproportionate problems may be experienced by older listeners when they must understand language spoken in multi-talker situations because of age-related changes in cognitive processing (for reviews see Pichora-Fuller, 2003; Pichora-Fuller & Singh, 2006; Schneider & Pichora-Fuller, 2000; Wingfield, 1996; Wingfield, Tun, McCoy, Stewart, & Cox, 2006). These age-related cognitive differences have been found for processes related to executive function, working memory capacity, speed of processing, inhibition, and attention. Some of the task-relevant attentional factors important for spatial listening include selective attention (i.e., attending to a single talker), inhibition (i.e., ignoring irrelevant talkers), divided attention (i.e., attending to multiple streams simultaneously), and auditory spatial attention (i.e., focusing listening resources along a spatial dimension). 2.4 Age-related changes in audition Audibility Evidence of the role of audibility in the listening difficulties experienced by older adults comes from studies of younger and older listeners with carefully matched hearing sensitivity (e.g., Humes, 1996; Humes & Roberts, 1990; Souza & Turner, 1994). Typically, these studies find that in listening conditions containing little background noise, performance between

25 11 younger and older listeners is quite similar, suggesting the importance of audibility in relatively quiet and/or simple listening environments. Various approaches for estimating speech intelligibility have been developed (e.g., the Articulation Index; ANSI, 1986 or the Speech Intelligibility Index; ANSI, 1997). In general, these approaches are based on calculating the amount of information that is audible to a listener. This is achieved by calculating the signal-tonoise ratio (SNR) in frequency bands that are summed based on a weighting that reflects each frequency band s relative contribution to word-identification performance. When these indices are used to model the performance for older listeners, they are generally quite successful in predicting speech intelligibility in quiet and/or simple listening environments (e.g., Dubno et al., 1984; Schum, Matthews, & Lee, 1991), thus providing further evidence that audibility is an important factor underpinning age-related deficits in speech perception. Unlike the strong relationship between audibility and word-identification performance in a quiet environment, loss of audibility in and of itself is a poor predictor of one s ability to understand language spoken in more complex listening environments such as those containing time-varying background noise or reverberation (Dubno et al., 1984; Duquesnoy, 1983; Frisina & Frisina, 1997; Gordon-Salant & Fitzgibbons; Nabelek & Robinson, 1982; 1993; Pichora- Fuller et al., 1995; Plomp, 1978; 1986). In addition to audibility, another important factor underpinning age-related deficits in speech perception in complex listening environments is the ability to process auditory temporal information (Bergman, 1971; Fitzgibbons & Gordon-Salant, 1996; Frisina & Frisina, 1997; Pichora-Fuller, 2003) Temporal processing Speech, by its very nature, unfolds over time and there is compelling evidence that temporal processing of the physical features of an auditory stimulus is important in auditory perception. Age-related differences in temporal processing have been observed at levels ranging from the small variations of the waveform within a single period of a periodic sound (i.e., fine structure cues) to longer durations in the overall amplitude (i.e., envelope level cues). Specifically, age-related differences favoring younger listeners have been observed on tasks of gap-detection (Divenyi & Haupt, 1997; Schneider & Hamstra, 1999; Schneider, Pichora-Fuller, Kowalchuk, & Lamb, 1994), auditory duration discrimination (Abel, Krever, & Alberti, 1990; Fitzgibbons & Gordon-Salant, 1993; Ostroff, McDonald, Schneider, & Alain, 2003), amplitude

26 12 modulation detection (He, Mills, Ahlstrom, & Dubno, 2008), word-identification of vocodedspeech manipulating envelope cues (Sheldon, Pichora-Fuller, & Schneider, 2008), temporal integration of pure-tone stimuli (Gleich, Kittel, Klump, & Strutz, 2007), and temporal order recognition (Gordon-Salant & Fitzgibbons, 1997; Szymaszek, Szelag, & Sliwowska, 2006). Consequently, in addition to the influence of audibility, there is evidence to suggest that agerelated deficits in speech perception may be attributed to impaired temporal processing. It should be noted that age-related changes in monaural auditory processing also have implications for binaural and spatial hearing and this will be discussed in Chapter Aging and cognition Age-related differences in cognitive processing may also contribute to the difficulties that older adults have in language spoken in challenging listening situations (for reviews see Pichora- Fuller, 2003; Pichora-Fuller & Singh, 2006; Schneider & Pichora-Fuller, 2000; Wingfield, 1996; Wingfield et al., 2006). Age-related slowing in information processing, reduced working memory capacity, inhibitory and other attentional deficits are some of the more likely cognitive mechanisms contributing to the difficulties of older listeners (for a review see Craik & Salthouse, 1992; 2000). Given that research exploring the link between cognition, audition, and aging must take into the account the possible influence of these alternative accounts of cognitive aging, a brief review of each of these topics will be provided. 2.3 Speed of processing Speed of processing refers to the rate at which individuals execute cognitive processes, and it represents one of the factors underlying fluid intelligence (Cattell, 1943). The ability to rapidly perform cognitive tasks is often used to benchmark cognitive efficiency and/or information processing capacity. Speed of processing is often invoked to explain impairments in cognitive functioning associated with aging and this topic has exerted considerable influence on research investigating cognitive aging phenomena (Salthouse, 1995; 1996). Evidence for the role of information processing speed in spoken language understanding comes from studies where age-related deficits in word identification are observed when the rate of speech is artificially increased (e.g., Tun & Wingfield, 1999; Vaughan & Letowski, 1997; Wingfield, Poon, Lombardi, & Lowe, 1985) or from studies observing correlations between performance on auditory tasks and measures of speed of processing (e.g., Tun & Wingfield, 1999; van Rooij,

27 13 Plomp, & Orlebeke, 1989; Wingfield et al., 1985). Finally, significant correlations between measures of processing speed and speech identification in noise, both with and without hearing instruments, have been reported (Hällgren, Larsby, Lyxell, & Arlinger, 2005; Lunner, 2003). 2.4 Working memory Working memory refers to the system that is involved in temporarily storing and manipulating information while carrying out cognitive tasks including those involving information processing, reasoning, learning, and comprehension (Baddeley & Hitch, 1974; Baddeley, 2000; Cowan, 1995). One of the defining features of the memory system is that it is capacity limited (Miller, 1956) and that its capacity serves as an index of overall processing resources (Cowan, 2001). Research investigating aging and working memory typically shows that when compared to younger adults, older adults exhibit a reduction in working memory capacity (e.g., Dobbs & Rule, 1989; Salthouse, 1994). This link between audition, cognition, and working memory capacity is highlighted by studies demonstrating that when auditory processing becomes more challenging, listeners attempt to maintain speech intelligibility by allocating working memory resources that draw upon prior knowledge (Murphy, Craik, Li, & Schneider, 2000; Pichora-Fuller, 2007; Pichora-Fuller et al., 1995; Rabbitt, 1968). 2.5 Inhibitory deficits The inhibitory deficit hypothesis proposed by Hasher and Zacks (1988; see also Hasher, Zacks, & May, 1999) suggests that declines in cognitive processing associated with age are, in part, observed because of failures to inhibit goal-irrelevant information such as when interfering stimuli should be ignored (see also Clapp & Gazzaley, 2010). Although evidence supporting this theory has been observed in cognitive domains such as memory, few studies have explored agerelated inhibitory processes in audition. In general, these studies have produced inconclusive findings. On the one hand, Frisina and Walton (2006) found support for the idea that in older compared with younger animals, central auditory neurons exhibit inhibitory neurotransmitter (i.e., glycine) deficits, and that these declines (along with synaptic and temporal processing changes) likely represent candidates for perceptual changes that develop with age such as declines in speech perception in background noise (see also Willott, 1991). Perceptual evidence supporting the inhibitory deficit hypothesis in audition has remained somewhat elusive, although a few studies have presented supporting evidence in studies of auditory speech communication

28 14 (e.g., Sommers & Danielson, 1999; Tun, O Kane, & Wingfield, 2002). For example, Murphy, McDowd, and Wilcox (1999) tested whether older adults demonstrated greater difficulty inhibiting distractor stimuli in a speech-on-speech listening situation and failed to observe an age-specific inability to ignore irrelevant information (see also Li, Daneman, Qi, & Schneider, 2004). Thus, it appears that while there is good evidence implicating the role of inhibition in cognitive aging, there is minimal support for the role of inhibitory deficits in auditory aging or during spoken language processing. 2.6 Attention Another possibility is that age-related speech intelligibility deficits arise from difficulties in managing attentional processes (for reviews see Kramer & Madden, 2008; McDowd & Craik, 1988). Given that the current series of experiments focuses on the role of attention in auditory processing for younger and older adults, the following sections will briefly review the influence of selective, divided, and auditory spatial attention on auditory perception Selective attention As described by Cherry (1953), auditory selective attention refers to one s ability to allocate attention to a subset of co-occurring sounds. Competing accounts of auditory selective attention (feature-based or object-based) describe the role of attention in auditory perception. Feature-based accounts suggest that attention serves to bind together auditory features, like frequency and intensity, in order to extract a meaningful and conscious representation (Treisman & Gelade, 1980). In contrast, object-based accounts suggest that auditory selective attention operates on representations (i.e., auditory objects) that are the output of a pre-attentive stage of analysis wherein auditory features are formed and grouped according to gestalt principles (Bregman, 1990; Woods & Colburn, 1992). In a review of meta-analyses, Verhaeghen and Cerella (2002) failed to observe age-related deficits on tasks of selective attention. Hence, it would appear that when listening to speech in the presence of competing speech, the difficulties that older adults report arise from difficulties unrelated to selective attention Divided attention In situations where more than one person speaks at the same time, listeners are often called upon to attend to more than one stream of auditory information. When a listener must

29 15 simultaneously follow multiple streams of information or respond to multiple task demands in a concurrent fashion, it is necessary to divide attention. Cognitive aging research has found that age-related differences are observed when multiple-tasks are performed concurrently (e.g., McDowd & Shaw, 2000; Tun, McCoy, & Wingfield, 2009). For example, in a typical auditory attention study, auditory streams are presented dichotically, and participants are asked to follow one or both auditory streams. The typical finding is that age-related deficits in speech understanding favoring younger listeners are observed under conditions requiring divided but not selective attention (e.g., Hällgren, Larsby, Lyxell, & Arlinger, 2001; Humes, Lee, & Coughlin 2006; Jerger, Chmiel, Allen, & Wilson, 1994). Age-related differences are also found when a switch in task is executed under conditions requiring the coordination of multiple task demands (i.e., global task-switching) (e.g., Kray, Li, & Lindenberger, 2000; Verhaeghen & Cerella, 2002). Multiple factors can influence dual-task performance, including task similarity (Treisman & Davies, 1973), task difficulty (e.g., McDowd & Craik, 1988), and task familiarity (McDowd, 1986), whereby dual-task performance improves when the component tasks are easier rather than harder (e.g., singing and juggling three balls, rather than singing and juggling five balls), if the component tasks are dissimilar (i.e., a dual-task where each task involves a different modality is easier than dual-tasks involving a single modality), and if the component tasks are more, rather than, less familiar (i.e., performance deficits are observed with novel tasks). Thus, unlike tasks that require the deployment of selective attention, age-related performance deficits are observed on tasks requiring a division of attentional resources, such as those concurrently requiring listeners to follow multiple streams of information or those requiring multiple task-demands Auditory spatial attention The study of visual attention has been highly influenced by research conducted using the cueing paradigm (Posner, Snyder, & Davidson, 1980) whereby an observer is asked to detect a signal at one of two (or more) locations, with a pre-trial cue indicating the most and/or least likely locations at which the signal will be presented. One common variant of the cueing paradigm is to provide observers with pre-trial cues indicating the probability of a target appearing at a particular location. For example, in a probabilistic spatial cueing study, a valid or likely trial occurs when a target is presented from a location with a high probability of

30 16 occurrence and an invalid or unlikely trial occurs when a target is presented from a location with a low probability of occurrence. By categorizing the data according to probabilistic spatial expectations, it is possible to assess the contribution of spatial expectations to performance. Despite the extensive use of the probability cueing paradigm in studies of visual attention, few studies have employed probability cueing in studies of auditory attention until recently (Hübner & Hafter, 1995). This is surprising for at least two reasons. First, a common complaint of many hard-of-hearing listeners is that they experience difficulty in hearing in situations where the location of a target is unknown. Second, a common hearing rehabilitation strategy that hearing healthcare practitioners provide to listeners seeking support for difficulty hearing in noise is to identify the location from where target sounds are being presented. Of those studies that have used the cueing paradigm, the effects of auditory attention on the accuracy of detection for targets presented in quiet backgrounds have typically been either nonexistent (Buchtel & Butter, 1988; Posner, 1978; Scharf, Quigley, Aoki, Peachey, & Reeves, 1987) or small (e.g., Spence & Driver, 1994), with moderately more robust effects of cue validity for response times on localization tasks (e.g., Mondor & Zatorre, 1995; Quinlan & Bailey, 1995; Spence & Driver, 1994). Interestingly, a different pattern of results is observed on measures of identification in cueing experiments where auditory targets are presented in listening environments containing sufficient energetic and/or informational masking (Greenberg & Larkin, 1968). Arbogast and Kidd (2000) found evidence for spatial tuning using tone sequences presented in a background of similar masker stimuli. Even larger effects have been found on measures of word identification in studies that use speech as the target and masker stimuli. For example, Ericson, Brungart, and Simpson (2004) measured word-identification performance under headphones for concurrently presented spoken sentences from the Coordinate Response Measure (CRM) corpus (Bolia, Nelson, Ericson, & Simpson, 2000) and found that a priori knowledge of the location, and of the callsign identifying the target sentence, yielded an 18 percentage point improvement in performance compared to when only the callsign cuing the identity of the target sentence was presented. These findings were extended in a study in which three CRM sentences were presented concurrently with each sentence being presented from one of three loudspeakers located at -54, 0, and +54 azimuth, with a priori cues specifying the location of the target

31 17 sentence which ranged from being completely certain to completely uncertain (Kidd, Arbogast, Mason, & Gallun, 2005). Importantly, presenting the target from a fixed, known location improved performance by as much as 40 percentage points in comparison to presenting the target sentence from a random location. Hence, in relatively simple listening conditions such as those containing minimal background noise, there appears to be only moderate benefit from deploying auditory spatial attention, but in more complex listening environments such as those containing higher levels of background noise, there appears to be significant benefit from deploying auditory spatial attention. Of note, no previous research investigating auditory spatial attention has examined the influence of age on word identification. 2.7 Summary of literature review There is evidence to suggest that the difficulties older adults experience in understanding language spoken in relatively quiet and simple listening environments arise from a lack of audibility of the speech signal. However, in acoustically more challenging listening environments, lack of audibility, by itself, is insufficient to explain why older adults experience difficulty understanding spoken language. Research demonstrates that age-related differences in auditory temporal processing are one reason for the difference in word-identification performance between younger and older adults in complex listening environments. In addition to age-related declines in auditory processing, cognitive aging research has observed widespread age-related declines in a number of cognitive processes including divided attention and task-switching, working memory, speed of information processing, and inhibitory processing. Although auditory factors account for most age-related variance in word identification, research from the past 25 years suggests that cognitive factors can contribute significantly to the difficulties that older adults experience when understanding spoken language. It is notable that the relative contribution of cognitive factors appears to be more important when listening to complex materials in complex listening situations. Most researchers have explored auditory and cognitive factors by presenting targets from fixed locations where listeners know in advance the location of the target and the masker. In reality, often listeners do not know the location of a target. Although there is considerable research in vision exploring age-related cueing effects (e.g., Folk & Hoyer, 1992; Hartley,

32 18 Kieley, & Slabach, 1990), there is a limited body of work examining the role of spatial cueing in audition, and even less investigating the role of aging in spatial listening. Currently, there is a need to better understand the auditory and cognitive processes that contribute to the ability to allocate auditory spatial attention. In particular, there is a poor understanding of the influence of age on the processes that contribute to the ability to focus auditory attention along a spatial vector. 2.8 Current research For all listeners, but especially those who are older and/or have impaired hearing, listening to speech with competing speech in the background is challenging. This difficulty is exacerbated in situations where the listener has imperfect information about the location of the target talker. Under conditions of spatial uncertainty, there is a limited understanding of how auditory and cognitive processes interact with spatial listening expectations and contribute to word-identification performance. Furthermore, there is a poor understanding with regards to the masking release attributable to auditory spatial attention in older adults. The goal of this research is to better understand auditory and cognitive mechanisms underlying auditory spatial attention in younger and older adult listeners. 2.9 Experiment I Although younger listeners can experience a significant benefit to word-identification performance from information that cues the location of a target talker in a multi-talker, multispatial listening environment, it remains unknown if older adults also experience a similar advantage. This question was addressed by comparing the influence of spatial listening expectations on word-identification performance in a group of younger and older adult listeners with normal audiometric thresholds below 4 khz in conditions where the target and two competitor sentences were each presented from a different spatial location. A second goal of Experiment I was to use the precedence effect to simulate spatial separation of the target and masker sentences in order to determine if possible age-dependent differences in spatial attention would be exacerbated when the monaural and binaural cues to spatial location were reduced.

33 Experiment II The goal of Experiment II was to investigate the time course of auditory spatial attention switching in a group of younger and older adult listeners with normal hearing sensitivity below 4 khz. Compared with the number of studies exploring the time course of visual spatial attention, relatively little is known about the time course of auditory spatial attention when listeners shift attention from one sound source to another. In everyday situations, listeners rapidly shift their focus of attention between different locations, and the availability of time may contribute to successful listening in several ways. For example, it takes time to switch attention (Shinn- Cunningham & Best, 2008), additional time could provide an opportunity to further process information (Barrouillet, Bernardin, & Camos, 2004), and/or time may facilitate the formation of auditory objects (Best, Ozmeral, Kopčo, & Shinn-Cunningham, 2008). Hence, the primary goal of Experiment II was to explore the effect of time on word-identification accuracy in a multitalker, multi-spatial listening environment in which the degree of uncertainty about the spatial location of a target is varied, with the secondary goal of investigating the influence of age on the time course of attention Experiment III The third goal of the dissertation was to investigate the effect of performing a single compared with multiple switches of auditory spatial attention on word-identification performance in a multi-talker, multi-spatial listening environment. Although there is limited research in audition, previous research in vision suggests that older adults exhibit poorer attentional control than younger adults, particularly in the presence of competitor stimuli that consume processing resources (Madden, Connelly, & Pierce, 1994). Specifically, it has been observed that, compared with younger adults, older observers experience more difficulty disengaging attentional resources when the task demands are such that participants must momentarily process competitor stimuli. Similar age-related performance deficits have also been reported in the literature on visuospatial search tasks that require the momentary processing of object features, such as those involving conjunction searches whereby observers repeatedly engage and disengage attention between multiple items presented in an array (Greenwood & Parasuraman, 1994; 1999). Hence, the motivation for conducting Experiment III was to explore

34 20 whether similar processes occur in an auditory task where listeners are instructed to momentarily process auditory masker sentence information Concluding remark In the following chapters, the logic, methods, results and discussion from three experiments are presented. These experiments have in common that they link previously disparate areas of research by attempting to simultaneously consider both cognitive and auditory factors as potentially important mediators of listening in complex listening situations and by attempting to experimentally capture some of the complexity characteristic of human communication.

35 21 Chapter 3 The Effect of Age on Auditory Spatial Attention in Conditions of Real and Simulated Spatial Separation 3.1 Introduction The relationship between auditory and cognitive factors in multi-talker listening situations is highlighted by work on the cocktail party problem (Cherry, 1953). In cocktail party situations, a listener selectively attends to and identifies the speech from a single talker among a mixture of background conversations in a multi-talker situation (for reviews see Bregman, 1990; Yost, 1997; Bronkhorst, 2000; Ebata, 2003; Haykin & Chen, 2005). Speech identification in the presence of background speech is a nontrivial challenge that results in a phenomenon called masking, which is the process by which the detection threshold for a sound (i.e., target) is made more difficult by the presence of another (i.e., masker) sound. When identifying target speech in the presence of speech maskers, the difficulties arise from two different types of masking. Whereas energetic masking (French & Steinberg, 1947) occurs when a target and masker compete for representation in a channel of information at the level of the auditory periphery (e.g., in a cochlear filter or on proximal portions of the auditory nerve), informational masking refers to the additional masking that is observed when there is competition for representation at higher or more central levels of processing (Freyman, Helfer, McCall, & Clifton, 1999; Durlach, Mason, Kidd, Arbogast, Colburn, & Shinn-Cunningham, 2003). Auditory scene analysis (Bregman, 1990) provides a useful framework within which to consider the cocktail party problem and the specific processes by which the human auditory system enables the decomposition of incoming complex signals into separate perceptual representations. When listening in noisy backgrounds, auditory input is partitioned based on the acoustic properties of the stimulus (i.e., bottom-up or data-driven processing) and/or categorized by making use of stored auditory object representations that are formed by prior knowledge and experience (i.e., top-down or prediction-driven processing). The assumption is that in order to distinguish a target from a masker, listeners will use a combination of auditory and cognitive processes.

36 Spatial separation Cherry (1953) suggested that spatially separating a target from a masker improves target identification. One benefit of spatial separation is that there is reduced competition between a target and masker in the auditory periphery (i.e., release from energetic masking) and because there is reduced competition at higher levels of processing (i.e., release from perceptual or informational masking) (e.g., Watson, 1987; Durlach et al., 2003; Hornsby, Ricketts, & Johnson, 2006; Li et al., 2004; Wu et al., 2005). Several mechanisms account for spatial unmasking in complex listening environments, and it may be useful to consider them as described in an information processing framework that incorporates both signal-driven bottom-up and cognitively-meditated top-down factors. Discussed next is a description of peripherally mediated, signal-driven auditory mechanisms. Spatial separation of a target and masker depends on a number of monaural and binaural auditory cues that could facilitate speech identification in noise. Foremost among the monaural cues is that separating a target and masker will change the SNR at an ear. Compare a situation in which both target and masker are located to the listener s left to a situation in which the target is on the right and the masker is on the left. Moving the target from left to right dramatically improves the SNR at the right ear due to the sound shadow cast by the listener s head. In addition, a shift in target position will change the spectral profile of the target at one ear due to diffractive and reflective properties of the pinna, head, and torso, but the role of monaural spectral profile cues in promoting intelligibility appears to be comparatively minor relative to monaural changes in SNR (Wightman & Kistler, 1997). Thus, spatially separating the target from the masker produces cues, which, when processed monaurally, could aid speech identification in noise. Speech identification could also be enhanced by binaural processing. Spatial separation of a target from a masker leads to an interaural level difference (ILD) that is different between target and masker because of the head shadow, as well as an interaural time difference (ITD) that is different between target and masker because of the different distances a sound must travel to reach each ear (for reviews see Blauert, 1997; Bronkhorst, 2000). Binaural processing of these interaural differences enables higher perceptual systems to take advantage of the subtle spectrotemporal differences between the target and masker signals arriving at each ear (e.g., Colburn,

37 23 Shinn-Cunningham, Kidd, & Durlach, 2006; Culling, Hawley, & Litovsky, 2004; Duquesnoy, 1983; Plomp, 1976). Hence, it is possible that age-dependent changes in either monaural or binaural processing could reduce the effectiveness of spatial separation in releasing a target from masking. Indeed, age-dependent differences in auditory binaural processing have been found in older listeners with normal hearing sensitivity in the speech range (for reviews, see Grose, 1996; Koehnke & Besing, 2001). In addition to bottom-up, signal-driven processes, cognitive, or prediction-driven, mechanisms are also likely to be engaged, or become more effective, when the target and masker are physically separated. One such top-down process (auditory spatial attention) may involve the listener s ability to focus attention along a spatial vector to a target (Arbogast & Kidd, 2000; Brungart & Simpson, 2007; Ericson et al., 2004; Kidd et al., 2005; Mondor & Zatorre, 1995; Spence & Driver, 1994). Evidence suggests that auditory spatial attention operates in a manner consistent with Broadbent s (1958) spotlight model of attention (Best et al., 2006) and that a gradient model best describes auditory spatial attention where performance declines with increasing distance from the spatial center of attentional focus (Mondor & Zatorre, 1995). Accordingly, physical separation of the target and masker should facilitate attentional processes by distancing the masker from the center of the spotlight. 3.3 Real and simulated spatial separation A listener s ability to benefit from spatial separation when listening to multiple sound sources could depend to varying degrees on different acoustical cues. The roles of ILD and ITD cues for the direct wavefront were described above. However, in realistic listening situations there are usually reflected wavefronts as well. Over short distances and brief time intervals, the precedence effect occurs, whereby the human auditory system fuses direct sound waves and early reflections into a single auditory event, rather than a percept followed by an echo (Wallach, Newman, & Rosenzweig, 1949; for reviews, see Zurek, 1980; Blauert, 1997; Litovsky, Colburn, Yost, & Guzman, 1999; Li & Yue, 2002). By manipulating the time delay between the presentation of a signal from two equidistant loudspeakers located in the right and left hemi-field of a listener, it is possible to simulate different spatial locations. By presenting a sentence simultaneously from each loudspeaker, a listener will perceive the sentence as originating from a location between the two loudspeakers. If a sound from one of the loudspeakers lags the sound

38 24 from the other loudspeaker by a few milliseconds, the resulting percept is localized near the location of the leading loudspeaker. In this way, one can simulate spatial separation between a target and masker when both sentences are presented from both loudspeakers. This methodology has been previously used to simulate spatial separation and is particularly well-suited to the study of auditory spatial attention insofar as the perception of spatial separation is achieved while minimizing the contribution of some of the acoustical differences (changes in monaural SNRs, interaural correlation, etc.) that are present when there is real spatial separation (e.g., Freyman et al., 1999; Rakerd, Aaronson, & Hartman, 2006 and the Appendix in Li et al., 2004). In essence, the precedence effect may bestow the benefits of true spatial separation even though head shadow and binaural interactions are reduced (Freyman et al., 1999, p. 3579). In the RSS case, the target is delivered only from a loudspeaker at one location, and each competitor is delivered only from a loudspeaker at a different location. Consequently, multiple interaural cues are available, and the average target-to-competitor ratio at one ear is not equal to that at the other ear because of head shadow effects. In contrast, when the precedence effect is used to simulate spatial separation, the target and competitors are delivered at equal presentation levels from each of two loudspeakers with time delays introduced to induce the perception of spatial separation. Consequently, interaural level difference cues due to head shadow are mimimal, and the average target-to-competitor ratio is the same at each ear. There are, of course, other differences between the spectra of signals presented in SSS conditions compared to those presented in RSS conditions, the most prominent of which are comb-filtering effects. However, the effects of comb-filtering are unlikely to affect performance when the onset delay between the left and right loudspeakers is long (> 2 ms) and tests are conducted in a non-anechoic soundattenuating chamber (Li et al., 2004; see also Brungart, Simpson, & Freyman, 2005). At the perceptual level, sounds whose perceived locations are induced using the precedence effect (playing the sounds over two loudspeakers with one sound leading the other) are perceived to be more diffuse, and are less precisely localized than the same sound played over a single loudspeaker (Blauert, 1997). The degree to which these perceptual differences between a precedence-induced spatially-located sound and a sound presented from single spatial location will affect performance will likely depend on the precision of localization demanded by a listening task.

39 Aging Older adults typically have more difficulty than younger adults when speech identification is tested using monaurally presented speech-in-noise tests, and their difficulties may be explained, at least partially, by age-dependent declines in temporal processing (for reviews see CHABA, 1988; Divenyi & Simon, 1999; Pichora-Fuller & Souza, 2003). For older listeners with near-normal pure-tone thresholds, their ability to use ILD cues may be less affected than their ability to use ITD cues. For example, Herman, Warren, and Wagener (1977) found ILD difference limens to be the same for younger and older listeners; however, Pichora-Fuller and Schneider (1992; 1998) found age-dependent binaural masking-level differences that were explained in terms of an age-dependent increase in the amount of temporal jitter in the binaural system (see also Dubno, Ahlstrom, & Horowitz, 2002; for a review see Grose, 1996). Thus, agedependent differences in some aspects of monaural and/or binaural auditory processing, especially those aspects involving temporal processing, could compromise the ability of older listeners to follow speech in complex multi-talker situations. 3.5 Current study: Dissertation study I The current investigation is designed to extend the findings from Kidd et al. (2005) and Li et al. (2004) to explore possible age-dependent differences in auditory spatial attention in a multi-talker situation where the acoustic cues are either fully available or reduced using the precedence effect. Using the precedence effect, Li et al. (2004) presented semantically meaningless target sentences at a location to the right of younger and older adult listeners, while manipulating the simulated location of a single masker (right, left, or in front). In their experiment, the location of target and masker sentences was 100% certain. Although older adults needed a better SNR to perform equivalently to younger adults, there were no age-dependent differences in masking release. Using RSS of target and maskers, the Kidd et al. study demonstrated that for younger listeners, certainty about the location of the target and prior information about the callsign cue to the target utterance are important factors contributing to the allocation of attention and improvement in word-identification performance. Age-dependent differences in unmasking may not have been found in the Li et al. study because the attentional demands placed on listeners were limited insofar as the location of the target was always known. In vision, previous research suggests that there is an effect of age when attentional demands are

40 26 high (for a review see M c Dowd & Shaw, 2000). Therefore, one possibility is that by manipulating the attentional demands of the listening task by varying location certainty, agedependent differences in performance might be revealed. By comparing the word-identification accuracy of younger and older age listeners in conditions of RSS (following Kidd et al., 2005), an attempt was made to determine if agedependent differences would be observed when the attentional demands of the task were varied. Furthermore, the precedence effect (following Freyman et al., 1999) was used to simulate spatial separation to determine if age-dependent differences would be exacerbated or reduced when the monaural and binaural cues to spatial location were reduced. By comparing the results between the two presentation methods, an attempt was made to gain a better understanding of how bottom-up processes interact with top-down processes to enhance word identification when a target is physically separated from a masker, and to determine the extent to which this interaction is age-dependent. Specifically, if age-related differences in performance are greater in the RSS than SSS conditions then the pattern of findings suggest the importance of bottom-up auditory processing. In contrast, if age-related differences in performance are greater in the SSS than the RSS conditions when the role of auditory cues has been minimized then the pattern of findings suggest the importance of top-down cognitive processing. If the age-related differences are the same in both RSS and SSS conditions then both types of processing are implicated. 3.6 Methods Participants Eight younger adults (aged years; mean = 24.38; SD = 3.02) and eight older adults (aged years; mean = 70.38; SD = 3.89) participated in the study. Participants were recruited from the local university and community. All participants were native English speakers and in good overall health. All listeners had clinically normal pure-tone air-conduction thresholds ( 25 db HL) from 0.25 to 3 khz in both ears (see Tables 1 and 2). All participants provided informed consent and were paid $10 per hour of testing Stimuli The stimuli consisted of sentences from the CRM corpus spoken by the four male talkers (Bolia et al., 2000). The sentences have the format: Ready [callsign] go to [color] [number]

41 27 now. The CRM corpus contains sentences with all possible combinations of eight callsigns (Arrow, Baron, Charlie, Eagle, Hopper, Laker, Ringo, Tiger), four colors (red, white, blue, green), and eight numbers (one, two, three, four, five, six, seven, eight). Table 1. Individual and group left- (L) and right-ear (R) audiometric thresholds (in db HL) for the eight younger (Y) adult participants in Experiment I. Frequency (khz) Participant L R L R L R L R L R L R L R Y Y Y Y Y Y Y Y Mean (Y) SD (Y) Equipment All testing was performed in a 3.3 x 3.3 m single-walled sound-attenuating Industrial Acoustics Company (IAC) sound booth. Presentation of the stimuli was controlled via custom software developed on a Visual Basic platform. All stimuli were routed from a computer to a Tucker-Davis Technologies System III to a Harmon/Kardon amplifier (model HK3380). Sentences were converted to analogue form using two TDT System III RP2.1s at a kHz sampling rate by a 24-bit Sigma-Delta digital-to-analogue converter. The analogue outputs were attenuated using two programmable attenuators (TDT System III PA5) for the SSS condition, and three attenuators for the RSS condition. Signals were then conditioned by stereo power amplifiers (TDT System III SA1), and presented over Grason-Stadler Inc. (catalog number ) loudspeakers. All loudspeakers were visible to the participants. Visual cueing and feedback was displayed on a Compar (model 1760nx) 17 NEC touch screen monitor on a 0.46

42 28 m high table located in front, but below the shoulder level of the listener. Calibration was performed using a Brüel & Kjær (B&K) Modular Precision Sound Level Meter (Type 2260) with a 0.5-inch B&K condenser microphone (Type 4189). The microphone was placed at the location that the listener s head would occupy, and measurements were collected using the A-weighting equivalent. Measurements were taken separately for each sentence presented alone from each loudspeaker. During calibration, concatenated sentences from the CRM corpus were presented at a level such that each loudspeaker, playing alone, would produce an average sound pressure of 60 dba at the location corresponding to the center of the listener s head Design The main dependent variable was accuracy of word-identification. Three independent variables were systematically manipulated in this experiment. Two of the variables, callsign cue and location certainty, were identical to those studied in the experiment of Kidd et al. (2005). A new variable, presentation method, was also manipulated. Word-identification was measured for all participants in each age group for 16 conditions (2 callsign cue conditions x 4 location probability specifications x 2 presentation methods). All participants completed every condition. The order of testing using the two presentation methods was counter-balanced, with half of the younger and older participants starting with one presentation method and the remainder starting with the other. For each presentation method there were 8 to 16 testing sessions, usually with two sessions completed in a one to two hour visit. The order of testing the two callsign cue conditions was counter-balanced for each participant; the data in each session were collected in eight blocks with a sequence of cue conditions alternating every two blocks (i.e., two blocks of one cue condition followed by two blocks of the other cue condition). Each block consisted of 30 trials. The starting callsign cue condition was randomly determined for each listener. Within each session, for each callsign condition, the four probability specifications were assigned randomly without replacement so that a different probability certainty condition was assigned for each block. Each participant minimally completed 3840 trials, with a minimum of 240 completed trials for each of the 16 conditions. Table 2. Individual and group left- (L) and right-ear (R) audiometric thresholds (in db HL) for the eight older (O) adult participants in Experiment I.

43 29 Frequency (khz) Participant L R L R L R L R L R L R L R O O O O O O O O Mean (O) SD (O) Procedures The listener s task was to identify the color and number in the target sentence that was presented simultaneously with two masker sentences. All sentences were presented at 60 dba, and were selected randomly from the CRM corpus. On a given trial, the target and masker sentences differed with respect to the color, number, callsign, and talker of the sentence. Listeners pressed on a touch screen to indicate which one of the four possible colors and which one of the eight possible numbers they heard in the target sentence. Both the correct color and number were required for a correct response. Feedback (correct or incorrect) was provided after every trial and the percentage of correct trials was provided at the end of each block. At the start of each visit, participants completed either one or two practice blocks using the presentation method to be tested. Within any given practice block, a randomly chosen callsign cue condition combined with a randomly chosen location certainty condition was completed. In the test phase, within any given block, participants were informed in advance of the probability specification for the block and they were cued to the identity of the callsign that began the target utterance either 1 s before (callsign cue before condition) or immediately

44 30 following (callsign cue after condition) each trial in the block. The probability specification appeared on the monitor 1 s before and throughout the block. There were four probability specifications indicating the proportion of trials that the target would be presented from the left, center, and right spatial locations ( , , , and ). For example, when participants were provided with a cue, they were instructed that there was a 10% chance that the target would be presented from the left location, an 80% chance that the target sentence would be presented from a center location in front of them, and a 10% chance that the target would be presented from the right location. Thus, a cue of indicated certainty that the target would be presented from a location in front and the cue indicated that the location of the target sentence would be determined randomly. On any given trial, the location of the target sentence was randomly selected from the left, center, and right locations, with the limitation being that the probability cue was accurate across a block of 30 trials. Participants were instructed to face directly ahead for the duration of the stimulus presentation. Two different presentation conditions were used, a RSS condition and a SSS condition. Following Kidd et al. (2005), in the RSS condition, three loudspeakers were used to present stimuli. The target and two masker sentences were presented simultaneously, but only one sentence was played from each loudspeaker. Each loudspeaker was located 1.83 m (approximately 6 ft) from the participant s head. As shown in Figure 1, they were positioned at 0 and ±54 azimuth in the horizontal plane.

45 31 Figure 1. Schematic of loudspeaker configuration for real (left) and simulated spatial separation (right). The circle indicates the position of the participant and the squares indicate the position of the loudspeakers. Two different presentation conditions were used, a RSS condition and a SSS condition. Following Kidd et al. (2005), in the RSS condition, three loudspeakers were used to present stimuli. The target and two masker sentences were presented simultaneously, but only one sentence was played from each loudspeaker. Each loudspeaker was located 1.83 m (approximately 6 ft) from the participant s head. As shown in Figure 1, they were positioned at 0 and ±54 azimuth in the horizontal plane. In the SSS condition, the precedence effect (e.g., Freyman et al., 1999) was used to achieve perceived spatial separation of the target and competitors by manipulating time delays between the presentations from only two loudspeakers. Each loudspeaker was located 2.08 m (approximately 6.8 ft) from the listener s head and was positioned at ±45 azimuth. The height of all the loudspeakers was approximately the same height as the listener s head when seated. Although all three sentences were presented from both loudspeakers, an utterance appeared to come from the left spatial location when the signal from the left loudspeaker led the signal from the right loudspeaker by 3 ms. Similarly, an utterance appeared to come from the right spatial location when the left loudspeaker lagged the right by 3 ms. Finally, an utterance appeared to

46 32 come from in front of the participant when a sentence was presented from both loudspeakers at the same time. Thus, perceived spatial separation of the target and competing sentences using the precedence effect was achieved, but all three sentences were presented from both the left and right loudspeakers (see Figure 1 for a schematic of the soundbooth configurations). 3.2 Results The overall results from the experiment are shown in Figure 2 for both age groups. In general, four main findings of interest were observed. First, although older adults performed worse than younger adults, there was a similar pattern of results across all conditions for both age groups. Second, performance improved when listeners were more certain about the location of the target. Third, performance was better when the callsign cue was known in advance. Fourth, the effect of location certainty and callsign cue were more pronounced in the RSS compared to the SSS. These effects were tested with a 2 x 2 x 2 x 4 repeated measures analysis of variance (ANOVA) where age (younger versus older) was a between-subjects variable and presentation method (RSS versus SSS), callsign cue (before versus after), and location certainty (1.0, 0.8, 0.6, and 0.33) were within-subjects variables. It should be noted that the same pattern of effects was observed when the ANOVA was conducted on either the mean percent correct data or when the data were transformed using the rationalized arcsine transformation (Studebaker, 1985). As such, all analyses will be reported for the mean percent correct data Age As shown in Figure 2, younger adults outperformed older adults in almost every condition of the study. This was confirmed by a significant main effect of age [F(1,14) = 7.60, p < 0.05]. Collapsing across all conditions, the mean word-identification accuracy of the younger adult group was approximately 8.2 percentage points better than that observed in the older adult group. Most importantly, significant interactions of age with presentation method, callsign cue, or location certainty were not observed (all p > 0.05), suggesting that younger and older adults did not differentially make use of information about target identity and location probability under either the RSS or SSS presentation methods.

47 Location certainty The single largest factor which improved performance was a priori target location information, as confirmed by the significant main effect of location certainty [F(3,42) = , p < 0.001]. On average, when listeners were most uncertain about the location of the target, correct identification was 41.6% whereas it was 85.4% when the target was presented at a fixed location, yielding an improvement of 44.2 percentage points. However, as depicted in Figure 2, the deleterious effect of location uncertainty was offset when participants had prior knowledge of the identity of the target callsign, particularly for presentations using RSS (i.e., in Figure 2, the slopes of the functions are shallower for the callsign-before compared to the callsign-after conditions). Nevertheless, even with this offset, performance improved by 27.5 percentage points when location certainty was increased from 0.33 (where performance was 61.7%) to 1.00 (where performance was 89.2%). This pattern of interactions was confirmed with a significant three-way interaction of presentation method, callsign cue, and location certainty [F(3,42) = 16.68, p < 0.001]. Significant two-way interactions of location certainty and callsign cue [F(3,42) = 47.80, p < 0.001], location certainty and presentation method [F(3,42) = 17.02, p < 0.001], and callsign cue and presentation method [F(1,14) = , p < 0.001] were also observed; however, the relationships between location certainty, callsign cue, and presentation method will not be further discussed as two-way interactions because the relationship between these variables can more precisely be characterized through consideration of the three-way interaction Callsign cue Word-identification improved with prior knowledge of the target callsign. Specifically, performance improved by 11.5 percentage points when the callsign cue was known prior to the start of a trial (67.8%) compared to when it was cued after stimulus presentation (56.3%). This was confirmed by a significant main effect of callsign cue [F(1,14) = 58.10, p < 0.001]. As illustrated in Figure 2, and confirmed by the significant three-way interaction discussed above, the effect of knowing the callsign cue prior to the start of the signal presentation was most pronounced when location certainty was reduced and when there was RSS. Whereas there was a 7 percentage point benefit associated with advance knowledge of the callsign for SSS, there was

48 Figure 2. Mean percent correct word-identification scores and standard errors of the mean for younger (top, unfilled symbols) and older (bottom, filled symbols) adults for the four location certainties. Scores on the y-axis were calculated as the percentage of trials where participants correctly identified both the color and number associated with the target callsign. Solid lines indicate RSS; dashed lines indicate SSS; circles indicate callsign cue before conditions; and triangles indicate callsign cue after conditions. 34

49 35 a 15 percentage point improvement with RSS, suggesting that the removal of some of the auditory cues associated with RSS lessens, but does not completely eliminate, the benefit of knowing the identity of the target callsign in advance of hearing the sentence Real vs. simulated spatial separation Relative to the listening situation where spatial location was simulated via the precedence effect, overall performance in the RSS condition was better by an average of 6.3 percentage points. This was confirmed by a significant main effect of presentation method [F(1,14) = 36.04, p < 0.001]. However, again evidenced by the significant three-way interaction, the benefit realized when sentences were generated from three independent spatial locations was most noticeable when the callsign cue preceded stimulus presentation and the target location was less certain. Notably, no benefit from RSS compared with SSS was observed when the callsign cue was presented after stimulus presentation Spatial listening expectations The significant three-way interaction discussed above indicates that the drop in performance with decreasing certainty is offset by prior knowledge of the target callsign when listening in RSS conditions. One possibility is that prior knowledge of the target callsign enables the listener to draw on the interaural cues available with actual spatial separation in order to selectively attend to targets originating from unexpected locations. To further investigate this possibility, a comparison of performance for targets appearing in expected and unexpected locations was conducted. For this analysis, the focus was on the callsign cue conditions where target identity was known before stimulus presentation and the conditions in which location certainty was less than 1.0 but more than chance (0.33). Whereas the callsign cue established listener expectations regarding target identity, by choosing the intermediate location certainty conditions it was possible to compare trials where the target was presented at the more likely central location or a less likely side location. For example, when the probability of the target being presented at the center location was 0.80 or 0.60, the listener would need to allocate some attentional resources from the expected central location to an unexpected side location on some trials. A likely trial would occur if the target callsign occurred at the center location and an unlikely trial would occur if the target callsign occurred at either the left or right spatial location. The ability of listeners to allocate spatial attention was gauged by comparing their

50 36 word-identification performance on trials where the target was presented at the likely central location or to an unlikely side location. The difference in performance between likely and unlikely trials should be smaller for listeners who are better able to re-allocate attention from expected to unexpected spatial locations. If there are age-related differences in the ability to reallocate spatial attention on this task, then there would be a greater difference in performance between likely and unlikely trials for older compared to younger listeners. As shown in Figure 3, for both age groups, word-identification was approximately 45 percentage points better on likely than unlikely trials. Importantly, although younger adults performed better than older adults by an average of 9 percentage points when the results are collapsed across the two presentation methods, the cost of re-allocating attention from a likely to an unlikely spatial location was similar for older (44%) and younger adults (46%). In order to statistically confirm these descriptions, a repeated-measures ANOVA was conducted with age (younger versus older) as the between subjects variable, and presentation method (RSS versus SSS) and cue validity (likely versus unlikely) as within-subjects variables. A significant main effect of cue validity [F(1,14) = , p < 0.001] was observed, indicating that performance was better for likely than unlikely spatial locations. Although a borderline main effect of age [F(1,14) = 4.14, p = 0.06] was observed, a significant cue validity x age interaction [F(1,14) = 0.05, p > 0.05] was not observed, suggesting that older adults did not exhibit more difficulty than younger adults re-allocating attentional focus The role of natural spatial cues The influence of the richness of the interaural cues on the cost of re-allocating attention from a likely to an unlikely side spatial location was examined by comparing performance in the RSS compared to the SSS conditions. As shown in Figure 3, the richness of the interaural cues strongly affected word-identification performance for the unlikely trials, with scores being 29 percentage points higher in the RSS compared to the SSS conditions. However, the availability of rich interaural cues did not affect performance on the likely trials, with there being less than a 1 percentage point difference between the RSS and SSS conditions. Again, the same pattern

51 37 Figure 3 Mean percent correct word-identification scores and standard errors of the mean for RSS (solid lines) and SSS (dashed lines) presentation conditions, depicted for the likely and unlikely spatial locations. Unfilled and filled circles represent data collected on younger and older adults respectively. Triangles represent the results of Kidd et al. (2005) for younger adults when performance was averaged over conditions in which the likely location was left, center, or right. Data are collapsed across location probabilities 0.8 and 0.6. was observed for both age groups. The description of this pattern was confirmed with a significant main effect of presentation method [F(1,14) = 98.33, p <0.001], as well as a presentation method by cue validity interaction [F(1,14) = 59.34, p < 0.001]. Post-hoc SNK testing of all possible comparisons revealed no significant difference between RSS and SSS conditions for likely trials (p > 0.05), but significant differences between the two presentation methods for unlikely trials (p < 0.01), suggesting that the benefit resulting from the availability of rich interaural cues in the RSS conditions is contingent upon a listener s spatial expectations. Finally, a significant presentation method x cue validity x age interaction [F(1,14) = 0.00, p >0.05] was not observed, suggesting that younger and older adults are equally disadvantaged from a reduction in auditory cues associated with SSS across both likely and unlikely spatial listening locations. No other significant effects were obtained.

52 Discussion The goal of the current investigation was to determine if there are age-dependent differences in word-identification that could be attributable to the use of acoustic cues and/or auditory spatial attention. First, we discuss our replication of previous findings (Kidd et al., 2005) for younger adults in the RSS conditions, and then we compare the results we found for younger and older listeners in the RSS and SSS conditions. Figure 4. Mean percent correct word-identification scores and standard errors of the means from the younger adults listening with RSS in the current study (solid lines) and in the study of Kidd et al. (2005; dashed lines) for the four location certainties. Circles indicate callsign cue before conditions, and triangles indicate callsign cue after conditions Performance of younger listeners On average, the results we found for younger adults tested in the RSS conditions are a very good replication of the findings reported previously (Kidd et al., 2005). Overall, the mean word-identification score was 3 percentage points higher for younger listeners in the present study than for those in the prior study. When the callsign cue was specified prior to sentence presentation, the scores for our participants and those in the previous study were 77% and 75% respectively. When the callsign cue was specified following sentence presentation, the

53 39 corresponding scores were 62% and 59%. Furthermore, as shown in Figure 4, the pattern of results was almost identical across the different conditions that were tested The effect of age on performance When there was RSS between the sound sources, in every condition the older adult group demonstrated significantly poorer word-identification performance compared to the younger adult group. Importantly, based on the lack of significant interactions between age and the other independent variables (presentation method, callsign cue, or location certainty), the pattern of findings indicates that there are general, but no condition-specific age-dependent differences in the ability to use auditory spatial attention to follow target speech when there are three simultaneous talkers. Furthermore, comparing the results obtained in the RSS and SSS conditions, we found that both younger and older adults performed significantly better when there was RSS than when there was SSS, but the older age group was not disproportionately impaired in the SSS condition. The finding that older listeners were not disproportionately disadvantaged in the SSS conditions indicates that there were no relevant age-dependent differences in the precedence effect. In one prior psychoacoustic study of the precedence effect in which loudspeaker delays were manipulated, older listeners had more difficulty fusing clicks compared to younger listeners for time delays less than 0.7 ms (Cranford, Boose, & Moore, 1990; Cranford, Andres, Piatz, & Ressig, 1993). However, our time delay was much longer (3 ms) and our findings are consistent with a previous psychoacoustic study that failed to observe age-dependent differences in the precedence effect for longer time delays (Schneider et al., 1994; Roberts & Lister, 2004). No doubt there are multiple mechanisms underpinning the overall poorer wordidentification abilities of the older adults compared to younger adults. One possible explanation that must be considered is that the older adults could be in the early stages of presbycusis. Although both age groups had normal audiometric thresholds for frequencies in the speech range, at higher frequencies (4 and 8 khz), the average differences between the thresholds of the older and younger groups were 14 and 17 db, respectively (see Tables 1 and 2). Note that of the 32 thresholds measured above the speech range in the older listeners (two ears x eight participants x two frequencies), only nine thresholds were classified as clinically abnormal, but all of these fell within the mild hearing loss category. Nevertheless, a high-frequency hearing

54 40 loss could have contributed to the main effect of age observed in this study. Previous research suggests that listeners with impaired high-frequency hearing fail to fully benefit from spatial separation between a target and noise because the improvement in SNR in the higher frequencies arising from head shadow effects is reduced or eliminated (Bronkhorst & Plomp, 1989), or because poorer binaural interaction reduced the benefit that would normally be gained from spatial separation when listening with one or two ears (Bronkhorst & Plomp, 1988). Consequently, the poorer performance observed in our older listeners may reflect an inability to take advantage of head shadow effects or binaural interaction. However, this explanation is unable to account for the overall pattern of results in the present study. If the better performance of the younger adults compared to the older adults were due to the superior high-frequency hearing thresholds and ability of the younger listeners to use high-frequency cues to take advantage of spatial separation, then there should have been a greater age-dependent deficit in the RSS conditions than in the SSS conditions where these cues were virtually eliminated. As shown in Figure 2, younger and older listeners exhibited similar reductions in performance when comparing the results of the RSS and SSS conditions, suggesting that the high-frequency audiometric deficits exhibited by our older adults cannot account for the pattern of results related to the benefit arising from spatial separation. Our finding of a main effect of age in the RSS condition is consistent with previous studies where speech-in-noise tests were conducted in conditions of spatial separation (Dubno et al., 2002; Gelfand, Ross, & Miller, 1988; Li et al., 2004). When considering the benefit from spatial separation, however, there is little evidence suggesting that older adults benefit less from spatial separation than younger adults. Gelfand et al. (1988) using the Revised Speech Perception in Noise test (Bilger, Nuetzel, Rabinowitz, & Rzeczkowski, 1984) found no age-dependent differences in benefit from spatial separation; however, Dubno et al. (2002) using the Hearing in Noise test (Nilsson, Soli, & Sullivan, 1994) found that the benefit from spatial separation was approximately 1 db less for older adults with normal hearing relative to younger adults. Compared to prior studies, in our study, we found that older adults benefited as much as younger adults from spatial separation either using real (Gelfand et al., 1988) or simulated (Li et al., 2004) spatial separation. Hence, the available evidence suggests there is little (or no) agedependent difference in the ability to use spatial separation for individuals with normal audiometric thresholds.

55 Auditory perception of spatial separation One possible explanation that may account for the observed differences in performance between the RSS and SSS conditions is the difference in the precise nature of the perceived location of sound sources between the two conditions. For example, the simulated spatial sound sources may have appeared to sound further or closer than the sentences presented in the corresponding RSS conditions. However, the pattern of results found by Li et al. (2004) suggests that this explanation is unlikely to account for the differences in question. In their study, the simulated location of the target was to the right of the listener and the simulated locations of the maskers were to the right, center, or left of the listener. Relative to the condition where the masker was simulated on the right, Li et al. (2004) observed a similar masking release of 4.8 db regardless of whether the simulated masker was located at the center or left position. Hence, it seems unlikely that the minor differences in perceived spatial location of the sound sources between the real and simulated conditions could explain the pattern of results we observed. Nevertheless, to our knowledge no previous experiment has attempted to simulate three simultaneous, spatially separated sentences using the precedence effect, and it would be prudent to interpret the results with caution. Future research should more carefully investigate to what extent differences observed between presentation methods using actual and perceived spatial separation are due to a loss of natural binaural cues, and which differences are due to failures of the precedence effect to assist in the formation of clearly localized auditory objects. In the current investigation, the precedence effect was used to reproduce spatial separation in order to isolate the contribution of the rich, natural binaural cues that are available with RSS. Although no benefit was observed when the callsign was cued after stimulus presentation, the advantage of full binaural cues became apparent when the callsign was cued before stimulus presentation, and was increasingly apparent when the certainty about the location of the target decreased for both age groups Selective auditory spatial attention In order to further examine the extent to which performance is governed by auditory and/or attentional factors, we conducted an analysis focusing on the relationship between spatial listening expectations and the benefit derived from auditory cues associated with spatial separation. Specifically, we compared the results obtained in the RSS and SSS conditions using

56 42 only trials on which the target was presented from a likely or unlikely spatial location. Performance was similar when stimuli were presented with RSS or SSS (< 1 percentage point difference) at likely locations, but there was a considerable performance difference (> 29 percentage points) between the RSS and SSS conditions for targets presented at unlikely spatial locations (see Figure 3). Seemingly, the cognitive influence of location expectation may modulate the importance of the natural auditory binaural cues associated with RSS. The usefulness of full binaural cues may be more critical in ambiguous and/or dynamic listening situations in which the listener is required to selectively re-allocate spatial attention because the availability of full binaural cues facilitates sound localization, thus allowing for greater precision with which to focus attentional resources. At unlikely listening locations, all listeners performed more poorly when listening with simulated rather than RSS. When the likely location of the target is at the center (as it always was in our study), and the target is presented from an unlikely side location, there is a monaural SNR advantage in the RSS conditions for side targets that is not available in the SSS conditions. The availability of this SNR advantage could explain why performance was poorer in the simulated than it was in the RSS conditions when the target appeared in an unlikely position. However, because we did not vary the location at which the target was likely to appear (as did Kidd et al., 2005), we were not able to determine the extent to which performance at the unexpected locations in the RSS condition varied as a function of the expected location of the target. Nevertheless, because Kidd et al. (2005) only reported results for likely and unlikely locations that were averaged over the different possible locations at which the target was expected, we can only compare their averaged data to our results. Figure 4 shows that our results in the RSS conditions are only slightly better than those of Kidd et al. (2005), with the difference between studies being only 3 percentage points at the expected and 4.5 percentage points when the target is presented at an unexpected location. Hence, although we cannot resolve whether the pattern of results found here for RSS conditions would vary with the location at which the target was expected, we note that our results replicate the data (averaged over the different locations at which the target was expected) of Kidd et al (2005). Assuming our view that the data truly reflect differences between likely and unlikely listening locations, and do not arise from a monaural SNR advantage for side locations in

57 43 conditions using RSS, then several mechanisms may potentially account for this pattern of results. One possibility is that the listener s probability of attending to a spatial position matches the probability that the target will appear at that spatial position (for a review see Vulkan, 2000). However, Kidd et al. (2005) rejected this alternative because if listeners are using a probability rule, then one would expect that the proportion of responses from an expected location in callsign-after conditions would closely approximate the associated location probability. Since Kidd et al. (2005) found that under such conditions listeners made substantially more responses from expected locations than would be predicted by a probability-matching strategy, they concluded that there is little evidence that listeners adopt such a strategy. A second possibility is that observers always begin each trial focused on the likely location and then switch attention to the unlikely location. A third possibility is that as uncertainty increases, listeners broaden the spatial extent of locations to which attention is allocated, or increase the width of the attentional beam. Based on the methodology used in this study, it is currently unclear if listeners a) switch a more tightly-focused attentional beam from likely to unlikely locations, b) more broadly tune their auditory attentional filter to encompass more spatial locations, or c) perform the listening task by incorporating elements of both attention switching and broadened attentional focus. Future research should more clearly delineate between these possible alternatives Interaction of auditory and attentional factors We have shown that the performance deficits observed in both age groups at unlikely spatial listening locations are modulated by the availability of full binaural cues, with the deficits being substantially larger in the simulated than in the real spatial location conditions. Currently, it is unclear whether the poorer performance at unlikely locations for simulated spatial locations relative to real spatial locations is primarily due to disruptions in a) monaural signal-tonoise cues, b) ILD cues, or c) ITD cues. Because the current methodology is unable to definitively identify the relative contributions of particular acoustic cues to word-identification performance at unlikely locations in a spatially-complex listening situation, future research will need to develop testing situations that are better able to do so. As noted above, the relative contributions of location expectations and the natural binaural cues associated with RSS were similar for younger and older adults. Although younger adults outperformed older adults by an average of approximately 9 percentage points on word-

58 44 identification performance across the likely and unlikely spatial listening locations, the loss of full binaural cues did not disproportionately impair the older listeners. To our knowledge no previous research has examined the role of binaural cues in auditory spatial attention in older adults. The general finding in vision studies of attention and aging is that disproportionate agedependent differences are found for slower, more controlled behaviors, but not for faster, reflexive behaviors that incorporate spatial attention switching (for a review see M c Dowd & Shaw, 2000). It may be that speech perception in auditory spatial displays follows the pattern of age-dependent results found in vision for faster, reflexive behaviors.

59 Introduction Chapter 4 The Time Course of Auditory Spatial Attention Compared to our understanding of visual attention, relatively little is known about the processes that contribute to the ability to allocate auditory attention to a position in space. For example, there is a substantial body of work exploring the time course of visual spatial attention (for a review see Egeth & Yantis, 1997), but there are very few studies exploring the time course of auditory spatial attention. On the one hand, the gap in our understanding regarding the time course of auditory spatial attention is not surprising because the conditions under which spatial cueing yields robust effects on listening abilities were only recently delineated (e.g., Arbogast & Kidd, 2000; Kidd et al., 2005). On the other hand, this gap is surprising given the long-standing observation that successful listening in noisy environments requires the rapid deployment of auditory attention to whatever sound source captures one s interest (Cherry, 1953). For example, when communicating over dinner with colleagues at a noisy restaurant, listeners must frequently switch their focus of attention from one talker to the next and time is likely to contribute to performance for several reasons: time may be required to switch attention (Shinn-Cunningham & Best, 2008); additional time could provide an opportunity to further process information (Barrouillet, Bernardin, & Camos, 2004); and/or time may facilitate the formation of auditory objects (Best et al., 2008). The primary goal of the current study was to explore the effect of time on wordidentification accuracy in a multi-talker, multi-spatial listening environment in which the degree of uncertainty about the spatial location of a target was varied. Using the cueing paradigm, the contribution arising from time was investigated by varying the interval between a callsign cue and the key words when a target sentence was presented concurrently with two competing sentences. In the original sentences, two words intervened between the earlier callsign cue and the following key words. In edited versions of the original sentences, the two intervening words were replaced by silent pauses of varying durations. The second goal of the study was to investigate the influence of age on the time course of attention switching. Age-related ability to allocate attention in a complex listening environment

60 46 was of interest because older adults, even those with little or no clinically significant hearing loss, often report difficulty following conversation in multi-talker situations (CHABA, 1988) and age-related cognitive changes in speed of processing, working memory, and attention may contribute to their difficulties. Importantly, slowing of a wide range of perceptual and cognitive processes occurs with age such that older adults may suffer disproportionately if attention switching requires rapid processing of information (Salthouse, 1996; Verhaeghen & Cerella, 2002). In addition, it has been demonstrated that working memory performance is impaired both in the presence of distracters (i.e., information that is irrelevant and should be ignored) and interrupting information (i.e., intervening information that interrupts the allocation of cognitive resources) (Clapp & Gazzaley, 2010). In studies of visual attention, a common finding is the failure to observe age-related cueing effects (Folk & Hoyer, 1992; Greenwood & Parasuraman, 1994; Hartley, Kieley, & McKenzie, 1992; Hartley, Kieley, & Slabach, 1990). A notable exception is Madden, Connelly, and Pierce (1994) who observed age-related cueing effects favoring younger adults in the presence of visual distractor information. To the extent that similar processes serve visual and auditory spatial attention, one possibility is that age-related differences in the effect of cueing on auditory attention may be more readily observed in attentionally demanding conditions. Indeed, the one study that has directly explored this issue in auditory attention failed to observe agerelated performance differences (Singh, Pichora-Fuller, & Schneider, 2008); however, one limitation of that study was that the cueing effects were tested in a fashion that did not systematically manipulate the time course of attention switching, and the relatively long time delay between the cue and target may have been sufficient to accommodate the possibly slower switching capacities of older compared with younger adults. Hence, in the current study, we tested younger and older adults with clinically normal audiometric thresholds up to 4 khz to determine if there are age-related differences in the time course of auditory spatial attention. By using sentence materials as both the target and distractor information, we hoped to gain insight into how the availability of time and the continuity/interruption of speech affects spoken language understanding in environments typical of everyday conversational interaction. The mechanisms that contribute to auditory spatial attention and the role of time in its deployment are poorly understood. Attending to an auditory object at a location can be

61 47 influenced by stimulus-driven or cognitively-driven processes (for a discussion see Pichora- Fuller & Singh, 2006). In most multi-participant conversations, different voices originate from distinct locations. Thus, there is the potential for listeners in such situations to allocate attentional resources according to the known spatial locations of the conversationalists. Moreover, because a to-be-attended source at a specified spatial location could be determined based on stimulus cues (e.g., monaural and binaural cues to location and/or voice cues) or on knowledge (e.g., expectations about who will say what), it is of interest to explore the mechanisms by which spatial location enhances selective auditory attention and the ways in which these mechanisms contribute to the processing of spoken information over time. Hence, in addition to our first two goals, a third goal was to better understand the relative contributions of the acoustic and cognitive mechanisms underlying auditory spatial attention. Specifically, if age-related differences in performance are greater in the RSS than SSS conditions then the pattern of findings suggest the importance of bottom-up auditory processing. In contrast, if age-related differences in performance are greater in the SSS than the RSS conditions when the role of auditory cues has been minimized then the pattern of findings suggest the importance of topdown cognitive processing. If the age-related differences are the same in both RSS and SSS conditions then both types of processing are implicated. 4.2 Methods The method used in the baseline condition of the present study was the same as that used in Experiment I. New conditions were tested in which the time between the cue and target were manipulated Participants Eight younger (mean = 22.4 yrs, SD = 4.4 years) and eight older adults (mean = 70.9 yrs, SD = 2.5 years) consented to participate in the study. All participants were native English speakers, in good overall health, and had clinically normal pure-tone air-conduction thresholds ( 25 db HL) from 0.25 to 3 khz bilaterally (see Tables 3 and 4). Nevertheless, there was approximately a 10 db HL difference in thresholds between the two listener groups in the mid and mid-low frequencies. All of the participants had participated in prior experiments concerning communication and aging. Two of the 16 participants (one younger and one older adult) had participated in our prior study (Singh et al., 2008). Participants provided informed consent and

62 48 were paid $10/hour. The experiment was approved by Office of Research Ethics of the University of Toronto. Table 3. Individual and group left- (L) and right-ear (R) audiometric thresholds (in db HL) for the eight younger (Y) adult participants from Experiment II. Frequency (khz) Participant L R L R L R L R L R L R L R Y Y Y Y Y Y Y Y Mean (Y) SD (Y)

63 49 Table 4. Individual and group left- (L) and right-ear (R) audiometric thresholds (in db HL) for the eight older (O) adult participants from Experiment II. Frequency (khz) Participant L R L R L R L R L R L R L R O O O O O O O O Mean (O) SD (O) Stimuli The stimuli consisted of sentences from the CRM corpus spoken by the four male talkers (Bolia et al., 2000). The sentences have the format: Ready (callsign) go to (color) (number) now, with all possible combinations of eight callsigns (Arrow, Baron, Charlie, Eagle, Hopper, Laker, Ringo, Tiger), four colors (red, white, blue, green), and eight numbers (1, 2, 3, 4, 5, 6, 7, 8). The callsign provided the cue to identify the target sentence on a given trial. The color and number were the key words that were scored. In addition to the original CRM sentences, listeners were also presented edited CRM sentences. For the original sentences, the words go to intervened between the callsign cue word and the key words to be reported by the listener. For the edited sentences, the words go to were replaced by a silent pause of 0, 150, or 300 ms. Each of the 1024 sentences was edited manually, with the editing points determined by visual examination of the time waveform. In order to minimize click artefacts, editing points occurred at zero crossing in the soundfiles of the sentence materials. The duration of the words go to

64 50 ranged from just over 225 to just under 300 ms in a random sample of 50 sentences. Thus, 300 ms was selected as the maximum pause duration because it was the longest duration of the words go to in the sample and it approximates the maximum time for listeners to switch attention on trials where the callsign cue in an original sentence was presented at an unlikely location Equipment All testing was conducted in a 3.3 x 3.3 m single-walled sound-attenuating booth. The stimuli were controlled and presented via custom software. The stimuli were routed from a Dell computer to a Tucker-Davis Technologies (TDT) System III to a Harmon/Kardon (model HK3380) amplifier. Sentences were converted to analog (RP2.1) at a sampling rate of khz by a 24-bit D/A converter, attenuated (PA5), conditioned (SA1), and presented over loudspeakers located at approximately the same height as the listener s head when seated. Visual cueing was displayed on a 17-inch touchscreen monitor on a table at a height of 0.46 m in front of the listener. The response choices were also displayed visually, as was feedback regarding whether or not each response was correct. Calibration was performed using a Brüel & Kjær (B&K) sound level meter (type 2260) with a 0.5-inch condenser microphone (type 4189) using the A-weighting equivalent to measure concatenated CRM sentences Procedures The listener s task was to identify the color and number key words in the target sentence that was presented simultaneously with two competing sentences. On a given trial, the target and masker sentences differed with respect to the color, number, callsign, and talker of the sentence. Each of the three sentences presented in a trial appeared to come from one of three possible locations: in front of, to the left, or to the right of the listener. On any given trial, the location of the target sentence was randomly selected from the left, center, and right locations, with the constraint that the probability cue was accurate across a block. The callsign word of the target sentence (e.g., Baron) was presented visually on each trial prior to the presentation of the spoken stimuli to cue the listener as to which of the three sentences in the trial was the target sentence. In addition to the callsign cue, one of four different probability specifications ( , , , ) was presented visually prior to a block of trials to indicate the likelihood of the targets being presented from the left, center, and right spatial locations. For example, indicated that the target would be presented from both the left and right spatial location for

65 51 10% of the trials and from the center location for 80% of the trials. Hence, indicated that the target would always be presented from the center location and indicated that the location of the target would be random. Both the callsign cue and probability specification were provided 1 s before each trial in the block and remained on screen throughout the trial. Both the correct color and number were required for a correct response. Feedback (correct or incorrect) was provided after every trial and summary scores were presented at the end of each block. At the start of each visit, participants completed practice trials. For each of the 4 sentence conditions (original and edited sentences containing 0, 150, and 300 ms pauses), the three sentences presented on any given trial were randomly selected with the condition that no two sentences could share the same talker, callsign, color, and number. For each of the four sentence conditions, listeners were presented with target and distractor sentences that were unedited or edited similarly. All sentences were presented at 60 dba. Participants were instructed to face directly ahead for the duration of the stimulus. Two presentation methods were used that differed in the availability of acoustic cues serving the perception of spatial location. For the first method, RSS, the target sentence was presented from one loudspeaker and the two competing sentences were presented from two different loudspeakers (one sentence per loudspeaker). Each loudspeaker was located 1.83 m from the participant s head at 0 and ±54 azimuth in the horizontal plane. The second method, SSS, used the precedence effect so that each of the three sentences was perceived to appear at one of three locations: in front, to the left, or to the right of the listener. The precedence effect was achieved by presenting all three sentences from two different loudspeakers and manipulating inter-loudspeaker time delays (following Freyman et al., 1999). For the SSS condition, each loudspeaker was located 2.08 m from the listener s head at ±45 azimuth. An inter-loudspeaker lead-lag time of 3 ms were used to simulate the left and right locations. For example, an utterance appeared to come from the left spatial location when the signal from the left loudspeaker lead the signal from the right loudspeaker by 3 ms. The center location was simulated by presenting the sentences simultaneously from both loudspeakers. Importantly, perceived spatial separation was achieved while minimizing the contribution of some of the monaural and binaural acoustical cues that were available in RSS presentations. Specifically, for the RSS presentation multiple interaural cues were available and because of the

66 52 influence of head shadow effects, the average target-to-masker ratio was different at the left and right ear. For the SSS presentations, because all three sentences were presented from both the left and right loudspeakers, the resulting interaural level differences resulting from head shadow were minimal and the target-to-masker ratio at each ear was the same. In addition to differences in interaural level difference between the two presentation methods, there were other differences between the spectra of signals presented via RSS and SSS conditions, such as comb-filtering effects and differences in perceived spatial width (for a detailed discussion of the difference between RSS and SSS conditions, see Experiment I, pages 22-23) Design Word-identification accuracy was measured using a 2 (presentation method: RSS and SSS) x 4 (target location certainty: , , , and ) x 4 (sentence: original and edited with 0, 150, and 300 ms pauses) design. All participants completed every condition. The order of testing was counterbalanced, with half of the younger and older participants starting with the RSS presentation method and the remainder starting with the SSS presentation method. For each presentation method, there were eight testing sets with four sets typically completed in a 2-hour visit. The testing order of the 8 sets was as follows: edited sentences with 0 ms, 150 ms, and 300 ms pauses and then the original sentences for half the listeners, with the order reversed for the other half of the listeners. Participants completed 2 sets for each of the sentence conditions. Each set consisted of four blocks of 30 trials. The four probability specifications were assigned randomly to a given block without replacement so that a different target certainty condition was assigned for each block. Each participant completed 60 trials for each of the 32 conditions. 4.2 Results Before presenting the results from the conditions most relevant to our stated goals, we will briefly compare the results from the baseline condition in the current study to the results from previous studies. The overall results will also be presented before focusing on the findings most pertinent to the goals of the current study.

67 Overall results Figure 5 depicts performance for younger and older adults for targets presented with RSS in the current and in two previous studies (Kidd et al., 2005; Singh et al., 2008). Overall, the results of the present study are in excellent agreement with previous findings. Figure 5. Mean percent correct word-identification scores for younger (left panel) and older (right panel) adults depicted across four location certainty conditions. The data presented are from Singh et al. (2008) (white bars), the current study (black bars), and from Kidd et al. (2005) (grey bars), with error bars representing the standard errors of the mean. In general, examination of the overall results revealed four patterns of interest (see Table 5). The most robust finding was that performance was better under conditions with less compared with more target location uncertainty. Collapsing the data across the two presentation methods, four timing conditions, and two age groups, performance was 80% correct when listeners were most certain about the location of the target and 51% correct under conditions where the location of the target was random. The second pattern, and which was also robust, was that performance was higher in RSS (mean performance = 72% correct) compared with SSS (mean performance was 55% correct) conditions. The third pattern was that performance differences were observed between the edited and unedited sentences; word identification was 67% correct for the unedited sentences and ranged from 61 to 63% correct for the 0 and 300 ms edited sentences, respectively. Finally, the effect of location certainty was more pronounced for stimuli presented in the SSS conditions compared to those presented in the RSS conditions. Whereas performance worsened by 19 percentage points as location certainty dropped from 0.8

68 54 to 0.33 when sentences were presented via SSS, across the same location certainty conditions performance worsened by only 11 percentage points when sentences were presented via RSS. In order to confirm our observations of the data, we conducted a 2 x 2 x 4 x 4 repeated measures ANOVA where age (younger versus older) was a between-subjects variable and presentation method (RSS versus SSS), target location certainty (1.0, 0.8, 0.6, and 0.33), and sentence condition (original and edited sentences with 0, 150, and 300 ms pauses) were withinsubjects variables. We found significant main effects of target location certainty [F(3, 42) = , p < 0.001], presentation method [F(1, 14) = , p < 0.001], and sentence condition [F(3, 42) = 8.30, p < 0.001]. For the main effect of target location certainty, post-hoc Student- Newman Kuels (SNK) analyses of all possible comparisons (i.e., six comparisons were conducted) revealed significant differences between each of the four target location certainty specifications (p < 0.05). We also found a significant two-way interaction of target location certainty and presentation method [F(3, 42) = 9.45, p < 0.001]. For the significant two-way interaction of target location certainty and presentation method, post-hoc Student-Newman Kuels (SNK) analyses of all possible comparisons revealed that whereas in the RSS conditions no significant differences were observed between the 0.8 and 0.6 and the 0.6 and 0.33 probability specifications, significant differences were observed between all four target location probability specifications when sentences were presented in the SSS conditions (p < 0.05). No other significant effects were observed (p > 0.05).

69 55 Table 5. Mean percent correct word-identification scores and standard errors of the mean (SEM) for RSS and SSS presentation conditions, presented for the four location certainty (1.0, 0.8, 0.6, and 0.33) and four sentence (original and 0, 150, and 300 ms) conditions for younger and older listeners in Experiment II. Location certainty Mean SEM Mean SEM Mean SEM Mean SEM Younger RSS 0 ms ms ms Original SSS 0 ms ms ms Original Older RSS 0 ms ms ms Original SSS 0 ms ms ms Original

70 Spatial listening expectations: Overall results Because the main goal of this experiment was to investigate the possible contribution arising from time when listeners switch auditory spatial attention, this section of the presentation of the results concentrates on and organizes the relevant subset of conditions. First, we focus on the intermediate (0.8 and 0.6) location certainty conditions where both likely and unlikely trials occurred. Similar to Experiment I, a comparable pattern of performance was observed for the 0.8 and 0.6 location certainty conditions, and thus for ease of presentation, the results across these two conditions have been collapsed. The results from this organization of the data are shown in Table 3. A 2 (likely versus unlikely) x 2 (younger versus older) x 2 (RSS versus SSS) x 4 (Original and 0, 150, and 300 ms sentences) repeated measures ANOVA was conducted where spatial listening expectations, presentation method, and sentence condition were within-subjects variables and age was a between-subjects variable. This analysis revealed a significant main effect of spatial listening expectations, where performance was better when targets were presented from likely rather than unlikely listening locations [F(1, 14) = , p < 0.001]. We also observed a significant main effect of sentence condition [F(3, 42) = 7.68, p < 0.001]. Furthermore, we found a significant two-way interaction of spatial listening expectations and sentence condition [F(3, 42) = 3.27, p < 0.01], which suggests that performance at the likely and unlikely listening locations was dependent on the type of sentence that was presented. Post-hoc SNK testing revealed that when targets were presented from the likely listening location, performance was better when listening to the original compared with the edited sentences (p < 0.05). When targets were presented at the unlikely listening location, however, post-hoc SNK testing revealed that performance was better when listening to the original or 300 ms edited sentences compared with the 0 or 150 ms edited sentences. In other words, whereas performance at the likely listening location was better when participants were presented continuous (i.e., original sentences) compared with interrupted (i.e., 0, 150, and 300 ms sentences) speech, performance at the unlikely listening location was better when participants were presented sentences containing longer duration intervals with either the words go to or 300 ms silent pauses compared with shorter duration intervals with 0 and 150 ms silent pauses.

71 57 Although we failed to observe a main of effect of age, we did observe a significant threeway interaction of age, presentation method, and spatial listening expectations [F(1,14) = 5.15, p < 0.05]. Post-hoc SNK testing of all possible comparisons revealed that whereas both age groups experienced a similar cost to performance when listening to sentences presented with SSS compared with RSS when targets when targets were presented from the unlikely listening location, older compared with younger adults experienced a slightly larger cost to performance when listening to sentences presented with SSS compared with RSS when targets were presented from the likely listening location (p < 0.05). There was also a significant main effect of presentation method, where performance was better when listening to RSS compared with SSS presentations [F(1, 14) = , p < 0.001]. In addition to this main effect, we observed significant two-way interactions of presentation method and sentence condition [F(3, 42) = 3.27, p < 0.01] and presentation method and spatial listening expectations [F(1, 14) = 17.24, p < 0.01]. Finally, we observed a significant three-way interaction of presentation method, sentence condition, and spatial listening expectations [F(3, 42) = 2.86, p < 0.05]. No other significant effects were observed (p > 0.05). In order to better understand the patterns of interactions and the influence of presentation method on performance, we conducted an additional 2 (spatial listening expectations) x 2 (age) x 4 (sentence condition) repeated-measures ANOVA for both the SSS and RSS conditions. For the SSS condition, the repeated-measures ANOVA revealed a significant main effect of spatial listening expectations [F(1, 14) = , p < 0.001], with better performance observed when targets were presented from the likely compared with unlikely listening location. We also observed a significant main effect of sentence condition [F(3, 42) = 8.14, p < 0.001], and posthoc SNK testing of all possible combinations indicated that performance was better when listening to the unedited compared with edited sentences (p < 0.05). No other significant effects were observed for the repeated-measures ANOVA conducted for the SSS conditions (p > 0.05). For the RSS condition, the repeated-measures ANOVA also revealed a significant main effect of spatial listening expectation [F(1, 14) = 77.46, p < 0.001], with better performance observed when targets were presented from the likely compared with unlikely listening location, and a significant main effect of sentence condition [F(3, 42) = 4.49, p < 0.01]. Post-hoc SNK testing of all possible combinations also indicated that performance was better when listening to

72 58 the unedited compared with edited sentences (p < 0.05). In addition, we also found a significant two-way interaction of spatial listening expectations and sentence condition [F(3, 42) = 4.10, p < 0.05]. No other significant effects were observed (p > 0.05) Spatial listening expectations: Effect of pause duration In order to better understand the interaction of sentence condition and spatial listening expectations and to more directly examine how performance was affected by the factor of primary interest, namely the possible benefit to word identification conferred by time as indexed by a comparison of shorter and longer pause durations between the callsign cue and the key words, we conducted three additional 2 (spatial listening expectations) x 2 (age) repeatedmeasures ANOVA. The three possible difference scores (i.e., 150 minus 0, 300 minus 150, and 300 minus 0 ms) served as the dependent variable for each of the repeated-measures ANOVAs, respectively. For the analysis comparing the difference scores between the 150 and 0 ms sentences, we failed to observe any significant effects (p > 0.05). In contrast, significant main effects of spatial listening expectations were observed when comparing the difference scores between the sentences containing pauses durations of 300 and 150 ms [F(1, 14) = 11.16, p < 0.001] and 300 and 0 ms [F(1, 14) = 7.36, p < 0.05], whereby performance was significantly better when listening to longer pause durations when targets were presented from the unlikely compared with likely listening location (see Figure 6). No other significant effects were observed (p > 0.05) Spatial listening expectations: Effect of filled and unfilled gaps The current study also examined the effects of sentence continuity on spoken language understanding for younger and older adults (see Table 6) by comparing performance when participants listened to unedited CRM sentences which contained the words go to to performance when they listened to edited sentences in which the words go to were replaced with a 300 ms silent pause that was about the same duration as the deleted words. Thus, in order to better understand the interaction of sentence condition and spatial listening expectations in conditions with RSS and to examine the influence of sentence continuity on word-identification

73 59 Figure 6. Mean percent correct word-identification difference scores for RSS presentation conditions, depicted for the likely and unlikely listening locations for younger and older adults. Differences in word-identification scores between four pairs of sentence conditions, 150 and 0 ms, 300 and 150 ms, 300 and 0 ms, and original and 300-ms, are shown in white, grey, striped, and black bars, respectively. performance, we conducted an additional 2 (spatial listening expectations) x 2 (age) repeatedmeasures ANOVAs, where the difference score between the original and 300 ms edited sentence was the dependent variable. The key result from the analysis comparing unedited (i.e., continuous) and similar duration but edited sentences (i.e., 300 ms pauses) was that the effect of listening to continuous compared with interrupted speech was minimal when targets were presented from the unlikely listening location (i.e., -2 percentage points), but significantly more beneficial when targets were presented from the likely listening location (i.e., 6 percentage points). The pattern was confirmed statistically by a significant main effect of spatial listening expectations [F(1, 14) = 6.33, p < 0.05].

74 60 Table 6. Mean percent correct word-identification scores and standard errors of the mean (SEM) for RSS and SSS presentation conditions, presented for the four sentence (original and 0, 150, and 300 ms) conditions at the likely and unlikely listening location for younger and older adults in Experiment II. Data are collapsed across location certainties 0.8 and 0.6. Likely Unlikely Group Presentation Sentence Mean SEM Mean SEM method condition Younger RSS 0 ms ms ms Original SSS 0 ms ms ms Original Older RSS 0 ms ms ms Original SSS 0 ms ms ms Original Discussion The primary goal of the current study was to investigate the time course of auditory spatial attention by exploring the possible benefit to word identification conferred by increasing pause duration in a complex multi-talker situation where listeners were required to switch auditory attention between different spatial locations. A related goal was to also consider the role of interruption on auditory spatial attention by comparing continuous versus interrupted speech. The second goal of the study was to determine if there are age-related differences in the ability to rapidly refocus spatial attention toward a target that is presented from an unexpected spatial

75 61 location. The third goal was to investigate the relative contributions of the acoustic and cognitive mechanisms underlying auditory spatial attention by exploring the extent to which the possible contribution of time is influenced by the physical properties of the stimulus when location is presented with actual or simulated spatial separation. Each of these goals is discussed in turn The time course of auditory spatial attention The pattern of findings suggests that the availability of additional time from increasing pause duration facilitates word identification in a multi-talker multi-spatial listening situation (see Figure 2). In conditions of RSS when targets were presented from unlikely compared with likely listening locations, whereas increasing the pause duration from 0 to 150 ms resulted in no benefit to performance (performance differed by < 1 percentage point), increasing the pause duration from 0 to 300 ms resulted in a benefit to performance of 9 percentage points. Two caveats are worth mentioning. First, the benefit to word identification from increasing pause duration beyond 150 ms was modest, albeit significantly so. Nevertheless, under different listening conditions the effect of pause duration could be even more robust; for example, had performance been measured systematically for different target-to-masker ratios rather than fixing the presentation level of each sentence at 65 dba, depending on the slope of the psychometric function, the benefit bestowed by additional time may have been considerably larger. The second caveat is that the benefit of increasing pause duration was observed under particular experimental conditions, namely when targets were presented from unlikely, but not when they were presented from likely spatial locations, and when sentences were presented with real and not simulated spatial separation. Why might the availability of more time enhance word-identification performance for targets presented from unlikely listening locations and not when target sentences are presented from the likely listening location? One possibility is that it takes time for listeners to switch the focus of auditory spatial attention from one location to another. This interpretation is consistent with previous research on auditory spatial attention showing that there is benefit from cueing when targets are presented from likely compared with unlikely listening locations (Arbogast & Kidd, 2000; Mondor & Zatorre, 1995; Spence & Driver, 1994; Singh et al., 2008). This interpretation is also consistent with the overall pattern of results in the current study, namely that performance was enhanced when targets were presented from an unlikely location if more

76 62 time was available, but that the availability of longer pause durations did not enhance performance when targets were presented from the likely location. Interestingly, research on spatial attention switching in vision suggests that the processes activated to disengage attention from one location and to refocus it on another location require from ms (Colgate, Hoffman, & Eriksen, 1973; Logan, 2005; Tsal, 1983). To the extent that time is required for eye movements, it is difficult to suggest an exact parallel in the auditory system, although it may take time for the efferent system to optimize processing at the level of the outer hair cells or at higher levels in the auditory system (Zhu et al., 2007). Thus, if attention switching operates similarly in vision and audition, then it would not be surprising to expect enhanced auditory spatial attention switching for durations up to 300 ms between a cue and a key word. One of the patterns observed from the analysis of the overall results was that performance was better in conditions where participants were presented unedited CRM materials containing the words go to compared with edited sentence materials. This difference was present even under conditions where a similar duration silent interval was inserted to replace the words go to (i.e., the sentences containing the 300 ms pause). This suggests that, despite having a similar duration over which to process each of the sentences, there also appear to be benefits to word identification associated with listening to continuous ongoing speech streams compared to when listening to interrupted speech (for a discussion of the effect of continuity see Bregman, 1990). In other words, this pattern of results suggests that there is a cost associated with disrupting sentence continuity. This cost remained even though the duration of the intervening time between the callsign cue and the target words in the edited sentences containing a 300-ms pause approximated the longest duration of the words go to in the original sentences. There are, however, two important clarifications that deserve mention. First, when targets were presented from the unlikely listening location, although word identification was better when listening to longer compared with shorter pause durations, the benefit bestowed by sentence continuity was minimal (see Figure 2), thus suggesting that shifting of auditory spatial attention takes time and disrupts the benefit from stream continuity. In contrast, in listening conditions where the location of the target was held constant at the likely location, although the available of longer duration pauses did not benefit word identification, performance improved when listening to a continuous compared with interrupted stream of information (see Figure 2). This suggests that the stream s

77 63 continuity, but not necessarily increased time, benefits the sustained allocation of auditory spatial attention, at least over the time intervals tested The influence of age on the time course of attention Age-related differences in word-identification performance from listening to sentences containing different pause durations were explored by comparing the performance of younger and older listeners. In conditions of RSS, performance differences between younger and older adults were not observed, even when the time between the callsign cue and the target words was reduced. From one perspective, this pattern of findings is not surprising given that several studies have demonstrated that younger and older adults exhibit similar patterns of performance when they are shifting attention in a complex visual environment (e.g., Folk & Hoyer, 1992; Greenwood & Parasurman, 1984; Hartley et al., 1992; Hartley et al., 1990). From another perspective, however, this is somewhat surprising given that age-related performance deficits have previously been observed in more complex visual environments containing distractor information (Madden et al., 1994). There are several possible explanations for our failure to observe age-related performance differences. First, the effects of the time course of auditory attention switching are important for older listeners, but they are equally important for younger listeners. Second, differences in performance between younger and older adults may emerge for other populations of older adults (e.g., adults > 80 years of age). Third, the listening conditions in which participants were tested were not sufficiently challenging and age-related differences may emerge if the degree of challenge were increased (e.g., at less favorable signal-to-noise ratios) Cognitive and acoustic mechanisms underlying auditory spatial attention The final goal of the experiment was to take into consideration the relative contributions of the acoustic and cognitive cues underlying auditory spatial attention by exploring differences in word-identification performance when sentences were presented via RSS and SSS. We found that whereas both groups experienced a similar cost to word-identification performance when listening to sentences presented with SSS compared with RSS when targets were presented from the unlikely listening location, older compared with younger adults exhibited a slightly larger cost when listening to sentences presented with SSS compared with RSS when targets were presented from the likely listening location. This suggests that older adults may have exhibited a

78 64 greater reliance on the availability of natural binaural cues than younger adults, but performance differences were small. At the likely listening location, whereas the cost to performance when listening to sentences presented with SSS compared with RSS was 8 percentage points for younger adults, in the same condition, older adult performance worsened by 14 percentage points. Nevertheless, this pattern of results underscores the importance of auditory processing and highlights that the availability of richer acoustical cues facilitates the allocation of auditory attention for both age groups Findings compared to the literature In the current and in previous studies (Kidd et al., 2005; Singh et al., 2008), similar baseline conditions were presented whereby participants heard unedited CRM sentences. Because a similar baseline condition was tested across studies, there is an opportunity to consider the reliability of word-identification performance by comparing findings across studies (see Figure 1). On the whole, the performance we observed in the current study was an excellent replication of previous findings reported in Kidd et al. and Singh et al. For the 8 conditions (4 target location certainties x 2 age groups) that were present in the current study and in Singh et al., performance differences did not exceed 5 percentage points except in one case (for younger adults where location certainty was 0.8) where performance was 7 percentage points worse in the current study. Overall, for each of the four probability conditions, the mean difference in performance between the two studies was < 1 percentage point (SD = 3.0 percentage points).

79 65 Chapter 5 The Effect of Single vs. Multiple Switches of Spatial Attention 5.1 Introduction The frontal lobes have an important role in performing complex tasks of executive function, such as those that involve updating working memory, inhibiting dominant responses, and shifting attentional resources (Miyake, Friedman, Emerson, Witzki, Howerter, & Wager, 2000). In cognitive aging research, it has been observed that aging is associated with declines in executive function (e.g., Braver & Barch, 2002; Kramer, Humphrey, Larish, Logan, & Strayer, 1994). One account that has not been explored in order to explain the listening difficulties experienced by older adults is that older compared with younger listeners may have more difficulty on tasks of executive function related to disengaging attentional resources after misdirecting spatial attention (the disengaging attention hypothesis). This idea is rooted in the notion that older adults exhibit greater perseverative behavior than younger adults (for a review see West, 1996). As attention shifts from one talker to the next, often there is uncertainty as to where a listener should direct attentional resources. The benefits of directing auditory spatial attention have been demonstrated in cueing experiments in which the typical finding is that intelligibility scores are higher for targets presented from likely compared to unlikely spatial locations (Arbogast & Kidd, 2000; Kidd, Arbogast, Mason, & Gallun, 2005; Mondor & Zatorre, 1995; Spence & Driver, 1994). However, in Experiments I and II, age-related deficits in performance when participants are required to rapidly shift attention from likely to unlikely locations were not observed, so it seems that relatively simple attention switching is not a problem for older listeners. In previous attempts to study how aging influences the misallocation of attention in complex environments, a number of studies of visual spatial attention have compared response performance for targets in the presence of competitor cues. Typically, it is observed that older adults exhibit poorer attentional control than younger adults, but that such deficits are observed for some and not other types of competitor stimuli. For example, compared to younger adults, older adults are less able to disengage their attention from exogenous cues, such as peripherally

80 66 presented abrupt onset cues (Juola, Koshino, Warner, McMickell, & Peterson, 2000), or from negatively arousing as opposed to neutral elements within a visual scene (Kensinger, Gutchess, & Schacter, 2007). There is also evidence to suggest that aging affects the influence of higher-level processes on the misallocation of attention. For example, in conditions where participants are instructed to process non-target information, Madden, Connelly, and Pierce (1994) observed age-related deficits in the shifting of focused attention. In this study, stimuli were visually presented at one of eight locations evenly distributed in a circle, and younger and older participants performed a cued choice response task where cue-validity was manipulated. Although no age-related differences were observed in conditions where the location of the spatial cue was accurate or displayed in a neutral fashion (i.e., at the center location), age-related deficits were observed when cues were presented in locations diametrically opposite from the location in which the target was presented, a finding that suggests age-related deficits when older adults disengage attention from a cued location in order to shift attention to a target presented at a different location. Similar age-related performance deficits have also been reported in the literature on visuospatial search tasks that require the momentary processing of object features, such as that involving conjunction search where observers repeatedly engage and disengage attention between multiple items presented in an array (Greenwood & Parasuraman, 1994; 1999). The research just described investigating the relationship between aging and attention switching in vision may have important implications for research on auditory aging. In complex listening environments, such as a dinner party at a restaurant, a listener s motivations and expectations contribute to how attention is allocated. One possibility is that in listening situations in which the spatial location of the target is uncertain, attention is frequently misdirected. Thus, listeners engage in attempts to repair by reallocating listening resources to different spatial locations. In cases where attention is misdirected and listeners must reallocate spatial listening resources, listeners may process and disengage attentional resources from an intermediate spatial position. Nevertheless, the literature exploring the role of aging on auditory spatial attention has focused on the role of relatively simple shifts of attention and placed less emphasis on exploring how more challenging shifts of attention influence perception. Specifically, to our knowledge, no research in audition has explored the costs of momentarily misdirecting spatial attention.

81 Current research For the current study, the effect of directing and misdirecting attentional resources was investigated using a cueing paradigm. Using a multi-talker, multi-spatial listening situation in which the location of the target was fixed or varied probabilistically, the effect of aging was explored by comparing a group of younger and older listeners with good hearing. In total, we compared word-identification performance on trials requiring no switch, a single switch, or multiple switches of auditory spatial attention. In no-switch trials, the target was presented from the loudspeaker with the highest a priori target location probability specification (i.e., the likely listening location). In single switch trials, the target was presented from a loudspeaker with the lowest a priori target location probability specification (i.e., one of the two unlikely listening locations) and listeners were required to switch attention from the likely to the cued unlikely listening location. In multiple switch trials, when the callsign was presented from an unlikely listening location, the listener s task was to report key words presented from the other unlikely location. Hence, by comparing the accuracy of word identification for key words presented at unlikely listening locations when single or multiple switches were required, we set out to determine the cost to word-identification performance when listeners momentarily misdirected spatial attention. Based on the literature, one would expect to observe better performance on trials requiring simple compared with complex switches (Eriksen & Yeh, 1985), better performance for younger compared with older adults (CHABA, 1988), no effect of performing a simple switch of attention (Singh et al., 2008), and an age x switch complexity interaction favoring younger adults (Madden et al., 1994). 5.3 Methods Participants Eight younger (M age = 21.1 years, SD = 2.0) and eight older (M age = 69.3 years, SD = 2.6) adults participated in the study. The participants were recruited from the local university and community. Criteria for eligibility were clinically normal pure-tone air-conduction hearing thresholds through the speech range (< 25 db HL from 0.25 to 3 khz bilaterally; see Table 7), self-reported good health and English as a first language. Participants were excluded if they selfreported a history of neurological impairment or stroke. Participants were monetarily

82 68 compensated to participate in the study at a rate of $10 per hour, and the research was approved by the University of Toronto Research Ethics Board. Table 7. Individual and group left- (L) and right-ear (R) audiometric thresholds (in db HL) for the eight younger (Y) adult participants from Experiment III. Frequency (khz) Participant L R L R L R L R L R L R L R Y Y Y Y Y Y Y Y Mean (Y) SD (Y) Stimuli The stimuli consisted of edited sentences from the CRM corpus spoken by the four male talkers (Bolia et al., 2000). The original sentences have the format: Ready (callsign) go to (color) (number) now., with all possible combinations of eight callsigns (Arrow, Baron, Charlie, Eagle, Hopper, Laker, Ringo, Tiger) which provided the cue to identify the target sentence on a given trial, four colors (red, white, blue, green), and eight numbers (1, 2, 3, 4, 5, 6, 7, 8). For all sentences, the words go to were replaced by a silent pause of 300 ms. Table 8. Individual and group left- (L) and right-ear (R) audiometric thresholds (in db HL) for the eight older (O) adult participants from Experiment III. Frequency (khz)

83 69 Participant L R L R L R L R L R L R L R O O O O O O O O Mean (O) SD (O) Equipment All testing was conducted in a 3.3 x 3.3 m single-walled sound-attenuating Industrial Acoustics Company booth. The stimuli were controlled and presented using custom software. The stimuli were routed from a Dell computer to a Tucker-Davis Technologies (TDT) System III to a Harmon/Kardon (model HK3380) amplifier. Sentences were converted to analog (RP2.1) at a sampling rate of khz by a 24-bit D/A converter, attenuated (PA5), conditioned (SA1), and presented over three loudspeakers located at approximately the same height as the listener s head when seated. Each loudspeaker was located 1.83 m from the participant s head at 0 and ±54 azimuth in the horizontal plane. Visual cueing was displayed on a 17-inch touchscreen monitor located on a table at a height of 0.46 m in front of the listener. Prior to each trial, the display changed to specify the callsign word that provided the cue to identify which sentence was the target sentence. Prior to each block of trials, the display changed to specify the probability of the target being presented at each of the three possible locations. The response choices were also displayed visually, as was feedback regarding whether or not each response was correct. Calibration was performed using a Brüel & Kjær (B&K) sound level meter (type 2260) with a 0.5-inch condenser microphone (type 4189) using the A-weighting equivalent to measure

84 70 concatenated CRM sentences. During calibration, concatenated sentences from the CRM corpus were presented at a level such that each loudspeaker, playing alone, would produce an average sound pressure of 60 dba at the location corresponding to the center of the listener s head Procedure On each trial, three sentences were presented from one of three possible locations: in front of, to the left, or to the right of the listener. Participants were instructed to face directly ahead for the duration of the stimulus. Two types of a priori cues were used in the study, callsign cues and location cues. Whereas location cues indicated the probability of the target being presented from the left, center, or right loudspeaker, the role of callsign cues depended on whether the participant was completing simple or complex conditions of the study. The location of the cued sentence was randomly selected from the left, center, and right locations, with the constraint that the probability cue was accurate across a block of 30 trials. On any given trial, the callsign cue of one of the sentences (e.g., Baron) was presented visually prior to the presentation of the spoken stimuli. In addition to the callsign cue, one of four different probability specifications ( , , , ) was presented visually prior to a block of trials to indicate the likelihood of the targets being presented from the left, center, and right loudspeakers. For example, indicated that the target would be presented from both the left and right spatial location for 10% of the trials and from the center location for 80% of the trials. Hence, indicated that the target would always be presented from the center location and indicated that the location of the target would be random. Both the callsign cue and probability specification were provided one second before each trial. Two sets of instructions were provided to the participants. For half of the trials (i.e., the simple switch condition), the listener s task was to identify the color and number key words from the cued sentence, regardless of the location where the callsign cue was presented. Thus, if the callsign cue was presented from 0 (i.e., the likely listening location), the task was to report the key words presented from the center loudspeaker, and if the cued callsign was presented from ±54 (i.e., one of the two unlikely listening locations), the task was to report the key words from the same location as the callsign cue. For the other half of the trials (i.e., the complex switch condition), the listener s task differed depending on the location of the cued sentence. When the cued sentence was presented from 0, the listener s task was identical to that in the simple

85 71 condition, namely to report the color and number key words from the sentence cued by the callsign. When the callsign cue was presented from either ±54, the task was to report the color and number from the sentence presented from the other unlikely listening location (i.e., the side location where the cued sentence was not presented). For example, when the cued sentence was presented from the loudspeaker located at +54 (right), the listener s task was to report the color and number from the sentence that was presented from the loudspeaker located at -54 (left). Both the correct color and number were required for a response to be scored as correct. Feedback (correct or incorrect) was provided after every trial and summary scores were presented at the end of each block. At the start of each visit, participants completed practice trials for a minimum of 30 minutes Design Word-identification accuracy was measured using a 2 (switch condition: simple and complex) x 4 (target location certainty: , , , and ) x 2 (age: younger and older) design. All participants completed every condition. For all conditions, three sentences were randomly drawn from the corpus of CRM sentences with the constraint that the callsign, color, and number associated with each sentence were all different from each other. The three sentences were presented simultaneously on a given trial. The order of testing was counterbalanced, with half of the younger and older participants starting with the simple switch condition and the remainder starting with the complex switch condition. Moreover, a repeatedmeasures ANOVA comparing order effects (completing the simple switch condition first vs. completing the complex switch condition first) failed to reveal a significant difference (p > 0.05) for either the younger or older adult group. For each type of switch, there were seven testing sets, usually with two sets completed in a 2-hour visit. Within each set, participants completed four blocks, each containing 30 trials. Each block was randomly assigned one of the four probability specifications without replacement so that a different probability specification was assigned for each block within a set. Each participant completed a total of 1680 experimental trials, with 210 trials completed for each level of probability specification for both the simple and complex switch conditions.

86 Results Overall results Mean word-identification performance is depicted in Figure 7. Overall, we observed three main patterns of interest: (i) listeners performed better in simple compared to complex switch conditions; (ii) performance was better in conditions with more compared to those with less location certainty; moreover, the benefit of increased location certainty was greater in complex compared with simple switch conditions; (iii) younger listeners outperformed older listeners; however, whereas similar between-group performance was observed for simple switch conditions, an age-related difference favoring younger listeners was observed for complex switch conditions. In order to confirm these descriptions, a series of statistical analyses was conducted. A repeated-measures ANOVA was conducted with age (younger vs. older) as a between-subjects variable and switch condition (simple vs. complex) and target location certainty (1.0, 0.8, 0.6, 0.33) as within-subjects variables. The analysis revealed significant main effects of age [F(1, 14) = 5.51, p < 0.05], switch condition [F(1, 14) = 7.75, p < 0.05], and target location certainty [F(3, 42) = 61.62, p < 0.001]. We also observed significant two-way interactions between switch condition and age [F(1, 14) = 5.23, p < 0.05], and switch condition and target location certainty [F(3, 42) = 6.17, p < 0.01]. Post-hoc SNK testing of all possible comparisons was conducted to further examine the interactions and revealed that similar between-group performance was observed for simple switches, but that younger listeners significantly outperformed older listeners when complex switching was required (p < 0.05). Finally, post-hoc SNK testing of all possible comparisons revealed that in conditions requiring simple switches of attention, performance was not significantly different when target location certainty was 0.8, 0.6, or 0.33; however, in conditions requiring complex switches, relative to when target location certainty was 0.8, performance was significantly poorer at 0.6 and 0.33 (p < 0.05). No other significant effects were observed (p > 0.05).

87 73 Figure 7. Mean percent correct word-identification scores and standard errors of the mean for younger (open circles) and older (filled circles) adults for the four location certainty conditions. Solid lines indicate conditions requiring simple switches and dashed lines indicate conditions requiring complex switches Spatial listening expectations For the purposes of the following analysis, the focus was on the intermediate target location probability specifications (0.8 and 0.6) in order to more directly consider the role of spatial listening expectations (i.e., trials where the target location was either perfectly certain or random were excluded from this analysis). One of the advantages of manipulating the certainty with which cued sentences are presented from 0 is that it is possible to distinguish between likely and unlikely listening locations, and thus, by categorizing the data by probabilistic expectations, it was possible to assess the contribution arising from spatial listening expectations. Furthermore, because a similar pattern of performance was observed for each location certainty condition (0.8 and 0.6), for ease of presentation, the results across these two location certainty conditions have been collapsed. The results from the analysis comparing performance at likely and unlikely listening locations are shown in Figure 8. Overall, the younger participant group outperformed the older

88 74 adult group, performance was better when simple compared to complex switches were required, and performance was better when targets were presented from the likely compared to the unlikely listening locations. When the cued callsign was presented from the likely (i.e., front) listening location, no differences in performance were observed between younger and older listeners. Similarly, in simple conditions when the cued callsign was presented from an unlikely (i.e., side) listening location and the task was to report the target keywords presented from the same location, we observed similar performance between younger and older adults. However, when complex switches were required (the cued callsign was presented from an unlikely (i.e., side) listening location and the task was to report the target keywords presented from the other unlikely listening location), performance was poorer for older compared to younger participants. When targets were presented from the unlikely listening location and listeners performed complex rather than simple switches, performance dropped by 19.0 percentage points for older adults and by 3.6 percentage points for younger adults. A repeated-measures ANOVA was conducted in order to statistically confirm the previously described pattern of results. For this analysis, age (younger vs. older) was a betweensubjects variable and switch condition (simple vs. complex) and spatial listening expectations (likely, unlikely) were within-subjects variables. The analysis revealed significant main effects of age [F(1, 14) = 5.36, p < 0.05], switch condition [F(1, 14) = 21.38, p < 0.001], and spatial listening expectations [F(1, 12) = 28.92, p < 0.001]. We observed a significant two-way interaction between switch condition and spatial listening expectations [F(1, 14) = 13.45, p < 0.01]. Post-hoc SNK testing of all possible comparisons indicated that although no statistical difference was observed when targets were presented from likely listening locations, at the unlikely listening location, performance was significantly better under conditions requiring simple compared with complex switches (p < 0.05). We also observed a significant two-way interaction between spatial listening expectations and age [F(1, 14) = 14.61, p < 0.01]. Post-hoc SNK testing of all possible comparisons revealed that both age groups performed similarly when targets were presented from the likely listening location, but that younger listeners outperformed older listeners when targets were presented from the unlikely listening location (p < 0.05). Finally, we observed a significant three-way interaction between switch complexity, spatial listening expectations, and age [F(1, 14) = 5.56, p < 0.05]. Post-hoc SNK testing of all possible comparisons revealed that when targets were presented from unlikely listening locations,

89 75 although no age-related deficits were observed under conditions requiring simple switching, younger adults outperformed older adults under conditions requiring complex switching (p < 0.05). No other significant effects were observed (p > 0.05). Figure 8. Mean percent correct word-identification scores and standard errors of the mean for younger and older adults in conditions requiring simple and complex switches for targets presented from the likely (white bars) and unlikely (black bars) listening location Error analysis In order to better understand listener errors, and to potentially shed light on the listening strategies adopted by the participants, an analysis of the incorrect responses was conducted. The goal of this analysis was to consider the location of the reported errors and to examine if listeners exhibited a different pattern of errors on trials requiring single and multiple shifts of attention or if there were age-related differences in the pattern of errors. The error analysis examined the critical conditions, namely the intermediate target location certainty conditions (0.8 and 0.6) when targets were presented from unlikely listening locations. A similar pattern of performance was observed and thus for ease of presentation, the results across the 0.8 and0.6 location certainty conditions were collapsed.

90 76 The cumulative proportion of errors for younger and older listeners is presented in Figure 9. The errors for color and number when simple (top) and complex (bottom) switches were required are shown as a function of the location where the incorrect response had been presented. In general, a similar pattern of errors was observed for both groups across both conditions. Most incorrect responses were reports of distractors presented from the center loudspeaker (i.e., the white portion of the bars). This part of the pattern of errors is consistent with Mondor and Zatorre s (1995) account of spatial attention whereby performance declines with increasing distance from a cued spatial location of attentional focus. It also appears that very little guessing occurred in any of the conditions (i.e., the black portion of the bars) insofar as most of the error responses were of color/number options presented in one of the distractor sentences. If listeners adopted a guessing strategy, one would expect 25% (1/4) of the color error responses and 62.5% (5/8) of the number error responses to have been from response options not presented on any given trial, but the cumulative proportion of non-presented color and number errors was always less than 10%. 5.3 Discussion The primary goal of Experiment III was to investigate the cost of misdirecting spatial attention in a multi-talker, multi-spatial listening environment for younger and older listeners. To explore this question, two conditions were tested and listeners were given a different set of instructions in each condition. In conditions requiring a simple switch of attention, when the callsign cue was presented from the unlikely listening location, the task was to report key words from the sentence containing the callsign cue, thus requiring listeners to perform a single shift of attention from the likely to the unlikely listening location. In conditions requiring a complex switch of attention, when the callsign cue was presented from the unlikely listening location, rather than reporting key words from the sentence containing the cued callsign, the task was to report key words from the other unlikely listening location, thus requiring multiple switches of attention. The key finding was that word-identification performance was considerably poorer for older adults compared with younger adults when the cue was presented at an unlikely listening

91 Figure 9. Cumulative proportion of color and number errors by source location for younger and older adults in conditions requiring simple (top) and complex (bottom) switches. White bars indicate the incorrect responses were reports of distractors presented from the center loudspeaker, grey bars indicate the incorrect responses were reports of distractors presented from the incorrect side loudspeaker, and black bars indicate the incorrect responses were reports of tokens that were not presented. 77

92 78 location and multiple switches of attention were required. To the extent that this experimental condition corresponds to real-world listening conditions where listeners momentarily misdirect spatial attentional resources, the results suggest that older compared with younger adults could experience greater costs of misdirecting attention. The results from Experiment III provide support for the account described by Madden et al. (1994) and Greenwood and Parasuraman (1999) who suggest that older compared with younger adults show more difficulty disengaging attentional resources from a cued location when shifting attention to a target at another location (the disengaging attention hypothesis). More generally, this account is consistent with theories of aging that suggest that there are failures of cognitive control on tasks of executive function (Braver & Barch, 2002; Kramer et al., 1994). Two points critical to this interpretation are worth mentioning. First, in the current and in the study of Madden et al. (1994), age-related differences were not observed in conditions where listeners use location-based cues to switch spatial attention from a likely to an unlikely location. This suggests that processes that facilitate attention switching per se are relatively unaffected by aging, and this finding is consistent with a number of studies in both auditory spatial attention (Singh et al., 2008) and visual spatial attention (Folk & Hoyer, 1992; Greenwood & Parasuraman, 1994; Hartley, Kieley, & McKenzie, 1992; Hartley, Kieley, & Slabach, 1990). Second, the disproportionately poorer performance of older compared with younger adults in the current study was only observed in conditions that required repeated engaging and disengaging of attentional resources, a finding that until now has only been reported in the literature on visual spatial attention (e.g., Greenwood & Parasuraman, 1994; 1999; Madden et al., 1994) Alternate explanations of the age x switch complexity interaction In order to conclude that differences in the ability to disengage attention account for the age-related difference in word-identification performance, it is necessary to consider how competing explanations might account for the overall pattern of data. The following six alternative explanations will be considered, as the results may have been influenced by agerelated differences in: (a) audiometric thresholds, (b) speed of attention switching, (c) learning of the experimental task, (d) divided attention, (e) task coordination, or (f) inhibitory processing. One of the factors that seemingly failed to influence the overall pattern of results is the slightly poorer audiometric thresholds of the older compared to the younger group (Tables 7 and

93 79 8). In light of the current study methodology, it is not possible to completely rule out the possibility that poorer hearing may be a source of the difficulty of the older listeners; however, the influence of group-based threshold differences on the age x switch condition interaction is considered to be negligible. It is unclear why threshold differences between the younger and older listeners would result in poorer performance only on trials requiring multiple switches of attention. Presumably, differences attributable to hearing thresholds would also emerge on trials requiring a single switch of attention when targets were presented at the unlikely listening location, but this was not the case. Yet another possible alternative explanation that may explain the pattern of age-related differences on trials requiring multiple shifts of attention is that older adults are slower than younger adults such that they may have difficulty with rapid spatial attention shifting. Time is clearly implicated in this process, because presumably, had listeners been provided an unlimited duration of time to shift spatial listening resources from one side loudspeaker to the other on trials requiring multiple shifts of attention, age-related performance deficits would have been attenuated. The more relevant issue, however, is whether older compared with younger adults exhibit deficits in the ability to rapidly shift auditory spatial attention or deficits in some other component process that also slows. This issue was investigated in Experiment II which examined the time course of attention switching by varying the duration of the time between the callsign cue and the onset of the key words. In this study, word-identification performance was measured in test conditions where filler words (i.e., the words go to ) from the CRM sentence corpus were replaced with silent pauses that were 0, 150, or 300 ms in duration. Although wordidentification performance was poorer at unlikely listening locations when listeners had less compared with more time between the callsign cue and key words, a significant age-related difference in performance was not observed; however, it is possible that age-related differences in the speed of attention switching may be observed with further testing of intermediate timing conditions. Nevertheless, the speed of attention switching hypothesis does not seem to offer an explanation for the present pattern of results, but further testing of this hypothesis is warranted. One of the possible confounds in the current study is that older compared with younger adults may have experienced more difficulty learning the experimental tasks (e.g., Salthouse & Somberg, 1982). In light of the potentially challenging attention switching demands required of

94 80 the participants in the current study, considerable care was taken to ensure that participants were provided with adequate training time. To this end, each participant completed a minimum of 30 minutes of practice and 3.5 hours of testing for each type of switching instruction. Nevertheless, in order to conclude that age-related learning differences did not account for the age-related performance difference for targets presented at the unlikely listening location when listeners made complex switches of attention, it would be useful to consider listener performance across the 7 sets of testing in the simple and complex switching conditions. Figure 10 shows that the performance for younger and older listeners making simple and complex switches of attention for targets appearing at unlikely listening locations was relatively consistent over time. In particular, across all testing sessions, younger adults always outperformed older adults and the performance of older adults was poorer when complex rather than simple switching was required. Across the 7 sets, the mean difference in performance between younger and older adults when they made complex switches of attention was 16.9 percentage points (SD = 9.0; range 7.6 to 34.3).

95 81 Figure 10. Mean percent correct word-identification scores and standard errors of the mean for younger and older adults in conditions requiring simple and complex switches of attention for targets presented from the unlikely listening location. Data are presented by testing session. Rather than adopting a strategy of listening to different sentence streams in a sequential fashion, another possibility is that listeners may have adopted a strategy of listening to multiple streams of information in a concurrent fashion, and that age-related differences in performing multiple switches of attention may arise from age-related differences in the ability to simultaneously following multiple streams of information (i.e., the divided attention hypothesis). The argument for this perspective is that on trials requiring multiple switches of attention listeners adopted a strategy of dividing their attention between two (or possibly three) locations; however, this does not appear to be the case. First, it is unclear why listeners would adopt such a strategy on trials requiring multiple switches of attention and not in conditions requiring a single switch of attention, insofar as a divided attention strategy could be beneficial in both conditions. As shown in Figure 7, the performance for the older compared with younger adults was poorer only in conditions requiring complex switches of attention. Nevertheless, even if one were to assume that listeners adopted a strategy of dividing their attentional resources only in conditions requiring complex switches of attention and not in conditions requiring simple switches, then performance deficits should be observed in the complex switching condition for older compared with younger adults for targets presented at both the likely and unlikely listening location. Based on Figure 8, this was not the case, as both age groups exhibited a similar pattern of performance for targets appearing at the likely listening location, but a different pattern for targets appearing at the unlikely listening location. Furthermore, based on modeling done by Kidd et al. (2005), who employed a similar three-talker multi-spatial listening task in younger adults using the CRM corpus stimuli, it appears that listeners adopt a strategy of focusing attentional resources on the most likely listening location and then switching attention to one of the other locations. For these reasons, the divided attention account for the observed age x switch complexity interaction does not seem to offer a satisfactory explanation. Another possible alternative explanation is that the poorer performance of older compared with younger adults in conditions requiring multiple compared with single switches of attention is due to an age-related reduction in the ability to coordinate multiple task-demands concurrently (i.e., task-switching) (e.g., Kray et al., 2002; Verhaeghen & Cerella, 2002).

96 82 According to this account, in complex compared with simple switch conditions, listeners were required to maintain two mental task sets (i.e., remain focused at the center loudspeaker if the target callsign is presented from the center location and shift to the opposite location if the target is presented from a side location). This alternative explanation does not seem to apply because our tasks in the simple and complex switching conditions only differ in a somewhat superficial manner insofar as the goal of the listener is always to repeat key color and number words. In the current study, listeners were asked to perform a word-identification task with the same closed set of key words as possible alternatives, albeit under conditions varying the extent of attentional disengagement. In contrast, in typical studies of task-switching, participants are asked to maintain two very different mental sets (e.g., when shown a series of digits, a participant is asked to report if the digit is odd or even if the color of the digit is red and to report if the number is larger or smaller than five if the color of the digit is blue). It is also important to consider the possibility that the poorer performance of older compared to younger adults on trials requiring multiple switches of attention arose from agerelated differences in processing information that is no longer relevant (i.e., the inhibitory-deficit hypothesis). From this perspective, when targets callsigns were presented from the left or right loudspeaker, older compared with younger adults should have had more difficulty inhibiting information from a side spatial position before shifting attention to the opposite loudspeaker position. One important distinction, however, is that whereas the disengagement hypothesis assumes a deficit in the ability to uncouple attentional focus from a spatial location, the inhibitory-deficit hypothesis assumes that there are deficits in the ability to delete information from the focus of attention (Lustig, Hasher, & Zacks, 2007). Hence, whereas the disengagement hypothesis remains agnostic on the role assigned to information, the inhibitory deficit hypothesis assumes that information gains access to processes involving attention and/or working memory. To explore whether information from the cued location remained in the focus of attention and/or working memory, we conducted an error analysis to investigate the possibility that if older adults had more difficulty down-regulating information from previously informative spatial positions, then a greater proportion of key word response errors would have been reported from the loudspeaker position where the callsign cue had been presented. Based on the error analysis (Figure 9) there is little evidence to support the inhibitory-deficit hypothesis in that we failed to

97 83 observe an age-related difference in the reporting of key words from the cued loudspeaker in listening conditions requiring complex switches of attention. In conclusion, the key finding from Experiment III was that whereas both age groups performed similarly in conditions requiring a single switch of attention, the performance of older but not younger listeners was reduced when multiple switches of spatial attention were required. Although questions remain unanswered regarding the factor responsible for this age-related difference in performance and potential interactions between some of the possibly contributing factors previously described, the findings suggest that difficulties disengaging attention may explain why older adults perform worse than younger adults in conditions requiring multiple switches of spatial attention.

98 84 Chapter 6 Aggregate Data Analysis In Experiments I, II, and III, data from all participants was measured in similar baseline conditions, namely when sentences were presented using RSS and when the probability of the target being presented from the center location was 1.0, 0.8, 0.6, and It should be noted, however, that there were minor differences in the number of trials collected per participant in each of the experiments. By comparing the patterns of data across the three experiments, the goal was to achieve a better understanding of the pattern and range of performance in listening conditions with target location uncertainty. 6.1 Results Overall data Figures 11 and 12 depict mean word-identification performance for the younger and older adult groups for each of the three experiments as a function of the four target location certainty conditions (the data are the same in both figures, but arranged differently to ease interpretation). Overall, four patterns of interest were observed. First, the most robust effect was that of target location uncertainty. Compared to when the location of the target was fixed (mean = 87%), performance worsened by 22 percentage points when the location of the targets was chosen at random (mean = 65%). Second, there was a modest difference observed between the group of younger and older adult participants, with the group of older adults (mean = 71%) performing more poorly than the group of younger adults (mean = 76%). Third, target location uncertainty had a different effect on older compared with younger adults. In location certainty conditions 1.0, 0.8, 0.6, and 0.33, younger adults outperformed older adults by 4, 3, 6, and 9 percentage points, respectively. Fourth, the effect of presenting targets with more uncertainty about their location was more pronounced in Experiments I and II than in Experiment III (see Figure 13). When the location of the target was no longer fixed but instead selected at random, word-identification scores in Experiments I, II, and III worsened by 27, 26, but only 14 percentage points, respectively. Finally, although the following pattern failed to reach statistical significance, there was also a trend in the data, whereby older adults, compared to younger adults, experienced less

99 85 of a cost to performance from increasing uncertainty about the location of the target in Experiments II and III, but not in Experiment I. Figure 11. Mean percent correct word-identification scores for younger (Y) and older (O) adults in Experiments (Expt) I, II, and III. Data are presented as a function of target location certainty. The patterns described above were confirmed with a 2 x 3 x 4 repeated measures ANOVA where age (younger versus older) and experiment (Experiments I, II, and III) were between-subjects variables and location certainty (1.0, 0.8, 0.6, and 0.33) was a within-subjects variable. We found a significant main effect of target location certainty [F(3, 126) = , p < 0.001], and a significant main effect of age [F(1, 42) = 7.08, p < 0.05]. A significant two-way interaction of age and location certainty was observed [F(3, 126) = 3.26, p < 0.05]. Post-hoc SNK testing of all possible comparisons indicated that the influence of target location uncertainty was less marked for younger compared with older adults (p < 0.05). We also found a significant two-way interaction of location certainty and experiment [F(6, 126) = 7.56, p < 0.001]. Post-hoc SNK testing of all possible comparisons indicated that the effect of target location uncertainty was significantly smaller in Experiment III compared with Experiments I and II (p < 0.05). No other significant effects were observed (p > 0.05). Finally, there was a non-significant three-way interaction of age, target location uncertainty, and experiment [F(6, 126) = 2.08, p = 0.06].

100 Figure 12. Mean percent correct word-identification scores for younger (top) and older (bottom) adults in Experiments (Expt) I, II, and III. Data are presented as a function of target location certainty. 86

101 87 Figure 13. Mean percent correct word-identification scores for Experiments I, II, and III. Data are presented as a function of target location certainty and collapsed across younger and older adults Spatial listening expectations Similar to the analyses conducted in each of the experiments, we conducted an analysis for targets presented from the likely and unlikely listening locations. For this purpose, the analysis focused on the collapsed data from the two intermediate target location uncertainty conditions (0.8 and 0.6). Figure 14 depicts mean word-identification performance for younger and older adults in each of the three experiments as a function of the target being presented at the likely and unlikely listening location. Similar to the influence of target location uncertainty on word-identification performance in the overall data, the most robust effect observed for the categorization of data according to likely and unlikely listening locations was that of spatial listening expectations. We found that performance increased by 24 percentage points when targets were presented from the likely (mean = 79%) compared with unlikely (mean = 55%) location. In addition to a main effect of target location uncertainty, an age-group difference was also present, with younger adults (mean = 69%) outperforming older adults (64%) by 5 percentage points. Finally, the benefit of presenting targets from likely compared with unlikely listening locations was less robust in Experiment III (mean = 11 percentage points) compared

102 88 with Experiments I (mean = 30 percentage points) and II (31 percentage points). This betweenexperiment difference in word-identification performance was a result of two contributing factors, namely poorer performance for targets presented from the likely location and better performance for targets presented from the unlikely listening location in Experiment III compared with performance in the same conditions in Experiments I and II (see Figure 15). Figure 14. Mean percent correct word-identification scores for younger and older adults for Experiments I, II, and III. Data are presented as a function of spatial listening expectation. The pattern of performance described above was confirmed by submitting the data to a 2 x 2 x 2 repeated measures ANOVA where age group (younger versus older) and experiment (Experiments I, II, and III) were between-subjects variables and spatial listening expectation (likely versus unlikely) was a within-subjects variable. This analysis revealed a significant main effect of spatial listening expectation [F(1, 42) = , p < 0.001] and a significant main effect of age [F(1, 42) = 4.34, p < 0.05]. In addition to the two main effects, the analysis revealed a significant two-way interaction of spatial listening expectation and experiment [F(1, 42) = 7.72, p < 0.001]. In order to better understand the pattern of interactions, post-hoc SNK testing of all possible comparisons was conducted. This analysis indicated two key findings: (i) for targets presented from the likely listening location, word-identification scores in Experiment III were significantly worse than word-identification scores obtained in Experiments I and II, and (ii) for