Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

The effect of acoustic cue redundancy on the perception of stop consonants by older and younger adults Carter, Nathaniel Ryan 2011

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2011_fall_carter_nathaniel.pdf [ 2.1MB ]
Metadata
JSON: 24-1.0072247.json
JSON-LD: 24-1.0072247-ld.json
RDF/XML (Pretty): 24-1.0072247-rdf.xml
RDF/JSON: 24-1.0072247-rdf.json
Turtle: 24-1.0072247-turtle.txt
N-Triples: 24-1.0072247-rdf-ntriples.txt
Original Record: 24-1.0072247-source.json
Full Text
24-1.0072247-fulltext.txt
Citation
24-1.0072247.ris

Full Text

THE EFFECT OF ACOUSTIC CUE REDUNDANCY ON THE PERCEPTION OF STOP CONSONANTS BY OLDER AND YOUNGER ADULTS  by  Nathaniel Ryan Carter  B.A., The University of British Columbia, 2006  A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  MASTER OF SCIENCE  in  THE FACULTY OF GRADUATE STUDIES  (Audiology and Speech Sciences)  THE UNIVERSITY OF BRITISH COLUMBIA  (Vancouver)  September 2011  © Nathaniel Ryan Carter, 2011  Abstract Speech recognition is known to become more difficult as aging progresses. Though age-related hearing loss accounts for a significant portion of this difficulty, changes in cognitive processing and in the central auditory nervous system are also thought to contribute. Age-related speech recognition declines become most apparent for complex speech signals in which acoustic cues may be degraded, missing, or misaligned temporally. Each phoneme normally contains multiple, redundant acoustic cues signaling its presence and identity. The redundancy hypothesis suggests that older listeners require this natural redundancy of acoustic cues to a greater extent than do younger listeners, and it is the paucity of redundant cues within complex signals that makes them especially difficult for older listeners. The main purpose of the present study was to determine whether age-related redundancy effects existed when only single or dual acoustic cues signaled the presence of a stop consonant. Closure gap and release burst amplitude were varied for two phoneme contrast pairs (/p/ in speed/seed and /t/ in steam/seam) constructed from natural recordings. Six older and 6 younger participants with normal hearing (better than 25 dB HL from 250-4000 Hz) were tested. Using a 2-alternative forced choice (AFC) paradigm, participants indicated whether they heard the word as containing the stop consonant or not. ANOVA of the results revealed a main effect of burst amplitude and inconsistent effects of age but no interaction between burst amplitude and age, p = .803 for /p/ and .232 for /t/. For those steam/seam contrast stimuli in which closure gap was the only cue to stop presence, older listeners reached threshold perception of /t/ as gap duration increased but younger listeners did not. Because they do not show an interaction between age and the presence of redundant acoustic cues, these ii  combined patterns of results do not support the redundancy hypothesis. They suggest rather that older and younger listeners with comparable hearing make similar use of the redundant presence of stop closure gap and consonant release bursts.  iii  Preface Nathan Carter recruited all participants from public venues with the assistance of Elissa Robb, except for two participants who were recruited through a screening conducted by Mary MacDonald for her Master of Science (Audiology and Speech Sciences) thesis (2011). Nathan Carter scheduled and conducted all testing sessions. All data was compiled and prepared by Nathan Carter, and analysis was conducted jointly with Dr. Lorienne Jenstad. Approval for this study was granted by the UBC Behavioural Research Ethics Board (certificate H09-03416). No part of this thesis has yet been submitted for publication.  iv  Table of Contents Abstract ............................................................................................................................... ii Preface................................................................................................................................ iv Table of Contents ................................................................................................................ v List of Tables .................................................................................................................... vii List of Figures .................................................................................................................. viii Acknowledgements ............................................................................................................. x 1. Introduction ..................................................................................................................... 1 1.1 Complexity................................................................................................................ 2 1.2 Redundancy............................................................................................................... 3 1.3 Cognitive declines in age .......................................................................................... 5 1.4 Time compression ..................................................................................................... 6 2. Specific Processing Deficits ........................................................................................... 9 2.1 Temporal processing ................................................................................................. 9 2.1.1 Processing of non-speech temporal cues ........................................................... 9 2.1.2 Processing of temporal cues in speech............................................................. 10 2.1.3 Temporal envelope processing ........................................................................ 11 2.1.4 Temporal fine structure processing .................................................................. 12 2.2 Spectral resolution .................................................................................................. 13 2.3 Processing of combined spectral-temporal cues ..................................................... 15 2.4 Cognitive processing ............................................................................................... 20 3. Integration of Redundant Cues .................................................................................... 23 3.1 Redundancy for younger listeners .......................................................................... 25 3.2 Redundancy for older listeners ............................................................................... 27 3.2.1 Visual cue redundancy in age .......................................................................... 27 3.2.2 Auditory cue redundancy in age ...................................................................... 29 3.2.3 Multimodal cue redundancy in age .................................................................. 30 3.2.4 Why the current study? .................................................................................... 31 4. Study Design and Method ............................................................................................. 35 4.1 Design overview ..................................................................................................... 35 4.2 Subjects (criteria for selection) ............................................................................... 35 4.3 Stimuli ..................................................................................................................... 36 4.4 Equipment and materials......................................................................................... 39 4.5 Data collection ........................................................................................................ 40 4.5.1. Hearing thresholds .......................................................................................... 40 4.5.2. Gap detection testing....................................................................................... 40 4.5.3. Cognitive status ............................................................................................... 41 4.5.4. Speech perception task .................................................................................... 41 4.5.5. Response confidence ....................................................................................... 43 5. Results ........................................................................................................................... 44 5.1 Statistical analysis and hypotheses ......................................................................... 44 5.1.1 Data requiring interpretation ............................................................................ 46 5.2 Participant descriptive data ..................................................................................... 47 5.2.1 Hearing thresholds ........................................................................................... 47 5.2.2 Gap detection thresholds .................................................................................. 48 v  5.2.3 Digit span ......................................................................................................... 49 5.2.4 General health self-report ................................................................................ 49 5.3 Results for the speed/seed contrast pair stimuli ...................................................... 50 5.4 Results for steam/seam contrast pair stimuli .......................................................... 53 5.4.1 Full burst and -12 dB burst conditions ............................................................ 56 5.4.2 No-burst condition ........................................................................................... 56 5.5 Response confidence ratings ................................................................................... 58 6. Discussion ..................................................................................................................... 60 6.1 Hypothetical patterns of results .............................................................................. 60 6.1.1 Possible result #1: no age-related difference ................................................... 60 6.1.2 Possible result #2: age-related differences in all low-redundancy conditions . 61 6.1.3 Possible result #3: age-related differences in all conditions ............................ 62 6.1.4 Possible result #4: age-related difference for specific stop consonants ........... 63 6.2 Speed/seed contrast stimuli ..................................................................................... 65 6.3 Steam/seam contrast stimuli ................................................................................... 67 6.4 Gap duration as a cue to the perception of stops by younger and older listeners ... 68 6.5 Relationship to time compression studies ............................................................... 80 6.6 Complexity and speech recognition for older listeners........................................... 81 6.5 Conclusions ............................................................................................................. 82 References ......................................................................................................................... 84 Appendices ........................................................................................................................ 97 Appendix A – Stimulus formant transition comparison ............................................... 97 Appendix B – Speech perception testing results for each participant ........................ 103 Appendix C – Participants‘ comments during block confidence rating ..................... 116 Appendix D – Spoken word frequency of stimulus words ......................................... 119 Appendix E – Participant consent form ...................................................................... 120 Appendix F – Questionnaire form .............................................................................. 123  vi  List of Tables Table 1. Frequency of stimulus words in spoken English. .............................................. 37 Table 2. Natural burst amplitude and gap duration for two words used in stimulus creation. ............................................................................................................................. 37 Table 3. Descriptive data of the study sample. ................................................................ 45 Table 4. Mean gap detection thresholds. ......................................................................... 49 Table 5. Mean digit span scores. ...................................................................................... 49 Table 6. Gap duration (ms) at stop perception threshold for the speed/seed contrast pair stimuli. .............................................................................................................................. 50 Table 7. Stop perception threshold (in ms of gap duration) for the full burst and -12 dB burst conditions of the steam/seam contrast pair. ............................................................. 55 Table 8. Within-subjects contrasts of stop perception rate for various gap pairs within the steam/seam contrast no-burst data. ................................................................................... 58 Table 9. Frequencies of formants F1-F4 at various pitch pulses (PP) within the vowel from speed. ........................................................................................................................ 97 Table 10. Frequencies of formants F1-F4 at various pitch pulses (PP) within the vowel from seed. .......................................................................................................................... 97 Table 11. Frequencies of formants F1-F4 at various pitch pulses (PP) within the vowel from steam. ....................................................................................................................... 99 Table 12. Frequencies of formants F1-F4 at various pitch pulses (PP) within the vowel from seam.......................................................................................................................... 99 Table 13. Frequencies of formants F1-F4 at various pitch pulses (PP) within the vowel from scum........................................................................................................................ 101 Table 14. Frequencies of formants F1-F4 at various pitch pulses (PP) within the vowel from some........................................................................................................................ 101 Table 15. Word frequency for speed (Davies, 2008). .................................................... 119 Table 16. Word frequency for seed (Davies, 2008). ...................................................... 119 Table 17. Word frequency for steam (Davies, 2008)..................................................... 119 Table 18. Word frequency for seam (Davies, 2008). ..................................................... 119  vii  List of Figures Figure 1. Effects of presence vs. absence of a release burst for younger listeners (based on....................................................................................................................................... 26 Figure 2. Hypothetical results for an acoustic cue redundancy study (Jenstad, 2006, reprinted with permission). ............................................................................................... 31 Figure 3. Waveform and spectrogram showing the nominal and actual gap for speed token with full burst and no burst. .................................................................................... 39 Figure 4. Participant 106's speed/seed contrast pair response curves for the three burst amplitudes. ........................................................................................................................ 46 Figure 5. Mean hearing thresholds and standard deviations by age group. ..................... 48 Figure 6. Mean stop perception curves for the speed/seed contrast pair stimuli. ............ 51 Figure 7. Mean gap durations at stop perception threshold for various burst conditions of ........................................................................................................................................... 53 Figure 8. Mean stop perception curves for the steam/seam contrast pair. ....................... 55 Figure 9. Median response confidence for each testing block and overall. ..................... 59 Figure 10. Possible result #1: no age-related difference. ................................................. 61 Figure 11. Possible result #2: age-related differences in all low-redundancy conditions. ........................................................................................................................................... 62 Figure 12. Possible result #3: age-related differences in all conditions. ......................... 63 Figure 13. Possible result #4: age-related difference for specific stop consonants. ........ 64 Figure 14. Spectra of the burst and fricative from the speed/seed contrast pair. ............. 76 Figure 15. Spectra of the burst and fricative from the steam/seam contrast pair............. 77 Figure 16. Spectrogram showing the formant structure in the vowel within speed. ....... 97 Figure 17. Spectrogram showing the formant structure in the vowel within seed. ......... 98 Figure 18. Comparison of formant transitions within the vowels of speed and seed. ...... 98 Figure 19. Spectrogram showing the formant structure in the vowel within steam. ....... 99 Figure 20. Spectrogram showing the formant structure in the vowel within seam. ........ 99 Figure 21. Comparison of formant transitions within the vowels of steam and seam. .. 100 Figure 22. Spectrogram showing the formant structure in the vowel within scum. ...... 101 Figure 23. Spectrogram showing the formant structure in the vowel within some. ...... 101 Figure 24. Comparison of formant transitions within the vowels of scum and some. ... 102 Figure 25. Speed/seed contrast pair response curves from participant 100 ................... 103 Figure 26. Steam/seam contrast pair response curves from participant 100 .................. 103 Figure 27. Speed/seed contrast pair response curves from participant 101. .................. 104 Figure 28. Steam/seam contrast pair response curves from participant 101. ................. 104 Figure 29. Speed/seed contrast pair response curves from participant 102. .................. 105 Figure 30. Steam/seam contrast pair response curves from participant 102. ................. 105 Figure 31. Speed/seed contrast pair response curves from participant 103. .................. 106 Figure 32. Steam/seam contrast pair response curves from participant 103. ................. 106 Figure 33. Speed/seed contrast pair response curves from participant 104. .................. 107 Figure 34. Steam/seam contrast pair response curves from participant 104. ................. 107 Figure 35. Speed/seed contrast pair response curves from participant 105. .................. 108 Figure 36. Steam/seam contrast pair response curves from participant 105. ................. 108 Figure 37. Speed/seed contrast pair response curves from participant 106. .................. 109 viii  Figure 38. Figure 39. Figure 40. Figure 41. Figure 42. Figure 43. Figure 44. Figure 45. Figure 46. Figure 47. Figure 48. Figure 49. Figure 50.  Steam/seam contrast pair response curves from participant 106. ................. 109 Speed/seed contrast pair response curves from participant 900. .................. 110 Steam/seam contrast pair response curves from participant 900. ................. 110 Speed/seed contrast pair response curves from participant 901. .................. 111 Steam/seam contrast pair response curves from participant 901. ................. 111 Speed/seed contrast pair response curves from participant 902. .................. 112 Steam/seam contrast pair response curves from participant 902. ................. 112 Speed/seed contrast pair response curves from participant 903. .................. 113 Steam/seam contrast pair response curves from participant 903. ................. 113 Speed/seed contrast pair response curves from participant 904. .................. 114 Steam/seam contrast pair response curves from participant 904. ................. 114 Speed/seed contrast pair response curves from participant 905. .................. 115 Steam/seam contrast pair response curves from participant 905. ................. 115  ix  Acknowledgements This research was supported by a Natural Sciences and Engineering Research Council (NSERC) Discovery Grant awarded to Dr. Lorienne Jenstad, as well as by an Alexander Graham Bell Canada Graduate Scholarship (M) awarded to Nathan Carter by the Natural Sciences and Engineering Research Council (NSERC). I would like to thank a number of people who helped me to complete this thesis. First, thank you to my supervisor, Dr. Jenstad, whose patience endured the long evolution of a usable stimulus set as well as the sudden and smoky destruction of the lab computer. If further academic work is in my future, it is likely because of the simultaneous humility and rigour you model in your academic life; your students do not fail to notice. Thank you to Dr. Ciocca and Dr. Small for help in mapping out my blind spots and for pointing me toward increasingly better questions. My gratitude to Krista, my wife, should be included in a footer on each page. Thank you for bringing home the non-NSERC bacon while I worked on this project, for endorsing my work on this project, and for endorsing my breaks from working on this project. Finally, I am thankful to the people who volunteered to be subjects. The trip to UBC was not a trivial matter for a number of the older participants, and their willingness to take part is appreciated.  x  1. Introduction This study is meant to contribute to answering the question of why difficulty in understanding speech commonly attends aging. Long reported subjectively, age-related speech recognition declines also have been demonstrated experimentally many times. The Committee on Hearing, Bioacoustics, and Biomechanics (CHABA), a group convened in 1988 to clarify the existing knowledge on speech recognition in age, identified three potential sources for reduced speech recognition with age: peripheral, central auditory, and cognitive declines (CHABA, 1988). The first, a peripheral account, has also been called the simple audibility model. Unsurprisingly, numerous studies have shown that reductions in the audibility of speech due to peripheral declines such as cochlear or conductive hearing losses are correlated with poorer recognition of speech (Fox et al., 1992; Jenstad, 2006). Evidence continues to emerge for CHABA‘s (1988) second potential source, age-related changes in central auditory processing (e.g. Wong et al., 2009). Likewise, empirical support for correlations between speech recognition and various cognitive changes in age, the third potential source of decline, has grown in the intervening decades (Akeroyd, 2008; Wingfield & Tun, 2001; Wingfield, Tun & McCoy, 2005). Though there is evidence for all three sources, it is not yet clear how they interact, and all three are still studied independently as well as in combination. The simple audibility model continues to explain the largest portion of variance in speech recognition studies to the present day (e.g. Akeroyd, 2008) but predominantly in simple (not complex) listening conditions where speech signals are presented in quiet without degradation or noise, and in which a number of redundant acoustic cues mark the presence of each phoneme (e.g. Nabelek, 1988; Dubno, Lee, Matthews, & Mills, 1997; 1  Gordon-Salant, 1986b; Gordon-Salant & Fitzgibbons, 1993; Humes, 1996). Once noise is added or the signal is degraded in some other way (reverberation, for instance), older listeners‘ speech recognition results are no longer predictable from their audiograms alone (e.g. Divenyi & Simon, 1999; Dubno, Dirks & Morgan, 1984; Elliot, Busse & Bailet, 1985; Divenyi & Haupt 1997; Gordon-Salant & Fitzgibbons, 1995a; 1995b; 1999; Ohde & Abou-Khalil, 2001; Schneider, Daneman & Murphy, 2005). This confirms the experience of many older people who report inordinately greater conversational difficulty in noisy restaurants or in echoing rooms, even when their audiograms indicate reasonably good pure tone thresholds. It seems likely that the older listener‘s recognition of speech signals depends to some degree on the complexity of those signals, and to factors in the listener beyond peripheral losses (Divenyi & Haupt, 1997), such as the central auditory or general cognitive changes discussed in CHABA (1988). The terms complexity and redundancy, as they apply to this discussion, are elaborated below.  1.1 Complexity Complexity in a speech signal can be defined as the extent to which it is altered away from an ideal simple form. Noise or degradation of any kind increases a signal‘s complexity and potentially reduces the redundancy of the acoustic information within it (Jenstad, 2006). Complexity in a speech signal can arise in a number of ways. It can be a characteristic of the sound source itself, as when speech is particularly rapid or poorly articulated (Jenstad, 2006; Pichora-Fuller & Singh, 2006; Sommers, Kirk & Pisoni 1997; Wingfield & Tun, 2001). It can be a characteristic of the surroundings in which a speech signal is generated, as is the case in reverberant, echoing or noisy environments (CHABA, 1988; Jenstad, 2006). Noise can also be thought of as arising from the 2  listener‘s auditory system; it is common to model sensorineural hearing loss as noise that originates within the listener, interfering with the signals from the periphery in an internal signal-to-noise ratio (e.g. Keefe, Schairer, Ellison, Fitzpatrick & Jesteadt, 2009; Lavandier & Culling, 2010). Therefore, it could be concluded that the listener him- or herself is a potential source of complexity. Regardless of the type or source of complexity, individual speech cues within a complex signal may be masked, deleted, suppressed, distorted, overlapped, shifted or otherwise altered. Sometimes these effects occur in combinations, yielding potentially greater complexity and potentially greater loss of redundancy (Jenstad, 2006). As was stated above, it is in complex, lowredundancy listening situations that the peripheral or simple audibility model of speech recognition declines fails to explain all the variance in speech recognition scores for older listeners.  1.2 Redundancy Redundancy of speech cues has been defined as ―any characteristic of the language that forces spoken messages to have on the average more basic elements per message, or more cues per basic element, than the barest minimum‖ (Coker & Umeda, 1974, p. 349). In speech perception, redundancy is not understood as waste but rather as a characteristic that makes the speech signal more reliable as a means of carrying information (e.g. Chiari, 2007; Hsia, 1977). This benefit of redundancy is exhibited both within and between each level of linguistic structure (Hsia, 1977); redundancy within and between the discursive (Bazzanella, 2011), semantic, prosodic (e.g. Aylett & Turk, 2004), syntactic, morphological and phonological (e.g. Bazzanella, 2011) channels increases the likelihood of successful transmission of meaning (Coker & Umeda, 1974; Hsia, 1977). 3  Likewise, redundancy of acoustic information can be understood to serve this function at the level of the peripheral input. This becomes especially important as the complexity of real listening situations reduces the integrity of individual cues: when a cue is lost or degraded to the point where it could no longer individually activate perception of its phoneme, other redundant cues or the summed effect of a number of degraded but redundant cues may still be sufficient (Hsia, 1977). However, the mere presence of redundant cues does not guarantee successful speech recognition. Especially in complex listening scenarios in which acoustic cues associated with a particular segment may end up standing in an ambiguous or even contradictory relationship, the mitigating influence of cue redundancy depends on the ability of the listener to integrate such cues effectively (Coker & Umeda, 1974). Thus, the redundancy of speech signals makes demands on the listener that go beyond his or her peripheral abilities. To identify, categorize, weight and integrate large numbers of speech cues arising from different aspects of the signal (i.e., the spectral and temporal content), domain-specific processing abilities as well as the means to integrate them for perception are likely to be required (Peelle, Troiani, Wingfield & Grossman, 2010). It is known that cognitive processing abilities of many kinds covary with age (Schneider, Daneman & Murphy, 2005; Sulzenbruck et al., 2010), and so it is that CHABA (1988) directs attention to cognitive factors when faced with the limited ability of the simple audibility model to predict age-related speech recognition declines from peripheral losses alone.  4  1.3 Cognitive declines in age Many cognitive processes that potentially affect speech and language are subject to change over the lifespan (Schneider, Daneman & Murphy, 2005; Sulzenbruck et al., 2010). As two examples out of many, the dynamic working memory necessary for success in reverse digit span tests (CHABA, 1988) and ―n-back‖ tests (e.g. Mattay, 2006) declines with age, as do listeners‘ reaction times in simple and complex cognitive tasks (Fozard et al., 1994; Hallgren et al., 2000; Mattay et al., 2006; Salthouse, 1996). Not all processes show age-related declines of the same magnitude; many are maintained undiminished throughout the lifespan and some even show age-related improvement. For example, it has been shown that healthy older listeners tend to have undiminished access to semantic memory and perhaps an age-related advantage in some top-down perceptual strategies like the use of linguistic constraint arguments (Wingfield & Tun, 2001), prosodic information (Wingfield, Lindfield & Goodglass, 2000; Wingfield, Wayland & Stein, 1992), and semantic context (Wingfield & Tun, 2001). Since CHABA (1988), working memory has been shown to be correlated (albeit weakly and inconsistently) to declining speech recognition in age (Bopp & Verhaeghen, 2009), but this correlation is seen almost exclusively when the testing involves long, complex spoken forms such as sentences and paragraphs and not when it involves only single words or phonemes (Akeroyd, 2008). Beyond memory of various types, some have argued that a more central cognitive change drives age-related speech recognition declines. Some researchers have argued that aging involves an overall reduction in processing speed (e.g. Salthouse, 1991; 1996). There is reasonable evidence for a general cognitive slowdown with increasing age, 5  according to a common interpretation of the extant research on specific cognitive abilities in age (Gordon-Salant & Fitzgibbons, 2001; Ohde & Abou-Khalil, 2001; Salthouse, 1991; 1996; Schneider, Daneman & Murphy, 2005; Vaughan & Letowski, 1997). Included as evidence for a central decline are the reaction time studies mentioned above, as well as selective attention studies (Glisky, 2007), to name two examples. A central decline like cognitive slowing could be expected to manifest itself in a number of more specific processing declines, of which speech recognition is one potential process affected detrimentally by general slowing. In speech recognition research, the most frequently used method for examining cognitive slowing has been the use of timecompressed speech (e.g. Fitzgibbons and Gordon-Salant, 1996; Gordon-Salant and Fitzgibbons, 1993, 2001; Vaughan & Letowski, 1997; Wingfield & Ducharme, 1999; Wingfield, Poon & Lombardi, 1985).  1.4 Time compression If generalized slowing accounts for age-related speech decline, older listeners should perform more poorly than their juniors in recognizing speech signals that have been time-compressed (Gordon-Salant & Fitzgibbons, 2001). However, a problem arises in preparing the stimuli to test this hypothesis: simply time-compressing a speech signal raises its pitch as well, introducing the confound of altered spectral content. In order to avoid this problem, time compression can be achieved by removing pitch pulses at regular intervals and reducing silent durations by the same proportion. When presented with stimuli of this kind, older listeners do show poorer speech recognition results than their juniors (Wingfield, Wayland & Stine, 1992), with age-related differences most apparent when time compression reduces the signal to approximately 40% - 60% of its 6  original duration (cf. Gordon-Salant & Fitzgibbons, 1993; Jenstad & Souza, 2007; Wingfield, Wayland & Stine, 1992). To reiterate, some researchers have interpreted findings like these as evidence of a generalized cognitive slowing but this conclusion has been challenged by some further experimental results. For example, Wingfield and Ducharme (1999), after replicating the expected results for time-compressed speech, allowed their subjects to adjust the presentation of the speech signal to a comfortable rate not by restoring the originally excised pitch pulses but by the addition of pauses on the order of milliseconds throughout the signal. With this new, slower time-restored signal all participants had improved recognition scores but, crucially, the older listeners did not benefit as much as the younger participants. This suggests that it was not purely the speed of time-compressed signals that had challenged their recognition abilities in the first place, but also the absence of the acoustic cues contained in the material excised during time compression (Jenstad & Souza, 2007). Gordon-Salant and Fitzgibbons (2001) drew similar conclusions after they tested recognition of speech passages varying by level of syntactic context and type of time compression, using older and younger participants with both normal and impaired hearing. They found that time compression methods that removed cues to consonant identity created special difficulties for older listeners. The imposition of further acoustic degradations on a time-compressed signal only serves to more clearly differentiate younger and older listeners in their speech recognition scores (Jenstad, 2006). This observation supplies support for the notion that degradation of specific acoustic elements as well as fast presentation rate provide independent challenges to older listeners in typical time compression studies. However, it is still unclear whether older listeners‘  7  good performance for simple speech is due to the richness of cue redundancy in these signals or to the abundance of processing time such signals afford in comparison to timecompressed and distorted speech. Stated theoretically, it is still not clear to what extent the challenge of complexity for older listeners is a peripheral or a cognitive challenge (Jenstad, 2005; Jenstad & Souza, 2007). Though this question remains, research into how older listeners process specific aspects of the acoustic signal has gone some way toward uncovering the contributions made by age-related changes in the auditory periphery. Below we look at temporal, spectral and dynamic cue processing abilities to show to what extent each might account for older listeners‘ performance in complex listening.  8  2. Specific Processing Deficits A number of global and specific processing deficits, beyond reductions in audibility, have been proposed as the basis for speech recognition declines seen in age. The most commonly explored have included temporal processing, spectral resolution, the combination of these in the processing of dynamic cues, and more global factors like declines in short-term memory and reductions in overall processing speed (e.g. reaction times, etc.).  2.1 Temporal processing The following sections outline age-related changes in the processing of speech and non-speech temporal cues. Processing of both the temporal envelope and the temporal fine structure are discussed.  2.1.1 Processing of non-speech temporal cues There is a well-established literature base on the decline of temporal processing in aging (e.g., Jenstad, 2006; Pichora-Fuller & Souza, 2003; Schneider & Pichora-Fuller 2001; Souza & Boike, 2006). Studies using non-speech stimuli have shown that older listeners, with or without hearing loss, are less able to detect short-duration gaps in pure tones or differences in the duration of tones of 250 ms and shorter, as compared to their juniors with similar hearing abilities (Fitzgibbons and Gordon-Salant, 1994). Lister and Tarver (2004), in a similar study with more spectrally complex stimuli, found that older adults had poorer temporal resolution than younger adults even when the groups had equally good hearing thresholds. Pichora-Fuller, Schneider, Benson, Hamstra & Storzer.  9  (2006) found older listeners to have higher gap duration difference limens, especially when these gaps were bounded by more complex (spectrally asymmetrical) sounds. They, as well as others (e.g. Haubert, 1999), found no correlation between decreased temporal resolution and hearing thresholds even in the higher frequencies, suggesting that age-related temporal processing decline is a phenomenon separate from age-related hearing loss. Studies like these show two things: first, that older listeners are less able to detect and compare very short sounds or silences than are younger listeners; and second, complexity of the test materials appears to matter more to older listeners. Because the accurate perception of speech requires listeners to detect and identify similarly brief acoustic events (stop consonant release bursts or differences in voice onset time for voiced versus voiceless consonants, for example), research that studies temporal processing of actual speech signals has been of value in the investigation of speech understanding declines in age.  2.1.2 Processing of temporal cues in speech Temporal processing has also been tested within numerous speech contexts, resulting in evidence that temporal processing declines are not only present at an older age, but that they have a negative impact on processing of speech sounds. The temporal characteristics of speech are often broken into two categories: the temporal envelope and the temporal fine structure. The temporal envelope describes the relatively slow changes in overall amplitude that mark syllabification and prosody at the suprasegmental level, and it incorporates the gap and duration cues that differentiate phonemes at the phonemic level (Divenyi & Haupt, 1997; Jenstad & Souza 2007; Lavandier & Culling, 2010; Lorenzi, Gilbert, Carn, Garnier & Moore, 2006; Pichora-Fuller & Singh, 2006; Schneider 10  & Pichora-Fuller, 2001; Souza & Kitch, 2001; Turner, 1995). The temporal fine structure refers to the temporal content found within the level of the segment in the frequency range of 600Hz –10 kHz (Rosen, 1992) and it carries information about, for example, timbre, formant transitions, and the minute interaural timing differences that allow for sound localization (Rosen, 1992; Schneider & Pichora-Fuller, 2001; Souza & Kitch, 2001). Both types of temporal processing have been studied for effects of aging.  2.1.3 Temporal envelope processing One investigation into temporal envelope processing was a study by Price and Simon (1984) that tested older and younger listeners on the perception of ‗rapid‘ vs. ‗rabid‘ where the middle consonant closure gap was varied between 35 and 125 ms, and the vowel duration was varied from 160 to 220 ms. Stop consonant closure gap and vowel duration are redundant cues for differentiating /p/ from /b/. Despite being nearly matched for PTAs (all had 25 dB HTL or better thresholds at 500, 1000, 2000 and 3000 Hz), older participants required longer duration closure gaps in order to perceive these stimuli as ‗rapid‘ than did their younger counterparts, suggesting a declining ability to process short duration speech envelope cues. Studies by Wingfield, Wayland & Stine (1992) and Haubert (1999) showed that temporal compression of the envelope (increased speech rate) taxes word recognition more for older listeners than for younger, and similar results have been found in subsequent studies (Pichora-Fuller & Souza, 2003). However, not all types of temporal envelope processing show evidence of decline in age: it appears that older listeners fully retain the ability to process the slowest temporal envelope cues, those at the level of prosody (Wingfield, Lindfield & Goodglass, 2000; Wingfield & Tun, 2001). 11  There has been some evidence (e.g. Lorenzi et al., 2006; Ohde & Abou-Khalil, 2001) for correlations between declines in temporal resolution and elevated hearing thresholds at frequencies higher than the 500 Hz to 3000 Hz screening that Price and Simon used to match their young and old participants raising questions about the independence of temporal processing as an age-related decline. However, several studies have failed to produce consistent confirmation of such a correlation (Gordon-Salant & Fitzgibbons, 1993; Souza, 2000; Strouse, Ashmead, Ohde, & Grantham, 1998; PichoraFuller et al., 2006).  2.1.4 Temporal fine structure processing Evidence for an effect of aging in the ability to process temporal fine structure has been equivocal. Lorenzi et al. (2006) compared older and younger normal-hearing and hearing-impaired adults on their ability to recognize consonants within an /aCa/ context where the signal was reduced to envelope cues only or to temporal fine structure cues only. Though correlations emerged between poorer temporal fine structure resolution and poorer hearing thresholds, no correlation was found between temporal fine structure resolution and aging. This contrasts with studies that have shown age-related difficulties in the resolution of low frequencies and in segregation of simultaneously presented vowels (Schneider & Pichora-Fuller, 2000). None of these declines was correlated to the participants‘ audiograms Note that the resolution of lower frequencies is categorized as a temporal rather than a spectral phenomenon because it is hypothesized to be the result of phase-locking difficulties which do not manifest above a certain low frequency range (Schneider & Pichora-Fuller, 2000). In summary, it is still unclear whether temporal fine structure processing is affected by aging. 12  Combining the evidence from speech and non-speech temporal processing studies at both the envelope and temporal fine structure levels, several provisional conclusions can be drawn. First, there is an age-related decline in listeners‘ ability to use certain aspects of the temporal envelope (Price & Simon, 1984; Wingfield, Wayland & Stine, 1992; Haubert, 1999; Pichora-Fuller & Souza, 2003). Second, it is possible that there exist age-related declines in the processing of temporal fine structure (Schneider & Pichora-Fuller, 2000). Third, these effects are not dependent on hearing loss and are therefore tied to the process of aging in some other way (Gordon-Salant & Fitzgibbons, 1993). Finally, as is the case with studies of hearing acuity‘s relationship to speech recognition, age-related temporal processing declines tend to become most apparent in listening situations characterized by increasing complexity (Gordon-Salant & Fitzgibbons, 1993; Haubert 1999; Pichora-Fuller & Souza, 2003; Schneider & PichoraFuller 2001). This suggests that the relationship between temporal processing ability and speech recognition performance is not straightforward and that other factors are at play in the complex everyday listening situations of older listeners (Gordon-Salant & Fitzgibbons, 1993). The ability of older listeners to resolve spectral content is a potential factor that has received attention in audiological research.  2.2 Spectral resolution Some studies have shown that the ability to discriminate fine differences in frequency, as is necessary for formant transitions in a speech signal, may decline in age (e.g. Alain, McDonald, Ostroff & Schneider, 2001; Konig, 1957; He, Dubno & Mills 1998; Moore & Peters, 1992; Patterson, Nimmo-Smith, Weber & Milroy, 1982). Using stimuli closer to a natural speech signal, Elliot, Busse and Bailet (1985) compared 13  younger and older listeners (all with PTAs at or better than 25 dB HL at 500, 1000 and 2000 Hz) to see whether these groups differed in their ability to identify and discriminate between synthesized 5-formant consonant-vowel syllables. Though they found some differences in performance that were correlated to hearing sensitivity differences at 4 kHz, the authors reported further differences in identification scores and in the slope of performance-intensity functions that they hypothesized might be due to an age-related reduction in the ability to resolve small differences in formant frequencies. Similarly, Lutman, Gatehouse and Worthington (1991), despite the large size of their study (n=240) found that there was only a minor correlation between increasing age and decreasing frequency resolution once hearing thresholds were accounted for: age accounted for approximately 4% of variance in frequency resolution scores. Cranford and Stream (1991) showed that older adults (mean age 73.3) had worse brief tone difference limens than younger adults (mean age 32 years). This decline in one aspect of spectral resolution was correlated with age but not with hearing thresholds. Though the study did not directly address it, there is a possibility that the challenge for older listeners was at least in part the brevity of the tonal markers in addition to their frequency differences. However, of the seven tone durations used, not all of Cranford and Stream‘s (1991) stimuli could be considered exceptionally brief. Their 500-ms tones could be considered to have long duration at least in comparison to the 6.4-ms and 250-ms markers used by Fitzgibbons and Gordon-Salant (1994) to show age-related temporal processing difficulties. Additionally, age differences proved significant for all of Cranford and Stream‘s (1991) tone durations (500, 200, 100, 50, 20, 5 ms), suggesting that the same difficulties were present for older listeners over a range of tone durations. Regardless, it  14  is not possible to attribute Cranford and Stream‘s (1991) results solely to spectral processing with certainty. Though studies like these indicate that some decrease in spectral resolution often accompanies aging, the relationship does not appear to be of a large enough magnitude to be useful as a primary explanation for the declines in speech recognition seen in age. Further, it does not provide an explanation for the differential performance of older listeners in simple vs. complex listening. Because hearing sensitivity, temporal processing, and frequency resolution ability are not able individually to account for the differences between older listeners‘ performance on simple and complex listening tasks, combinations of these factors have been investigated.  2.3 Processing of combined spectral-temporal cues Temporal and spectral processing abilities are employed simultaneously in the processing of spectral changes over time; for example, in the accurate perception of formant transitions at vowel-consonant boundaries and formant changes within diphthongs. Though it has not been studied as extensively as temporal resolution, the processing of combined cues like these has received substantial attention over at least the last six decades. Liberman et al. (1954) had begun to identify the importance of formant transitions in the recognition of consonants in CV contexts. Since then, research on combined spectral-temporal cues has been extended to the study of older populations (e.g. Dorman, Marton, Hannley & Lindholm, 1985) and to the use of these cues as a basis for automatic phoneme recognition (Nossair and Zahorian, 1991). Fox, Wall and Gokcen (1992) recorded whole CVC words and created modified versions from which a center portion of the vowel was removed. In this silent-center condition, the combined spectral15  temporal cues inherent in the remaining vowel on- and offset become more important for identifying vowel quality. The formant transitions in vowel on- and offset are generally considered to be secondary cues to vowel identity (Hedrick, 1997), with the vowel‘s steady-state formant structure providing the primary cues (Ohde & Abou-Khalil, 2001). In Fox, Wall and Gokcen‘s (1992) study, older and younger participants with normal audiograms were compared in their ability to identify the vowel based on primary plus secondary cues in the unadulterated stimuli or based on only the secondary cues remaining in the silent-center stimuli. The results showed that older listeners were significantly poorer at identifying excised vowels and surrounding consonants in the tokens that required them to rely on secondary (combined spectral-temporal) cues. Note that the older participants had significantly worse hearing sensitivity for tones of 8 kHz, but that their data were not excluded because the experiment‘s stimuli had little or no spectral content above 4.5 kHz. In a similar study, Dorman et al. (1985) gave a variety of phonetic identification tasks to younger participants with normal hearing and older participants with and without sensorineural hearing loss. In one task, they removed some static identity cues (defined by Ohde & Abou-Khalil, 2001, as cues that change very little over time)—release bursts in this case—forcing the participants to rely on modified second formant shapes (combined spectral-temporal cues) as identity cues. In this condition, performance was poorer for older listeners regardless of their hearing acuity. To summarize, reliance on secondary spectral-temporal cues alone was problematic for the older listeners. In contrast, a study by Ohde and Abou-Khalil (2001) claimed to have found little evidence for the effects seen by Dorman et al. (1985). Their study investigated the  16  importance of static, dynamic, and integrated cues for identification of consonants and vowels in CV syllables. Their static cues were those that did not change over time: straight formants and stop release bursts. Their dynamic cues were those whose spectra did change over time: formant transitions, in this case. Ohde and Abou-Khalil‘s (2001) dynamic cues are equivalent to what we have classified as combined spectral-temporal cues. Finally, their integrated cues were amalgamations of dynamic and static properties that together formed what could be considered a separate acoustic property. Their examplar for the integrated cue is a consonant-initial word‘s onset spectrum; it consists of a static burst and the combined spectral-temporal cue of a formant transition. For Ohde and Abou-Khalil (2001), this integrated cue is considered a single primary cue to the identity of the initial consonant. The results of their experiment showed that ―…onset information provided by moving formant transitions [was] usually a sufficient cue for correct place-of-articulation identification across age groups.‖ (Ohde & Abou-Khalil, 2001, p. 2160, italics mine). Clearly, older listeners were successful in using secondary, combined spectral-temporal cues to aid speech recognition. However, the question of what comprised the ―usual‖ conditions in which they had this success is important to the current study. Though they performed quite well overall, the older listeners made poorer use of dynamic cues than the younger and middle-aged listeners, specifically in the condition of having to rely on these cues exclusively; i.e., when other cues were absent. Put another way, they made best use of dynamic cues only when they had the opportunity to integrate these cues with other redundant cues. Once it is seen that Ohde and AbouKhalil‘s (2001) older listeners‘ successful use of combined cues depended to some degree on the presence of redundant cues, the difference in outcomes between their study and  17  Dorman et al.‘s (1985) is superficial. Both confirm that older listeners retain much of the ability to use secondary combined spectral-temporal cues with the constraint that redundant cues must be present. As was the case with changes in temporal processing and spectral resolution, it appears that for processing of combined cues, it is in situations of low redundancy (or high complexity) that aging is correlated with poor performance. Further details of Ohde and Abou-Khalil‘s two-part study are germane to the design of this study, and are outlined below. The first half of the study was an investigation of stop consonant perception. Four types of synthetic CV syllable were constructed using presence/absence of a release burst and duration (10 or 46 ms) of dynamic and static vowel formant transitions as the stimulus parameters. The participants belonged to a young (mean age 22:7), middle-aged (mean age 54:9), or old (mean age 73:7) group. All had hearing thresholds better than 30 dB HL except for four subjects, three of them being in the older group, who met criterion thresholds for one ear and had thresholds between 30 and 40 dB HL for the other. Participants were required to identify the initial consonant for each token by selecting it in a forced choice from three options [b d g]. When burst and long voiced vowel segments (46 ms) were present, all participants identified initial consonants with high accuracy. When the burst was absent, younger and middle-aged participants had reduced identification scores with short vowel portions (10 ms) but returned to high accuracy when the stimulus had a long vowel portion (46 ms). In contrast, the longer vowel duration had almost no positive impact on the identification scores of the older group in the absence of a release burst. This amounted to an approximately 20% difference in identification accuracy scores between age groups, and the difference held for both static  18  and dynamic vowel formant conditions. This half of the experiment confirms what was outlined in section 2.2: that spectral content is a speech cue that remains accessible to older listeners, and that its usefulness depends on its agreement with other cues. Ohde & Abou-Khalil (2001) extended their experiment to investigate vowel recognition using dynamic spectral cues to vowel identity. The second half of Ohde and Abou-Khalil‘s (2001) study investigated vowel identification in CV syllables. Synthetic stimuli were constructed using voiced stop consonants with appropriate bursts and three vowels [i, a, u], each varying by duration (10, 30 or 46 ms) and presence/absence of formant movement. The method of testing was similar to that used in part one of their study. Results showed that identification performance was generally high for all groups in all conditions. The exception was the performance of the oldest group on the non-dynamic formant version of [u]. In this condition, no increase in voicing duration enabled the older listeners to match the scores they achieved with dynamic formants. This result was inconsistent in that it was observed for only one of the three vowels, but despite this variability in the results—a common theme in older adult speech perception literature (e.g. Humes, 1996; Jenstad 2006; Jenstad & Souza, 2007; Schneider & Pichora-Fuller, 2001)—it nonetheless aligns with the observation from part one of their study that if older listeners exhibit speech recognition deficits, it is likely to be in situations of low redundancy (or high complexity). Clearly, the older adults in this study were not ―blind‖ to any particular type of speech cue (i.e., static, dynamic or integrated) but they could achieve their best perception scores only when multiple cues were present in a reinforcing relationship.  19  Results like these, showing speech cue variable interactions that depend so strongly on age and on the number and clarity of provided cues, suggest that though there may be demonstrable decrements in hearing sensitivity and in specific aspects of auditory processing (like the temporal, spectral, and combined spectral-temporal processing discussed in the preceding sections), the underlying reason for decreasing speech recognition in age may not belong to any one of these domains, but may rather be explainable by changes in the way higher-level systems integrate all of them.  2.4 Cognitive processing Changes in more general cognitive factors have also been suggested as potential explanations for poorer speech recognition in age. These include changes in motivation, fatigue, memory capacity, memory processing, linguistic processing (Fox, Wall, & Gokcen, 1992), as well as more fundamental neurological declines such as general cognitive slowing (e.g. Salthouse, 1991; 1996). Some studies of speech recognition have looked for correlations with cognitive changes by including standard cognitive tests in their methodologies, and some correlations of this type have indeed been found. For example, Elliot, Busse and Bailet (1985) saw a significant correlation between their older participants‘ ability to detect JNDs (just noticeable differences) on a place-of-articulation continuum and scores on block design and concept formation subtests from the Wechsler Adult Intelligence Scale-Revised and Woodcock-Johnson Psycho-Educational Battery, respectively. A more recent review (Akeroyd, 2008) of 20 studies of cognitive function and speech recognition published since CHABA (1988) reported inconsistent correlations between cognitive test scores and speech recognition. There was no single cognitive test that could predict speech recognition difficulty in quiet or in noise reliably, but some tests 20  showed correlations with speech recognition more often than others. Working memory test scores, particularly for reading span, were the most likely to account for some variance in speech recognition scores, at least for sentence-length and longer speech signals, and the more general measures (e.g. I.Q.) were least likely to do so. It must be remembered that these results are correlations and do not reveal the nature or extent of any causal links between speech recognition and the constructs being tested. Though evidence continues to accumulate for various correlative connections between cognitive change and speech recognition, no domain-specific cognitive decline has provided a strong model explaining speech recognition declines in aging (Akeroyd, 2008). Among the more fundamental cognitive declines investigated has been generalized cognitive slowing, as outlined in the introduction. An advantage of establishing a high-level cognitive processing account like this is that such global changes could presumably be causing many of the more specific processing decrements discussed above, leading to a more parsimonious explanation for age-related effects. The cognitive slowing hypothesis as it relates to speech recognition has been repeatedly tested using time compression. To summarize material from section 1.4, these studies have indeed found some evidence for cognitive slowing in that older listeners‘ speech recognition did not tolerate increased presentation speed as well as did younger listeners‘ (e.g. Wingfield & Ducharme, 1999). This matches evidence for such slowing in other domains. However, the studies have also shown that cognitive slowing fails to account for the fact that speech recognition difficulties persist once processing time has been restored without concurrent restoration of acoustic cue redundancy (e.g. Schneider, Daneman & Murphy, 2005). Additionally, cognitive slowing does not explain why time  21  compression of consonants is particularly damaging for older listeners in comparison to vowels (e.g. Gordon-Salant & Fitzgibbons, 2001). For these reasons, we are led to the alternate hypothesis proposed in some time-compression studies (e.g. Jenstad, 2006; Jenstad & Souza, 2007): that a large part of the difference between younger and older adults in complex listening is the extent to which they rely on acoustic cue redundancy. This possibility is detailed in the sections that follow.  22  3. Integration of Redundant Cues The redundancy hypothesis, as put forward by Jenstad (2006), and implied more or less strongly by others (e.g. Gordon-Salant & Fitzgibbons 2001, Jenstad & Souza, 2007; Ohde & Abou-Khalil, 2001; Souza & Kitch, 2001), is that age-related declines in complex recognition tasks reflect an increased need for acoustic cue redundancy in speech signals. This hypothesis does not propose that the use of redundancy is a discrete ability that changes with age. Rather, it proposes that redundancy of acoustic cues in speech signals is particularly important to older listeners given other developmental changes they are undergoing. The redundancy hypothesis provides a feasible explanation of time-compression study results (Jenstad, 2006) and, if it finds further confirmation, may lend support to an account of age-related speech recognition declines that is centered on effects of age occurring above the level of the periphery. The following sections provide more detail on the kinds of redundant cues available in speech, the evidence for an increased reliance on them in age, and the nature of their relationship to complexity of listening conditions. It is rarely, if ever, the case in speech signals that the identity of a phoneme is signified by a single acoustic property. For example, the stop consonant [g] is characterized by a range of acoustic cues including the structure and dynamics of formants before, during and after velar contact; the duration and spectral content of the noise burst associated with release of velar contact; the relative timing of velar release burst and voice onset; the relative amplitude of the burst in relationship to the vocalic nucleus of the syllable in which it occurs; the duration of the following vowel; and any coarticulatory influences on surrounding phonemes that may provide further evidence of 23  its presence (Hedrick & Carney, 1997). The hypothesis emerging from time-compression studies suggests that older listeners would need a larger number of these cues to be present in order to perceive this phoneme. To use the current example in the framework of the acoustic cue redundancy hypothesis, complex listening conditions are those that weaken or degrade any of the cues to the point where a listener must perceive the presence of [g] based on fewer, and potentially contradictory, acoustic cues. Souza and Kitch (2001) and Ohde and Abou-Khalil (2001) imply this redundancy hypothesis to some extent when they posit an age-related increase in reliance on speech cue integration. The alternative to relying on integration, exhibited by younger participants (Ohde & Abou-Khalil, 2001), was the ability to accomplish high speech recognition scores (>90%) using a solitary brief cue. Fortunately for elderly listeners there is evidence that this increasing reliance on cue integration is matched in age by an increasing facility with cue integration, and an expanded ability to apply a wider array of stored knowledge more effectively than can be done by younger listeners. The evidence for mitigating age-related developments is discussed in section 3.2 below. Finally, if acoustic cue redundancy is really the mechanism by which older listeners overcome some degree of complexity in speech signals, it provides a potential explanation for the difference between older listeners‘ relatively good performance in essentially simple listening situations (typically possessing higher redundancy) and their relatively poor performance in more complex listening tasks (typically possessing low redundancy). It may also provide a descriptive framework for the interaction of the various sub-process declines discussed in the sections above: each has the effect of lowering redundancy in different ways and to different extents. To paraphrase Jenstad  24  (2006) and summarize the acoustic redundancy hypothesis: older listeners have a greater need for acoustic cue redundancy than their juniors. If redundancy is supplied in the speech signal older listeners are able to perform well, but their speech recognition performance declines more quickly as redundancy is removed and complexity is increased. This hypothesis appears to align with further evidence from numerous cueweighting studies over the last decades, as described in the following sections. We first review the results of younger listeners.  3.1 Redundancy for younger listeners Repp (1984) undertook a cue-weighting study that examined the perception of presence of velar, alveolar, or labial stops in C(C)CVC monosyllabic words (eg, splat vs. sklat vs. stlat). The stimuli were constructed from recorded natural speech segments for younger listeners. He varied the presence/absence of a release burst and the duration of the preceding closure gap (from 0 to 100 ms in 20 ms steps). His participants were himself (eventually excluded because his results were significantly different from the younger participants‘) and nine paid student volunteers whose age and hearing acuity were not reported, but who were presumably young and possessed subjectively normal hearing. The results (see Figure 1) showed that without a release burst in the stimulus, a longer gap duration (on average, 60 ms without a burst as opposed to 35 ms with a burst) was needed for perception of the stops. That is, a normal-hearing younger listener recognized speech best in high-redundancy conditions but could understand a lowerredundancy speech signal provided that a remaining cue was made more salient. Note that in Figure 1 the dashed curve represents nominal gap in the no-burst condition and  25  that the arrows indicate actual gap duration (closure gap plus the silent duration of the missing burst).  Figure 1. Effects of presence vs. absence of a release burst for younger listeners (based on results from Repp, 1984).  If these effects were found only for the cues that Repp was manipulating in his study, the results would suggest the interaction of only those cues and would not necessarily support a redundant cue hypothesis for speech recognition in general. However, this is not the case, as similar results are found in studies that manipulate other cues. Ohde and Abou-Khalil (2001), in a cue-weighting study (described more fully in section 2.3), varied the presence of formant transitions, vowel duration, and burst presence for synthesized CV syllables, and found results similar to Repp‘s (1984) for their young participants: the younger listeners showed increased stop consonant 26  recognition when redundant cues to speech segment identity were present, and any decrement in performance for low-redundancy conditions could be mitigated by strengthening one of the remaining cues. In the particulars of their study, an increase in vowel duration from 10ms to 46ms (from one glottal pulse to five glottal pulses), even in the absence of a consonant release burst or formant transitions, allowed the young listeners to move from approximately 75% to 95% identification of stop consonants. The above studies show that acoustic cue redundancy increases recognition scores for young normal-hearing listeners, and that their ability to make use of strong remaining secondary cues allows them to maintain high levels of performance even in relatively complex listening situations where this redundancy is diminished. These results, though shown to hold across cue types, do not hold across age groups when young or middle-aged subjects are compared to elderly subjects.  3.2 Redundancy for older listeners Older listeners appear to rely on redundancy more heavily than younger listeners do, and this effect is not limited to the auditory modality. It has been demonstrated in visual perception and in multimodal perception as well as in audition. The following sections describe the findings of the age-related redundancy literature for several domains, all indicating that older people require a larger amount of redundancy, especially as complexity increases.  3.2.1 Visual cue redundancy in age Older adults have been shown to make more use of visual redundancy than their juniors. The second part of a recent study by Bucur and Madden (2005) compared older 27  adults (mean age 68.7) and younger adults (mean age 19.0) on their response times in recognizing certain visual features (predetermined colours and letters) presented in adjacent areas of the visual field. When both target features were presented as a unified figure (a purple letter K, in this case), reaction times shortened for all participants but more so for older viewers than for young. Older viewers had a 10% improvement over single-feature reaction times, as compared to an 8% improvement for younger viewers, a small but significant effect. The fact that participants had to attend simultaneously to one area to determine colour presence and to an adjacent area to determine letter presence is considered an increase in task complexity (Bucur & Madden, 2005). This complexity was not present in the first part of their study, where competing features (letter, orientation, and colour) were always combined in a single figure, and in which no significant age-related difference in the redundant signals effect was found. Explaining the difference in results between the two parts of the experiment, the authors conclude that beyond some threshold of task complexity, the presence of redundant information differentially benefits older adults (Bucur & Madden, 2005). This confirmed results from an earlier study by Allen et al. (1992) in which participants were asked to make a quick two-way forced choice decision as to the presence of either half of a two-letter pair presented either singly or in redundant multiples on a screen. Complexity was varied with the presence of simultaneous distractor letters. Redundancy was shown to significantly shorten reaction times for all viewers, with older viewers showing a significantly larger benefit of redundancy in the most complex, distracting conditions when compared to the younger viewers.  28  3.2.2 Auditory cue redundancy in age As was seen in section 3.1, auditory cue redundancy benefits young listeners. The current section reviews literature on the effects of such auditory cue redundancy for older listeners. In the study by Ohde and Abou-Khalil (2001) discussed in section 3.1, a small group of older participants (mean age 73:7) required the presence of multiple, redundant speech cues to perform as well as their juniors could with a single strong cue. In fact, once a redundant stop consonant release burst was removed, increasing the salience of the remaining cue by more than quadrupling the vowel length from 10 ms to 46 ms yielded almost no increase in identification scores for this group, showing that the integration of redundant cues was more important than the strength of the individual cues for older listeners. Ohde and Abou-Khalil (2001) termed this as ―less efficient‖ listening, in that older listeners required a larger number of redundant cues in order to do the work of recognizing a phoneme. Limitations of this study include a small sample size (8 people in the older listener group) and a certain amount of audiometric asymmetry between age groups (3 of the 8 older participants exceeded the study‘s hearing threshold criterion at 4000 Hz by up to 10 dB in one ear). Though Ohde and Abou-Khalil stated that this hearing loss appeared inconsequential given the participants‘ scores in testing, it would have been desirable to have had a cohort of older listeners that better met the study‘s hearing criterion. Regardless of these drawbacks, the study‘s results replicated earlier research by Dorman et al. (1985) and support the conclusion that the performance declines seen in their older participants were not due to a deficit in hearing sensitivity, nor in temporal or spectral processing per se, but were rather due to an inability to identify speech sounds when redundant cues were missing.  29  3.2.3 Multimodal cue redundancy in age The age-related effects seen in visual and auditory redundancy appear to hold between, as well as within, these modalities. It is well known that visual and auditory cues interact in the normal perception of speech (e.g. McGurk, 1976). In a recent reaction time study, Keifer (2009) used simple tones and visual symbols as stimuli to show that younger listeners (mean age 20:0) benefited from combined auditory/visual redundancy as well as they did with redundancy in either modality on its own. These findings confirm those from a 2005 study by Laurienti et al., in which younger listeners (mean age 28) and older listeners (mean age 71) discriminated the appearance of circles of certain colours and verbalizations of the colours‘ names in the presence of distractors. All participants scored with high accuracy (above 94%), and all their reaction times were shorter when a visual and auditory manifestation of a target occurred simultaneously. The magnitude of enhancement for redundant presentations was greatest for the older listeners (a reduction from 87.5 ms to 53.2 ms over the best single modality reaction time). Laurienti et al. (2005) concluded that multimodal redundancy was differentially beneficial to older adults, that older adults were possibly more able to exploit redundancy in all modalities, and that the benefit of multimodal redundancy was so great that it sometimes enabled older adults to achieve performance comparable to that of younger adults who were using cues from their fastest single modality. There is reasonable evidence then that older adults make increased use of redundancy in many modalities, and that complexity of tasks tends to draw out agerelated differences in the use of redundancy in each modality. There is therefore evidence from a wide range of research streams for testing the redundancy hypothesis.  30  Such research is of special theoretical value since it may bear on the potential sources of age-related speech recognition declines that have been discussed since CHABA (1988); results showing significant age-related differences in use of acoustic cue redundancy would lend support for a central auditory or cognitive account as opposed to a peripheral account.  3.2.4 Why the current study? Jenstad (2006) has combined the pattern of results from Repp‘s (1984) younger listeners and Ohde and Abou-Khalil‘s (2001) younger and older participants into a hypothesis of how older listeners like those tested in the latter study will respond in an experiment that manipulates two redundant cues to the presence of a stop consonant (see Figure 2).  Figure 2. Hypothetical results for an acoustic cue redundancy study (Jenstad, 2006, reprinted with permission).  31  A study designed to establish evidence for an age-related redundancy hypothesis needs to overcome certain limitations in some of the studies discussed so far. First, the participants must be carefully selected for hearing acuity in order to better isolate the effects of aging from those of age-related hearing loss. Second, the old participants must be sufficiently old that an age effect is likely to be seen (Jenstad, 2006). Several studies suggest that age effects in speech recognition or processing ability begin to be most clearly observable close to age 80. Magnusson (1996) found age-related effects for subjects at 83 years and older, and Sherbecoe and Studebaker (2003) found sharp declines beginning after 70 and especially after 75, as did Humes (1996), so participants of this age are desirable. Third, the stimuli should be created along the lines established by Repp (1984) to allow for direct comparison to the results he obtained for younger listeners (see Figure 1), thereby showing that any differences in the use of redundancy are due to age and not to substantial differences of stimulus preparation. Gap duration and burst presence, the cues Repp (1984) used, are desirable too because they can be manipulated with precision even in recorded as opposed to synthetic stimuli. Using them would allow for discussion of results within the well-established literature on temporal resolution in age. Fourth, the older group should be compared to an audiometrically similar younger group using the same test materials and methodology. Fifth, if voiceless consonants are used, the salient dynamic effects of formant transitions during voicing onset could be minimized or eliminated. The absence of salient formant transitions limits the amount of overall redundancy, and thereby makes age-related differences more likely to be observed.  32  In light of the literature reviewed above, the following methodological decisions were made in the design of the current study:  1. Because age-related redundancy effects are generally not seen with simple speech, complexity of the stimuli was increased by the removal of all but two acoustic cues (presence of a release burst and closure gap) for the target phonemes. 2. To isolate the effects of acoustic redundancy, cues from other levels of the linguistic signal (i.e. syntactic, semantic, discursive, etc.) were minimized by the selection of single word stimuli with no sentential context. Use of single words also lessens the impact of cognitive factors associated with aging as these are more likely to play a role in phrase or sentence length stimuli (Akeroyd, 2008). 3. Because they are known to be more problematic for speech perception in aging populations (Gelfand, Piper & Silman, 1985; 1986; Gordon-Salant & Fitzgibbons, 2001), consonants were selected for investigation. 4. Because age-related hearing loss can obscure other age-related factors, both older and younger listeners were required to have normal hearing thresholds.  The current study tests the redundancy hypothesis with the following experimental question: do listeners 70 years and older with normal hearing (thresholds at or better than 25 dB HL at 500, 1000, 2000 and 4000 Hz) require the redundant presence of both a natural release burst and a substantial closure gap to perceive a voiceless stop  33  consonant in a /s/(C)VC real word in quiet, as compared to audiometrically similar listeners aged 30 and younger? Support for the redundancy hypothesis would be found in an interaction between age and cue redundancy. Several patterns of results could be anticipated. First, there could be no statistical interaction, with older and younger listeners using redundancy of cues in the same way. Second, a redundancy effect could be found for older listeners, supporting the redundancy hypothesis. Third, an age-related decrement in performance, unrelated to level of redundancy, could be found. Finally, age-related redundancy effects could be seen only for subsets of the data.  34  4. Study Design and Method This chapter provides details about the subjects, stimuli, equipment and testing procedures used in conducting the current study.  4.1 Design overview This is an acoustic cue manipulation study using a two-alternative forced choice paradigm to estimate stop consonant perception thresholds for older and younger age groups. The stimuli are recordings of monosyllabic words in which the stop consonants have been modified along two parameters: closure gap duration and amplitude of a stop release burst. The threshold is defined as the closure gap duration at which a voiceless consonant is heard 50% of the time.  4.2 Subjects (criteria for selection) Two groups were recruited for this study – a young normal-hearing group comprising 7 adults younger than 31 years (mean age 23.6, SD 4.1), and an old-old (Jenstad 2006) normal-hearing group comprising 6 adults older than 70 years (mean age 74.1, SD 2.1). The total number of participants was 13. All participants were required to: - pass standard audiometric testing without use of hearing aids, exhibiting a threshold of 25dB HL or better for octaves between and including 500 and 4000 Hz, and no air-bone gaps greater than 10 dB at any frequency. - have vision adequate for using a touch-screen computer - score at least 28 out of 30 on the Mini-Mental State exam (Folstein, Folstein & 35  McHugh, 1975), a brief cognitive screen often used in hospitals. - have English as their first language since at least the age of 12. - be ambulatory for access to the sound booth  4.3 Stimuli The stimuli were generated using four real words (steam, seam, speed and seed) recorded monophonically using an AT3035 30 series AudioTechnica microphone routed through an Ultralite MOTU D/A pre-amplifier, and using the Praat software (Boersma & Weeninck, 2002) on an Apple iMac. The microphone was maintained approximately 10 cm from the speaker‘s lips during recording. Recordings were made at a sampling rate of 44100 Hz with the input level set to prevent clipping. Stimulus words were spoken by the co-investigator without carrier phrases and in a randomized order intermixed with words not used in the study. Each of the two words with the stop consonant present (steam and speed) contains a different place of articulation for English stop consonants (alveolar and bilabial, respectively), and each remains a word in English without the presence of the stop consonant (seam and seed) to avoid any semantic confounds due to the real/nonsense word distinction. The frequency of these four words in spoken English, as reflected in the Corpus of Contemporary American English (Davies, 2008), is reported in Table 1 and more fully in Appendix D. Though word frequency is not perfectly matched between the two stimulus word pairs, in both cases the relationship between the pair halves is qualitatively the same: the consonant-present word is significantly more frequent than the consonant-absent word.  36  Steam  Seam  Speed  Seed  6.98  0.31  31.31  6.43  Spoken frequency per million words  Table 1. Frequency of stimulus words in spoken English.  Words were chosen that began with /s/ rather than a vowel in order to minimize the presence of dynamic stop identity cues available in the material preceding the stop consonant. Words were also chosen whose nucleic vowels were monophthongs similar in place of articulation to the stop consonants, to reduce the effect of formant transition cues on the perception of a stop. Recordings of the consonant-present versus consonantabsent words were compared to determine the extent to which they differed in the formant dynamics of the vowel. Appendix A explains the perceptual differences discovered, using speed vs. seed and steam vs. seam. Perceptual differences were also noted in the quality of the /s/ frication preceding the stop consonant. These differences motivated the decision to use the consonant-absent tokens as the basis from which to build all tokens, assuring that no added cues to the stop‘s presence could be gleaned from either formant transitions or fricative transition sounds surrounding the stop consonant. Natural gap duration and arbitrary burst amplitude measures (measured with the Praat software after recording all stimuli with the same input level settings) are displayed in Table 2 for the two original recordings.  Word  Closure Gap  Arbitrary Burst Amplitude (dB)  Steam:  59 ms  60.17  Speed:  69 ms  57.23  Table 2. Natural burst amplitude and gap duration for two words used in stimulus creation.  37  Using the Praat software (Boersma & Weenink, 2002), release bursts were identified and excised from the consonant-present waveforms using Francis, Ciocca and Yu‘s (2003) criteria for determination of voicing onset, and visual inspection and perceptual measures were used to determine burst onset. Eighteen stimuli were generated from each of the two words using a matrix of gap duration and burst amplitude. First, the appropriate excised burst was inserted into the consonant-absent token. From this point, the gap was modified by adding silence before the inserted burst (see Figure 3). Modified gap durations ranged from 0 to 100 ms in 20 ms steps, as in Repp (1984). The upper gap duration of 100 ms was decided upon based on findings by Best, Morrongiello and Robson (1981) that a gap of 100 ms was sufficient to ensure that all participants reached high scores for perception of a voiceless stop consonant. Burst amplitude was varied by attenuating natural bursts in separate files using the Praat software (Boersma & Weenink, 2002) and then inserting them back into the appropriate stimuli. Burst was infinitely attenuated rather than excised from the waveform for the no-burst conditions in order to maintain the same overall stimulus duration between burst and no-burst conditions (see Figure 3). As the no-burst stimuli essentially include an extra gap portion equal to the duration of the burst, results are generally reported in terms of actual, rather than nominal, gap duration for each stimulus (cf. Figure 1, taken from Repp (1984)).  38  ―Speed‖ with full release burst and 40 ms nominal gap  ―Speed‖ with silenced release burst and 40 ms nominal gap (51 ms actual gap)  Figure 3. Waveform and spectrogram showing the nominal and actual gap for speed token with full burst and no burst.  4.4 Equipment and materials Stimuli were presented and responses recorded using a PC computer running SykofizX 2.0 software. Sounds were routed from the computer through a TDT S3 ZBus interface, TDT RP2.1 Enhanced Realtime Processor, TDT PA5 Programmable Attenuator, and TDT HB7 Headphone Driver to Sennheiser HD280 Pro 64Ω headphones. Response options were displayed, and responses recorded, through a touch computer screen in a sound-attenuated booth.  39  4.5 Data collection Details of data collection for hearing thresholds, gap detection thresholds, cognitive status, consonant perception, and response confidence are provided below.  4.5.1. Hearing thresholds Hearing thresholds for all subjects were tested using accepted clinical procedures (ANSI S3.6 1996). For the testing procedure, the participant was seated in a soundtreated room and insert earphones were placed in the ears after ear canals were visualized with an otoscope. Tones were presented to one ear at a time at different levels and the listener was asked to push a button when a tone was heard. Air conduction thresholds were recorded at octave intervals from 500 Hz to 4000 Hz. Bone conduction thresholds were tested at the same frequencies using a Radioear B71 50 Ω bone oscillator.  4.5.2. Gap detection testing A clinical measure of gap detection was conducted (custom software, Valter Ciocca, © 2004) in which silent gaps of 0, 2, 5, 10, 15, 20, 25, 30 and 40 ms, bounded by 17 ms broad-band noise bursts, were presented binaurally over headphones at comfortable listening levels. Broad-band markers were chosen because they more closely resembled the release bursts adjacent to closure gaps in natural speech than do pure-tone markers. Each participant performed a practice session in the presence of the experimenter in which 10 tokens each of the 0 ms gap and the 40 ms gap stimuli were presented. Once the participant‘s understanding of the protocol was confirmed, the testing began. Stimuli were presented in a randomized order with each stimulus being presented 10 times for a total of 90 responses. A gap detection threshold was determined 40  by observing the smallest gap duration for which gap detection scores were above chance (50%).  4.5.3. Cognitive status Clinical screening measures of health status, cognition, and memory were performed using the Mini Mental Status Exam (Folstein, Folstein & McHugh 1975) and the forward and reverse digit span subtests from the Wechsler Adult Intelligence Scale.  4.5.4. Speech perception task Testing, including a practice session, took approximately 45 minutes. Practice: Each participant performed a practice session in the presence of the experimenter. Stimuli containing either maximal (full burst amplitude and longest gap duration) or minimal (no burst amplitude and shortest closure gap duration) consonant cues for each word pair were presented 10 times each, in a randomized order determined by the Sykofizx 2.0 software, for a total of 40 stimulus presentations. Stimulus presentation was identical to that of the full speech perception task described below. The experimenter monitored the participant‘s accuracy on practice tokens by comparing responses against a hidden visual display of stimulus parameters. Accuracy was defined as a stop-present response for stimuli containing maximal cues and a stop-absent response for stimuli containing minimal cues. Feedback was provided verbally only if the participant exhibited difficulty with the practice stimuli. Practice responses were recorded for each participant. Once the participant‘s understanding of the protocol was confirmed and accuracy with practice material had been established at a criterion of 90%  41  correct classification, the testing began.  Repetitions: The 36 stimuli comprise 2 words (steam and speed) X 3 burst attenuation levels (0 dB, -12 dB, and infinite attenuation) X 6 gap durations (0, 20, 40, 60, 80 and 100 ms of nominal gap). Each stimulus was presented 10 times through the session for a total of 360 stimulus presentations. The stimuli were presented in 5 blocks of 72 tokens, each block containing 2 repetitions of each of the 36 tokens, all presented in a computer-randomized order.  Presentation method: All stimuli were presented monaurally to each participant‘s better ear (or to the right ear if no difference was indicated by the audiogram) through headphones at 70 dBA peak using Sykofizx 2.0 software to present stimuli and record results. Monaural presentation to either ear is considered acceptable because right ear advantage has only been demonstrated in the recall of lengthy stimulus sets under conditions of dichotic competition (Fry, 1974). Additionally, right ear advantages are evident in reaction times and not in accuracy measures (Samuel & Tartter, 1986).  Test Environment: The subject sat in a chair in a double-walled sound-isolated suite. After each stimulus presentation, the participant indicated the word he or she heard by touching that word on a touch-screen computer monitor. Neither response button was visible while the stimulus audio file was playing; this eliminated any priming effects that might occur due to the participant‘s reading one or the other of the response options while listening to the stimulus. Monaural presentation was chosen to take advantage of  42  any minor asymmetry in the older listeners‘ audiometric thresholds. Anticipating that older listeners might exhibit more asymmetry and slightly poorer thresholds even within the hearing criteria, using each listener‘s better ear helped to minimize within-criterion age-related threshold differences between the two age groups. Each token was followed by up to a 30 second response window (the response window terminated once the subject responded, allowing for as much time as the subject needed to respond). A two-way intercom system was activated at all times while the subject was in the sound-isolated suite. The subject was free to leave the suite at any time and was encouraged to take frequent breaks as needed.  4.5.5. Response confidence Five measures of response confidence were recorded during testing, with one confidence rating gathered after each block of 72 responses (2 phonemes X 3 burst amplitudes X 6 gap durations X 2 repetitions). Participants were asked to express their confidence on a 5-point scale where 5 represented absolute confidence, and 1 represented no confidence, in the accuracy of stimulus categorization. The response was elicited with the following query: ―On a scale of one to five, where five is completely confident and one is not at all confident, how confident did you feel about your responses for the last block?‖  43  5. Results Results from all test procedures, and the statistical methods used to analyze them, are reported below.  5.1 Statistical analysis and hypotheses The dependent variable in most analyses was the Stop Perception Threshold, the gap duration (in ms) at which the threshold of stop perception was achieved (50% stop responses). The between-subjects independent variable was Age (two groups: older and younger). The within-subjects independent variables were Phoneme (two phonemes: /p/ and /t/ in speed/seed and steam/seam contrast pairs, respectively) and Burst Amplitude (three levels: full or natural burst amplitude, -12 dB burst amplitude, and no-burst amplitude). A repeated measures ANOVA with one between-subjects factor (Age) and one within-subjects factor (Burst Amplitude) was performed for all burst conditions in the speed/seed contrast pair data, and for the full burst and -12 dB burst conditions in the steam/seam contrast pair data. Because the stop perception threshold was not reached for a substantial number of participants in the no-burst condition for the steam/seam contrast pair, these data were analyzed using a repeated measures ANOVA with one betweensubjects factor (Age) and one within-subjects factor (Gap Duration), with Stop-Present Response Rate as the dependent variable . The purpose of the study was to determine whether redundancy of speech cues affected stop consonant perception differently for older and younger listeners with normal hearing. The main question was whether older listeners would require greater cue redundancy than would younger listeners in perceiving stop consonants. Secondary goals 44  included replication of Repp‘s (1984) findings for young listeners, and determination of whether age-related results differed by phoneme. One younger participant (subject 105) was removed from the analysis because the gap durations required for her to reach stop perception threshold were more than two standard deviations away from mean gaps for her age group. Data from all other participants met this criterion for admissibility, and are provided in Appendix B. Descriptive data of the study‘s participants are provided in Table 3 below.  Participant #  Age  Health SelfReport (/10)  MMSE score (/30)  Gap detection threshold (ms)  Combined forward and reverse digit span scores  100  27:8  7  30  10  13  101  20:3  8  30  5  14  102  20:10  7  30  5  12  103  29:0  9  30  10  13  104  19:2  7.5  30  5  15  105*  25:2  9  30  10  15  106  24:11  8  30  5  15  900  70:9  8  29  10  10  901  74:5  9  29  10  11  902  73:8  4  29  5  12  903  73:11  5  30  10  9  904  77:2  7  29  10  10  905  74:8  6  29  5  8  Table 3. Descriptive data of the study sample. Note: *Participant 105 was excluded from analysis because she had stop perception thresholds two standard deviations away from the mean values for her age group. Older participants are indicated with boldface type.  45  5.1.1 Data requiring interpretation One participant in the younger group (subject 106) achieved stop perception threshold at two gap durations for the no-burst speed/seed contrast pair stimuli, the first being a down-going crossing of the threshold and the second being an up-going crossing (see Figure 4). Because the down-going crossing falls at 15 ms actual gap duration, which is more than 2 SD from the group mean (younger group mean: 62 ms, SD 8 ms), while the up-going crossing (61 ms actual gap duration) falls near the group mean, the second crossing was taken as the threshold. The reasons for this participant‘s suprathreshold stop-present response rate with such limited acoustic cues (11 ms actual gap and no burst) are not apparent.  Figure 4. Participant 106's speed/seed contrast pair response curves for the three burst amplitudes. Note: Horizontal line indicates the threshold of stop perception. Arrows indicate the two threshold crossings considered in determining a data point for the no burst condition.  46  5.2 Participant descriptive data Descriptive analysis of data concerning hearing and gap detection thresholds, digit span scores, and general health self-report scores are reported below.  5.2.1 Hearing thresholds Mean hearing thresholds are displayed by age group in Figure 5. An ANOVA was performed with Hearing Threshold (dB HL) as the dependent variable and Age (two groups: younger and older) and Test Frequency (five levels: 250 Hz, 500 Hz, 1000 Hz, 2000 Hz, and 4000 Hz) as the between-subjects and within-subjects independent variables, respectively. There was a marginal main effect of Test Frequency, p = .079, and there was no interaction between Age and Test Frequency, p = .976. There was a main effect of Age, p = .003. These findings show that older listeners had worse hearing thresholds regardless of test frequency for the frequencies tested. When the data were averaged across test frequencies, older listeners had a mean hearing threshold of 15.17 dB HL (SD = 2.86) and younger listeners had a mean hearing threshold of 4.83 dB HL (SD = 5.78).  47  Figure 5. Mean hearing thresholds and standard deviations by age group.  5.2.2 Gap detection thresholds The mean gap detection thresholds for the two age groups are shown in Table 4. This difference of approximately 1.6 ms (SD 2.6 ms) is less than is generally expected between young and old groups, and a two-tailed t-test shows that it is not statistically significant, p = .290. Fitzgibbons and Gordon-Salant (1996) quote studies showing gap detection thresholds for older listeners that were roughly double those of younger listeners. The gaps tested in this study were 0, 2, 5, 10, 15, 20, 25, 30 and 40 ms and all participants had thresholds of either 5 or 10 ms. These results only suggest that the groups are not grossly different in this one measure of temporal processing.  48  Age Group  Mean gap detection threshold (ms)  SD  Younger  6.67  2.58  Older  8.33  2.58  Table 4. Mean gap detection thresholds.  5.2.3 Digit span Digit span scores were collected as a measure of working memory. The mean results for each age group are reported in Table 5. Group scores were compared using two-sample t-tests assuming equal variance. The age groups showed significant differences in forward digit span scores, p = .002, two-tailed; in backward digit span scores, p = .003, two-tailed; and in combined digit span scores, p = .001, two-tailed. These scores indicate that the age groups differed significantly on a measure of working memory capacity.  Age group  Digits forward  Digits backward  Combined digit span  (SD)  (SD)  (SD)  Younger  7.67 (0.52)  6.00 (0.89)  13.67 (1.21)  Older  5.83 (0.98)  4.17 (0.75)  10.00 (1.41)  Table 5. Mean digit span scores.  5.2.4 General health self-report Self-reports of general health for the week leading up to testing were collected for each participant. Younger participants reported a mean health score of 7.75/10 (SD = 0.52). Older participants reported a mean health score of 6.50/10 (SD = 1.87). Because self-reported health is an ordinal variable, a non-parametric method of analysis was used. 49  The Mann-Whitney U-test showed that the difference in self-reported general health was not significant, p = .240, two-tailed.  5.3 Results for the speed/seed contrast pair stimuli The mean stop perception curves for the speed/seed data are shown in Figure 6. Gap durations (in ms) at stop perception thresholds are shown for each participant in Table 6.  Participant  Full  100 101 102 103 104 106 Mean SD  30.0 10.0 11.1 16.0 4.0 0.0 11.85 10.51  900 901 902 903 904 905 Mean SD  0.0 31.1 12.5 30.0 14.3 30.0 19.65 12.74  Burst Amplitude -12 dB Younger 34.3 20.0 28.9 24.0 25.7 5.0 22.98 10.04 Older 40.0 48.6 27.5 30.0 44.0 34.3 37.40 8.22  No burst (actual gap) 71.0 59.0 51.0 57.7 71.0 61.0 61.78 7.89 104.3 71.0 62.4 57.7 91.0 54.3 73.45 20.01  Table 6. Gap duration (ms) at stop perception threshold for the speed/seed contrast pair stimuli. Note: Older participants are indicated with boldface type.  50  Figure 6. Mean stop perception curves for the speed/seed contrast pair stimuli.  In Figure 6, data from older listeners are represented with black lines, and data from younger listeners are represented with grey lines. Solid, dashed, and dotted lines represent full burst, -12 dB burst, and no-burst conditions, respectively. Recall that actual gap refers to the nominal duration of the closure gap plus the duration of the silenced burst. The shape and relationship of the curves agree with Repp‘s (1984) findings (see Figure 1): as burst amplitude is decreased, longer gaps are required for listeners to meet stop perception threshold. To determine the effect of age in the speed/seed stop perception data, an ANOVA was conducted with Stop Perception Threshold (in ms of gap duration) as the dependent variable and Age and Burst Amplitude as the between-subjects and within-subjects independent variables, respectively. The analysis of variance showed that there was a significant main effect of Age, F(1,30) = 7.62, p = .010, partial η2 effect size = .429. As  51  shown in Table 6, stop perception thresholds were longer for older than for younger listeners in all burst conditions. The analysis showed that there was a significant main effect of Burst Amplitude in the speed/seed data, F(2, 30) = 57.51, p = .000, partial η2 effect size = .852. This effect replicates Repp‘s (1984) finding that longer stop perception thresholds were observed as burst amplitude decreased. To further investigate the Burst Amplitude effect, post-hoc paired samples t-tests were conducted for each possible pair of burst conditions with a p-value set to .016 for each comparison to maintain a family-wise p-value of .05. Significant effects of Burst Amplitude existed for all possible combinations of the three levels: full burst vs. -12 dB burst, p = .001; full burst vs. no burst, p = .000; and -12 dB burst vs. no burst, p = .000. In all cases, increased burst amplitude resulted in shorter stop perception thresholds. Results are displayed with 95% CIs in Figure 7.  52  Figure 7. Mean gap durations at stop perception threshold for various burst conditions of the speed/seed contrast pair.  It was hypothesized that there would be a significant interaction between Burst Amplitude and Age, giving evidence of an age-related redundancy effect. Specifically, it was predicted that older listeners would have stop perception thresholds that were increasingly worse than their juniors as burst amplitude, the redundant cue, decreased. The ANOVA analysis showed no significant interaction between Age and Burst Amplitude for the speed/seed contrast pair, F(2,30) = .22, p = .803, partial η2 effect size = .022.  5.4 Results for steam/seam contrast pair stimuli The mean stop perception curves for the steam/seam contrast pair data are shown in Figure 8. Data from older listeners are represented with black lines, and data from younger listeners are represented with grey lines. Solid, dashed, and dotted lines  53  represent full burst, -12 dB burst, and no-burst conditions, respectively. Stop perception thresholds are shown for each participant in Table 7. Because a majority of participants did not reach stop perception threshold in the no burst condition, only the full burst and -12 db burst data could be analyzed with Stop Perception Threshold (in ms of gap duration) as the dependent variable. Therefore, in the analysis for the no burst condition, Stop-Present Response Rate was used as the dependent variable. Age and Gap Duration were the between-subjects and withinsubjects independent variables, respectively, in both analyses.  54  Participant  Full  100 101 102 103 104 106 Mean SD  10.0 8.9 11.1 10.0 11.1 10.0 10.18 0.83  900 901 902 903 904 905 Mean SD  0.0 10.0 10.0 5.7 5.7 32.0 10.57 11.13  Burst Amplitude -12 dB Younger 30.0 10.0 23.3 16.7 14.3 14.3 18.10 7.28 Older 10.0 43.3 20.0 30.0 12.5 36.7 25.42 13.43  No burst (actual gap) 111.0 N/A N/A 97.7 91.0 84.3 N/A N/A  Table 7. Stop perception threshold (in ms of gap duration) for the full burst and -12 dB burst conditions of the steam/seam contrast pair. Note: Older participants are indicated with boldface type.  Figure 8. Mean stop perception curves for the steam/seam contrast pair.  55  5.4.1 Full burst and -12 dB burst conditions A repeated measures ANOVA was conducted on the full burst and -12 dB burst data. Box‘s Test of Equality of Covariance Matrices returned a significant value, p = .001, indicating that the data were heteroscedastic. Thus, a Greenhouse-Geisser correction factor was used to adjust the ANOVA results. The analysis of variance showed that there was no overall difference between younger and older listeners in terms of stop perception thresholds for tokens containing full bursts or -12 dB bursts (main effect of Age, F(1,20) = 0.662, p = .435, partial η2 effect size = .062). The analysis showed that there was a significant main effect of Burst Amplitude, F(1,20) = 8.69, p = .002, partial η2 effect size = .635. This means that increased Burst Amplitude resulted in better stop perception thresholds in this subset of the steam/seam contrast pair data. It was hypothesized that there would be a significant interaction between Burst Amplitude and Age, giving evidence of a redundancy effect. Specifically, it was predicted that older listeners would have stop perception thresholds that were increasingly worse than their juniors as burst amplitude, the redundant cue, decreased. In comparing the full and -12 dB burst for the steam/seam contrast pair data, the analysis showed no interaction with Age, F(1,20) = .81, p = .232, partial η2 effect size = .139.  5.4.2 No-burst condition The analysis for the no-burst condition used Stop-Present Response Rate as the dependent variable (because this provides data points in the absence of threshold crossings), and Age and actual Gap Duration (11 ms, 31 ms, 51 ms, 71 ms, 91 ms, and 111 ms) as the between-subjects and within-subjects factors, respectively. 56  Mauchly‘s Test of Sphericity returned a significant value, p = .000, indicating that the sphericity assumption for ANOVA was violated, so Greenhouse-Geisser corrections were applied in the analysis. The ANOVA showed a significant Age X Gap Duration interaction, p = .027, partial η2 effect size = .324. To further explore the effects of Age and Gap Duration within the interaction, the simple main effect of Gap Duration was determined for each age group. Gap Duration had a significant effect on Stop-Present Response Rate for older listeners, p = .007, partial η2 effect size = .699, but not for younger listeners, p = .140, partial η2 effect size = .341. The simple main effect of gap duration was further explored using orthogonal and independent comparisons between gap durations with 11 ms actual gap duration as the reference category (i.e., every gap duration was compared to the 11 ms gap condition). Results of this analysis are displayed in Table 8. Older listeners showed significant effects only in the comparison of the shortest gap duration of 11 ms to the two longest gap durations of 91 ms, p = .014, partial η2 effect size = .733, and 111 ms p = .010, partial η2 effect size = .765. For the younger listeners, no comparison of stop-present response rates for any pair of Gap Duration conditions produced significant results. A trend towards significance was seen only in the comparison of the shortest and longest gap duration conditions for this group, p = .069, partial η2 effect size = .517.  57  Gap contrast  p value  31 ms vs. 11 ms 51 ms vs. 11 ms 71 ms vs. 11 ms 91 ms vs. 11 ms 111 ms vs. 11 ms  .363 .175 .111 .259 .069  31 ms vs. 11 ms 51 ms vs. 11 ms 71 ms vs. 11 ms 91 ms vs. 11 ms 111 ms vs. 11 ms  .363 -* .110 .014 .010  Partial η2 effect size Younger .167 .333 .429 .245 .517 Older .167 -* .430 .733 .765  Table 8. Within-subjects contrasts of stop perception rate for various gap pairs within the steam/seam contrast no-burst data. Note: Significant results (p < .05) are in boldface. *Older group had a mean score of 0 at 51 ms, precluding contrast calculation.  Analysis of within-subjects effects showed no significant simple main effect of Age at any Gap Duration, p = .099, partial η2 effect size = .249. This means that though there appears to be a visible difference between the younger and older groups' response curves, this difference is not statistically significant although it does trend towards significance. The two age groups did not significantly differ in their Stop-Present Response Rates for this subset of the data.  5.5 Response confidence ratings Response confidence ratings on a five-point scale were collected after each block of 72 responses (a random ordering of 36 tokens X 2 repetitions each) during testing for a total of five ratings per participant. A rating of 1 was ―no confidence in my responses‖ and a rating of 5 was ―complete confidence in my responses‖. Younger listeners expressed complete confidence for only three of the 30 possible confidence ratings (5  58  blocks X 6 younger participants). Older listeners expressed complete confidence for 14 of the 30 possible confidence ratings (5 blocks X 6 older participants). The median response confidence ratings for each group, both within and across the five blocks of testing, are presented in Figure 9. Median values for all participants in both age groups were analyzed using the Mann-Whitney U-test for nonparametric data. Results showed that the difference in confidence between the groups trended towards significance, p = .076, two-tailed.  Figure 9. Median response confidence for each testing block and overall.  59  6. Discussion The main objective of this study was to test the hypothesis that older listeners rely more on acoustic cue redundancy than do younger listeners in speech perception tasks. Specifically, older listeners were expected to require the redundant presence of high amplitude release bursts as well as long duration closure gaps in order to perceive voiceless stop consonants. .  6.1 Hypothetical patterns of results Several potential patterns of results for this study were anticipated. Four possibilities are discussed below.  6.1.1 Possible result #1: no age-related difference If no difference in stop consonant perception as a function of the manipulated variables were found between old and young listeners in this study, it could suggest that older and younger listeners with similar hearing thresholds use speech cues and speech cue redundancy in the same way and to a similar extent. This pattern is represented in Figure 10 below. These results might provide evidence against an age-related decline in peripheral auditory system function. On the other hand, it might indicate a ceiling effect that hides a true but small difference between older and younger listeners. This would lead to a reworking of the variables in further experiments in order to catch more subtle differences.  60  For all stimuli: Younger  Older  Figure 10. Possible result #1: no age-related difference.  6.1.2 Possible result #2: age-related differences in all low-redundancy conditions The results of this study may confirm the hypothesis outlined in Jenstad (2006). That is, the older participants may show a significant decrease in speech recognition performance in low-redundancy conditions as compared to the younger participants. This result is represented in Figure 11 below. The hypothesized pattern of results would provide evidence that older adults use acoustic cue redundancy to enhance speech recognition in complex listening conditions, and that they require such redundancy to a greater extent than do their juniors.  61  For all stimuli: Younger  Older  Figure 11. Possible result #2: age-related differences in all low-redundancy conditions.  6.1.3 Possible result #3: age-related differences in all conditions A third possibility is that of an age-related decrement in performance in all conditions regardless of level of cue redundancy, as presented in Figure 12 below. This result would be evidence against the redundancy hypothesis, and would contradict the accepted idea that older listeners perform remarkably well in the simpler listening contexts. It might also indicate a floor effect where the high-redundancy stimuli created for this experiment are not redundant enough. This is a real possibility because of the decision to avoid formant transitions and fricative transitions in stimulus creation in favour of using only static cues to consonant identity.  62  For all stimuli: Younger  Older  Figure 12. Possible result #3: age-related differences in all conditions.  6.1.4 Possible result #4: age-related difference for specific stop consonants It may occur that significant age-related consonant recognition declines are found only for tokens with stop consonants at certain places of articulation or with certain following vowels. These results are represented in Figure 13 below. In Ohde and AbouKhalil (2001), older listeners showed increased difficulty in identifying stop-consonants followed by a low vowel as opposed to a high vowel when all other variables were held constant. The authors did not suggest a reason for this effect. A similar pattern of results in the current study would suggest that older and younger adults depend on different cues for the recognition of different consonants. This result might be of interest to those researchers working on optimizing hearing aid design and programming, and would suggest the necessity of further cue-weighting studies that looked more carefully at the older listener‘s perception of particular sounds or sound combinations.  63  For ―Speed/Seed contrast‖ Young er  Older  For ―Steam/Seam contrast‖ Younger Older  Figure 13. Possible result #4: age-related difference for specific stop consonants.  The actual results have some resemblance to the fourth pattern discussed above (Figure 13). There were inconsistent main effects of burst amplitude and gap duration. Increases in either acoustic cue tended to increase stop-present response rates. The exception is the no-burst condition of the steam/seam contrast in which increases in gap duration result in significant stop perception improvements for older listeners only. A main effect of age is seen for the speed/seed contrast but not for the steam/seam contrast. The only age interaction found was with gap duration for the steam/seam contrast no-  64  burst condition, in which older listeners showed higher stop perception rates as gap duration increased. No interaction between age and burst amplitude was found. The lack of an interaction between age and cue redundancy in the current findings provides evidence against the redundancy hypothesis. The results are separated by phoneme contrast pair and discussed more fully below.  6.2 Speed/seed contrast stimuli First, we discuss the speed/seed contrast data. As was found in previous cue weighting studies (e.g. Hedrick & Jesteadt, 1996), burst amplitude had a significant effect on stop perception. The presence of a full-amplitude burst accounted for an approximate 50-ms shortening of stop perception thresholds for both age groups with this data set. The gap duration of 61.8 ms at which younger listeners perceived a /p/ in the absence of a release burst was nearly identical to the value of 60.5 ms found by Repp (1984) for similar conditions. However, when the /p/ release burst was present, Repp‘s participants and ours showed markedly different thresholds. With a full /p/ burst present, younger listeners in the current study had mean stop perception thresholds at approximately 12 ms. Repp‘s young listeners had stop perception thresholds near 40 ms. There are some difficulties in making a direct comparison between these data sets however, as Repp summarized his data differently and had different stimulus characteristics. Firstly, he combined data from /p/, /t/, and /k/ stimuli, and the burst attenuation/amplification levels (two per phoneme) were not the same for all phonemes. Secondly, Repp truncated the /p/ burst to roughly half the duration of the original stimulus recordings, from 17 to 10 ms. This truncation likely contributed to lengthening the stop perception thresholds for burstpresent stimuli. The present study used 11-ms bursts that were of similar duration to 65  Repp's, but they were largely unmodified from their original recordings, with only a few pitch pulses excised during stimulus assembly to remove overlap with the fricative portion of the burst. Thirdly, Repp low-pass filtered his stimuli at 4.8 kHz, whereas our stimuli were not filtered spectrally and contained energy above 10 kHz. Therefore, it is possible that the bursts of /p/ tokens in the present study resulted in stop perception thresholds at shorter gaps because of their greater spectral and temporal similarity to natural bursts. To summarize, the difference between the present results and Repp‘s are likely resulting from the strong manner cues that are contained in natural, largely unmodified release bursts in comparison with substantially truncated, low-pass filtered release bursts. Notwithstanding the difference in stop perception thresholds, the current study extends Repp‘s (1984) findings by showing that the overall pattern of improved stop perception with increasing burst amplitude also holds for older listeners. For the speed/seed contrast data in this study, burst amplitude changes had an effect on stop perception that was similar in both direction and magnitude regardless of age. Age was associated with differences in stop perception thresholds for the speed/seed contrast across all burst conditions: older listeners showed an overall lengthening of stop perception thresholds on the order of 10 – 15 ms. This effect is not surprising, given the fact that older adults typically have longer gap detection thresholds (Pichora-Fuller & Singh, 2006). Though gap detection testing in this study showed a trend toward confirming these longer thresholds for older participants, this trend did not reach significance. Details of the gap detection testing are worth considering. All participants in this study exhibited gap detection thresholds of either 5 or 10 ms when presented with  66  a range of stimuli that included 0, 2, 5, 10, 15, 20, 25, 30 and 40 ms gap durations. The use of finer increments of gap duration and a larger sample size during threshold testing might have resulted in a finer resolution of thresholds and in significant differences being found between age groups. Evidence for the redundancy hypothesis would have appeared in this speed/seed contrast data set as an interaction between burst amplitude and age. That is, the difference between the low redundancy curve (no burst) and the high redundancy curves (full burst and -12 dB burst) would be greater for the older listeners. The results did not show such an interaction, withholding support for the redundancy hypothesis.  6.3 Steam/seam contrast stimuli Results for the steam/seam contrast showed different patterns from those seen for the speed/seed contrast. Recall that the analysis of the steam/seam contrast data set was split to accommodate a lack of threshold crossings in the no-burst condition. The steam/seam contrast analysis for bursts-present was done in the same way as for the speed/seed contrast data but with the exclusion of the no-burst condition. When bursts were present, an age effect was not seen. That is to say, for /t/ tokens that included either full or partially attenuated bursts, older and younger listeners had similar stop perception thresholds. This is in contrast to the age effect seen in the speed/seed contrast data. Results for the steam/seam contrast showed a main effect of burst amplitude, as was the case with the speed/seed contrast. The increase from -12 dB to full burst amplitude shortened stop perception thresholds on the order of approximately 10 – 15 ms. Repp‘s (1984) discussion of a trading relationship between release burst and gap duration cues fits these data well: increases in burst amplitude can offset the effects of shorter duration 67  closure gaps. Evidence for the redundancy hypothesis would have appeared in the steam/seam contrast data set as an interaction between burst amplitude and age. There was no interaction between age and redundancy and therefore no evidence for the redundancy hypothesis. A significant and unexpected interaction between age and gap duration was revealed in the steam/seam contrast no-burst data. Older listeners showed increasing stop perception rates as the gap duration increased, whereas younger listeners did not. Even with the longest of actual gaps (111 ms, roughly twice as long as the natural closure gap recorded for /t/), younger listeners did not meet stop perception threshold or show a main effect of gap duration. It is not likely that the younger listeners were unable to detect the closure gaps. The gap detection results in this study did not show any significant difference between the groups in this regard, and previous studies (e.g. Lister & Tarver, 2004; Pichora-Fuller, Schneider et al., 2006; Strouse, Ashmead,Ohde & Grantham, 1998) have found that older, not younger, listeners tend to have poorer gap detection thresholds. It is more likely that the difference is one of perceptual processing of the cue rather than of cue detection. The redundancy hypothesis would predict younger listeners to make more use of gap duration as a lone cue to the presence of a stop than older listeners. Possible reasons for the actual pattern of results are discussed below.  6.4 Gap duration as a cue to the perception of stops by younger and older listeners Previous discussion in the literature may help in explaining younger listeners' perception of /p/, but not /t/, based on gap duration alone. Repp (1984), and Bailey and Summerfield (1980), concluded that closure gap duration was the predominant cue to  68  place of articulation, and that the spectral and temporal characteristics of placeappropriate bursts were secondary. Specifically, it has been suggested that a long gap will tend to elicit a perception of /p/. Repp (1984) found that listeners perceived /p/ based on a long gap cue far more than they perceived /t/ or /k/, even when the gap was accompanied by the burst from a different phoneme (/t/ or /k/). Bailey and Summerfield (1980) found that for syllables of /s/ - silence – vowel, /p/ and /k/ were perceived between the fricative and the vowel depending on the second-formant frequency of the vowel, but that /t/ was rarely perceived with any second formant frequency. These findings would suggest that participants listening to the no-burst steam/seam contrast tokens in the current study may have been perceiving 'speam' and not 'steam'. There was qualitative support for this in the comments made by some participants who expressed more dissatisfaction with the forced choice alternatives for the steam/seam contrast than with those for the speed/seed contrast. Some requested a ‗speam‘ response button for the steam/seam tokens, but no parallel requests were made for the speed/seed contrast. Previous research also suggests that older listeners employ more liberal response criteria in word recognition tasks (e.g. Gordon-Salant, 1986a). If this is the case, older listeners may have been liberal in applying the label of ‗steam‘ to what they had perceived as ‗speam‘ given the forced choice paradigm. The younger listeners, with more conservative response criteria, may have been reserving ‗steam‘ responses only for tokens that that did not present this ambiguity. Trends in the response confidence ratings and a survey of comments made during testing, discussed below, support this interpretation.  69  Though it did not reach statistical significance, there was a trend towards higher response confidence in older listeners throughout all blocks of testing (see Figure 6). The response confidence trend is supported qualitatively by a theme in the comments participants made after stating their confidence scores (See Appendix C for a listing of all comments). Younger participants‘ comments showed that they were aware of cue manipulation in the experimental stimuli, whereas older participants seemed to think that the stimuli were naturally-spoken tokens. For example, the younger listeners made the following individual comments:  Participant 101: ―The space between the /s/ and the next letter was getting smaller.‖ Participant 105: ―If you think about it too long, like, ‗was that just a gap or actually a /p/?‘ you can confuse yourself.‖ ―Sometimes it cuts out to make you think there‘s a /t/.‖ Participant 106: ―A few I was suspicious of because they didn‘t sound like a total ‗seam‘ or ‗seed‘."  Other young participants made similar comments, suggesting not only that they were aware that the stimuli had been manipulated along various parameters but also that they were responding carefully, sometimes waiting for what they felt to be proper or ―total‖ tokens (see last quote above) before they indicated the presence of a stop  70  consonant. Comments from the older listeners were less likely to suggest awareness that the tokens were anything but recordings of natural speech. Below are selected examples of comments made by older listeners:  Participant 900: ―I made a few errors in hurrying.‖ ―I am confident, but interested to know whether results bore this out.‖ Participant 901: ―I made a keystroke error.‖ Participant 902: ―Upgraded from four out of five, damn it! I‘m not sure I got even one wrong.‖ ―There was one, or a few, that were vague. I missed one, I think.‖ Participant 904: ―Maybe one or two [were wrong]. I felt I heard them properly.‖ ―A couple I was weak on.‖ Participant 905: ―99% confident: well, actually, 100%, but…‖  Note that in the comments above, as well as in the numerical confidence results, older participants quite often reported complete confidence, feeling that they ―didn‘t miss a single one.‖ After the completion of testing, one of the older participants asked whether someone had had to read out the long lists of stimuli in one recording session, indicating a belief that the experiment was in some sense a continuous natural recording  71  and not a set of tokens that had been individually manipulated. It may be that younger listeners are as a cohort more familiar with manipulation of auditory media. Once they became aware that the tokens for this study were likely to have been manipulated, which their comments corroborate, they may have intentionally adopted a more conservative response bias to avoid giving stop-present responses for obviously modified tokens (see again the comments made by participant 105 above). In contrast, older listeners as a cohort may be more likely to assume that each token is a recording of natural speech. They therefore appear to have had fewer reservations about their forced choice options and about delivering consonant-present responses. To sum up, Gordon-Salant‘s (1986a) concept of more liberal response criteria for older listeners, along with possible generational differences in exposure to manipulated auditory material, appears to describe both the data and the experience of the participants. This is most apparent for the steam/seam no-burst tokens because they are especially ambiguous: as was discussed above, the long gap tends to create the percept of /p/, but the forced choice options given allowed only responses indicating /t/ or the absence of any stop consonant altogether. Older listeners were more confident in liberally applying the /t/ label, whereas younger listeners more conservatively withheld the label, waiting for a "total" token. This is not to suggest that liberal or conservative response criteria were entirely conscious, though the comments listed above indicate that at least some of the participants were using explicit strategies. A related theme in the participants‘ comments is the frequency with which older listeners attributed errors to details of task execution (i.e. replying too hastily or making keystroke errors), whereas the younger listeners were more likely to discuss  72  characteristics of the stimuli that posed difficulty and reduced their confidence. If older listeners mentioned problematic tokens at all, they tended to label them as sounding "vague" or "off", and none of the older listeners expressed a lack of confidence about the token set as a whole. Future studies may benefit from measures of response confidence taken for each token type or for sets of tokens rather than for randomized blocks including all tokens. It would be valuable to know whether response confidence differences were greater for the no-burst steam/seam contrast data where the greatest stop perception difference between age groups was seen. Likewise, it would be interesting to track the comments made by younger and older listeners when presented with cue variations constrained to a more natural range (e.g. gap durations that do not exceed canonical or recorded values). Would such a stimulus set, designed to avoid over-caution in younger listeners, still have a wide enough range of cue levels that both groups could meet stop perception threshold? Especially with the most natural sounding stimulus sets, matching token-specific confidence ratings to the kinds of comments collected in this study would provide a connection between objective and subjective aspects of speech perception tasks. It is not yet clear whether these subjective differences generalize from repetitive speech tasks to perception of speech in everyday life. Acoustic differences between the phonemes, and age-related changes in the ability to hear them, may also go some way in explaining the differences seen between phoneme contrast pairs. The hearing screening that was used to select normal-hearing participants for this study tested thresholds at octaves from 250 Hz through 4 kHz. It is possible that there were acoustic differences between /p/ and /t/ that existed in frequencies beyond the screened thresholds. If the age groups differed in their hearing  73  thresholds above 4 kHz, cues to one or both of the phoneme contrast pairs may have been more accessible to one age group. Age-related hearing loss typically affects higher frequency thresholds earlier and to a greater extent than lower frequencies. Making the assumption that our older participants' thresholds for frequencies higher than 4 kHz would be worse than their juniors', we would expect higher frequency acoustic cues to be more difficult for older listeners to perceive. An analysis of the spectral content of each burst (see Figures 14 and 15) shows that /t/ contained more energy between 4 kHz and 8 kHz than did /p/. Older listeners‘ potentially poorer hearing thresholds in this frequency range may have been adequate for hearing the more intense high-frequency components of /t/ but not the less intense components of /p/ falling within the same frequency band. This would provide a potential explanation for the fact that age effects are seen for burstpresent /p/ tokens but not burst-present /t/ tokens. On the other hand, if older listeners‘ hearing thresholds were poorer still, the high frequency components of both phonemes may have been inaccessible. If that were the case, we might expect the older listeners to perceive /p/ more readily than /t/ because its burst contains more spectral energy in the lower frequencies for which their hearing thresholds are known to be normal. This, however, does not fit the data, as age effects appear only for /p/. Without establishing hearing thresholds at 8 kHz, we cannot be certain to what extent age-related hearing losses at the highest audiometric frequency explain the difference in results for /p/ and /t/. Older listeners' poorer hearing thresholds (approximately 10 dB poorer than for the younger listeners, regardless of test frequency) do not appear to be a decisive factor in these results either. Poorer overall thresholds would be expected to have the greatest effect for the least intense cues. Results from the least intense burst-present condition of  74  the steam/seam contrast (-12 dB burst) show no significant difference between younger and older listeners. It appears then that neither the possibility of unmeasured age-related hearing loss in the higher frequency thresholds nor the screened threshold differences between age groups adequately explain our results. This is not surprising, given that the same conclusion has been reached in previous speech perception studies with similar participants. Nabelek (1988), for instance, found that age was not correlated with performance on a vowel identification task for normal-hearing young and old groups separated by a difference in hearing thresholds similar to ours (older participants had mean thresholds 9.1 dB higher for test frequencies not exceeding the 4 kHz used in his study). Spectral similarity between bursts and their preceding fricatives does not appear to be a problem with this stimulus set. If the spectra of /s/ and the burst had been too similar, the presence of the burst would have been largely imperceptible for tokens in which the intervening gap was short enough. If /s/ and the bursts were indistinguishable spectrally, tokens in which burst presence was the only cue would have held no detectable cues. However, spectral analysis of the bursts and fricatives of the current study‘s stimuli shows that spectral differences existed even within the frequency range used in screening the participants for normal hearing. Figures 14 and 15 show spectra of the burst and fricative of the speed/seed and steam/seam contrasts, respectively. These spectra were created using a 100 Hz analysis bandwidth and no pre-emphasis. The burst spectra were calculated from the entire burst duration, and the fricative spectra were calculated from a large steady-state portion (>90%) of each fricative.  75  Speed /p/ burst spectrum: 4000 Hz  Sound pressure level (dB/Hz)  40  20  0  -20  -40 100  200  500  1000 2000 Frequency (Hz)  5000  104  Speed /s/ fricative spectrum: 4000 Hz 40  Sound pressure level (dB/Hz)  20  0  -20  -40 100  1000 2000 5000 104 Frequency (Hz) Figure 14. Spectra of the burst and fricative from the speed/seed contrast pair. Note: Vertical line indicates the maximum frequency for which the participants‘ hearing thresholds are known. 200  500  76  Steam /t/ burst spectrum: 4000 Hz  Sound pressure level (dB/Hz)  40  20  0  -20  -40 100  200  500  1000 2000 Frequency (Hz)  5000  104  Steam /s/ fricative spectrum: 4000 Hz  Sound pressure level (dB/Hz)  40  20  0  -20  -40 100  1000 2000 5000 104 Frequency (Hz) Figure 15. Spectra of the burst and fricative from the steam/seam contrast pair. Note: Vertical line indicates the maximum frequency for which the participants‘ hearing thresholds are known. 200  500  77  Both instances of /s/ were characterized by highest energy above approximately 4 - 5 kHz, and both bursts had highest energy below these frequencies, ensuring that there was a distinct spectral change in the transition from the fricative to the burst, regardless of whether the intervening short-duration gaps were detected. Even when the spectral comparison was limited to frequencies below the highest screening frequency (4 kHz), fricatives and bursts were spectrally distinct. Though relative word frequency within the steam/seam and speed/seed contrast pairs was considered in the selection of stimuli for this study, it is possible that there may have been differences between the age groups in the extent to which each forced-choice option (and its homophones, in the case of seed and seam, whose pronunciations are the same as cede and seem, respectively) activated facilitative effects. There may have been cohort differences in exposure to the stimulus words. Though this would be useful in explaining, for example, why younger listeners did not perceive 'seam' as frequently in the no-burst steam/seam contrast conditions, we do not have the data to establish either the actual exposure each group has had to each word, or to measure the facilitative effects such exposure might have had. Future studies might incorporate a method for determining the familiarity of each participant with the stimulus words. Whatever the causes behind the phoneme-based differences seen in this study, further testing should expand on our methodology by including other phoneme contrast pairs, and by placing /p/ and /t/ in different stimulus words, to determine to what extent our findings can be generalized. Central to the discussion of the redundancy hypothesis is the question of how this study operationalized redundancy and non-redundancy. The redundancy hypothesis is  78  meant to explain the difficulty experienced by older listeners in complex listening conditions. To reiterate, it is hypothesized that they are challenged, in particular, by the reduction in acoustic cue redundancy that such conditions often entail. In this study, complex experimental conditions were created by editing natural speech stimuli so that participants had to rely on only one or two acoustic cues to detect the presence of stop consonants. That no age/redundancy interaction was found could suggest either that the hypothesis does not hold in general, or that this study did not provide conditions conducive to observing a redundancy effect. Ohde and Abou-Khalil (2001) drew a distinction between cue types. They noted that an age effect was seen particularly in the integration of secondary cues, which in their case included voicing duration and formant transitions. Likewise, Fox, Wall, and Gokcen (1992) showed that older listeners were particularly challenged by tokens containing only unsupported secondary cues. Repp (1984) held that closure gap duration, one of the cues used in the current study, was a primary cue to stop identity. It appears possible that the presence of primary cues reduces the necessity of integrating secondary cues, regardless of age. Thus, determining the hierarchy or relative weighting of cues for each phoneme should be an on-going aim of cue-weighting studies. Though changes in cue weighting due to effects of hearing loss and amplification have received attention (e.g. Hedrick & Carney, 1997; Hedrick & Younger, 2001; Wang & Humes, 2008), the separate effect of age on cue weighting needs further investigation, as it will affect the design of studies testing the redundancy hypothesis. Future studies might operationalize redundancy differently than we have, using varying numbers of cues proven to be secondary for older listeners rather than the one primary and one secondary cue used in this study.  79  6.5 Relationship to time compression studies Time compression studies indicated that both reduced processing time (e.g. Wingfield, Wayland & Stine, 1992) and removal of redundant acoustic cue material (Jenstad & Souza, 2007; Wingfield & Ducharme, 1999) were problems for older listeners, and that these difficulties were most apparent for consonants (Gordon-Salant & Fitzgibbons, 2001). If so, why were such findings not borne out in the current study? Some differences between time compression studies and the current cue weighting study are worth noting. First, time compression has frequently been accomplished by the excision of large numbers of 20 ms sections from the speech signal, often because the same Lexicon Varispeech compression equipment was used (e.g. Wingfield, Poon, Lombardi, & Lowe, 1985; Wingfield & Ducharme, 1999). Gordon-Salant and Fitzgibbons (2001), using other equipment, excised 5 – 15 ms sections from multiple locations within the signal. In the current study, though gap durations were incremented in similar 20 ms steps, burst amplitude manipulations were always temporally bounded by the release burst‘s natural 11 ms duration. Compared to a typical time-compression study, these manipulations are of short duration, are aligned exactly with the cue‘s temporal boundaries, and are single instances rather than pervasive throughout the stimulus. Crucially, in no condition in the current study was only part of a burst attenuated. The pervasive excisions typically used in time compression studies cannot be presumed to have fallen with such precise timing that only a single acoustic cue was affected in each case. Gordon-Salant and Fitzgibbons (2001) pointed out that time compression techniques may be creating added complexity by removing numerous small fragments in close temporal proximity to each other, each of which could constitute a 80  whole cue, a number of cues, or a portion of a cue. They also remarked that their compression techniques may have been removing non-redundant as well as redundant cues. The current study, in respecting the temporal integrity of every burst cue and in attenuating bursts always in their entirety, could be considered to be less complex. Second, typical time compression studies judge speech recognition not by the detection of individual phonemes, but by verbal recall (e.g. Wingfield, Wayland & Stine, 1992; Wingfield & Ducharme, 1999) or written recall of words, phrases, or sentences (see, e.g., Gordon-Salant & Fitzgibbons, 2001). Such responses are more cognitively demanding because they rely to a greater extent on memory, and because the pool of potential responses is not bounded by a two-alternative forced choice as it was in the current study. These are yet further ways in which time compression studies could be considered to provide more complexity than the current study.  6.6 Complexity and speech recognition for older listeners What can be concluded about speech recognition declines in age, then? To summarize, we know that they are reported subjectively by older listeners with and without age-related hearing loss in everyday listening situations. We also know that some level of speech signal complexity is needed for age-related declines to become apparent beyond any effects of hearing loss. This is borne out both in subjective reports and in experimental conditions (CHABA, 1988). We know that impairments in cognitive/perceptual processing abilities, as well as some improvements, are associated with aging (Akeroyd, 2008; Schneider, Daneman & Murphy, 2005; Sulzenbruck et al., 2010; Wingfield, Lindfield & Goodglass, 2000; Wingfield & Tun, 2001; Wingfield, Wayland & Stein, 1992). Time compression studies show that a lack of processing time 81  and degraded speech input are both challenges to older listeners in complex listening conditions (Gordon-Salant & Fitzgibbons, 2001; Jenstad & Souza, 2007). We know that age-related redundancy effects have been found across many modalities (Allen et al., 1992; Bucur & Madden, 2005; Laurienti et al, 2005). Combining these facts with the results of this study, it seems likely that age-related redundancy effects do exist, but that they are unlikely to be seen in conditions that provide little cognitive load. In lessconstrained situations, such cognitive loads grow to a critical level in crowded, noisy, reverberant listening conditions. In time compression studies, such cognitive loads amass in the layered difficulties posed by temporal distortions, limited processing time, and the partial or full removal of redundant and non-redundant acoustic cues. The current study intentionally provided listening conditions with lower cognitive loads in order to determine whether cue redundancy alone was enough to elicit an age-related effect. That it was not able to do so suggests that the difficulty experienced by older listeners is not likely to be explained fully by peripheral or cognitive declines alone, but rather by an interaction of the two. Future studies may benefit from designs that allow for the manipulation of cognitive loading along with levels of cue redundancy.  6.7 Conclusions The results of this study allow for the following inferences. For the stimuli in this study: i.  Higher burst amplitude increases the likelihood that /p/ and /t/ will be perceived by both younger and older listeners with normal hearing.  ii.  Longer closure gap durations generally increase the likelihood that stop 82  consonants will be perceived. For older listeners this was true for both /p/ and /t/. For younger listeners, it held everywhere except for /t/ in the steam/seam no-burst conditions. iii.  Older listeners exhibit longer stop perception thresholds than younger listeners for /p/ but not /t/.  iv.  Age and burst amplitude do not interact in the perception of stop consonants. The absence of this interaction is evidence against the redundancy hypothesis.  v.  Only older adults make use of increasing gap duration in the perception of /t/ if closure gap is the sole cue to the consonant‘s presence.  vi.  Older listeners appear to be less conscious of cue manipulation and employ less conservative response criteria in reporting the presence of stop consonants.  83  References Akeroyd, M. A. (2008). Are individual differences in speech reception related to individual differences in cognitive ability? A survey of twenty experimental studies with normal and hearing-impaired adults. International Journal of Audiology, 47(S2), S53.  Alain, C., McDonald, K. L., Ostroff, J. M., & Schneider, B. (2001). Age-related changes in detecting a mistuned harmonic. The Journal of the Acoustical Society of America, 109, 2211.  Allen, P. A., Madden, D. J., Groth, K. E., & Crozier, L. C. (1992). Impact of age, redundancy, and perceptual noise on visual search. Journal of Gerontology, 47(2), P69.  Aylett, M., & Turk, A. (2004). The smooth signal redundancy hypothesis: A functional explanation for relationships between redundancy, prosodic prominence, and duration in spontaneous speech. Language and Speech, 47(1), 31.  Bazzanella, C. (2011). Redundancy, repetition, and intensity in discourse. Language Sciences, 33(2), 243-254.  Best, C. T., Morrongiello, B., & Robson, R. (1981). Perceptual equivalence of acoustic cues in speech and nonspeech perception. Perception & Psychophysics, 29(3), 191– 211.  84  Boersma, P., & Weenink, D. (2005). Praat: doing phonetics by computer (Version 4.3.01) [Computer program]. Retrieved from http://www.praat.org/  Bopp, K. L., & Verhaeghen, P. (2009). Working memory and aging: Separating the effects of content and context. Psychology and Aging, 24(4), 968-980.  Bucur, B., Madden, D. J., & Allen, P. A. (2005). Age-related differences in the processing of redundant visual dimensions. Psychology and Aging, 20(3), 435.  Chiari, I. (2007). Redundancy elimination: The case of artificial languages. Journal of Universal Language, 8, 7-38.  Committee on Hearing, Bioacoustics and Biometrics (CHABA). (1988). Speech understanding and aging. The Journal of the Acoustical Society of America, 83(3), 859-895.  Cranford, J. L., & Stream, R. W. (1991). Discrimination of short duration tones by elderly subjects. Journal of Gerontology, 46(1), P37.  Daneman, M., & Carpenter, P. A. (1980). Individual differences in working memory and reading. Journal of Verbal Learning and Verbal Behavior, 19(4), 450-466.  Davies, M. (2008-). The corpus of contemporary american english (COCA): 410+ million words, 1990-present. Retrieved 01/07, 2011, from http://www.americancorpus.org  85  Divenyi, P. L., & Haupt, K. M. (1997). Audiological correlates of speech understanding deficits in elderly listeners with mild-to-moderate hearing loss. I. age and lateral asymmetry effects. Ear and Hearing, 18(1), 42.  Divenyi, P. L., & Simon, H. J. (1999). Hearing in aging: Issues old and young. Current Opinion in Otolaryngology & Head and Neck Surgery, 7(5), 282.  Dorman, M. F., Marton, K., Hannley, M. T., & Lindholm, J. M. (1985). Phonetic identification by elderly normal and hearing-impaired listeners. The Journal of the Acoustical Society of America, 77(2), 664.  Dubno, J. R., Dirks, D. D., & Morgan, D. E. (1984). Effects of age and mild hearing loss on speech recognition in noise. The Journal of the Acoustical Society of America, 76, 87.  Dubno, J. R., Lee, F. S., Matthews, L. J., & Mills, J. H. (1997). Age-related and genderrelated changes in monaural speech recognition. Journal of Speech, Language, and Hearing Research, 40(2), 444.  Elliott, L. L., Busse, L. A., & Bailet, L. L. (1985). Identification and discrimination of consonant-vowel syllables by younger and older adults. Perception & Psychophysics, 37(4), 307-314.  Fitzgibbons, P. J., & Gordon-Salant, S. (1996). Auditory temporal processing in elderly listeners. Journal-American Academy of Audiology, 7, 183-189.  86  Fitzgibbons, P. J., & Gordon-Salant, S. (1994). Age effects on measures of auditory duration discrimination. Journal of Speech & Hearing Research, 37(3), 662.  Folstein, M. F., Folstein, S. E., & McHugh, P. R. (1975). "Mini-mental state". A practical method for grading the cognitive state of patients for the clinician. Journal of Psychiatric Research, 12(3), 189-198.  Fox, R. A., Wall, L. G., & Gokcen, J. (1992). Age-related differences in processing dynamic information to identify vowel quality. Journal of Speech and Hearing Research, 35(4), 892-902.  Fozard, J. L., Vercruyssen, M., Reynolds, S. L., Hancock, P. A., & Quilter, R. E. (1994). Age differences and changes in reaction time: The baltimore longitudinal study of aging. Journal of Gerontology, 49(4), P179.  Francis, A. L., Ciocca, V., & Yu, J. M. C. (2003). Accuracy and variability of acoustic measures of voicing onset. The Journal of the Acoustical Society of America, 113, 1025.  Fry, D. B. (1974). Right ear advantage for speech presented monaurally. Language and Speech, 17(2), 142.  Gelfand, S. A., Piper, N., & Silman, S. (1985). Consonant recognition in quiet as a function of aging among normal hearing subjects. The Journal of the Acoustical Society of America, 78, 1198.  87  Gelfand, S. A., Piper, N., & Silman, S. (1986). Consonant recognition in quiet and in noise with aging among normal hearing listeners. The Journal of the Acoustical Society of America, 80, 1589.  Glisky, E. L. (2007). Changes in cognitive function in human aging. In D. R. Riddle (Ed.), Brain aging: Models, methods, and mechanisms. (pp. 3-20). Boca Raton (FL): CRC Press.  Gordon-Salant, S. (1986a). Effects of aging on response criteria in speech-recognition tasks. Journal of Speech and Hearing Research, 29(2), 155.  Gordon-Salant, S. (1986b). Recognition of natural and time/intensity altered CVs by young and elderly subjects with normal hearing. The Journal of the Acoustical Society of America, 80, 1599.  Gordon-Salant, S., & Fitzgibbons, P. J. (1993). Temporal factors and speech recognition performance in young and elderly listeners. Journal of Speech and Hearing Research, 36(6), 1276.  Gordon-Salant, S., & Fitzgibbons, P. J. (1995a). Comparing recognition of distorted speech using an equivalent signal-to-noise ratio index. Journal of Speech and Hearing Research, 38(3), 706.  Gordon-Salant, S., & Fitzgibbons, P. J. (1999). Profile of auditory temporal processing in older listeners. Journal of Speech, Language, and Hearing Research, 42(2), 300.  88  Gordon-Salant, S., & Fitzgibbons, P. J. (2001). Sources of age-related recognition difficulty for time-compressed speech. Journal of Speech, Language, and Hearing Research, 44(4), 709.  Gordon-Salant, S., & Fitzgibbons, P. J. (1995b). Recognition of multiply degraded speech by young and elderly listeners. Journal of Speech and Hearing Research, 38(5), 1150-1156.  Hällgren, M., Larsby, B., Lyxell, B., & Arlinger, S. (2001). Cognitive effects in dichotic speech testing in elderly persons. Ear and Hearing, 22(2), 120.  Haubert, N. (1999). Speech perception by older listeners: Contributions of auditory temporal processes. University of British Columbia).  He, N., Dubno, J. R., & Mills, J. H. (1998). Frequency and intensity discrimination measured in a maximum-likelihood procedure from young and aged normal-hearing subjects. The Journal of the Acoustical Society of America, 103, 553.  Hedrick, M. (1997). Effect of acoustic cues on labeling fricatives and affricates. Journal of Speech, Language, and Hearing Research, 40(4), 925.  Hedrick, M. S., & Carney, A. E. (1997). Effect of relative amplitude and formant transitions on perception of place of articulation by adult listeners with cochlear implants. Journal of Speech, Language, and Hearing Research, 40(6), 1445.  Hsia, H. J. (1977). Redundancy: Is it the lost key to better communication? Educational Technology Research and Development, 25(1), 63-85. 89  Humes, L. E. (1996). Speech understanding in the elderly. Journal-American Academy of Audiology, 7, 161-167.  Jenstad, L. M. (2005). The effect of temporal envelope changes on recognition of normal rate and time-compressed speech by young-old and old-old hearing-impaired listeners. Unpublished doctoral dissertation, University of Washington.  Jenstad, L. M. (2006). Speech perception and older adults: Implications for amplification. Hearing Care for Adults: An International Conference, Chicago, IL. 13.  Jenstad, L. M., & Souza, P. E. (2007). Temporal envelope changes of compression and speech rate: Combined effects on recognition for older adults. Journal of Speech, Language, and Hearing Research, 50(5), 1123.  Keefe, D. H., Schairer, K. S., Ellison, J. C., Fitzpatrick, D. F., & Jesteadt, W. (2009). Use of stimulus-frequency otoacoustic emissions to investigate efferent and cochlear contributions to temporal overshoot. The Journal of the Acoustical Society of America, 125, 1595.  Keifer Jr, O. P. (2009). Redundancy gain: Correlations across s sensory modalities from a neurologically normal population.  König, E. (1957). Pitch discrimination and age. Acta Oto-Laryngologica, 48(5-6), 475489.  Laurienti, P. J., Burdette, J. H., Maldjian, J. A., & Wallace, M. T. (2006). Enhanced multisensory integration in older adults. Neurobiology of Aging, 27(8), 1155-1163. 90  Lavandier, M., & Culling, J. F. (2010). Prediction of binaural speech intelligibility against noise in rooms. The Journal of the Acoustical Society of America, 127, 387.  Liberman, A. M., Delattre, P. C., Cooper, F. S., & Gerstman, L. J. The role of consonantvowel transitions in the perception of the stop and nasal consonants American Psychological Association.  Lister, J., & Tarver, K. (2004). Effect of age on silent gap discrimination in synthetic speech stimuli. Journal of Speech, Language, and Hearing Research, 47(2), 257.  Lorenzi, C., Gilbert, G., Carn, H., Garnier, S., & Moore, B. C. J. (2006). Speech perception problems of the hearing impaired reflect inability to use temporal fine structure. Proceedings of the National Academy of Sciences, 103(49), 18866.  Lutman, M. E., Gatehouse, S., & Worthington, A. G. (1991). Frequency resolution as a function of hearing threshold level and age. The Journal of the Acoustical Society of America, 89, 320.  Magnusson, L. (1996). Predicting the speech recognition performance of elderly individuals with sensorineural hearing impairment A procedure based on the speech intelligibility index. Scandinavian Audiology, 25(4), 215-222.  Mattay, V. (2006). Neurophysiological correlates of age-related changes in working memory capacity. Neuroscience Letters, 392(1-2), 32-37.  McGurk, H., & MacDonald, J. (1976). Hearing lips and seeing voices. Nature, 264, 746748. 91  Moore, B. C. J., & Peters, R. W. (1992). Pitch discrimination and phase sensitivity in young and elderly subjects and its relationship to frequency selectivity. The Journal of the Acoustical Society of America, 91, 2881. Nábělek, A. K. (1988). Identification of vowels in quiet, noise, and reverberation: Relationships with age and hearing loss. The Journal of the Acoustical Society of America, 84, 476.  Nossair, Z. B., & Zahorian, S. A. (1991). Dynamic spectral shape features as acoustic correlates for initial stop consonants. The Journal of the Acoustical Society of America, 89, 2978.  Ohde, R. N., & Abou-Khalil, R. (2001). Age differences for stop-consonant and vowel perception in adults. The Journal of the Acoustical Society of America, 110, 2156.  Patterson, R. D., Nimmo‐Smith, I., Weber, D. L., & Milroy, R. (1982). The deterioration of hearing with age: Frequency selectivity, the critical ratio, the audiogram, and speech threshold. The Journal of the Acoustical Society of America, 72, 1788.  Peelle, J. E., Troiani, V., Wingfield, A., & Grossman, M. (2010). Neural processing during older adults' comprehension of spoken sentences: Age differences in resource allocation and connectivity. Cerebral Cortex, 20(4), 773.  Pichora-Fuller, M. K., Schneider, B. A., Benson, N. J., Hamstra, S. J., & Storzer, E. (2006). Effect of age on detection of gaps in speech and nonspeech markers varying  92  in duration and spectral symmetry. The Journal of the Acoustical Society of America, 119, 1143.  Pichora-Fuller, M. K., & Singh, G. (2006). Effects of age on auditory and cognitive processing: Implications for hearing aid fitting and audiologic rehabilitation. Trends in Amplification, 10(1), 29.  Pichora-Fuller, M. K., & Souza, P. E. (2003). Effects of aging on auditory processing of speech. International Journal of Audiology, 42, 11-16.  Price, P., & Simon, H. (1984). Perception of temporal differences in speech by ‗‗normal‐hearing‘‘adults: Effects of age and intensity. The Journal of the Acoustical Society of America, 76, 405.  Repp, B. H. (1984). Closure duration and release burst amplitude cues to stop consonant manner and place of articulation. Language and Speech, 27(3), 245-254.  Rosen, S. (1992). Temporal information in speech: Acoustic, auditory and linguistic aspects. Philosophical Transactions: Biological Sciences, 336(1278), 367-373.  Salthouse, T. A. (1991). Theoretical perspectives on cognitive aging. Hillsdale, NJ: Lawrence Erlbaum.  Salthouse, T. A. (1996). The processing-speed theory of adult age differences in cognition. Psychological Review, 103(3), 403-427.  93  Samuel, A. G., & Tartter, V. C. (1986). Acoustic-phonetic issues in speech perception. Annual Review of Anthropology, 15, 247-273.  Schneider, B. A., Daneman, M., & Murphy, D. R. (2005). Speech comprehension difficulties in older adults: Cognitive slowing or age-related changes in hearing? Psychology and Aging, 20(2), 261.  Schneider, B. A., Pichora-Fuller, K., & Daneman, M. (2010). Effects of senescent changes in audition and cognition on spoken language comprehension. In S. GordonSalant, R. D. Frisina, A. N. Popper & R. R. Fay (Eds.), The aging auditory system (pp. 167-210) Springer.  Schneider, B. A., & Pichora-Fuller, M. K. (2000). Implications of perceptual deterioration for cognitive aging research. The handbook of aging and cognition [The Handbook of Aging and Cognition] (2nd ed., pp. 155–219) Lawrence Erlbaum.  Schneider, B. A., & Pichora-Fuller, M. K. (2001). Age-related changes in temporal processing: Implications for speech perception. Seminars in Hearing, , 22(3) 227240.  Sherbecoe, R. L., & Studebaker, G. A. (2003). Audibility-index predictions of normalhearing and hearing-impaired listeners' performance on the connected speech test. Ear and Hearing, 24(1), 71.  Sommers, M. S., Kirk, K. I., & Pisoni, D. B. (1997). Some considerations in evaluating spoken word recognition by normal-hearing, noise-masked normal-hearing, and  94  cochlear implant listeners. I: The effects of response format. Ear and Hearing, 18(2), 89.  Souza, P. E. (2000). Older listeners' use of temporal cues altered by compression amplification. Journal of Speech, Language, and Hearing Research, 43(3), 661.  Souza, P. E., & Boike, K. (2006). The psychoacoustics of aging. considerations for amplification. Hearing Care for Adults: An International Conference, Chicago, IL. 47.  Souza, P. E., & Kitch, V. (2001). The contribution of amplitude envelope cues to sentence identification in young and aged listeners. Ear and Hearing, 22(2), 112.  Strouse, A., Ashmead, D. H., Ohde, R. N., & Grantham, D. W. (1998). Temporal processing in the aging auditory system. The Journal of the Acoustical Society of America, 104, 2385.  Turner, C. W., Souza, P. E., & Forget, L. N. (1995). Use of temporal envelope cues in speech recognition by normal and hearing‐impaired listeners. The Journal of the Acoustical Society of America, 97, 2568.  Vaughan, N. E., & Letowski, T. (1997). Effects of age, speech rate, and type of test on temporal auditory processing. Journal of Speech, Language, and Hearing Research, 40(5), 1192.  95  Wingfield, A., & Ducharme, J. L. (1999). Effects of age and passage difficulty on listening-rate preferences for time-altered speech. The Journals of Gerontology Series B: Psychological Sciences and Social Sciences, 54(3), P199.  Wingfield, A., Lindfield, K. C., & Goodglass, H. (2000). Effects of age and hearing sensitivity on the use of prosodic information in spoken word recognition. Journal of Speech, Language, and Hearing Research, 43(4), 915.  Wingfield, A., Poon, L. W., Lombardi, L., & Lowe, D. (1985). Speed of processing in normal aging: Effects of speech rate, linguistic structure, and processing time. Journal of Gerontology, 40(5), 579.  Wingfield, A., & Tun, P. A. (2001). Spoken language comprehension in older adults: Interactions between sensory and cognitive change in normal aging. Seminars in Hearing, 22(3) 287-302.  Wingfield, A., Tun, P. A., & McCoy, S. L. (2005). Hearing loss in older adulthood. Current Directions in Psychological Science, 14(3), 144.  Wingfield, A., Wayland, S. C., & Stine, E. A. L. (1992). Adult age differences in the use of prosody for syntactic parsing and recall of spoken sentences. The Journal of Gerontology, 47(5), P350.  Wong, P., Jin, J. X., Gunasekera, G. M., Abel, R., Lee, E. R., & Dhar, S. (2009). Aging and cortical mechanisms of speech perception in noise. Neuropsychologia, 47(3), 693-703.  96  Appendices Appendix A – Stimulus formant transition comparison Below is a comparison of vowel formants from recordings of tokens both with (i.e., speed, steam, and scum) and without (i.e., seed, seam, and some) each stop consonant present. All pitch pulses are numbered from the first vowel pulse from which Praat 2.0 could gather complete formant data for a given token.  Speed vs. seed F1 (Hz) F2 (Hz) F3 (Hz) F4 (Hz)  PP1 296 1907 2659 3257  PP4 314 2249 2913 3392  PP8 301 2257 3042 3412  PP12 290 2275 3075 3483  PP16 289 2246 3009 3467  Table 9. Frequencies of formants F1-F4 at various pitch pulses (PP) within the vowel from speed.  Figure 16. Spectrogram showing the formant structure in the vowel within speed.  Note: Dotted red lines indicate formants as calculated by Praat 2.0.  F1 (Hz) F2 (Hz) F3 (Hz) F4 (Hz)  PP1 326 1873 2735 3995  PP4 331 2118 2894 3462  PP8 317 2268 2915 3404  PP12 304 2340 3063 3512  PP16 289 2330 3007 3498  Table 10. Frequencies of formants F1-F4 at various pitch pulses (PP) within the vowel from seed.  97  Figure 17. Spectrogram showing the formant structure in the vowel within seed.  Note: Dotted red lines indicate formants as calculated by Praat 2.0.  Figure 18. Comparison of formant transitions within the vowels of speed and seed.  Note: Black lines indicated formants for speed; grey lines indicate formants for seed.  98  Steam vs. seam F1 (Hz) F2 (Hz) F3 (Hz) F4 (Hz)  PP1 327 2212 2985 4008  PP4 325 2313 3010 3603  PP8 337 2435 3006 3600  PP12 358 2440 3008 3703  PP16 390 2340 2847 3762  Table 11. Frequencies of formants F1-F4 at various pitch pulses (PP) within the vowel from steam.  Figure 19. Spectrogram showing the formant structure in the vowel within steam. Note: Dotted red lines indicated formants as calculated by Praat 2.0.  F1 (Hz) F2 (Hz) F3 (Hz) F4 (Hz)  PP1 283 2114 2769 3659  PP4 319 2345 2970 3469  PP8 324 2367 3049 3400  PP12 320 2393 2977 3588  PP16 335 2356 2892 3513  Table 12. Frequencies of formants F1-F4 at various pitch pulses (PP) within the vowel from seam.  Figure 20. Spectrogram showing the formant structure in the vowel within seam. Note: Dotted red lines indicate formants as calculated by Praat 2.0.  99  Figure 21. Comparison of formant transitions within the vowels of steam and seam. Note: Black lines indicated formants for steam; grey lines indicate formants for seam.  100  Scum vs. some  F1 (Hz) F2 (Hz) F3 (Hz) F4 (Hz)  PP1 443 1700 2524 3717  PP4 566 1467 2473 3473  PP8 619 1270 2688 3398  PP12 613 1214 2918 3039  PP16 462 1135 2442 3071  Table 13. Frequencies of formants F1-F4 at various pitch pulses (PP) within the vowel from scum.  Figure 22. Spectrogram showing the formant structure in the vowel within scum. Note: Dotted red lines indicate formants as calculated by Praat 2.0.  F1 (Hz) F2 (Hz) F3 (Hz) F4 (Hz)  PP1 473 1346 2850 3297  PP4 617 1266 2751 3035  PP8 694 1232 2932 2999  PP12 666 1209 2959 3267  PP16 477 1204 2725 3169  Table 14. Frequencies of formants F1-F4 at various pitch pulses (PP) within the vowel from some.  Figure 23. Spectrogram showing the formant structure in the vowel within some. Note: Dotted red lines indicate formants as calculated by Praat 2.0.  101  Figure 24. Comparison of formant transitions within the vowels of scum and some. Note: Black lines indicated formants for scum; grey lines indicate formants for some.  Note: The scum/some contrast pair was considered for use in this study, but was rejected because pilot testing showed it to produce ceiling effects in the burst-present condition and floor effects in the no-burst condition. This comparison (cf. Figures 18, 21, 24) shows that the presence of /k/ had a greater effect on the vowel formant structure than was seen in tokens containing /t/ or /p/.  102  Appendix B – Speech perception testing results for each participant Participant 100  Figure 25. Speed/seed contrast pair response curves from participant 100  Figure 26. Steam/seam contrast pair response curves from participant 100  103  Participant 101  Figure 27. Speed/seed contrast pair response curves from participant 101.  Figure 28. Steam/seam contrast pair response curves from participant 101.  104  Participant 102  Figure 29. Speed/seed contrast pair response curves from participant 102.  Figure 30. Steam/seam contrast pair response curves from participant 102.  105  Participant 103  Figure 31. Speed/seed contrast pair response curves from participant 103.  Figure 32. Steam/seam contrast pair response curves from participant 103.  106  Participant 104  Figure 33. Speed/seed contrast pair response curves from participant 104.  Figure 34. Steam/seam contrast pair response curves from participant 104.  107  Participant 105  Figure 35. Speed/seed contrast pair response curves from participant 105.  Figure 36. Steam/seam contrast pair response curves from participant 105.  108  Participant 106  Figure 37. Speed/seed contrast pair response curves from participant 106.  Figure 38. Steam/seam contrast pair response curves from participant 106.  109  Participant 900  Figure 39. Speed/seed contrast pair response curves from participant 900.  Figure 40. Steam/seam contrast pair response curves from participant 900.  110  Participant 901  Figure 41. Speed/seed contrast pair response curves from participant 901.  Figure 42. Steam/seam contrast pair response curves from participant 901.  111  Participant 902  Figure 43. Speed/seed contrast pair response curves from participant 902.  Figure 44. Steam/seam contrast pair response curves from participant 902.  112  Participant 903  Figure 45. Speed/seed contrast pair response curves from participant 903.  Figure 46. Steam/seam contrast pair response curves from participant 903.  113  Participant 904  Figure 47. Speed/seed contrast pair response curves from participant 904.  Figure 48. Steam/seam contrast pair response curves from participant 904.  114  Participant 905  Figure 49. Speed/seed contrast pair response curves from participant 905.  Figure 50. Steam/seam contrast pair response curves from participant 905.  115  Appendix C – Participants’ comments during block confidence rating Listed are the participants‘ responses to the question, ―On a scale of one to five, where five is completely confident and one is not at all confident, how confident did you feel about your responses for the last block?‖ Each block contained 72 responses. When a participant mentions, for example, ―missing one,‖ he/she is referring to one response out of a block of 72 responses. Each comment was made immediately after the participant had given a numeric confidence rating.  Part. Blk# Conf. Comment Younger listeners 100 1 2 3 4 5  4/5: 4/5: 3/5: 4/5: 4/5:  1 2 3 4  4/5: 4/5: 3/5: 3.5/5:  Are they all the same? Some seem different. About the same. I made one keystroke error. Back up to [confidence rating] four. [No comment]  101  5  Fairly confident. Still the same. The space between the /s/ and the next letter was getting smaller. Same reason [for a lowered confidence score: space between /s/ and the next letter getting smaller]. It was maybe a bit easier. 3.5/5: [No comment]  1 2 3 4 5  4/5: 4/5: 4.5/5: 4.5/5: 4.5/5:  No problems. [No comment] [No comment] [No comment] [No comment]  1 2 3 4 5  5/5: 3.5/5: 5/5: 4.5/5: 5/5:  Speed! [Imitating token] [This block was] harder. [No comment] [No comment] [No comment]  102  103  116  104 1 2 3 4 5  4/5: 2.5/5: 2/5: 4/5: 4/5:  There were some that got me. Fairly dec[ent]. Like a three or a two. Sound like ―feed‖ a lot of the time. [No comment] [No comment]  1 2  5  3.6/5: You could tell some of them were drawn out. 4/5: Sometimes it cuts out to make you think there‘s a /t/. You get used to it and to looking at the words maybe. It gets easier. 4/5: Getting easier. If you think about it too long, like ―was that just a gap or actually a /p/?‖ you can confuse yourself. Trying to answer fast but not too fast. 3.8/5: [I‘m] thinking too much: ―Am I getting these right?‖ equals less confidence. 3.8/5: I can psych myself out.  1  4/5:  105  3  4  106  2 3 4 5  Five for almost all. A few I was suspicious of because they didn‘t sound like a total seam or seed. I erred toward [consonant present] in those with three out of five confidence. 4/5: About the same. 4/5: [No comment] 4.5/5: A little bit better. 4.5/5: [No comment]  Older listeners 900 1 2 3 4 5  5/5: 5/5: 5/5: 5/5: 5/5:  I made a few errors when I was hurrying [No comment] Better than before No different Confident [in overall results for all blocks], but interested to know whether results bore this out.  1 2 3 4  5/5: 4/5: 5/5: 5/5:  [No comment] I made one keystroke error. [Confident] except for the ones that sound ―off‖ Upgraded from four out of five, damnit! I‘m not sure I get even  901  117  5  one wrong. 5/5: OK, good, hell! **Note: This participant noted ambiguity and dissatisfaction with options at various other points in testing.  902 1 2 3 4 5  4/5: 4/5: 4/5: 4/5: 4/5:  1  4/5:  I might have missed a few. Might have missed one. Some sounded halfway between. Missed one again. There was one, or a few, that were vague. I missed one, I think. Same.  903  2 3 4 5  I missed a few… when you‘re going fast. [Note that in instructions, participant was told that responses were not timed, and that speed was not a factor in the results] 4/5: I made a few mistakes. About the same. 4/5: I know I made a few mistakes because of hurrying. 4.5/5: That one [block] was a little better. 4/5: I didn‘t get it all right.  1 2 3 4 5  5/5: 3.5/5: 4/5: 4/5: 4.5/5  1  4.25/5: I made a few booboos. I pressed one then decided to press the other. I guess I can‘t do that. 85 percent. 5/5: 97 percent! 5/5: 97 percent. Actually 100 percent, but… 5/5: 99 percent. 5/5: 99 percent.  904 Maybe one or two [were wrong]. I felt I heard them properly. A couple I was weak on. A little more confident. Maybe four. About the same [No comment]  905  2 3 4 5  118  Appendix D – Spoken word frequency of stimulus words This word frequency report was generated using The Corpus of Contemporary American English (Davies, 2008). Each table shows the total occurrences and frequency per million words in a number of modalities (i.e., in spoken word, written fiction, magazine articles, etc.).  Speed Occurrences Frequency per million  Spoken Fiction Magazine Newspaper 2728 4869 11892 5651 31.31 59.58 136.44 67.66  Academic 4427 53.39  Table 15. Word frequency for speed (Davies, 2008).  Seed Occurrences Frequency per million  Spoken 560 6.43  Fiction Magazine Newspaper Academic 1247 3663 2428 1657 15.26 42.03 29.07 19.98  Table 16. Word frequency for seed (Davies, 2008).  Steam Occurrences Frequency per million  Spoken 608 6.98  Fiction Magazine Newspaper Academic 2181 1761 1163 999 26.69 20.20 13.93 12.05  Table 17. Word frequency for steam (Davies, 2008).  Seam Occurrences Frequency per million  Spoken 27 0.31  Fiction Magazine Newspaper Academic 368 336 101 66 4.50 3.86 1.21 0.80  Table 18. Word frequency for seam (Davies, 2008).  119  Appendix E – Participant consent form The University of British Columbia School of Audiology and Speech Sciences Friedman Building, 2177 Wesbrook Mall Vancouver, B.C. V6T 1Z3 Phone: (604) 822-5591, Fax: (604) 822-6569  Consent Form Quantifying Hearing Impaired Listeners’ Use of Acoustic Information in Speech Principal Investigator: Dr. Lorienne Jenstad, School of Audiology and Speech Sciences, UBC. Phone: 604-822-4716 Co-Investigator(s): Dr. Valter Ciocca, Director, School of Audiology and Speech Sciences, UBC. Phone: 604-822-5795. Nathan Carter, M.Sc. candidate, School of Audiology and Speech Sciences, UBC. Phone: 604-267-1747. This research supports the completion of a Masters thesis. Information in the study is protected by standard confidentiality measures. Sponsor: This research has been sponsored by an NSERC (National Sciences and Engineering Research Council of Canada) Discovery Grant awarded to L. Jenstad and by an NSERC fellowship awarded to the student coinvestigator. Purpose: The purpose of this study is to see how aging affects people’s ability to understand speech. You are being invited to participate in this study because you are younger than 31 or older than 74 and have indicated you think you have normal hearing. Study Procedures: When you arrive at the amplification lab, you will take part in a brief hearing screening (5 minutes), a test to determine your ability to detect very short sounds (5 minutes), and a generic questionnaire and mental state test (10 minutes). You will then be seated in a soundproof booth in front of a touch computer screen. You will be fitted with headphones and 120  the volume will be adjusted to suit you. You will be able to stop testing at any point for any reason. The main testing involves listening to words one at a time over the headphones. For each word, two choices will appear on the touch screen and you will touch the one that seems most like the word you heard (40-60 minutes). After each of your responses, the next word will automatically play and new choices will appear on the screen. Potential Risks: Known risks are boredom and discomfort with the earphones, which can be adjusted for comfort at any time. Potential Benefits: You will gain information about your own hearing and speech understanding. This research may contribute to hearing aid processing and design. If you are interested in receiving a copy of the study’s results, please contact Dr. Jenstad at 604-822-4716. Confidentiality: Your identity will be kept strictly confidential at all times. Data will be stored in computerized files and a locked filing cabinet in the Amplification Research Lab in the basement of the Woodward library at UBC. Files saved on the computer will not include your name or personal identifying information. Your information will be identified only by a code number and all personal information will be stored as hard copy in a locked filing cabinet within the lab, which has 2 locking doors and an alarm system. Remuneration/Compensation: In order to make up for the costs of transportation/inconvenience, you will receive an honorarium of $15. You are free to withdraw from the study at any time and for any reason without a reduction in the honorarium. Contact for information about the study: If you have any questions or would like further information about the study or about your involvement in it, please contact Dr. Jenstad at 604-8224716.  121  Contact for concerns about the rights of research subjects: If you have any concerns about your treatment or rights as a research subject, you may contact the Research Subject Information Line in the UBC Office of Research Services at 604-822-8598 or if long distance e-mail to RSIL@ors.ubc.ca or toll free 1-877-822-8598. Consent: Your participation in this study is entirely voluntary and you may refuse to participate or withdraw from the study at any time. This will not jeopardize your $15 honorarium or our appreciation of your involvement. Your signature below indicates that you have received a copy of this consent form for your own records. Your signature indicates that you consent to participate in this study. May we contact you for related studies in the future? Yes____ No____  ____________________________________________________ Subject Signature Date  122  Appendix F – Questionnaire form General health status scale During the past week, what was the health-related quality of your life? (circle a number) 1 Very poor  2  3  4  5 OK  6  7  8  9  10 Very good  MINI-MENTAL STATUS EXAMINATION Instructions: To be completed by tester. See attached directions. MAX 5 5  3  SCORE ORIENTATION (1 point for each correct answer) 1. What is the (year) (season) (date) (day) (month) 2. Where are we (province) (regional district) (city) (department) (floor) REGISTRATION (1 point for each correct answer) 3. Name 3 objects: 1 second to say each. Then ask the subject all 3 after you have said them. Give 1 point for each correct answer at the first repetition. Continue to repeat them until he learns all 3. (Up to 6 trials). If he does not learn all three, RECALL (#5) cannot be meaningfully tested. ATTENTION AND CALCULATION (1 point for each correct answer)  5  3  4. Serial 7‘s: have subject begin with 100 and count backward by 7. Stop after 5 answers. 1 point for each one correct. Or alternatively: spell ―world‖ backwards. Score by number of letters in correct order. RECALL (1 point for each correct answer) 5. Ask for the names of the 3 objects repeated above in REGISTRATON (#3). Give one point for each one correct. LANGUAGE (1 point for each correct answer)  2  6. Point to the following objects and have the subject name each: a pencil and a watch  1  7. Have the subject repeat the following: ―No ifs, ands ,or buts‖  123  3  8. Have the subject follow a 3-stage command: ―Take a paper in your right hand, fold it in half, and put it on the floor‖  1  9. Show the subject the statement: ―Close your eyes‖ (Item A, attached) and have her read and obey what it says.  1 1  10. Have the subject write a sentence with a subject and verb. 11. Have the subject copy the figure (item B, attached).  30  TOTAL SCORE  TOTAL SCORE ( / 30) ITEM B  ITEM A Item A is a sheet of paper with ―CLOSE YOUR EYES‖ written on it in large black capital letters.  Assess level of consciousness along a continuum: 1 = alert; 2= drowsy; 3=stupor; 4=coma  124  Digit span subtest of the WAIS-III Digits forward  Score  Digits backward  Score  6-4-3-9 7-2-8-6  4 4  2-8-3 4-1-5  3 3  4-2-7-3-1  5  3-2-7-9  4  7-5-8-3-6  5  4-9-6-8  4  6-1-9-4-7-3  6  1-5-2-8-6  5  3-9-2-4-8-7  6  6-1-8-4-3  5  5-9-1-7-4-2-3  7  5-3-9-4-1-8  6  4-1-7-9-3-8-6  7  7-2-4-8-5-6  6  5-8-1-9-2-6-4-7  8  8-1-2-9-3-6-5  7  3-8-2-9-5-1-7-4  8  4-7-3-9-1-2-8  7  Digits forward =  Draw a line through any series failed. Circle score for maximum number repeated correctly.  Digits backward = Digits total =  125  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0072247/manifest

Comment

Related Items