Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Effects of noise and hearing loss on the conversational behaviour of seniors Roodenburg, Kristin E. J. 1996

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

Item Metadata

Download

Media
831-ubc_1996-0486.pdf [ 5.39MB ]
Metadata
JSON: 831-1.0099176.json
JSON-LD: 831-1.0099176-ld.json
RDF/XML (Pretty): 831-1.0099176-rdf.xml
RDF/JSON: 831-1.0099176-rdf.json
Turtle: 831-1.0099176-turtle.txt
N-Triples: 831-1.0099176-rdf-ntriples.txt
Original Record: 831-1.0099176-source.json
Full Text
831-1.0099176-fulltext.txt
Citation
831-1.0099176.ris

Full Text

EFFECTS OF NOISE AND HEARING LOSS ON THE CONVERSATIONAL BEHAVIOUR OF SENIORS By Kr ist in E. J . Roodenburg M.A., University of Antwerp, 1992 A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE in THE FACULTY OF GRADUATE STUDIES THE FACULTY OF MEDICINE THE SCHOOL OF AUDIOLOGY AND SPEECH SCIENCES We accept this thesis as conforming to the required standard THE UNIVERSITY OF BRITISH COLUMBIA August, 1996 ® Kr is t in E. J . Roodenburg, 1996 In presenting this thesis in partial fulfilment of the requirements for an advanced degree at the University of British Columbia, I agree that the Library shall make it freely available for reference and study. I further agree that permission for extensive copying of this thesis for scholarly purposes may be granted by the head of my department or by his or her representatives. It is understood that copying or publication of this thesis for financial gain shall not be allowed without my written permission. Department The University of British Columbia Vancouver, Canada DE-6 (2/88) 11 ABSTRACT. The purpose of this study was to investigate the effects of hearing loss and adverse listening conditions on the conversational patterns of seniors. A second goal was to determine what measures best describe these conversational patterns. Results were compared to those of a case study of a hard-of-hearing senior presented by Pichora-Fuller & Johnson (submitted). One normal-hearing and one hearing-impaired senior were selected for this study. Each subject participated in two conversations, one in an advantageous condition and one in an adverse, noisy listening condition. Participants were interviewed about their ability to understand the conversations. Further measures of comprehension included free and recognition recall tasks, a conversational fluency rating, and a detailed discourse analysis based on a transcript of the conversations. Of all the comprehension measures used, the recognition task appeared to be the most sensitive indicator of perceptual diff iculty and comprehension of conversational detail. Free recall results were a good measure of relative memory for detail and gist across the different signal-to-noise ratio conditions. Results of the discourse analysis provided quantitative evidence that was able to support or contradict the comprehension diff icult ies indicated on the conversational fluency ratings. The discourse analysis procedure was too time-intensive to be- a viable clinical tool.' However, some aspects of the i i i analysis, including production of new content, overt repair requests, and subtopic management, could be singled out as particularly helpful in identifying comprehension problems, indicating that a more selective analysis may suffice for clinical purposes. No marked differences in comprehension between the hearing-impaired and normal-hearing senior were observed for the conversations in advantageous listening conditions. In the adverse listening conditions, the normal-hearing participant reported greater effort, but his comprehension did not appear to be significantly affected. Results were consistent with the Pichora-Fuller & Johnson (submitted) study in that both hard-of-hearing seniors showed reduced comprehension of the conversations in adverse listening conditions on all of the comprehension measures. However, individual variables appeared to have a significant effect on conversational behaviour. In particular, the subject of the Pichora-Fuller & Johnson (submitted) study tended to conceal her hearing problems by feigning understanding, while the hearing-impaired senior of the present study appeared to initiate repairs whenever they were needed. It was evident from the results presented here that individual baseline information is essential to the assessment of conversational behaviour and comprehension difficulties. iv TABLE OF CONTENTS ABSTRACT i i i LIST OF TABLES ix ACKNOWLEDGEMENTS x i i CHAPTER 1 1 INTRODUCTION AND LITERATURE REVIEW 1 Overvi ew 1 Impairment, Disabi l i ty and Handicap . . . 2 Measures of Comprehension 4 Self-Report 5 Recall 7 Memory and Comprehension 7 Influences on Recall 10 Working Memory Constraints 10 Recognition recall 11 . Free recall 13 Narrative Schemas and Scripts . . . . 14 Affective Value 15 Limitations of Memory Tasks 16 Discourse Analysis 17 Discourse Analysis and Comprehension 18 Indirect Measures of Comprehension 19 Turntaking 19 Topic 20 Direct measures of comprehension 21 Limitations of Discourse Analysis 22 Research Questions 24 CHAPTER 2 25 METHOD 25 Subjects 25 The Subject Pool 25 Preliminary Measures 26 Basic audiometry 26 Revised Speech Perception in Noise 26 V Hearing Handicap Inventory for the Elderly . . . 27 Hearing Performance Inventory 27 Mill House Vocabulary Test 28 Working Memory Span Test 28 Group data 29 SPIN-R 29 Mill House Vocabulary and Working Memory Span tests 30 Hearing Handicap Inventory for the Elderly and Hearing Performance Inventory 30 Individual Subjects 32 Subject 1: EL 32 Subject 2: ZW 36 Subject 3: GK 40 Procedure 45 Comprehension Measures 47 Verbal Protocol 48 Free Recall 48 Recognition Recall 50 Conversational Fluency Ratings 51 Discourse Analysis 52 Analysis 54 CHAPTER 3 56 RESULTS 56 Subject 1: EL 56 Verbal Protocol 56 Recall 57 Recognition Recall 57 Free Recall 58 Conversational Fluency Rating 60 Discourse Analysis 62 Subject 2: ZW 71 Verbal Protocol 71 Recall 72 Recognition Recall 72 Free Recall . 73 Fluency Ratings 74 Discourse Analysis 74 vi Subject 3: GK 83 Verbal Protocol 83 Recall 84 Recognition Recall 84 Free Recall 84 Fluency Ratings 86 Discourse Analysis 86 CHAPTER 4 97 DISCUSSION 97 Comparison of Results 97 Verbal Protocol 97 Recall 98 Recognition Recall 98 Free Recall 99 Fluency Ratings 103 Discourse Analysis 105 General Discussion 108 Evaluation of Comprehension Measures 108 Hearing Loss and Effects of Noise 112 Individual Differences 114 Interactions between Conversational Participants 115 Concluding Remarks 117 REFERENCES 120 APPENDIX A 125 Responses to the Hearing Handicap Inventory for the Elderly (HHI/E) 125 APPENDIX B 126 Verbal Protocol 126 APPENDIX C 130 Definition of Discourse Analysis Categories 130 APPENDIX D 133 Lists of Facts 133 APPENDIX E 136 Recognition Recall Test 136 VI1 APPENDIX F 148 Results of Free and Recognition Recall Tests 148 LIST OF TABLES v i i i TABLE 1. Average of estimated S:N in dB corresponding to 50$ SPIN-R score for each subgroup ± one standard deviation 30 2. Average Mill House Vocabulary and Working Memory Span scores per subgroup ± one standard deviation .. 31 3. Average HPI and HHI/E scores per subgroup ± one standard deviation 31 4. Pure-tone thresholds in dBHL for EL 33 5. SPIN-R percent correct score as a function of S:N condition for EL 34 6. HPI subscores for EL and experimental subgroups 35 7. Pure-tone thresholds in dBHL for ZW 37 8. SPIN-R percent correct score as a function of S:N condition for ZW 38 9. HPI subscores for ZW and experimental subgroups 39 10. Pure-tone thresholds in dBHL for GK 41 11. SPIN-R percent correct score as a function of S:N condition for GK 43 12. HPI subscores for GK and experimental subgroups 44 13. Percentage of content turns, content clauses and subtopics recalled by EL for each participant and each conversation 59 14. Conversational fluency ratings of EL's conversations 61 15. Content contributed by KF and EL in each conversation with EL . . . 63 16. Frequency of occurrence of each subtype of contentless clause per speaker in conversations with EL • 65 17. Evidence of comprehension per speaker in conversations with EL . . 66 18. Evidence of failed comprehension per speaker in conversations with EL 67 19. Ambiguous evidence of comprehension per speaker in conversations with EL 69 20. Evidence of comprehension offered per speaker in reponse to prosodic cues in conversations with EL 70 21. Percentage of content turns, content clauses and subtopics recalled by ZW for each participant and each conversation 73 22. Conversational fluency ratings for conversations with ZW 75 23. • Content contributed by KF and ZW in each conversation with ZW . . . 76 24. Frequency of occurrence of each subtype of contentless clause per speaker in conversations with ZW 79 25. Evidence of comprehension per speaker in conversations with ZW . . 80 26. Evidence of failed comprehension per speaker in conversations with ZW 80 ix 27. Ambiguous evidence of comprehension per speaker in conversations with ZW 82 28. Evidence of comprehension offered per speaker in response to prosodic cues in conversations with ZW 83 29. Percentage of content turns, content clauses and subtopics recalled by GK for each participant and each conversation 85 30. Conversational fluency ratings for conversations with GK 87 31. Content contributed by KF and GK in each conversation with GK . . . 88 32. Frequency of occurrence of each subtype of contentless clause per speaker in conversations with GK 91 33. Evidence of comprehension per speaker in conversations with GK . . 92 34. Evidence of failed comprehension per speaker in conversations with GK 92 35. Ambiguous evidence of comprehension per speaker in conversations with GK 94 36. Evidence of comprehension offered per speaker in response to prosodic cues in conversations with GK 95 37. Percentage of content clauses and content turns recalled by each subject for each conversation 101 X ACKNOWLEDGEMENTS I would l ike to express my thanks to the many people who have made the timely completion of this project possible. Thanks are due f i rs t of all to Dr. Kathy Pichora-Fuller and Dr. Carolyn Johnson, for their generous sharing of ideas and materials, and for the luxury of thoughtful feedback. I would also l ike to thank Noelle Lamb for her helpful comments on the various drafts of this thesis as well as her assistance with the data preparation. Special mention is due to my family, for their patience and long-distance support. CHAPTER 1 1 INTRODUCTION AND LITERATURE REVIEW Overview This study wi l l focus on procedures for describing the conversational behaviour of normal-hearing and hard-of-hearing seniors. Conversation is probably the single most important communication.activity in everyday l i f e . Furthermore, conversations frequently take.-piace under less than ideal , noisy l istening conditions, which are known to cause d i f f icu l ty for both normal-hearing and hearing-impaired seniors. A variety of assessment measures are available to evaluate the hard-of-hearing c l ien t ' s degree of impairment in everyday conversation. The purpose of this thesis is to determine which measure or combination of measures is most l i ke ly to give an accurate description of the c l ient 's ab i l i ty to comprehend conversation. Conversations were held between a c l in i c ian and individual participants from three d is t inct groups, both in favourable and adverse signal-to-noise ratio conditions. Comprehension of the conversations was tested using both recall and recognition tasks. Also, conversations were recorded to allow for in-depth discourse analysis. In addition, participants were interviewed about their ab i l i t y to understand the conversations. 2 The present study extends a case study of GK, a hearing-impaired senior, conducted by Pichora-Fuller and Johnson (submitted). Results of this case study indicated a discrepancy between the various measures of comprehension used, suggesting that no single measure is l ikely to provide an accurate description of the subject's conversational abil i t ies (see Chapter 3 for a detailed summary). This thesis will present the results obtained from one additional hearing-impaired and one normal-hearing senior and compare them to those obtained from GK, to determine which data and conversational patterns may be generalisable indicators of the effects of hearing loss on conversational behaviour. A complete description of the method used in this study is given in Chapter 2. Chapter 3 presents the results and Chapter 4 discusses these results. In the remainder of this chapter, I will review the concepts needed to address the issues raised in this thesis and present the research questions that wil l be central to the discussion of the results outlined in Chapter 3. Impairment, Disability and Handicap For the purposes of this study, i t wil l be useful to distinguish between impairment, disability and handicap, following definitions presented by the World Health Organization (1980). Impairment refers to physical, physiological, or anatomical loss, or a reduction in normal function that is commonly measurable in objective terms. In the context of audiology, a hearing impairment 'refers to a reduced ability to preceive sound and may be described by standard audiometric measures. Disability is the direct impact of an impairment on task-related function. A hearing loss, for instance, may result in reduced ability to extract information presented in conversation, depending on such factors as familiarity with the topic of conversation, ability to make use of redundancy in discourse, or speaker in te l l ig ib i l i ty (Erber, 1988, Ch. 2). The effects of a disability on a person's social and emotional function and well-being define the experience of handicap. For instance, reduced abil ity to comprehend information in everyday conversation may result in avoidance of social and communicative situations that rely heavily on adequate hearing and comprehension. Comprehension is the primary indicator of conversational competence used in the present study. In other words, the efficacy and accuracy of information transfer and processing was considered central to a subject's abil ity to participate in the conversation. In this sense, the measures described are indicators of hearing disabil i ty, or a possible reduction in task-related function. Similarly, assuming that the listener's abil ity to extract information is compromised, this may or may not be experienced as a handicap. While most comprehension measures are unreliable indicators of the hearing-impaired listener's subjective experience of the conversation, self-report provides one means of obtaining information on hearing handicap. 4 Measures of Comprehension Frederiksen, Brackwell, Breuleux, and Renaud (1990) and Renkema (1993) provide overviews of the methods most commonly used in the study of language and discourse comprehension. These include product-oriented methods such as free and probed recall, question/answer paradigms, and multiple choice tasks. Other methods are more process-oriented and include reading times, eye fixation times, and on-line interpretation tasks. Many of these process-oriented tasks have been devised for the study of reading comprehension. Since the present paper is a study of oral discourse processing, its methods fal l mainly into the f irst category. One advantage of studying reading comprehension over listening comprehension is that i t is much easier to develop methods for isolating on-line discourse processing from its product (e.g. recall). Researchers have measured reading times and eye fixation times in self-paced reading tasks to determine the relative processing load of specific aspects of written language. Kintsch (1981), for instance, discusses how some propositions need to be frequently recalled and reviewed for the purpose of establishing text cohesion; the corresponding memory searches are typically associated with regressive eye movements to the appropriate portions of the text. In the processing of oral discourse, on the other hand, eye movements obviously do not apply; whether 5 time constraints allow a regressive mental memory search on-line is less easily documented. The following sections will review each of the methods used in the present study (self-report, recall, and discourse analysis) and discuss their relative advantages and shortcomings. Self-Report Traditionally, self-report by means of questionnaires such as the Hearing Handicap Inventory for the Elderly (HHI/E, Ventry & Weinstein, 1982) has been used by audiologists as a relatively easy and inexpensive measure of hearing disability or handicap. Increased use of these questionnaires stems from a growing awareness among clinicians that clients' needs must be considered within the context of their everyday experiences. One concern regarding hearing handicap questionnaires relates to the need for experimentation and refinement of existing instruments (Giolas, 1990), although some such attempts have been made. Mulrow, Tuley and Aguilar (1990), for instance, reported on a comparison of the psychometric value of the HHI/E and the Revised Quantified Denver Scale of Communication Function (RQDS). The authors concluded that while both scales were equally sensitive measures for correctly identifying individuals with hearing loss, the HHI/E was a superior instrument for measuring changes in response to audiological rehabilitation. 6 A second limitation relates to the inevitably subjective nature of the results obtained from such questionnaires. Rather than describing the dif f icult ies actually encountered by hearing-impaired clients, they are an indicator of clients' perceived problems and their willingness to report on them. Erdman (1994) l is ts several of the barriers to obtaining accurate information from self-assessment tools, including the influence of social desirability on response patterns, overly cautious and deliberated responses, and a possible bias towards responses in an agreeable or positive mode. In addition, results of a study on the communicative activities of hard-of-hearing residents in a home for the aged (Pichora-Fuller & Robertson, 1994) showed that hearing-impaired residents may frequently have avoided social situations that are likely to create communication problems, while for activities in which participation was mandatory, a much wider range of handicap was reported. The study underscores the fact that "there may be only a narrow time window in which residents might realize that their participation in activities is affected by hearing loss" (p. 286). When used with care, however, self-report may provide valuable information. As Noble (1993) points out: One impulse for the advocacy of self-assessment is that i t enables the perspective of the person affected to be appreciated. The very individuality of that perspective (no two of us are alike) makes i t vital 7 that self-assessment be a flexible and open-ended procedure. Yet professional concern, bringing the expertise of psychometrics to bear on the job, could produce the opposite outcome, (p. 301) In other words, even though the psychometric value of specific tools may be questionable, clinicians are likely to benefit from allowing individual clients to share their individual perspectives. Only by giving individuals the opprortunity to state their specific concerns wil l clinicians be able to address the issues that affect clients in their everyday lives. The frequent discrepancy noted between results obtained from self-assessment tools and standardised audiometry indicates that a variety of factors determines the extent to which a hearing impairment does or does not constitute a barrier for any individual client and that measures of impairment alone do not adequately predict clients' rehabilitative needs (Erdman, 1994). Recall Memory and Comprehension Memory tasks give a second indication of conversational abil ity, because they provide an indirect measure of discourse comprehension and yield more objective information than self-assessment tools. 1 The rationale behind using memory 1 For an overview of issues in the study of memory for conversation, see Cohen (1989), Ch. 7. 8 tasks to test discourse comprehension is that the process of semantic decoding and interpretation is l ikely to be reflected as superior recall. Several researchers have explored this relation between discourse processing and memory. A distinction is commonly made between semantic and episodic memory (Cohen, 1989, pp. 114-15). Semantic memory represents a store of concepts, facts, generalisations, and world knowledge. Episodic memory, on the other hand, is memory for specific events, objects, or people. Its representations may incorporate concepts from semantic memory, which are related and interpreted by means of schemas and scripts. The specific and personal nature of episodic memory may be contrasted with the abstract and general character of semantic memory. Memory for conversation may be characterised as a form of episodic memory, as many conversations are accounts of events and personal experiences, and involve a dynamic interaction between two or more individuals. Nonetheless, general concepts and statements may frequently be extracted from verbal interactions and become a part of semantic memory. Kintsch (1981) claimed that episodic memory is a reliable indirect measure of discourse processing: "memory for text is closely associated with comprehension" (p. 188). Kintsch's argument is that since texts are meaningful, generally of interest to us, and often structured in predictable ways, our understanding and therefore our memory of them is better than for 9 random word l i s ts . The implication of this statement is that comprehension precedes and facilitates memory. Other researchers have discussed some of the factors that may faci l i tate comprehension and memory. Craik and Lockhart (1972) and Lockhart, Craik and Jacoby (1976) proposed a model of memory in which "retention is a function of depth, and various factors, such as the amount of attention devoted to a stimulus, its compatibility with the analyzing structures, and the processing time available, will determine the depth to which i t is processed" (Craik & Lockhart, 1972, p. 676). Within this model, the "deeper" a stimulus has been processed, the more likely i t wil l be committed to memory. "Depth" is defined as the duration and quality of the attention devoted to any given stimulus. In addition, the more easily the incoming information may be integrated into an existing knowledge framework, the more readily i t wil l become part of that knowledge system. Since listeners are more l ikely to devote attention to conversations than to random arrangements of stimuli, and since conversations frequently extend from and expand existing knowledge, memory of them should be superior. The following section wil l elaborate on some of the factors affecting comprehension and recall. 10 Influences on Recall Working Memory Constraints The relation between on-line comprehension processes and recall has been developed in more recent theories on working memory (Daneman & Carpenter, 1980; Just & Carpenter, 1992). The Capacity Theory conceives of working memory as a pool of limited operational resources that has two functions. One is to execute language processes, which are equivalent to the performance of computations. The second is the storage of the intermediate and/or final products of comprehension. Working memory becomes active when readers or listeners construct and integrate ideas from text or spoken discourse. With regards to the present study, i t is to be expected that hearing loss, as well as adverse signal-to-noise ratio conditions, impose demands on working memory that interfere with full semantic processing and encoding. In other words, i f the resource pool, of working memory is in fact limited, the number of resources available for higher-order linguistic and cognitive processing should decrease when more of the resource pool is required for perceptual processing. Similar effects may be expected when fatigue and reduced attention preclude adequate processing (such as near the end of a lengthy conversation in noise). This contention is supported by Pichora-Fuller, Schneider and Daneman (1995), who presented data comparing the performance of young and old adults on tasks requiring the identification and recall of speech heard in noise. 11 Specifically, the study found that all experimental groups were adversely affected by noise. Old and presbycusic subjects, however, were more severely affected. Also, both old groups appeared to derive more benefit from supportive context than did the young group. One reason postulated by the authors is that because of an age-related decline in perceptual processing, old subjects are more experienced at deriving information from context. On the other hand, old subjects performed poorly on recall tasks, l ikely because working memory resources were mostly dedicated to listening and recovering information from context, and therefore less available for storage and retrieval. In the present study, working memory limitations are l ikely to affect the results of both recognition and free recall tasks. Recognition recall. Strunk Sachs (1967) reported on a study indicating that recognition memory for formal sentence characteristics declines much more rapidly than recognition memory for their meaning. However, i f deeper level processing cannot take place because of competing demands, differences in relative performance on specific recognition recall items may be observed depending on whether inferencing is required or not. Subjects may be more l ikely to select answers on multiple choice tasks whose surface form is identical to, or closely resembles that of the original text. Answers 12 containing information which does not appear directly in the original, but may be derived from i t , would then be less automatically chosen. Craik (1983) presented evidence in support of this view. He argued that the memory deficits observed in older people reflect reduced comprehension. More specifically, he claims that "the 'processing resources' necessary to energize mental operations decline with age" (p. 111). If the resources required for meaningful processing are insufficient, then the subject must fal l back on "effortless routine encoding operations" (ibid.). Along the same lines, Craik (1983) cited research indicating that older subjects frequently fai l to draw inferences from text materials. In other words, older subjects may perform based on surface form rather than semantic content. This is possible because recognition tasks are easier than recall tasks, in that they do not require subjects to produce surface forms themselves, and less information is typically required to generate a correct response (Brown, 1976). This is consistent with the argument presented by Craik (1986), namely that age-related differences in memory frequently disappear when the environmental context supports mental operations and imposes fewer demands on subjects' internal resources. Kintsch (1991), in a more recent addition to his original model, specified that language processing is largely bottom-up. In the context of this study, this would mean that the hard-of-hearing listener may not be able to proceed 13 beyond the surface form of the discourse to the level of integration. Similarly, Pichora-Fuller and Kirson (1994) suggested that comprehension is the result of a balance of top-down and bottom-up processes and may be affected equally by external circumstances (such as noise affecting signal clarity) and listener characteristics (such as cognitive abi l i ty). Free recall. Excessive on-line demands on working memory should also be reflected as impoverished free recall and content integration. This process of integration has been described by Kintsch and van Dijk (1983) as the creation of a "situational model" or macrostructure. The situational model describes the state of affairs described by a text. Surface form is typically erased in the process of language understanding and encoding. What is retained is a propositional text base consisting of explicit ly stated propositions, bridging inferences needed for coherence, and macropropositions representing inferences about the topic of the text. The "integration" phase is based on spread activation around the recipient's knowledge system. Propositions are assigned activation weights, depending on their relevance to the linguistic context, the text topic, and the situation model constructed thus far. Age-related differences in the ability to process and integrate verbal information in this way have been reported (Adams, 1991; Hess, 1992; Wingfield & Stine, 1992), with older subjects consistently showing superior use of contextual, 14 top-down information and production of more abstract and interpretive text summaries. These findings appear to contradict the evidence presented by Craik (1983), who found that older subjects frequently failed to draw inferences from text materials. However, Craik's findings pertained to inferences that required the processing of novel information at sentence level. The studies showing that older subjects produced more abstract summaries made use of narrative materials with an overall "theme" which could well have been derived without perfect sentence-level processing. Even though older subjects show a reduced capacity for assimilating new information, they appear to be more skil led at using their existing knowledge base to assign thematic and metaphorical meaning to discourse. Narrative Schemas and Scripts Some types of conversation may maximally faci l i tate integration and comprehension: "the texts we remember are often simple narratives for which we are very well prepared indeed, whereas descriptive texts tend to be harder to remember" (Kintsch & van Dijk, 1983, p. 361). In other words, narrative elements can frequently be related to prior knowledge structures, which include structural as well as thematic conventions. It is therefore possible to predict that material embedded in texts (and, by extension, conversations) which most closely adhere to narrative conventions and structural scripts 15 should be most easily processed and remembered. Conversely, texts and conversations that have a less obvious structural and thematic framework should be harder to process and remember. The powerful effect of pre-existing knowledge structures is confirmed by Neisser's (1982) discussion of John Dean's recall of his conversations with President Nixon, which demonstrates that script-based intrusions in recall increase as memory fades. Affective Value Another characteristic of text or discourse that may influence level or depth of processing is its affective value and interest to the listener/reader (Pichora-Fuller & Kirson, 1994; Yoder & Elias, 1992). Affective value may be increased by evaluative expressions of the speaker's positive or negative attitudes. Long (1994), for instance, found that subjects exhibited significantly greater memory for evaluative and formulaic expressions than for statements that were neutral with respect to the speaker's positive and negative attitudes. Johnston and Strickland (1985) compared recall of both affect-laden and more factual narration in both face-to-face and audio-only communication modes. The authors anticipated that recall of affect-free prose would be enhanced by the audio-only mode, because nonverbal cues would not be available to distract the listener from the essential (factual) meaning components of the passage. Conversely, they postulated that memory for affect-16 laden narration would benefit from face-to-face presentation. Results of their experiment, however, demonstrated that subjects showed better recall for both types of passages in the face-to-face.mode, which is likely to support a higher level of listener involvement and attention. Also, the additional context and information provided by visual cues in the face-to-face mode may partially account for these results. A'significant effect of passage type was noted, with consistently superior recall for the affect-laden prose compared to the factual passage. Thus, affective value appears to be a function both of the intrinsic nature of the material presented, as well as the presentation mode. In an attempt to control for effects of affective value, topics chosen for conversation in the present study were those ranked as of high interest and familiarity by the subjects. The presentation mode however, was not chosen to support maximal emotional involvement, as subjects were seated with their backs to the experimenter (see Chapter 2 for details). Limitations of Memory Tasks While memory tasks may reflect the level of semantic processing that has taken place during the course of the conversation, deficient recall does not necessarily imply impaired on-line comprehension. Listeners may well have adequately perceived the incoming signal and correctly assigned meaning to i t , but nonetheless failed to encode i t for future recall. 17 A second limitation is that results of memory tasks do not specify the details of the conversational process. They tap into listeners' abil ity to access the semantic components and information structures that constitute the conversation, once completed. Conversations, however, typically do not have a predetermined set of meaning components and relations, but involve ongoing negotiation of contents and participant roles. Discourse Analysis The efficiency with which contents and participant roles are negotiated in the course of the conversation provides a third indicator of discourse comprehension, and can be tapped into by means of discourse analysis, which provides quantitative measures of conversational behaviour. Other, more qualitative measures such as TOPICON and QUEST7AR (Erber, 1988; Erber & Lind, 1994) are available to assess a cl ient's use of conversational patterns and communicative strategies. The TOPICON procedure allows the clinician to e l i c i t a conversation sample from a hard-of-hearing client and evaluate the contents and fluency of the conversation (see Chapter 2 for a more detailed discussion of this procedure). QUEST?AR consists of a l i s t of questions presented by the clinician to guide simulated conversations with hard-of-hearing clients. However, because neither measure provides normative data, their results rely heavily on clinical intuition for interpretation and do not very easily allow 18 for comparison across subjects. Discourse Analysis and Comprehension While the reason we are able to manage conversations is that we have some understanding of conversational rules, many aspects of the conversation are not previously known. Nonetheless, i t is possible to specify some of the rules that constitute conversational interaction. For instance, speakers have intuitions about appropriate procedures for turntaking, timesharing, and topic management. A given speaker may not observe these procedures, either by dominating or by contributing minimally to the conversation. This lack of symmetry between speaker contributions may be an indication of poor comprehension, resulting in one speaker's inability to monitor his own conversational behaviour as he otherwise would. However, personality is a key factor in conversational style and is. certain to affect a speaker's degree of active participation. Similarly, social roles and hierarchies frequently impose limitations on conversational structure. Other aspects of the conversational interaction, such as breakdown and repair sequences, are more direct and unequivocal indicators of failed comprehension. 19 Indirect Measures of Comprehension Turntaking One aspect of the negotiation process that takes place during conversation is the mechanisms conversational participants use for turntaking. Duncan (1972) discusses some of the signals used by speakers and listeners to indicate their readiness to yield or take the floor. These include body motion, grammatical and prosodic features, and some stereotyped conclusion sequences. While turntaking usually operates smoothly, these signals are subtle enough that any competing attentional demands may adversely affect this process of "impression formation" (p. 292). On the other hand, many of these turntaking signals are nonverbal and may be picked up by hearing-impaired listeners, despite impaired perception of the conversational content. In other words, i t is possible to insert turns at the appropriate points in the conversation without the benefit of complete comprehension. Any contributions made in this way, however, may s t i l l reveal failed comprehension, either because they do not contribute meaningful content, or because they contribute content that does not follow smoothly from the previous speaker's turn. 20 Topi c Topic management is a second area requiring negotiation between speakers. Brown and Yule (1983) insist that the term "topic" qualifies speakers and writers, not texts (p. 68). In other words, speakers come to the conversation with their own ideas and. agendas,' which can only become the topic of conversation i f other participants agree to acknowledge them as such. In a fluent everyday conversation, in which all participants have equal status, the dynamics of topic negotiation result in a balance between the introduction and the integration of ideas. No one speaker dominates the conversation, nor is its interactional nature compromised because all speakers are speaking on the same topic, but not "speaking topically" (Brown & Yule, 1983; Sacks, 1995). The act of "speaking topically" creates cohesion in a given dialogue. In other words, speakers typically interact by considering the intention and meaning of immediately preceding contributions by other speakers and formulating their own contributions as contingent responses. They do not commonly hold parallel monologues on a topic, except perhaps in such formal contexts as conferences and meetings. The number of possible elaborations on any given topic is largely undetermined and depends on what the listener considers an appropropriate response to the speaker's contributions. This is consistent with Bublitz (1988), who defines "topic" as a hearer-oriented notion, which is assigned 21 through an on-line process of ascription. What one speaker considers an approppriate contribution is partly a factor of what he perceives the current topic of the conversation to be, and this is not necessarily what the other speaker has in mind. Some degree of mismatch between participants' perceptions of what constitutes the conversational topic is therefore to be expected. Even though topics for the conversations in the present study were preselected, the negotiation process s t i l l applies to the transition between subtopics and the level of cohesion between speaker turns on different subtopics. Theoretically, two patterns of topic development that are posssible indicators of failed comprehension may be observed. Subjects may fai l to introduce a significant number of new ideas and subtopics, or they may frequently elaborate on content words of the prior speaker's while ignoring the turn's actual intent, thus shifting the focus of the conversation. The latter pattern might result in a relatively disjointed conversation consisting of a large number of subtopics that are only briefly touched upon. Direct measures of comprehension. The least ambiguous indicator of failed understanding is the presence of overt breakdowns and repair sequences. However, these depend on the listener's willingness to reveal a lack of comprehension and thereby interrupt the conversational flow. 22 In the more formal context of research, subjects may be expected to perceive a tendency toward experimenter control and be less assertive than they otherwise might be. This is particularly relevant in the case of hard-of-hearing subjects, whose comprehension of the conversation may depend on their level of assertiveness and for whom self-initiated repair is frequently not perceived to be an option. Subjects may either feel less inhibited, since less is at stake in a staged conversation, or more reluctant because of the unfamiliar setting and conversation partner. A second issue that may be raised here is when failure to seek repair becomes psychologically and cl inical ly significant. As noted in Linell (1995), all cultures have "rules of etiquette and politeness which serve to mitigate or even inhibit attempts at thematisation and/or repairing some kinds of possible misunderstandings. Utterances are produced not only 'with respect to' the other but also 'in respect for' the other (and oneself)" (p. 183). The addition "and oneself" is important, as i t cannot be assumed that hard-of-hearing participants' only goal in conversation is to comprehend its contents as fully as possible; some conversations may simply not warrant that kind of effort and concentration. Limitations of Discourse Analysis The absence of hard-.and-fast conversational rules implies that, even though quantitative measures of conversational competence may be obtained, they need 23 to be interpreted with respect to the specific context in which they occur. Also, even direct measures of comprehension such as the use of repair strategies may fai l to give an adequate account of a listener's understanding of the conversation. Participants' personalities or perceived role in the conversation may result in a more passive conversational style that does not explicitly reveal problems with auditory processing and comprehension. Taking turns, negotiating topics, and monitoring understanding are operations essential to the development of conversation, but all require some degree of assertiveness on the speaker's part. For clinicians, an additional drawback to this approach to assessment of conversational competence is the time investment required. Tye-Murray (1994) raises a more fundamental concern, which is a limitation of the way clinical conversations have typically been set up to date rather than of the discourse analysis approach per se. She notes that assessments based on staged interactions do not take into account that "clients' use of communication strategies may vary as a function of conversational partners and settings" (p. 196). 24 Research Questions The purpose of this study is to investigate methods for evaluating the conversational behaviour of hearing-impaired seniors, specifically the procedures of self-report, recall and discourse analysis described above. Specific research questions addressed are: 1. What measure or combination of measures is most helpful in describing hearing disability and handicap? In other words, what pertinent information can or cannot be obtained from the various measures? 2. What are the cl inical ly significant differences between the conversational behaviour of normal-hearing and hearing-impaired seniors? 3. How do the listening conditions and signal-to-noise ratio affect the participants in the conversation? CHAPTER 2 25 METHOD This study is based on the analysis of four conversations between KF, a rehabilitative audiologist, and two seniors, one of whom is hearing-impaired. Pichora-Fuller & Johnson's (submitted) discussion of the conversations held with GK, a hard-of-hearing senior from the same subject pool, will provide a basis of comparison for the data presented here. A discussion of the results of this case study onGK wil l be presented in Chapter 3. 2 Subjects The Subject Pool Data were collected from a total of 24 subjects, participating in a project designed to determine how everyday communication is affected by hearing impairment. The subject pool included a group of eight presbycusic elderly subjects (mean age = 75 years, SD = 6). In addition, a group of eight normal-hearing seniors (mean age = 70 years, SD = 4) participated in this experiment, as well as eight normal-hearing young adults (mean age = 24, SD = 3). For the purposes of this study, "normal-hearing" was defined as pure-tone air-information on GK, as well as the method used for data collection and analysis was adapted and reprinted with permission from the authors. 26 conduction thresholds less than or equal to 25 dBHL from 250 to 3000 Hz, with no conductive component, and with no interaural threshold asymmetry greater than 15 dB at more than two of the standard audiometric test frequencies. Preliminary Measures All subjects participated in a series of preliminary tests intended to assess their personal and medical background, hearing status, and cognitive abi l i t ies. These included the following: Basic audiometry. Basic audiometric measures obtained from all participants included a pure-tone audiogram, as well as speech recognition thresholds in quiet (CID Auditory Test W-l). Word discrimination scores were obtained using the CID Auditory Test W-22. Revised Speech Perception in Noise (Bilger, Neutzel, Rabinowitz & Rzeczowski, 1984). The Revised Speech Perception in Noise (SPIN-R) test is a speech discrimination test using sentence-level stimuli,, which are presented in competing 8-talker babble noise with a standard signal-to-noise ratio of +8 dB. The level of presentation of speech was set at 50 dB above subjects' babble thresholds. For the purposes of this study, sentences were also presented in a range of nonstandard signal-to-noise ratio conditions. Subjects were asked 27 to repeat the test item, which is the final word in each sentence. In half of the test sentences, target words are highly predictable from the preceding context. The other half of the sentences do not provide a predictable context for the test items, so that subjects are required to rely entirely on bottom-up processing. Hearing Handicap Inventory for the Elderly (Ventry & Weinstein, 1982). The Hearing Handicap Inventory for the Elderly (HHI/E) was designed to assess the impact of hearing-impairment on the emotional and social adjustment of elderly people. The comprehensive version consists of 25 items. A shortened version consisting of ten items, most often used as a screening tool, was administered in the present study. High scores on this inventory are correlated with a low degree of hearing handicap. Hearing Performance Inventory (Giolas et a l . , 1979). Consisting of 131 items to be rated on a 5-point scale, the Hearing Performance Inventory (HPI) requires much more administration time than the HHI/E, but as a result is l ikely to be more informative. It yields a total score, as well as four subscores, on subscales rating social adjustment, understanding of speech in noise, understanding of speech in quiet, and response to auditory failure, respectively. Contrary to the HHI/E, the HPI was not designed specifically for 28 use with the elderly. High scores on this questionnaire correspond to a high degree of reported handicap. Mill House Vocabulary Test,(Raven, 1946). This multiple-choice task was designed to test the vocabulary ski l ls of normal-functioning adults. Test items are 20 words ranging from more to less common or frequent. Subjects select the word closest in meaning from a l i s t of six alternatives. Working Memory Span Test (Daneman & Carpenter, 1980). As discussed in the previous chapter, comprehension of spoken language is strongly influenced by working memory span, insofar as i t governs the subject's abil ity to integrate new auditory information with previously acquired knowledge, and to process or manipulate this information. Daneman and Carpenter (1980) found working memory span to be highly correlated with individual differences in the ability to comprehend both written and spoken language. As in the SPIN-R, the test items are sentence-final words; however, in the working memory task, subjects are required to recall test items from up to six successive sentences. Both a reading and listening version of this test are available. Subjects of the present study participated in the reading version of the working memory span test, using the Daneman and Carpenter's (1980) materials. 29 Group data Following audiometric testing, all subjects participated in the SPIN-R, Mill House Vocabulary, and Working Memory Span tests described in the previous section. In addition, each subject completed both the HPI and HHI/E questionnaires. Averaged results on these measures for each of the three experimental groups are given below. SPIN-R. Based on the range of scores obtained for a range of signal-to-noise ratio conditions on the SPIN-R, i t was estimated under what condition subjects were expected to recognise 50ft of the test items. These predicted 50ft correct scores are presented in Table 1. It is obvious from these results that performance on the SPIN-R is least compromised by increasing noise levels in the young group. Also, old subjects generally score better than presbycusic subjects. In particular, presbycusic subjects appear to be much more strongly affected by adverse signal-to-noise ratios in the low-context conditions than either of the other groups (see also Pichora-Fuller, Schneider & Daneman, 1995). 30 Table 1. Average of estimated S:N in dB corresponding to 50$ SPIN-R score for each subgroup ± one standard deviation Presbycusic Old Young High-context Sentences -0.3 ± 1.5 -3.0 ± 1.5 -5.0 ± 1.0 Low-context Sentences 8.2 ± 1.6 1.5 ± 2.9 -1.7 ± 1.3 Mill House Vocabulary and Working Memory Span tests. The results of each participant subgroup on the Mill House Vocabulary and Working Memory Span tests are presented in Table 2. Young subjects showed lower vocabulary scores (possibly as a function of lesser l i fe experience) but higher working memory span than either of the other groups. Given the relation between comprehension of spoken language and working memory span discussed above, both groups of old adults may therefore be expected to show greater diff iculty understanding new information embedded in conversation. The difference noted between the old and presbycusic groups is unexpected, and may be attributed to individual differences. Hearing Handicap Inventory for the Elderly and Hearing Performance Inventory. In addition to the tests described above, each subject completed both the HHI/E and HPI questionnaires. Subgroup averages for these results are given in Table 3. 31 Lower scores on the HPI indicate better performance. On the HHI/E, higher scores indicate better performance. As expected, both groups of normal-hearing participants report equivalent and minimal levels of hearing handicap. Presbycusic subjects score almost twice as 'poorly on the HHI/E and all HPI subscales, with the exception of "Response to Auditory Failure." Table 2. Average Mill'House Vocabulary and Working Memory Span scores per subgroup ± one standard deviation Presbycusic Old Young Mill House Vocabulary 16.5 ± 2.3 15.3 ± 2.5 14.3 ± 1.8 Working Memory Span 3.21 ± 0.9 2.75 ± 0.7 4.0 ± 0.8 Table 3. Average HPI and HHI/E scores per subgroup ± one standard deviation Presbycusic Old Young HPI Understanding Speech in Quiet 2.0 + 0 6 1 2 + 0.2 1 1 + 0 2 Understanding Speech in Noise 3.1 + 0 1 1 6 + 0.4 1 6 + 0 7 Response to Auditory Failure 3.1 + 0 5 2 9 ± 0.8 2 6 ± 0 7 Social Subscale 3.0 + 0 6 1 8 + 0.4 1 8 + 0 5 Total 2.7 + 0 5 1 7 ± 0.4 1 6 + 0 4 HHI/E 2.9 + 0 7 3 8 + 0.6 3 9 ± 0 2 32 One possible explanation for these scores on the latter HPI subscale is that normal-hearing subjects may in fact respond to auditory breakdowns in ways similar to hearing-impaired subjects. Another factor may be that this particular subscale does not take into account the frequency of situations in which poor hearing is a factor. Individual Subjects The present study will focus on two subjects, selected from the old and presbycusic groups, respectively. In addition, the Pichora-Fuller and Johnson (submitted) data on GK, a hard-of-hearing senior, wil l be presented for the purpose of comparison. Subject 1: EL EL belonged to the group of eight normal-hearing seniors. He was a monolingual English speaker who had 13 years of formal education, which was below the average of the normal-hearing old group (16 years). However, at the time of this study, EL was taking university courses in Asian studies. He had previously worked as an accountant. EL obtained a score of 17/20 on the Mill House Vocabulary Test, which was above the average score for the old group. EL's reading working memory span 33 of 2.33 items on the Daneman and Carpenter (1980) task was below the average span measured for the presbycusic, the old, and the young groups. His general health was good and he was taking no medications that would affect communication. He reported no problems with hearing or vision. Objective measures of hearing impairment included basic audiometry, as well as standard and nonstandard measures of speech perception in noise. EL was found to have hearing within normal limits bilaterally (see Table 4). Consistent with a pure-tone average within normal limits, speech recognition thresholds in quiet for the right and left ear were both obtained at 10 dBHL. Speech discrimination in quiet was excellent, with CID W-22 scores of 96$ when speech was presented at 45 dBHL to the right ear and 96$ when i t was presented at 45 dB HL to the left ear. On the SPIN-R test, in the standard +8 dB S:N condition, with speech presented at 45 dBHL and 8-talker competing babble presented at 37 dBHL, EL correctly recognised 100$ of the sentence-final target Table 4. Pure-tone thresholds in dBHL for EL Frequency (kHz) .25 .5 1 2 3 4 6 8 Right Ear (Air Conduction) 10 10 5 10 10 10 30 25 Left Ear (Air Conduction) 10 10 15 10 10 15 20 20 Left Ear (Bone Conduction) 5 5 5 10 10 15 34 words presented in high-predictability contexts and 84ft of those presented in low-predictability contexts. His percent correct scores for words in high- and low-predictability sentences on l ists presented over a range of nonstandard S:N conditions are shown in Table 5. Based on the range of scores obtained, the conditions in which EL is expected to recognise 50ft of the words on the SPIN-R test correctly are .14 dB S:N and -5.1 dB S:N for low- and high-predictability contexts, respectively. His abi l i ty to perceive sentence-final words in competing babble is therefore slightly better than the average for the group of normal-hearing seniors. His performance is comparable to that of the group of normal-hearing young listeners and is well above that of the presbycusic group. Both the HHI/E and HPI questionnaires were administered. EL's score of 4 on the HHI/E was slightly higher than the average scores of both normal-hearing groups, indicating a low degree of perceived hearing handicap. Similarly, his Table 5. SPIN-R percent correct score as a function of S:N condition for EL S:N (dB) -6 -4 -2 0 +2 +4 +6 Low-Context Sentences 28 76 96 92 100 100 100 High-Context Sentences 12 36 52 48 76 84 100 35 HPI total score of 1.69 is about average for both of these groups, again corresponding to a low degree of handicap. EL's HPI subscale scores (see Table 6) fall within one standard deviation of those obtained for the normal-hearing groups. All of his HPI and HHI/E scores are better than those of the presbycusic group. Overall, EL appears to be a typical member of the old group in terms of his hearing ability and speech perception sk i l ls . Considering his years of education, his vocabulary score, and his working memory span combined, EL is l ikely to have average cognitive abil i t ies compared to most elderly individuals. Table 6. HPI subscores for EL and experimental subgroups EL Presbycusic Old Young Understanding Speech in Quiet 1,00 2.0 ± 0.6 1.2 ± 0.2 1.1 ± 0.2 Understanding Speech in Noise ' 2.00 3.1 ± 0.1 1.6 ± 0.4 1.6 ± 0.7 Response to Auditory Failure 2.82 3.1 ± 0.5 2.9 ± 0.8 2.6 ± 0.7 Social Subscale 1.81 3.0 ± 0.6 1.8 ± 0.4 1.8 ± 0.5 36 Subject 2: ZW ZW belonged to the group of eight presbycusic subjects. He was a monolingual English speaker who had l ived in several countries and Canadian provinces during his adult l i f e . He was previously employed in the mil i tary forces and had also worked as an off ice administrator. He had 15 years of formal education, which was the average for the presbycusic group, but s l ight ly below the average of the normal-hearing old group (16 years). ZW obtained a score of 18/20 on the Mil l House Vocabulary Test, which was above the mean scores measured for a l l three groups. Taking this vocabulary score as an indicator of intel l igence, ZW was a highly intel l igent individual. ZW's reading working memory span of 3 items on the Daneman and Carpenter (1980) task was below the mean span measured for the young subjects and the presbycusic subjects, but better than the average span of the old group. ZW's general health was good and he was taking no medications that would affect communication. He did, however, report problems with vision. He also reported problems with hearing and tinnitus in the right ear. ZW had had his hearing tested but was not wearing a hearing aid at the time of this study. In fact , the sharply sloping configuration of his hearing loss (see Table 7) made him a poor candidate for amplification. 37 Objective measures of hearing impairment included basic audiometry, as well as standard and nonstandard measures of speech perception in noise. ZW was found to have a bi lateral sloping sensorineural hearing loss consistent with presbycusis and/or noise-induced hearing loss (see Table 7). Consistent with a pure-tone average within normal l imi ts , speech recognition thresholds in quiet for the right and le f t ear were measured at 5 and 0 dBHL, respectively. Speech discrimination in quiet was good, with CID W-22 scores of 80$ when speech was presented at 40 dBHL to the right ear and 88$ when i t was presented at 35 dB HL to the lef t ear. On the SPIN-R test, in the standard +8 dB S:N condition, with speech at 50 dBHL and 8-talker competing babble at 42 dBHL, ZW correctly recognized 100$ of the sentence-final target words presented in high-predictabil i ty contexts and 40$ of those presented in low-predictabi l i ty contexts. His percent correct scores for words in high- and low-predictability sentences on l i s ts presented over a range of S:N conditions, Table 7. Pure-tone thresholds in dBHL for ZW Frequency (kHz) .25 .5 1 2 3 4 6 8 Right Ear (Air Conduction) 10 0 5 20 80 90 95 95 Left Ear (Air Conduction) 10 0 5 10 60 85 95 95 Left Ear (Bone Conduction) 5 0 0 15 70 75 --38 Table 8. SPIN-R percent correct score as a function of S:N condition for ZW S:N (dB) -4 -2 0 +4 +12 +16 +20 Low-Context Sentences 4 36 76 100 100 100 100 High-Context Sentences 0 0 32 32 60 . 80 72 in which the speech was held constant and the noise level varied, are shown in Table 8. Even given very favourable S:N ratios,' ZW's best score on the low-context sentences is only fa i r , and s l ight ly lower than his speech discrimination scores in quiet. Based on the range of scores obtained, the condition in which ZW is expected to recognise 50ft of the words on the SPIN-R test correctly are 9.14 dB S:N and -1.3 dB S:N for low- and high-predictabil i ty contexts, respectively. His abi l i ty to perceive sentence-final words in competing babble is therefore within one standard deviation of the scores obtained for the presbycusic group. However, i t is much poorer than that of the group of normal-hearing seniors, or the group of normal-hearing young l isteners. Both the HPI and the HHI/E were administered. On the HHI/E, ZW scored 2.2, which is at the upper margin of the range obtained for the prebycusic group, but poorer than the average scores of both normal-hearing groups. In 39 part icular, he indicated that he occasionally does have d i f f i cu l ty hearing, part icularly when l istening to whispered speech or in noisy conditions. He also responded that his hearing loss appears to create conf l ic t with family members and causes frustration (see Appendix A for a complete l i s t of responses). Similar ly, ZW's total score on the HPI of 2.75 is equivalent to the average score of the presbycusic group, but worse than the average scores of the old and young groups. Of the HPI subscale scores (see Table 9), a l l were within one standard deviation of the average scores for the presbycusic group, except for the "Understanding Speech in Noise" subscale, on which ZW scored worse. Table 9. HPI subscores for ZW and experimental subgroups ZW Presbycusic Old Young Understanding Speech in Quiet 2.48 2.0 ± 0.6 1 2 ± 0.2 1.1 ± 0 2 Understanding Speech in Noise 3.58 3.1 ± 0.1 1 6 ± 0.4 1.6 ± 0 7 Response to Auditory Failure 2.84 3.1 ± 0.5 2 9 ± 0.8 2.6 ± 0 7 Social Subscale 3.43 3.0 ± 0.6 1 8 ± 0.4 1.8 ± 0 5 40 Overall, ZW appears to be a fa i r l y typical member of the presbycusic group. His re lat ively normal pure-tone average suggest that ZW's hearing is l ike ly to be adequate for quiet l istening situations. Reduced speech discrimination is particularly evident in noise and when the redundancy of the speech signal is minimal, consistent with the sharply sloping loss in the high frequencies. Considering his years of education, his vocabulary score and his working memory span combined, ZW is l ike ly to have cognitive ab i l i t i es comparable to most elderly individuals. Subject 3: GK Like ZW, GK belonged to the group of eight presbycusic subjects 3. She was an active woman who l ived alone in her own apartment but maintained regular contact with her extended family. She was a monolingual English speaker who had l ived in a number Canadian provinces during her adult l i f e . She was a trained musician who had worked as a singing teacher for many years. She had 19 years of formal education, which was above the average for a l l three groups. GK obtained a score of 19/20 on the Mil l House Vocabulary test, which was above the average scores measured for al l three groups. Taking this vocabulary score 3This description is taken from Pichora-Fuller and Johnson (submitted), with permission 41 as an indicator of intel l igence, GK was a highly inte l l igent individual. GK's reading working memory span of 4 items on the Daneman and Carpenter (1980) task was equivalent to the average span measured for the young group and better than that of the presbycusic group or the old group. GK's general health was good and she was taking no medications that would affect communication.' She did, however, report problems with v is ion. She also reported problems with hearing. She had no signif icant medical history that would indicate any etiology of hearing loss other than presbycusis. She reported that people sometimes complained about her hearing, but she had never had a hearing test prior to her participation in the study. Objective measures of hearing impairment included basic audiometry, as well as standard and nonstandard measures of speech perception in noise. GK was found to have a bi lateral sloping mild-to-severe sensorineural hearing loss consistent with presbycusis (see Table 10). Table 10. Pure-tone thresholds in dBHL for GK Frequency (kHz) .25 .5 1 2 3 4 6 8 Right Ear (Air Conduction) 35 30 40 45 55 50 80 75 Left Ear (Air Conduction) 35 30 40 55 55 60 75 80 Left Ear (Bone Conduction) 30 35 40 55 50 60 42 Consistent with GK's audiogram, speech recognition thresholds in quiet for the right and le f t ear were measured at 30 dBHL. Speech discrimination in quiet when speech was presented at" 65 dB HL was good, with CID W-22 scores of 84ft for the right ear and 76ft for the le f t . On the SPIN-R test, in the standard +8 dB S:N condition, with speech presented at 65 dBHL and 8-talker competing babble presented at 57 dBHL, GK correctly recognised 100ft of the sentence-final target words presented in high-predictabil i ty contexts and 44ft of those presented in low-predictability contexts. Her percent correct scores for words in high- and low-predictabil ity sentences on l i s t s presented over a range of S:N conditions, in which the speech was held constant and the noise level varied, are shown in Table 11. Even given very favourable S:N ratios, GK's best score on the low-context sentences is only fa i r , and lower than her speech discrimination scores in quiet. Based on the range of scores obtained, the condition in which GK is expected to recognise 50ft of the words on the SPIN-R test correctly are 8.75 dB S:N and -1.0 dB S:N for low- and high-predictability contexts, respectively. Her ab i l i ty to perceive sentence-final words in competing babble is therefore about average for the presbycusic group. However, i t is much poorer than that of the group of normal-hearing seniors, as well as the group of normal-hearing young l isteners. 43 Both the HPI and the HHI/E were administered. On the HHI/E, GK scored 2.8, which was equivalent to the average of the presbycusic group, but poorer than the average scores of both normal-hearing groups. This score is consistent with a degree of reported hearing handicap similar to that of the hearing-impaired group, and higher than that of the normal-hearing participants. In particular, she indicated that she occasionally does have d i f f i cu l ty hearing, such as when listening to whispered speech. She does not, however, report that her problems are of any emotional or social consequence, except that she sometimes feels embarrassed when meeting new people (see Appendix A for a complete l i s t of responses). Similar ly, GK's total score on the HPI of 2.6 is equivalent to the average score of the presbycusic group, but worse than the average scores of the old and young groups. Of the HPI subscale scores (see Table 12), a l l were within one standard deviation of the average scores for the presbycusic group, except for the Response to Auditory Failure subscale, on which she scored much worse. Table 11. SPIN-R percent correct score as a function of S:N condition for GK S:N (dB) -2 0 +2 +4 +8 +10 +12 +16 Low-Context Sentences 44 56 76 92 100 100 96 100 High-Context Sentences 16 8 24 52 44 60 64 72 44 Overall, GK appears to be a typical member of the presbycusic group in terms of her degree of hearing impairment and d isab i l i t y . Reduced speech discrimination is particularly evident in noise and when the redundancy of the speech signal is minimal. Considering her hearing impairment, i t is noteworthy that she reports l i t t l e psychological and social handicap. Despite a higher degree of hearing loss, GK's scores on three of the four HPI subscales suggest that she experiences less hearing handicap than ZW. Considering her years of education, her vocabulary score and her working memory span combined, GK is l i ke l y to have cognitive ab i l i t i es comparable to those of young subjects and superior to many elderly individuals. Table 12. HPI subscores for GK and experimental subgroups GK Presbycusic Old Young Understanding Speech in Quiet 1.70 2.0 ± 0.6 1.2 ± 0.2 1.1 ± 0.2 Understanding Speech in Noise 2.70 3.1 ± 0.1 1.6 ± 0.4 1.6 ± 0.7 Reponse to Auditory Failure 3.80 3.1 ± 0.5 2.9 ± 0.8 2.6 ± 0.7 Social Subscale 2.80 3.0 ± 0.6 1.8 ± 0.4 1.8 ± 0.5 45 Procedure Each of the three subjects engaged in two conversations with KF, lasting roughly between 10 and 20 minutes. These two conversations were held one month apart and were conducted in an IAC sound-attenuating booth with KF seated behind the subjects so that no visual cues were available to them. Competing babble noise was presented to both ears under earphones. This noise was the eight-talker babble recording used to determine babble threshold in the SPIN-R test (Bilger et a l . , 1984) and was routed from a taperecorder through a Madsen OB802 c l in ica l audiometer. For the f i r s t conversation with ZW, the babble level was set to 40 dBHL; for the second, the noise level was increased to 50 dBHL. The f i r s t conversations with EL and GK also took place with a noise level of 40 dBHL; for the second conversation, i t was f i r s t set to 50 dBHL and then increased to 55 dBHL about halfway through. Assuming that KF was talking at an average conversational level of 45 dBHL, the f i r s t conversations took place in a S:N condition of about +5 dB, a level at which al l three subjects obtained perfect scores on the high-context sentences of the SPIN-R test. At a noise level of 50 dBHL, or a S:N condition of about -5 dB, EL's estimated score on the high-context sentences l ies between 28 and 76 percent correct, and ZW's below 4% percent correct. At a noise level of 55 dBHL, or about -10 S:N, EL's performance would have been expected to drop below 26% correct. The S:N 46 conditions of about -5 and -10 dB were levels more adverse than those at which GK had missed more than half of the target items even in high-context sentences. Thus, the f i r s t conversations were intended to take place in a relat ively favourable l istening condition, whereas the second conversations were intended to take place in a relat ively adverse l istening condition. A l l speakers wore lapel microphones and the conversations were audiotaped for later analysis. The conversational task was a modified version of Erber's (1988) TOPICON procedure, in which c l ient and c l in ic ian engage in a 5- to 10-minute conversation on a topic previously selected by the c l ient from a set of topics supplied by the c l in i c ian . The set consisted of the following topics: Cats, Restaurants, Family Parties, Real estate, Holidays, Cigarette Smoking, Moving, Unemployment, Tax, Shopping, Diets, Education, Cooking, Computers, Weather, Babies. For each subject, two topics were selected based on a fami l iar i ty and an interest ranking of this set of topics. On a scale of 1 to 16, 1 indicated the topic of highest fami l iar i ty and interest, respectively. The topics retained for ZW were "Family Part ies," rated 5/16 for interest and 3/16 for fami l iar i ty and "Babies," rated 2/16 for interest and 1/16 for fami l iar i ty . Thus the latter topic, chosen for the conversation with 50 dBHL noise (the less favourable condition) was sl ightly more familiar and interesting to ZW than the 47 former. Topics for EL were "Real Estate," rated 2/16 for both interest and famil iari ty, and "Holidays," rated 3/16 for interest and 1/16 for fami l iar i ty . In EL's case, then, overall interest and familiarity rating were equal for both conversations. GK rated the f i r s t topic, "Cats," 2/16 for interest and 7/16 for famil iari ty, and the second, "Restaurants," as 5/16 for interest and 10/16 for famil iari ty. Thus, topics were generally of lower interest and fami l iar i ty to GK than they were to ZW and EL. The TOPICON procedure was modified in this study to allow for comprehension testing in the form of recognition reca l l . A l i s t of ten personal facts which could not have been part of or derivable from the subjects' world knowledge was prepared by KF prior to the conversation. KF embedded these facts in the conversation in a relatively natural fashion. It was then possible to conduct a postconversation comprehension test to determine how well subjects had retained this sample of the conversation. Comprehension Measures Five different measures of comprehension were taken, the results of which were later compared. Immediately following the conversation, the subjects completed three tasks: 1. a verbal protocol, 2. free recall of the conversation, and 3. recognition recall of the ten facts embedded in the conversation. Based on the 48 taperecordings, two additional measures were obtained: 1. fluency ratings and 2. detailed discourse analysis. Verbal Protocol A modified version of Erber's (1988) verbal protocol was administered following a l l conversations (except "Family Parties" for ZW). The verbal protocol consists of a brief questionnaire (see Appendix B for a complete l i s t of questions) which provides a framework for the c l i n i c ian , who interviews the hard-of-hearing person with regards to how much of the conversation was understood and why. These interviews were taperecorded for later transcription and analysis. Free Recall A free recall test was administered following al l four conversations. Subjects were given written and verbal instructions to write a verbatim transcript of the conversation that had taken place in the session. Written instructions were as follows: Imagine that on your way out of the building today you are stopped by an RCMP of f icer who te l l s you that Dr. Fuller is suspected of being a spy. The off icer believes that you may have been mistaken for an undercover 49 contact and wants you to report as accurately as possible exactly what was said by Dr. Fuller and you during the conversation that took place in the lab. Please write the report in the following format: Dr. F.: Let's have a conversation. Me: O.K. That's fine with me. Subjects were told that they could take as much time as they wanted to write their report. Specifically, they were told to include any part of the conversation they could remember even i f the order in which they remembered these parts was not the same as in the original conversation. The free recall test was later compared to the original conversation to determine how much and which parts of the conversation had been remembered. Specifically, each conversation was divided into subtopics, turns and clauses (see Appendix C for definitions of the italicised terms used in the detailed discourse analysis). Both clauses and turns were categorized as to whether or not they contributed new information or content to the conversation. Only clauses'and turns conveying content were considered conversational material to be remembered. That i s , clauses such as "oh, really?" and "I didn't get that" were not counted. To determine how much of the conversation had been understood and encoded, the following measures were calculated: 1. the number of subtopics spoken on with content by each participant, 2. the number of content turns and 50 clauses contributed per subtopic by each participant, and 3. the number of content clauses and turns by each participant from which the subject was able to recal l material. Correctly recalled content was taken as evidence of comprehension; conversely, incorrectly recalled content was taken as evidence of lack of comprehension. Of course, this recall task is not a reflection of subjects' abi l i ty to process and encode information on-line. Failure to recall information does not necessarily imply lack of comprehension in the course of the conversation. Recognition Recall As mentioned in the previous chapter, memory l imitations compromise the value of free recall as a measure of comprehension. Recognition recall places fewer demands on subjects' resources because no information has to be generated. Simply, a judgement needs to be made as to whether or not information has been previously encountered. On the other hand, the recognition test is selective and only tests a small sample of the content to be understood and remembered. The recognition test was administered following al l four conversations in the form of a multiple choice task with questions pertaining to each of the facts embedded in the conversations (see Appendix D for the l i s t s of embedded facts and Appendix E for the recognition recall tests for the four topics). 51 Conversational Fluency Ratings Fluency ratings were provided by an expert c l in ic ian (NL) who had previously rated over 100 conversations in collaboration -with Norman Erber using his Conversational Fluency Rating Scale (1988). Upon completion of the conversation, or after l istening to a taperecording of i t , the cl in ican uses a five-point scale ranging from "low" to "high" to rate the overall fluency of the conversation, as well as 14 specif ic aspects defining subcomponents of conversational fluency. These are: presupposition, receptive ab i l i t i e s , expressive ab i l i t i es , motivation/attention, turntaking, specif icity/accuracy, new vs. old information, nonverbal communication, topic maintenance, cooperation, time sharing, ver i f icat ion, independent repair, and metacommunication ab i l i ty (for futher def in i t ion, see Erber, 1988, p 82-83). Rather than considering the contributions of the hard-of-hearing participant only, ratings are based on the c l in i c ian 's global impression of the whole conversation. Therefore, conversational fluency may be high i f the d i f f i cu l t i es of the hard-of-hearing partner are offset by the c l i n i c ian 's sk i l l at using compensatory strategies. Erber (1988) defines fluency as an "attribute of a conversation in which information, ideas, feelings, and attitudes are exchanged in an ef f ic ient and coherent manner; such a conversation contains few diversions, l i t t l e explanation or se l f - c la r i f i ca t ion , and not very much meta-52 communication" (p. 82). A conversation is rated as high or low depending on the amount of information that appears to be understood. Speci f ica l ly , a score of 1 (low) would mean that only 20% of the conversation has been understood, as evidenced by the fact that repetition and c lar i f i ca t ion are required v i r tua l l y a l l of the time. Conversely, a score of 5 (high) is given when repetition and c lar i f ica t ion are required v i r tual ly none of the time, or when more than 80% of the conversation is understood. Intermediate ratings correspond to intermediate levels of comprehension. A rating of 2 is assigned i f 21-40% of spoken sentences are understood, or repetition and c lar i f i ca t ion are required most of the time. A rating of 3 is assigned i f 41-60% of the spoken sentences are understood, with repetition and c lar i f i ca t ion being required half of the time. A rating of 4 is assigned i f 61-80% of the conversation is understood, with repetition and c lar i f i ca t ion being required l i t t l e of the time (Erber, 1995). Discourse Analysis Discourse analysis quantifies aspects of the conversation that can be interpreted as indices of comprehension, though this interpretation is not always straightforward. As mentioned above, conversations were f i r s t divided into subtopics, turns and clauses. It was then determined how much of the 53 conversation consisted of actual content to be understood, i .e. which clauses conveyed new content, which turns contained at least one content clause and which subtopics included at least one content turn. Second, measures of rate of information flow were calculated, such as the mean number of content clauses per turn, the mean number of content turns per topic, content turns and clauses per minute, and the mean number of topics per minute. Presumably, a less fluent conversation wil l have a lower rate of information flow than a more fluent one. In addition to indicating failed comprehension, a higher instance of breakdowns and repair sequences will slow down the rate of information exchange. Third, the patterns of turntaking and topic management were analyzed. It was determined how many subtopics were introduced by each participant and on how many of these subtopics each participant spoke with content. Finally, features indicating comprehension (or lack thereof) were identified. Possible indices of comprehension are shared subtopics, smooth subtopic transitions, high use of connectives across turns, overlapping turns, requests for verification when comprehension was confirmed, acknowledgements, backchannel responses, and information questions relating to the content of the prior speaker's turn. Conversely, lack of comprehension is indicated by abrupt subtopic shifts, failure to contribute to a subtopic, overt statements about 54 failed comprehension or comments about hearing, spontaneous repairs by the current speaker, requests for repair, responses to requests for verification confirming failed comprehension, interruptions, and gaps over three seconds between turns. While some of these measures are clear indicators of comprehension or failed comprehension, many are ambiguous and need to be related to the specific context, intonational contours, their exact time of occurrence, and other variables to be interpreted fair ly. For instance, while statements about poor hearing are clear indicators of failed information flow, backchannel responses and acknowledgements are not so clearly interpretable, as their function is partly social. Their use could merely be an expression of the listener's willingness to remain engaged in the conversation despite impaired comprehension, or his/her reluctance to interrupt the flow of the conversation with repair requests. Analysis Both the conversations and verbal protocols were transcribed for further analysis. For each subject, a detailed discourse analysis was completed on the transcripts of all conversations. Results of the Discourse Analysis were then compared and related to those obtained from the verbal protocols, memory tasks, 55 and conversational fluency ratings. The specif ic focus of this study was on effects of hearing abi l i ty , noise level, and individual style on conversational behaviour. Details of the data analysis wi l l be provided in Chapter 3. 56 CHAPTER 3 RESULTS In this chapter, the data obtained by means of the measures discussed above wil l be presented for the three subjects individually (EL, ZW, and GK4). An attempt wil l also be made to isolate which of the factors of hearing, noise level, and individual style are most l ikely to account for any effects observed. Subject 1: EL Verbal Protocol For the conversation held under favourable listening conditions ("Real Estate"), EL reported having understood all of the conversation, and assumed the same of the experimenter. Correspondingly, he did not find the conversation very effortful, nor did he think the experimenter did. EL stated that the effects of the noise became negligible after he had habituated to i t . In the second conversation (with a higher noise level), he s t i l l reported having understood most of its content, but that i t required much more effort, particularly after the noise level was turned up and towards the end. The 4Adapted from Pichora-Fuller & Johnson, submitted. 57 babble noise made i t d i f f i cu l t to concentrate and was distracting to a degree that EL stated he would not have been able to continue the conversation for very much longer. He was aware that asking for repetition or c la r i f i ca t ion might have been beneficial ; on the other hand, he did not feel compelled to make sure he had understood every detail of the conversation. He reported that both topics were of personal interest to him and estimated that both speakers had shared their time equally. On the other hand, he considered i t the experimenter's role to make sure the conversation went smoothly, and correspondingly reported that she asked more questions than he did, in order to fac i l i t a te the conversational process (see Appendix B). Recall Tables F1-F3 in Appendix F summarise the analyses of EL's free and recognition recall results. Recognition Recall On the recognition recall test, EL scored 9/10 correct on the conversation about Real Estate and 6/10 on the conversation about Holidays. Most errors occurred on questions pertaining to the second half of the conversation (with a noise level of 55 dBHL), resulting in a partial score of 2/5. In the verbal 58 protocol, EL indicated having missed one of the facts in the second half of the conversation ("in 1985, we went to an a l l - inc lus ive resort called Eden II in Jamaica for a week"), where the information requested was a factual detail (the date of the holiday, i .e . 1985). Other errors related to facts introduced during an abrupt topic shif t . For instance, with respect to the fact that "at University of Toronto, the second week of February is the 'reading week' holiday," the test item required EL to recall when "reading week" took place. However, this fact was mentioned at the start of the conversation, without a great deal of preceding context that would have allowed EL to incorporate this information more readily. In addition, this particular detail ("the second week of February") appeared to be secondary to the main topic, namely the fact that at the University of Toronto, students get a week off in the early spring, and the reasons why the holiday was inst i tuted. A third category of errors pertained to information requiring some degree of inferencing (e.g. "Lake Superior" phrased as "Northern Ontario" on the multiple choice task). Free Recall The amount of each conversation that was recalled was measured in terms of the number of content clauses, the number of content turns and the number of subtopics contributed by each participant that was at least par t ia l ly 59 remembered. Results for EL are given in Table 13. Overall, EL showed approximately equal memory for his own content turns and clauses compared to KF's. The fact that recall of his own subtopics dropped across the three conditions may or may not be a function of the adversity of the l istening condition. He introduced so few topics overall that the significance of this effect is questionable. Overal l , there does not appear to be a signif icant difference in recall between the two conversations, with the possible exception of a s l ight ly lower number of subtopics recalled for the second conversation. This suggests that the signal-to-noise ratio did not affect EL's comprehension, even though the verbal protocol indicates i t affected the degree of effort required to process Table 13. Percentage of content turns, content clauses and subtopics recalled by EL for each participant and each conversation KF EL Clauses Turns Subtopics Clauses Turns Subtopics Real Estate 17 30 ' 53 ' 22 32 71 Holidays 1 18 26 50 15 15 50 Holidays 2 22 31 50 29 34 40 60 the information. Also, the markedly higher recall percentages noted for the 55 dBHL condition as opposed to the 50 dBHL condition confirms that the increase in noise level was not suff icient to result in decreased performance on this measure of comprehension and suggests that other factors (such as level of interest or involvement) may have been at play in the half of the conversation for which scores were relat ively low. It is interesting to note that the distribution of the information recalled is relat ively random in terms of recency and primacy effects. It is also evident from these results that memory for larger information units was better than memory for deta i l . Thus, memory for subtopics was superior to memory for turns, which in turn was superior to memory for clauses. Conversational Fluency Rating As shown in Table 14,. both the conversation on Real Estate and the conversation on Holidays were rated as highly fluent overal l . In addition, both conversations were rated as high on specif ic aspects of conversation. Ratings did not drop as a function of the adversity of the l istening condition. Speci f ica l ly , ver i f icat ion was minimally, needed during either conversation. The coder also noted that in the f i r s t conversation, the experimenter appeared to talk more, whereas the reverse seemed true for the second conversation. 61 Table 14. Conversational fluency ratings of EL's conversations Rating Low (1) (2) (3) (42 High (5) Overall Fluency R.H1.H2 Factors Related to Fluency: Presupposition R.H1.H2 Receptive Ab i l i t ies R.H1.H2 Expressive Ab i l i t ies R.H1.H2 Motivation. Attention R.H1.H2 Turntaking R.H1.H2 Speci f ic i ty . Accuracy R.H1.H2 New vs. Old Information R.H1.H2 Non-verbal Communication Topic Maintenance R.H1.H2 Cooperation R.H1.H2 Time-Sharing R.H1.H2 Veri f icat ion R.H1.H2 Independent Repair R.H1.H2 Meta-Communication Note: Non-verbal communication and meta-communication could not be rated. R = Real Estate; HI = Holidays, part 1; H2 = Holidays, part 2 62 Discourse Analysis A l l conversations were divided into clauses, turns and subtopics. Each turn was then categorized according to whether or not i t conveyed content (see Appendix C). A summary of the information distr ibution in these conversations is presented in Table 15. For the conversation on Holidays, separate counts are given for the 50 and 55 dBHL noise conditions. KF and EL produced an equivalent number of turns for a l l conversations. E l , however, produced more clauses in the conversation on Real Estate (henceforth referred to as R), but not in the conversation on Holidays (henceforth referred to as HI and H2 for the 50 and 55 dBHL noise conditions, respectively). This difference is mainly accounted for by the high number of contentless clauses in EL's contributions, indicating that this is a prevalent characterist ic of his conversational sty le. The percentage of contentless clauses used by EL went up in HI but decreased again in H2, suggesting that the overall increase in contentless clauses in conversation H was not related to the increase in noise levels. Across both conversation, KF produced more content than EL did. Despite the more adverse l istening condition, this is more noticeable in conversation H, in which only 12$ of KF's turns were contentless; in contrast, she produced about three times as many contentless turns in conversation R. 63 Table 15. Content contributed by KF and EL in each conversation with EL Real Estate Holidays. Part 1 Holidays. Part 2 KF EL KF EL KF EL Number of Turns Total # Content!ess Turns X Contentless Turns Number of Clauses Total # Contentless Clauses X Contentless Clauses Topic Shifts Init iated Topics Spoken on with Content Information Distribution Content Clauses/Content Turns Content Turns/Subtopic Content CIauses/Subtopic Transmission rate Total Time (minutes) Clauses/Minute Content Clauses/Minute Turns/Minute Content Turns/Minute Subtopics/Minute 66 21 32 114 25 22 17 21 1.8 2.1 4.2 67 33 49 141 54 38 6 17 2.6 2.0 5.1 13:35 8.4 10.4 6.6 6.4 4.9 4.9 3.3 2.5 1.5 1.2 40 3 63 13 20 7 9 1.1 4.1 4.4 46 23 50 42 61 2 9 1.2 2.5 2.9 6:15 10.0 10.8 6.4 4.2 6.4 7.3 5.9 3.7 1.4 1.4 45 7 16 87 22 25 7 12 1.7 3.2 5.4 56 31 56 89 43 48 5 10 1.9 2.4 4.6 8:48 9.9 10.1 7.4 5.2 5.1 6.2 4.3 2.7 1.4 1.1 64 Conversations R, HI and H2 varied slightly in duration. An analysis of the information transmission rate per minute shows that while the information transmission values measured for EL remain constant or drop slightly (with the exception of the number of turns per minute), those measured for KF increase across the two conversations. As predicted by the results of the verbal protocol, KF also introduced the majority of the subtopics. This difference between conversations R and H may have been a function of EL's interest in the the conversational topic and his personal style rather than the increase in noise levels (since EL generally performed better in H2 than in HI). As shown in Table 16, the contentless clauses were further divided into the following subcategories: acknowledgements, expressions of uncertainty (e.g. "well," "I don't know" etc.), incomplete or interrupted clauses, repetitions of prior content, comments on hearing, comments on the experiment i tsel f , requests for verification or repair, and "asides" by one of the speakers (see also Appendix C for a definition of "contentless clause"). In both conversations, EL used more acknowledgements than KF, most of which fell into the category of stand-alone acknowledgements. In other words, they appeared as a turn in their own right and were not used as a transition to new content. Their total number for EL increased from R to H, whereas for KF, a drop was noted. KF used relatively few acknowledgements in both conversations. 65 Compared to EL, a smaller percentage of her acknowledgements were stand-alone acknowledgements. Acknowledgements account for by far the majority of contentless clauses in these conversations. The second most important subtype consists of interrupted and incomplete utterances, which appear at similar levels in both conversations Table 16. Frequency of occurrence of each subtype of contentless clause per speaker in conversations with EL Real Estate Holidays. Part 1 Holidays. Part 2 KF EL KF EL KF EL Acknowledgements Stand-alone 13 35 0 33 3 32 Turn-Init ial 7 11 5 8 7 8 Turn-Final 0 1 0 0 0 0 Well (I don't know) 0 1 0 0 0 0 Incomplete/Interrupted 7 9 4 1 4 6 Repeated Information 2 3 2 2 1 0 Comments on Hearing 0 . 0 0 0 0 0 Comments on Experiment 4 2 4 . 0 1 0 Repair Requests . 0 0 0 0 0 0 Minsunderstood Information 0 0 0 0 0 0 Asides 0 0 0 0 0 0 • 66 and for both speakers. Notably, no repair sequences, comments on poor hearing, or requests for ver i f icat ion were produced. Tables 17 and 18 outline direct evidence for or against comprehension of the content conveyed in the conversation. Table 17. Evidence of comprehension per speaker in conversations with EL Real Estate Holidays! Part 1 Holidays. Part 2 KF EL KF EL KF EL Shared Subtopics 17/23 9/9 10/12 Smooth Subtopic Shifts 10 5 3 2 3 4 Turns with Lexical Cohesion 4 9 1 2 0 4 Veri f icat ion Requested with 0 0 0 0 0 0 Overt Question and Comprehension Confirmed In EL's case, we can simply say that no direct evidence against comprehension was found in either conversation. Also, the number of smooth subtopic shifts and shared subtopics remained equivalent for both speakers in both conversations, indicating a similar level of comprehension. No difference in the use of lexical cohesion was noted in either conversation. 67 Table 18. Evidence of fa i led comprehension per speaker in conversations with EL Real Estate Holidays, Part 1 Holidays. Part 2 KF EL . KF EL KF EL Comments on Hearing 0 0 0 0 0 0 Misunderstanding Stated 0 0 0 0 0 0 Spontaneous Repair 0 0 0 0 0 0 Repair Requests 0 0 0 0 0 0 Veri f icat ion Requested with 0 0 0 0 0 0 Overt Question and Misunderstanding Confirmed Table 19 presents evidence that is ambiguous with respect to the subjects' comprehension of the content embedded in the conversation. Its face value with respect to comprehension is indicated but open to further interpretation. In most cases, the ambiguity arises from the partially social nature of conversations. While backchannel responses and acknowledgements, for instance, indicate the listener's intent to acknowledge the content of the speaker's contributions, they allow the listener to do so without making any overt statements with regards to that content. In this way, listeners can remain a part of the conversational interaction without contributing to i t or having processed'all of its contents. Also,- some of the features listed in Table 19, such as the use of connectives, "well,"and abrupt topic shifts may be - • 68 characterist ic of the speaker's style rather than evidence for or against comprehension. Overlaps usually occur because the second speaker was able to anticipate the transition-relevant point, but this need not be the case. Large gaps may be a sign of fa i led comprehension or may simply indicate that both speakers have temporarily run out of conversational material. Nonetheless, the conversation in more advantegeous l istening conditions provides a baseline on the basis of which any changes may be evaluated. As mentioned before, the frequent use of acknowledgements is an important feature of EL's conversational sty le, and need not be interpreted as evidence of comprehension d i f f icu l t ies. He also used (overall) a s l ight ly higher number of overlaps than KF for both conversations, but approximately the same number of interruptions. This is tentative evidence that comprehension was not significantly affected, which is confirmed by the results previously discussed. However, an important decrease in the number of connectives was evident for conversation H. EL's use of connectives in conversation R was similar to KF's but dropped by about 50$ in conversation H, while KF's remained constant. Connectives are less essential to the main meaning of discourse, in that they increase the level of cohesion but do not contribute signif icant content. Thus, we see that while EL's comprehension remained relat ively unaffected throughout, there may have been a cost associated with the increased effort 69 involved at higher noise levels which could have resulted in decreased attention to the subtler aspects of discourse production. Table 19. Ambiguous evidence of comprehension per speaker in conversations with EL Evidence Face Value* Real Estate Hoiidays. Part 1 Holidays. Part 2 KF EL KF EL KF EL Unshared Subtopics - 6/23 0/9 2/12 Abrupt Topic Shifts - 7 1 4 0 4 1 Topic Named + 3 2 9 2 0 2 Connectives + 30 24 23 9 25 13 Well - 0 1 0 0 0 0 Information + 1 3 1 2 2 0 Requests Backchannel + 7 8 2 1 3 7 Acknowledgements + 20 47 5 41 10 40 Overlaps + Content Turns 11 9 3 3 6 3 Backchannel 5 8 1 1 2 4 Acknowledgements 8 12 0 6 1 9 Interruption - 6 6 2 3 4 3 Gaps> 3 seconds - 0 0 0 0 0 0 * Note: "+" = evidence in favour of comprehension; " - " = evidence against comprehension 70 An additional indicator of comprehension is provided by subjects' responses to KF's intonational patterns (see Table 20). Speci f ical ly , in addition to marking information questions, KF used r is ing intonation at the end of a statement to e l i c i t confirmation and verify whether the content of the statement had been understood. In EL's case, responses fe l l mainly within the category of acknowledgements, and are therefore ambiguous with respect to comprehension. Table 20. Evidence of comprehension offered per speaker in reponse to prosodic cues in conversations with EL Real Estate Holidays, Part 1 Holidays. Part 2 KF EL KF EL KF EL Evidence For Response Confirms/Repeats 0 0 0 0 0 1 Other Appropriate Responses 5 2 7 3 4 3 Ambiguous Evidence Backchannel 0 0 0 0 0 0 Acknowledgements 0 3 0 9 1 10 Evidence Against Misunderstanding Stated 0 0 0 0 0 0 Repair Requests 0 0 0 0 0 0 No Response 2 1 0 1 1 3 71 Subject 2: ZW Verbal Protocol Like EL, ZW (the hard-of-hearing subject) reported trouble with the conversation in adverse S:N conditions. He thought that he had understood most of the conversation, but reported a much higher level of effort than EL (the normal-hearing senior) did. In addition to impaired hearing, he reported that the high incidence of medical jargon in the conversation contributed to this increase in effort. In his defini t ion of "effort ," he also included the use of repair strategies, indicating that he was very aware of those as interrupting the flow of the conversation. He also stated that in real l i f e , he might not have requested repair or c lar i f i ca t ion as frequently as he did during the experiment, depending on the type of conversation and the perceived importance of the information.. He fe l t that these requests were an imposition on conversational partners and might not be received very well i f they occurred too often. Like EL, he considered the noise to be very distract ing, although he considered the babble noise used in the experiment to be easier to di f ferent iate from the experimenter's voice than ordinary background noise might be. In addition, he noted that he was aware of noise in rea l - l i f e settings most of the time. He thought that the experimenter had done most of the talking and, l i ke EL, considered i t her role to introduce areas for 72 discussion and keep the conversation going (see Appendix B). Recall Tables F4-F6 in Appendix F summarise the analyses of ZW's free and recognition recall results. Recognition Recall On the conversation about Family Part ies, ZW scored 9/10. On the conversation about Babies, his total score was 7/10. Even though the noise level in this conversation was kept constant at 50 dBHL, a l l three errors occurred in the second half, suggesting that fatigue may have been a factor in the decreased performance. This conversation was over 20 minutes long and took place under S:N conditions which, given ZW's performance on the SPIN-R test, could be expected to create a signif icant disadvantage. A second pattern may account for ZW's decreased performance on the second conversation--all three errors pertained to information embedded in lenghtier turns (five successive content clauses or more) or in sequences of shorter turns by the c l in ic ian where ZW only contributed brief acknowledgements. A third factor may have been the nature of the information, which ZW identi f ied during the verbal protocol as rather technical and d i f f i cu l t to understand. 73 Free Recall Table 21 shows the amount of each conversation ZW recalled in terms of the number of content clauses, the number of content turns, and the number of subtopics contributed by each participant that were at least part ia l ly remembered. Overall, ZW showed poorer memory for his own content turns and clauses compared to KF's, possibly because he contributed fewer content turns and clauses. He recalled a far greater percentage of KF's subtopics than of those he introduced. As for EL, this may be a function of the fact that he in i t iated very few subtopics. Memory for subtopics was better than memory for turns. Overall memory for turns was superior to memory for clauses. In other words, the gist of the conversation was retained better than specif ic detai ls . Table 21. Percentage of content turns, content clauses and subtopics recalled by ZW for each participant and each conversation KF ZW Clauses Turns Subtopics Clauses Turns Subtopics Family parties 34 34 73 15 25 0 Babies 1 12 21 70 14 31 0 Babies 2 13 22 83 5 10 33 74 Fluency Ratings As shown in Table 22, the conversation on Family Parties was rated as highly fluent overall, whereas the second conversation on Babies was rated between midway and high on the fluency scale. Both conversations were rated as high on specific aspects of conversation. A mid-to-high score was given on two aspects of the f i rs t conversation--receptive abi1ti es and turntaking. In particular, the coder noted the presence of silent intervals, which she suspected ZW may have used to buy planning time. Ratings dropped as a function of the adversity of the listening condition on only two aspects; verification and repair were needed in the second conversation but not in the f i rs t . Interestingly, receptive abil i t ies were rated to be equivalent in both conversations. Discourse Analysis A summary of the information distribution in these conversations is presented in Table 23. For the conversation on Babies, a split-half rel iabi l i ty count was completed on salient aspects of the analysis, to determine possible effects associated with the greater length of this conversation. Results of these counts are included under a separate heading in Table 23. Specifically, aspects of the conversation that appeared to change as a function of the 75 Table 22. Conversational fluency ratings for conversations with ZW Rating Low (1) (2) (3) (4) High (5) Overall Fluency B F Factors Related to Fluency: Presupposition F.B Receptive Ab i l i t i es F.B Expressive Ab i l i t i es F.B Motivation. Attention F.B Turntaking F B Speci f ic i ty , Accuracy F.B New vs. Old Information F.B Non-verbal Communication Topic Maintenance F.B Cooperation F.B Time-Sharing F.B Veri f icat ion B Independent Repair B Meta-Communication Note: Nonverbal Communication and Meta-Communication could not be rated. No evidence of Verification/Independent Repair noted in F; F = Family Parties: B = Babies 76 Table 23. Content contributed by KF and ZW in each conversation with ZW Family Parties KF ZW Babies KF ZW Babies. Part 1/ 2 ZW Number of Turns Total # Contentless Turns X Contentless Turns Number of Clauses Total # Contentless Clauses % Contentless Clauses Topic Shifts Init iated Topics Spoken on with Content Information Distribution Content Clauses/Content Turns Content Turns/Subtopic Content Clauses/Subtopic Transmission rate Total Time (minutes) Clauses/Minute Content Clauses/Minute Turns/Minute Content Turns/Minute Subtopics/Minute 36 11 31 70 25 36 11 13 1.8 1.8 3.2 39 16 41 93 26 28 2 12 3 2 5.1 9:00 7.8 10.3 5 5.5 4 4.3 2.8 2.6 1.4 1.3 27 31 210 57 27 16 19 2.5 3.6 8.0 110 70 64 192 95 49 3 16 2.3 1.9 5.1 22:29 9.3 8.5 6.8 4.3 3.9 4.9 2.7 1.8 0.8 0.7 47/63 34/36 72/57 87/105 40/55 46/53 9:53/12:36 8.8/8,3 4.8/4.0 4.8/5.0 1.3/2.1 77 l istening condition, such as the number of acknowledgements and contentless clauses used by ZW, or measures of information transmission, were counted separately for both halves of the conversation. No such effects of length (or fatigue) were apparent, however. KF and ZW produced an equivalent number of turns for both conversations. ZW, however, produced more clauses in the conversation on Family Parties (henceforth referred to as F), but not in the conversation on Babies (henceforth referred to as B). In B, but not in F, this difference can be accounted for by the high number of contentless clauses in ZW's contributions. In other words, while the percentage of contentless clauses was similar for both speakers in F. In the higher level of background noise in B, ZW produced 75$ more contentless clauses. The number of contentless clauses noted for the experimenter, on the other hand, remained constant from conversation F to B. The same pattern holds for the distribution of contentless turns. While ZW was apparently affected by the noise level , as shown by an increase in contentless turns, KF was not. In addition, ZW produced more content clauses but fewer content turns-than KF in F, indicating his turns were longer than KF's. In B, he produced fewer content clauses and fewer content turns. In other words, the relative length of ZW's turns, decreased across the conversations. 78 Measures of information distribution show that while ZW produced more content per turn and per topic than KF in F, this relation was reversed in B. An analysis of the information transmission shows that while KF and ZW introduced content clauses and turns at similar rates in F, ZW produced fewer content clauses and turns per minute than the experimenter in B. As predicted by the results of the verbal protocol, KF introduced the majority of a l l subtopics for both conversations. Table 24 shows the frequency of occurrence of the different types of contentless clauses. Contrary to EL, a more drastic difference between both conversations was noted for ZW, who produced a total of 19 acknowledgements in F but 57 in B. In addition, whereas the majority of ZW's acknowledgements in F were of the connective, turn-init ial type, over 80% of those in B were stand-alone and did not contribute additional content. KF used approximately the same number of acknowledgements in both conversations, and none of them were stand-alone. The increased noise level in the second conversation resulted in an increased incidence of repeated information (presumably as an effort towards c la r i f i ca t ion) , part icularly by KF. Also, there were instances of misunderstood information, repair requests, and comments about hearing by ZW, which were absent in conversation F. Tables 25 and 26 show direct evidence of fa i led or successful comprehension. 79 As mentioned above, ZW produced a number of stated misunderstandings, comments on hearing, and repair- requests in B, but not in F. KF and ZW shared approximately the same proportion of subtopics in both conversations. No differences in the use of lexical cohesion were noted across the conversations. Table 24. Frequency of occurrence of each subtype of contentless clause per speaker in conversations with ZW Real Estate Babies KF ZW KF ZW Acknowledgements Stand-alone 0 0 0 46* Turn-Init ial 9 15 5 4 Turn-Final 9 4 10 7 Well (I don't know) 0 0 1 1 Incomplete/Interrupted 4 4 9 9 Repeated Information 1 2 19 9 Comments on Hearing 0 0 0 4 Comments on Experiment 2 1 12 7 Repair Requests 0 0 0 6 Misunderstood Information 0 0 0 1 Asides 0 0 1 3 * Note: 21 of these occurred in Part 1 of the conversation on Babies, occurred in Part 2. 80 Table 25. Evidence of comprehension per speaker in conversations with ZW Family Parties Babies KF ZW KF ZW Shared Subtopics 12/13 16/19 Smooth Subtopic Shifts 4/11 2/2 9/16 2/3 Turns with Lexical Cohesion 3 5 4 4 Verif icat ion Requested with Overt 0 0 0 0 Question and Comprehension Confi rmed Table 26. Evidence of fa i led comprehension per speaker in conversations with ZW Family Parties Babies KF ZW KF ZW Comments on Hearing 0 0 0 4 Misunderstanding Stated 0 0 0 1 Spontaneous Repair 0 0 0 0 Repair Requests 0 0 0 6 Veri f icat ion Requested with 0 0 0 0 Overt Question and Misunderstanding Confirmed 81 Table 27 presents evidence from discourse analysis that is ambiguous with respect to comprehension. As discussed previously, ZW produced an increased number of acknowledgements in conversation B, which could be interpreted as decreased involvement in the conversational process, l ike ly as a result of poor hearing and comprehension. As for EL, a 50$ decrease in the use of connectives was noted for ZW. KF, on the other hand, produced more connectives in B than in F. She also introduced the topic name over seven times as frequently as ZW did. There does not appear to be a signif icant difference between the number of overlaps and interruptions in i t ia ted by either speaker. Table 28 presents ZW's responses to KF's prosodic cues. In part icular, KF made frequent use of r ising intonation to e l i c i t confirmation or ver i f icat ion of understanding. She used this pattern markedly more in the conversations with ZW (a total of 102) than in those with EL (a total of 56), indicating a compensatory strategy on the experimenter's part to fac i l i t a te or verify comprehension for the hard-of-hearing participant. As a result, perhaps, ZW did not respond verbally to 33 of these requests for confirmation, which frequently occurred in midsentence and may have been disruptive to the flow of the conversation. A l l of the repair requests in conversation B occurred in response to prosodic cues. One possible implication of this observation is that in everyday contexts, with naive conversational partners, ZW may have been 82 less likely to init iate repair sequences (as confirmed by his comments on the verbal protocol). For both EL and ZW, the majority of responses fel l into the category of acknowledgements, and were therefore ambiguous with respect to comprehension. Table 27. Ambiguous evidence of comprehension per speaker in conversations with ZW Evidence Face Value* Family Parties Babies KF ZW KF ZW Unshared Subtopics - 1/13 3/19 Abrupt Topic Shifts - 6 0 7 1 Topic Named + 4 1 45 6 . Connectives + 10 14 25 9 Well - 0 0 1 1 Information Requests + 2 0 3 2 Backchannel + 10 1 8 8 Acknowledgements + 19 19 15 55 Overlaps + Content Turns 2 3 8 6 Backchannel 2 0 2 3 Acknowledgements 1 5 4 7 Interruption - 2 1 6 7 Gaps> 3 seconds - 0 0 0 0 *Note: "+" = evidence in favour of comprehension; " - " = evidence against comprehension 83 Table 28. Evidence of comprehension offered per speaker in response to prosodic cues in conversations with ZW Family Parties Babies KF EL KF EL Evidence For Response Confirms/Repeats 0 1 2 5 Other Appropriate Responses 1 3 . 7 7 Ambiguous Evidence Backchannel 0 0 0 2 Acknowledgements 0 3 0 30 Evidence Against Misunderstanding Stated 0 0 0 1 Repair Requests 0 0 0 6 No Response 0 1 0 33 Subject 3: GK Verbal Protocol GK, the subject discussed by Pichora-Fuller & Johnson (submitted), associated the hardest parts of the conversation in higher noise levels with perceptual d i f f icul t ies and the need for repetition. This conversation required much more effort than the conversation in more advantageous S:N ratio conditions, and she reported having understood only half of i ts contents. She also admitted that 84 in everyday l i f e she frequently pretends to understand rather than reveal that she has trouble hearing. Contrary to ZW, she stated having talked more than the experimenter, which is consistent with her observation that the hardest parts of the conversation were those where she was not the speaker. Recall Tables F7-F9 in Appendix F summarize the analyses of GK's free and recognition recall results. Recognition Recall GK's scores on both conversations were considerably lower than ZW's, a difference which may in part be explained by a higher degree of hearing loss, as evidenced by a higher pure tone average and SRT score. She scored 6/10 and 4/10 correct on the conversations in favourable and adverse l istening conditions, respectively. Errors in the latter condition were distributed equally across both halves of the conversation. Free Recall Table 29 shows the amount of each conversation GK recalled in terms of the number of content clauses, the number of content turns and the number of 85 subtopics contributed by each participant that were at least par t ia l ly remembered. In contrast to ZW and EL, GK showed consistently better recall for her own content turns and clauses compared to KF's. She also recalled more information from subtopics she herself introduced. Possibly, this pattern relates to the fact that GK took more control of the conversation and introduced more new content than either of the other subjects did. As was the case for both EL and ZW, memory for gist (subtopics and turns) was better than memory for detail (clauses). GK's free recall results were superior to ZW's. They also dropped much less dramatically as a function of the adversity of the l istening condition, even though an overall decrease in recall was evident as the noise level increased. Table 29. Percentage of content turns, content clauses and subtopics recalled by GK for each participant and each conversation KF GK Clauses Turns Subtopics Clauses Turns Subtopics 24 46 100 50 71 100 23 42 66 37 57 75 23 32 63 54 53 92 Cats Restaurants 1 Restaurants 2 86 Fluency Ratings As shown in Table 30, the conversation on Cats and the f i rs t part of the conversation on Restaurants were rated as highly fluent overall, whereas the second part of the conversation on Restaurants was rated as being only midway on the fluency scale. All three conversations were rated as high on specific aspects of conversation. Ratings dropped as a function of the adversity of the listening condition on three aspects only--receptive abi l i t ies, topic maintenance, and verification. Thus, these ratings are consistent with GK's claim on the verbal protocol that she often pretends to understand even though she doesn't. Correspondingly, few attempts at repair were made during the conversations with the experimenter. Discourse Analysis A summary of the information distribution in these conversations is presented in Table 31. Both participants .produced an equivalent number of turns per conversation. However, in all three conversations, GK produced more content turns and clauses than KF, although the difference is larger for the Cat conversation (henceforth referred to as C) than for either part of the Restaurant conversation (henceforth referred to as RI and R2). Although GK produced more content turns overall, a larger percentage of KF's turns conveyed 87 content. In other words, GK's content turns were on average longer than KF's Table 30. Conversational fluency ratings for conversations with GK Low (1) (2) Rating (3) (4) High (5) Overall Fluency Factors Related to Fluency: Presupposition Receptive Ab i l i t i es Expressive Ab i l i t i es Motivation. Attention Turntaking Speci f ic i ty . Accuracy New vs. Old Information Non-verbal Communication Topic Maintenance Cooperation Time-Sharing Veri f icat ion Independent Repair Meta-Communication R2 R2 R2 C.R1 C.R1.R2 R2 CR1 R l Rl C.R1.R2 C.R1.R2 C.R1.R2 R1.R2 C.R1.R2 C.R1.R2 C.R1.R2 C.R1.R2 Note: Nonverbal Communication and Meta-Communication could not be rated C = Cats: Rl = Restaurants, Part 1; R2 = Restaurants. Part 2 88 Table 31. Content contributed by KF and GK in each conversation with GK Cats KF GK Restaurants, Part 1 Restaurants. Part 2 KF GK KF GK Number of Turns Total 23 22 # Contentless Turns 1 5 X Contentless Turns 4 23 Number of Clauses Total 64 121 # Contentless Clauses 10 11 X Contentless Clauses 16 9 Topic Shifts Init iated 5 2 Topics Spoken on with Content 7 7 Information Distribution Content Clauses/Content Turns . 2.5 6.5 Content Turns/Subtopic 3.1 2.4 Content Clauses/Subtopic 7.7 15.7 Transmission rate Total Time (minutes) 9:11 Clauses/Minute 20.2 Content Clauses/Minute 17.9 Turns/Minute 4.9 Content Turns/Minute 4.3 Subtopics/Minute 8.0 24 5 21 46 11 24 6 9 1.8 2.1 3.9 22 8 36 63 12 19 4 9 3.6 1.6 5.7 5:4 21.5 17.0 9.1 6.5 2.0 60 22 37 100 31 31 8 14 1.8 2.7 4.9 59 29 48 127 37 29 12 18 3.0 1.6 5.0 10:41 21.3 14.9 11.1 6.4 1.9 89 For both KF and GK,' the proportion of turns that conveyed meaningful content decreased as the listening condition became more adverse. However, even in the least favourable listening condition, GK produced about 50% more content clauses per content turn than KF did. Measures of information distribution show that the number of unshared topics increased markedly as the noise level rose, with GK contributing on the majority of subtopics. While there was a slight increase in the number of subtopics initiated by KF as the listening conditions grew more adverse, the number of subtopics introduced by GK increased much more markedly, from 2 in C to 4 and 12 in Rl and R2, respectively. This shift is l ikely related to GK's increased difficulty in following the conversation as the competing noise became more obvious. An average of about 2 or 3 content turns are contributed per subtopic spoken on with content by each participant in all three of the conversations. The number of content clauses spoken on with content per subtopic was more variable. In C, GK contributed almost twice as many content clauses per subtopic as KF. This difference between GK and KF decreased as the noise level became more adverse, with both participants contributing about equally in R2. An analysis of the information transmission shows that the total number of clauses spoken per minute was constant across the three conversations. 90 However, since the proportion of clauses that conveyed content decreased as the adversity of the l istening condition increased, i t follows that the number of content clauses spoken per minute decreased likewise. The number of turns and subtopics per minute actually increased, but the number of content clauses per turn and subtopic decreased. In other words, turns became shorter as the noise level increased. Table 32 breaks down the different subtypes of contentless clauses. In the least adverse condition, both participants used about the same percentage of contentless clauses, but in the most adverse condition, GK used about 30$ more than did KF. Considering the types of contentless clauses, KF used more acknowledgements than GK in C, but the reverse is true for Rl and R2. In addition, whereas the majority of KF's acknowledgements were of the connective, tu rn - in i t ia l type, GK's were mostly stand-alone and did not contribute additional content. The increased noise level in the second conversation resulted in an increased incidence of repeated information by KF (presumably as an effort towards c la r i f i ca t ion) . Furthermore, there were instances of misunderstood information, repair requests and comments about hearing by GK, which were absent in conversation C. Tables 33 and 34 show direct evidence of adequate or fai led comprehension. GK produced a number of stated misunderstandings, comments' on hearing and 91 repair requests in C, although not in R. As mentioned above, the number of shared subtopics decreased from C to R. GK's use of lexical cohesion actually increased across the three conditions. However, while this shows that she understood the target word, i t does not imply fu l l understanding of the other speaker's intent. In other words, appropriating a content word from the other speaker's turn may in fact have helped GK conceal a lack of comprehension. Table 32. Frequency of occurrence of each subtype of contentless clause per speaker in conversations with GK Cats Restaurants. Part 1 Restaurants. Part 2 KF GK KF EL KF EL Acknowledgements Stand-alone 1 4 1 5 11 22 Turn-Init ial 7 0 4 3 8 5 Turn-Final 2 0 0 0 2 0 Well (I don't know) 0 3 0 0 1 1 Incomplete/Interrupted 0 1 0 0 1 0 Repeated Information 0 3 3 1 1 3 Comments on Hearing 0 0 0 .0 2 4 Repair Requests 0 0 0 3 0 2 Misunderstood Information 0 0 0 3 0 3 Table 33. Evidence of comprehension per speaker in conversations with GK 92 Cats Restaurants. Part 1 Restaurants. Part 2 KF GK KF GK KF GK Shared Subtopics 7/7 8/18 12/20 Smooth Subtopic Shifts . 5 1 6 4 6 10 Turns with Lexical Cohesion 0 3 1 3 1 7 Veri f icat ion Requested with . 0 , 0 0 0 0 0 Overt Question and Comprehension Confirmed Table 34. Evidence of fa i led comprehension per speaker in conversations with GK Cats Restaurants, Part 1 Restaurants. Part 2 KF GK KF GK KF GK Comments on Hearing 0 0 0 0 2 4 Misunderstanding Stated 0 0 0 0 0 3 Spontaneous Repair 0 0 0 0 0 0 Repair Requests 0 0 0 3 0 2 Veri f icat ion Requested with 0 0 0 0 0 2 Overt Question and Misunderstanding Confirmed 93 Table 35 presents evidence that is ambiguous with respect to comprehension. As discussed previously, GK produced an increased number of stand-alone acknowledgements in conversation R, which could be interpreted as an attempt at maintaining her involvement in the conversation despite poor hearing and comprehension. In addition, GK produced s l ight ly more abrupt topic shifts and shared fewer of KF's subtopics. Interestingly, KF, too, showed a decrease in shared topics, possibly related to GK's tendency to seek increased control of the conversation. There does not appear to be a significant difference between the number of overlaps and interruptions in i t ia ted by either speaker. Evidence observed in the conversations with GK indicated a sl ight ly different pattern of comprehension di f f icul ty than appeared to be the case for both other subjects. Overall, GK's use of connectives, expressions of uncertainty and requests for information did not di f fer from KF's. GK used an increasing number of acknowledgements and backchannel responses in the more adverse l is tening condition, indicating greater comprehension d i f f i cu l ty . This is consistent with GK's report in the verbal protocol that she often pretended to understand. In addition, a greater number of overlapping acknowledgements, interuptions, and slow turn in i t ia t ions suggest increased comprehension d i f f i cu l ty in the second conversation. 94 Table 35. Ambiguous evidence of comprehension per speaker in conversations with GK Evidence Face Value* Cats Restaurants. Part 1 Restaurants. Part 2 KF EL KF EL KF EL Unshared Subtopics - 0 0 1 1 6 2 Abrupt Topic Shifts - 0 1 0 0 2 2 Topic Named + 11 21 4 5 6 6 Connectives + 6 9 13 10 14 13 Well 0 3 0 0 1 1 Information Requests + 3 1 2 1 1 3 Backchannel + 8 0 5 6 11 3 Acknowledgements + 8 4 5 8 19 27 Over!aps + Content Turns 1 1 1 1 0 0 Backchannel 3 0 2 5 0 1 Acknowledgements 1 0 1 1 0 3 Interruption - 4 0 1 1 0 4 Gaps> 3 seconds -Turn Uptake 1 0 1 1 0 3 Turn Continued 1 0 1 1 1 1 * Note: "+" = Evidence in favour of comprehension; " - " = evidence against comprehension As mentioned above with regard to subjects' responses to intonational cues, KF used rising intonation to e l i c i t verification markedly more in the 95 conversations with ZW (a total of 102) than in those with EL (a total of 56). Interestingly, in the conversations with GK, she only used a total of 39 (see Table 36). Table 36 Evidence of comprehension offered per speaker in response to prosodic cues in conversations with GK Cats Restaurants. Part 1 Restaurants. Part 2 KF GK KF GK KF GK Evidence For Response Confirms/Repeats 0 0 0 1 0 1 Other Appropriate Responses 0 0 0 0 0 3 Ambiguous Evidence Backchannel 0 0 0 0 0 1 Acknowledgements 0 0 0 4 0 12 Evidence Against Misunderstanding Stated 0 0 0 0 0 2 Repair Requests 0 0 0 1 0 1 No Response 0 4 0 1 0 9 96 Possibly, GK's tendency to conceal her lack of conprehension led the experimenter to reduce the use of this conversational strategy. In contrast to ZW, not all of GK's requests for clarif ication were responses to prosodic cues. This is consistent with observations made earlier that she was a more assertive and dominant speaker. This does not necessarily imply that she would be more l ikely .to admit failed comprehension'. . ZW and GK made approximately the same number of overt statements about hearing and lack of comprehension. However, given GK's greater degree of hearing impairment, her comprehension was much more likely to be compromised and more opportunities for repair sequences were expected to arise. As was the case for both EL and ZW, the majority of GK's responses fal l into the category of acknowledgements and are therefore ambiguous with respect to comprehension. 97 CHAPTER 4 DISCUSSION Based on the results described in the previous chapter, a comparison of the three subjects and their respective conversational styles and strategies wil l be presented. In addition, some of the implications of these results for the practice of audiological rehabilitation will be discussed. Comparison of Results Verbal Protocol While all subjects reported greater diff iculty under adverse signal-to-noise ratio conditions, the noise level affected the hearing-impaired participants much more strongly, as evidenced by their higher effort ratings. Also, the hard-of-hearing subjects acknowledged a need to resort to compensatory strategies (repair requests or increased control of the conversation) under less favourable conditions, whereas the normal-hearing subject did not consider such strategies to be essential. From the verbal protocol, we may infer that while ZW did seek repair, he was also more likely to withdraw from the conversation as an active participant because of the level of effort and the risk of exposing hearing problems 98 involved in requests for c la r i f i ca t ion . Along the same l ines, both EL and ZW assigned a position of control to the experimenter, which was reflected as minimal in i t ia t ive in topic development on their part. GK, on the other hand, suggested that she was more l i ke ly to deal with perceptual d i f f i cu l t i es by increasing her own contributions to the conversation, thus reducing the need for repair and allowing her to remain an active participant in the conversation. Recall Recognition Recall For a l l three participants, performance on the recognition recall tasks was part ia l ly predictable from the immediate context in which the test items or "facts" were embedded. Specific factors could be identif ied that made the avai labi l i ty of information for recognition recall less l i ke ly . For instance, information might pertain to a factual detail (such as a date), be introduced during an abrupt topic shi f t , or require some degree of inferencing. Other errors related to information embedded in lenghtier turns (five successive content clauses or more) or in a sequences of shorter turns by the c l in ic ian during which the subject only contributed brief acknowledgements. Facts introduced in any of these contexts were less l i ke ly to be recalled, 99 The number of errors made on the multiple choice task appears direct ly related to participants' hearing status and abi l i ty to perceive the information presented. Interestingly, this is the one task on which the EL, the normal -hearing subject appears to be signif icant ly affected by the noise level , suggesting that the recognition recall test may be a sensitive indicator of perceptual d i f f i cu l ty . GK's scores on both conversations were considerably lower than ZW's, a difference which may be explained by a higher degree of hearing loss, as evidenced by a higher pure tone average and SRT. Free Recall A summary of the free recall results as the overall percentage of content clauses and content turns recalled by each subject for both speakers combined is presented in Table 37. EL's scores were relat ively low for the f i r s t half of his conversation in unfavourable listening conditions (50 dBHL noise level ) , compared to his f i r s t conversation. However, the markedly higher recall percentages noted for the 55 dBHL condition as opposed to the 50 dBHL condition suggests that the increase in noise level was not l i ke ly to be responsible for EL's decreased performance in the f i r s t half of the second conversation. Other factors, such as level of interest or involvement, may have been at play. 100 In contrast, the conversations with ZW showed a signif icant overall difference in recall between the different l istening conditions, l i ke ly indicating that the decreased signal-to-noise ratio adversely affected his ab i l i ty to process the information conveyed in the second conversation. The significance of this difference for ZW needs to be evaluated with some caution, as his second conversation was approximately twice as long as his f i r s t and therefore introduced much more information for encoding. By comparison, GK's overall performance on this task was generally higher and much more stable across l istening conditions than ZW's, although i t did drop s l ight ly , too. ZW and GK's relative performance on the free recall task is consistent with, and interpretable with respect to their relat ive auditory s k i l l s and working memory scores. Despite an overall decrease in performance, GK's free recall scores remain superior to ZW's under poor l istening conditions. One factor that may have led GK to report more information during free recall resides in task interpretation. ZW may have valued conciseness more than inclusion of de ta i l , while GK may have attempted to be as comprehensive as possible. To what extent this difference is a reflection of culturally endorsed gender roles that expect women to be more fu l ly descriptive in their language use (see Tannen, 1994) remains an open question. 101 Table 37. Percentage of content clauses and content turns recalled by each subject for each conversation. Conversation 1 Conversation 2, Part 1 Conversation 2, Part 2 Clauses Turns Clauses Turns Clauses Turns EL 20 31 17 22 25 33 ZW 22 29 13 24 9 17 GK 41 56 31 49 41 41 A second possible' explanation of these results relates to GK's superior working memory span, allowing her to process and assimilate more of the incoming information. GK's poorer auditory sk i l l s do not allow her to perceive as much of the conversation as ZW, resulting in poor memory for detail (which rel ies largely, on bottom-up processing), and low recognition recall scores. ZW, on the other hand, scored much lower than GK on the working memory span test , and correspondingly showed poorer encoding and integration of conversational material. However, his superior perceptual s k i l l s are reflected as higher scores on the recognition recall task. For a l l three subjects, memory for subtopics was superior to memory for content turns, which i t se l f was superior to memory for content clauses. In other words, despite individual differences in processing ab i l i t y , subjects 102 preserved a gross outline (or, in terms of Kintsch, 1991, a "situation model") of the conversation, but performance declined on the recall of more detailed information. This is not suprising in view of psycho!inguistic literature demonstrating that in the process of language comprehension, general meaning is retained, but form is quickly lost (Strunk Sachs, 1967). In contrast to GK, neither EL nor ZW showed superior recall for shared subtopics as opposed to unshared subtopics. This is l ikely to be a function of GK's tendency towards a high rate of subtopic shifts in the conversation with adverse noise conditions, used to establish control over the direction of the conversation. EL and ZW both introduce relatively few subtopics overall. In GK's case, then, shared subtopics would be those on which actual dialogue took place and which therefore involved a deeper level of processing. EL and ZW, on the other hand, may have had relatively good perceptual and processing abil i t ies even for unshared subtopics. A related difference between GK and the two other participants concerns the source of the information that was most frequently recalled. EL recalled an equivalent number of turns and clauses from both his own and KF's contributions and for both conversations, indicating that he was relatively unaffected by the noise levels on this task. ZW recalled fewer of his own content clauses and turns for both conversations. While overall performance dropped for the more 103 adverse listening condition, the proportion of clauses and turns recalled from KF's and ZW's contributions remained similar. GK showed similar performance for both conversations, with the exception of recall for her own content clauses, which dropped by about 50$. For both conversations, she showed markedly superior recall for her own content clauses and turns as opposed to KF's . This is not surprising, considering that GK contributed more content, part icular ly under adverse l istening conditions, than either of the other subjects. It is interesting that despite th is , GK's recall for her own content clauses decreased as dramatically as i t did. Her decreased performance on this aspect of the free recall task, however, supports the contention that GK's contributions became less contingent and meaningful to the previous speaker's turns as the l istening conditions grew more adverse. As a result, the process of information integration that constitutes comprehension and precedes effective encoding was unable to take place fu l l y . Fluency Ratings Overall fluency ratings were high for a l l three participants in the conversations with favourable l istening conditions. In the least favourable listening condition, scores ranged from 3 (mid) for GK to 4 for ZW and 5 (high) for EL. Thus, the overall scores appear to ref lect the relative differences 104 in subjects' perceptual ab i l i t ies in predictable ways. However, the overall rating is minimally informative in terms of subjects' specif ic conversational d i f f i cu l t ies and compensatory strategies. On specif ic aspects of conversational ab i l i t y , EL scored high for both conversations. ZW's scores were s l ight ly lower (4/5) on two aspects of the f i r s t conversation (receptive ab i l i t i es , turntaking), and on three aspects of the second conversation (receptive ab i l i t i es , ver i f icat ion, and independent repair). Thus, receptive ab i l i t ies were judged to be mildly impaired for both conversations. While GK's scores are generally high for the f i r s t conversation and the f i r s t half of the second conversation, they drop signi f icant ly for the most adverse listening condition. Verif ication and independent repair are the only aspects to be scored low to low-mid for a l l three conditions, indicating that GK was not judged to take principal resonsibi l i ty for c la r i f i ca t ion and resolution of conversational d i f f i cu l t i es . This is consistent with results from other measures, suggesting that GK frequently pretended to understand. In addition, GK's topic maintenance scores are lower than those of either of the other participants, and appear direct ly related to the adversity of the listening condition. This again confirms GK's use of topic shif ts to establish and maintain control over the conversation. 105 Discourse Analysis No s igni f icant differences in rate of information flow and information distr ibut ion were found for the normal-hearing participant. By contrast, conversations with the hard-of-hearing subjects under adverse l istening conditions are characterized by a decreased rate of information flow and an increase in contentless contributions. In addition, length of turn diminished as the noise level increased. It is interesting to note the results obtained for GK in this regard. While in her case, too, the rate of information flow and the percentage of contributions with content decreased in the less advantageous condition, i t did so for both speakers. More notably, GK introduced fewer subtopics than the experimenter in the f i r s t conversation, but 50% more than KF in the second. She also used fewer content turns than KF but more content clauses in both conversations. In other words, she used fewer, but longer turns than the experimenter. This marks an important difference in conversational style between the hard-of-hearing subjects. ZW resorted to contentless contributions to maintain his participation in the conversation. GK's contributions, on the other hand, are l i ke ly motivated by an attempt to hold the f loor. Correspondingly, the c l in i c ian 's active participation in the conversation and contribution of new content increased for the conversations with EL and ZW, but decreased in the conversations with GK. 106 While the conversations with EL showed variations in the number of acknowledgements that appeared unrelated to the level of.background noise, conversations with the hearing-impaired subjects showed a difference in both the number and type of acknowledgements used. That i s , a decrease in the use of connective and turn- in i t ia l acknowledgements and an increase in the number of stand-alone acknowledgements were observed for the conversations in less advantageous l istening conditions. Overt breakdowns and/or repairs were the clearest indicators of hearing di f f icul ty. Direct evidence of failed comprehension was noted for the hard-of-hearing subjects, . but was absent for the normal-hearing participant. Interestingly, comments on hearing and-repair requests only appeared in the noisier l istening conditions. Both ZW and GK used an equivalent number of these overt repair strategies (a total of 10 and 9, respectively). However, both the degree of GK's hearing loss and her performance on the recognition recal l task suggest that more than nine opportunities for repair were l i ke ly available and that a signif icant amount of specif ic information was not understood. However, KF decided to increase the noise level to 55 dBHL in the second conversation, despite these presumed comprehension d i f f i cu l t i es , suggesting that GK did in fact succeed in concealing her hearing problems. Even though ZW used approximately the same number of overt repairs, KF 107 perceived these to be indicators of significant perceptual diff icult ies and chose to keep the noise level in the second conversation constant, whereas for GK she did not. In other words, even though explicit repair requests and comments on hearing by the hearing-impaired speaker are the most unequivocal indicators of comprehension diff icult ies, they appear to be interpreted by the conversational partner as only part of a general profile and conversational style. GK presented herself as a confident, involved, and dominant speaker. In comparison to ZW, perhaps, her overt attempts at repairing breakdowns were of minor impact during the course of the conversation. In the conversations with EL and ZW, KF introduced more subtopics and therefore used a larger number of smooth and abrupt subtopic shifts. As a result, the significance of any difference in subtopic shifts is hard to evaluate for both EL and ZW, as neither assumed an active role in topic management at any point in the experiment. GK, on the other hand, showed a clear increase in smooth topic shifts as the noise level increased. Given what we know about her conversational strategies, this is unlikely to indicate superior comprehension in the second conversation. Partial understanding, such as comprehension of key words, may have been sufficient to keep the conversation flowing. As a result, perhaps, KF did not resort to prosodic cues 108 in order to e l i c i t veri f icat ion to nearly the same degree she did in the conversations with ZW. General Discussion Evaluation of Comprehension Measures Based on the results discussed above, we can return to the research questions outlined in Chapter 1. The f i r s t pertained to the effectiveness of the various measures used. In other words,' what measure or combination of measures provides the most accurate prof i le of hearing handicap? It is important to note that results obtained by means of the different measures applied in this study were in fact largely consistent. In other words, differences between the results of recall tasks, fluency ratings, verbal protocols and discourse analyses were mostly quantitative in nature. However, al l of the measures used are limited in scope. We therefore need to determine what information is obtained using different evaluation methods and, conversely, what information is not tapped into by specif ic measures. For al l subjects, the verbal protocol was a good indicator of the degree of effort involved in the different conversations. Also, i t provided information on participants' perspectives on their own conversational strategies. However, i t proved to be a less rel iable indicator of how much of the conversational 109 content was understood, particularly for ZW. This is not surprising, as subjects are not necessarily evaluating their own understanding of the conversation unless conditions are extremely poor, and may not always be aware when misunderstandings do occur. After a l l , phenomena such as digressions and topic shifts are part of everyday conversation and need not be the result of failed comprehension. More objective measures may be more appropriate, then, to effectively evaluate comprehension. Free recall and recognition results were shown to be complementary, and supply information on different aspects of comprehension. Specifically, the recognition recall task appeared to be particularly sensitive to deficits in bottom-up processing and was therefore a good indicator of the listener's receptive ski l ls. Despite the impaired auditory abilities seen in the hard-of-hearing subjects, however, their ability to extract the gist of the conversation was found to be relatively intact, and appeared to be related to working memory span. In other words, subjects were able to make use of top-down processes to construct a global representation of the essential information even when the incoming signal was impoverished. The free recall task provides valuable information in combination with discourse analysis, because it. has the potential of revealing areas of comprehension diff iculty that were concealed in the course of the conversation. 110 Both recall tasks appeared to be rel iable measures of the overall encoding of the conversation, in that no effect of primacy or recency was seen. Subjects' memory for parts of the conversation can therefore be interpreted as the result of superior processing and encoding per se. Factors other than comprehension, however, may have operated to affect the results of the recall tasks. One is the relative.length of the conversations. Speci f ica l ly , the second conversation with ZW, which took place under more adverse l istening conditions, was twice as long as the f i r s t one. This is l i ke ly to have affected the free recall task in part icular, on which ZW performed worse for part 2 of the second conversation, and worse overall than either of the other subjects. A second factor is the internal structure of the conversation. Some topics, such as Family Parties, lend themselves to a more scripted type of organisation, which is known to fac i l i t a te encoding and reca l l . The results of discourse analysis are less l ikely to be compromised by either length or internal structure of the conversation. Also, this measure provides more detailed information regarding specif ic aspects of comprehension. Speci f ica l ly , i t par t ia l ly accounts for the results of the memory tasks in terms of the organisation of the conversation. As discussed earl ier in this chapter, memory for discourse information can largely be predicted by the local I l l environment in which i t occurs. Information presented at an abrupt topic sh i f t , occurring when the subject was not actively participating or embedded in long experimenter turns, was less l i ke ly to be recalled than information on which the subject had requested c lar i f i ca t ion or contributed new content. Discourse analysis also allowed for a more detailed description of the differences between subjects' conversational ab i l i ty in terms of parameters of comprehension, information distr ibut ion, and rate of information flow than any of the other measures, and showed in what ways the experimenter contributed to the development of the conversation. Which of these comprehension measures, then, are l i ke ly to be viable tools for the evaluation of conversational ab i l i ty in c l in ica l practice? While discourse analysis is perhaps more informative than any of the other measures, i t is also far too time-intensive to be c l i n i ca l l y valuable. However, the results presented here suggest ways in which the analysis may be adapted for c l in ical purposes. First, many of the parameters included in the analysis did not appear to be sensitive indicators of the comprehension d i f f i cu l t i es , in that they showed no or minimal change across the different l istening conditions. It would be more effective, then, only to focus on those aspects of the analysis which appear to y ield signif icant results. These include the proportion of contentless contributions (particularly acknowledgements), the 112 number of statements about hearing and repair requests, and the number of smooth or abrupt subtopic shifts. Second, since i t appears that the introduction of "facts," or specific detail for recognition call is a good indicator of hearing ability, clinicians may be able to capitalise on this task by introducing these facts in contexts known to cause diff iculty to hard-of-hearing listeners. For instance, information may purposely be embedded for recognition recall at abrupt topic shifts or as part of a lengthier turn, both of which are conditions requiring more extensive processing on the part of the listener. In the f irst, the discontinuous nature of an abrupt topic shift may make i t more diff icult for the listener to predict the content of the facts on the basis of the preceding context, in other words, to supplement the degraded signal by means of top-down information. In the second condition, more information is presented all at once, so that i f the listener is experiencing hearing problems, fewer working memory resources may be available for semantic processing, and some information may be lost for lack of processing time. Hearing Loss and Effects of Noise Noisy listening conditions may cause difficulty for all listeners. The adverse effects of noise on seniors are well documented. Hearing-impaired seniors, in particular, frequently complain that they have trouble understanding spoken 113 language, especially when the l istening environment is less than ideal (for a review of speech perception in adverse l istening conditions, see Nabelek & Nabelek, 1994 and CHABA, 1988). The second question raised in Chapter 1 concerned the c l i n i ca l l y signif icant differences between the conversational ab i l i t y of normal-hearing and hard-of-hearing seniors. The results of this study clearly indicate that i t cannot be answered without reference to the third issue, namely, how the l istening conditions and signal-to-noise ratio affect the participants in the conversation. In other words, differences between the hard-of-hearing and normal hearing participants arise mainly in relation to variations in the signal-to-noise rat io. The redundancy in regular conversation appears to be such that comprehension d i f f i cu l t ies do not become apparent until greater demands are placed on participants' perceptual and cognitive resources. Despite individual differences, a l l subjects showed decreased production of new content and slower rate of information flow across the two conversations. Second, and only for the hearing-impaired participants, the less advantageous condition resulted in the appearance of overt evidence of fai led comprehension, such as misunderstandings, comments on hearing, and repair requests. Also, both hearing-impaired seniors showed a steady decline in recall performance as a result of increased noise levels, partly reflected as decreased recall 114 scores, and reported markedly higher effort ratings than the normal-hearing participant on the verbal protocol. One implication for c l in ica l practice is that because these differences do not become apparent until the level of the competing noise is suff ic ient ly challenging, assessment of conversational ab i l i t y needs to be based on a baseline indicating the c l ien t ' s performance in advantageous l istening conditions, to which his or her performance in adverse l istening conditions may be compared. Individual Differences Results of this study also point to the need for a second type of baseline information. It is apparent that some features of the conversations are characterist ic of the speaker rather than of impaired hearing. EL, for instance, used a great deal more acknowledgements and other contentless clauses than either of the other subjects, regardless of his hearing status or the experimental listening condition. Also, i t was discussed that some indicators of fai led comprehension, such as repair requests, need to be considered within the global context of the c l ient 's conversational style and strategies. Thus, the absolute frequency of any conversational feature is meaningless without regard for the speaker's individual prof i le. 115 Interactions between Conversational Participants Another implication for c l in ical practice relates to the interactions that take place between any two conversational partners. In the conversations with EL and ZW, the experimenter appeared unaffected by the l istening condition and frequently showed an increase in the production of new content. In the case of GK, who took an active role in a great part of the conversation, the experimenter showed an analogous decrease in contributions with content and rate of information flow. It is not entirely clear whether the f i r s t two subjects i n i t i a l l y took a more passive stance, to which KF reacted by taking increased responsibil i ty for the conversational flow, or whether the experimenter set a pattern by taking more control of the conversation from the onset. A second example of interactions between conversational partners is KF's use of prosodic cues to e l i c i t confirmation of understanding, which was most prominent when the subject in question showed overt signs of hearing di f f icul ty. This emphasises the point raised by Watzlawick et a l . (1967), that "the very elusiveness and f l ex i b i l i t y of th[e] system-environment or system-subsystem concept in no small way accounts for the power of systems theory in the study of l iv ing organisms" (p. 122): Communication needs to be defined as an open-ended, ongoing process of interrelations occurring over time. The results reported in this study-, therefore, may not be preserved with different 116 conversational partners. In particular, one wonders i f both male participants might have taken more in i t ia t ive and shown more active participation had their conversational partner been a fr iend, a spouse, or another male speaker. This question raises the issue of gender and i ts possible effects on the conversational interactions observed in the context of this study. The issue of gender differences in communication style has been given l i t t l e prior consideration in aural rehabil i tat ion l i terature, but may nonetheless be an important factor. Both male participants deferred to the experimenter by allowing her to lead the conversation. This appeared to. constitute cooperative behaviour on their part. GK, on the other hand, was apparently motivated by a desire to remain an active participant and join equally in the conversation. This dist inct ion between hierarchy and symmetry in subjects' respective conversational positions is equivalent to the male-female dichotomy discussed by Tannen (1994). While she cautions against hasty c lass i f icat ion of conversational patterns, which are often culturally determined and usually open to multiple interpretations, she describes an indirect conversational style based on sol idari ty that is commonly perceived to be typical of female speakers. Within this framework, women are said to "bond" in conversation, while men are more often thought to use conversation primarily as a forum for discussing ideas. GK's tendency to conceal her hearing d i f f i cu l t i es may then 117 be interpreted as an attempt to avoid disruptions to the conversational flow which might be perceived as intrusive. ZW, on the other hand, appeared to value information above social interaction and request clarif ication whenever i t was needed. It is apparent from the results of this study that, despite the quantitative measures i t has focused on, participants do not usually consider the transfer of information to be the sole purpose of conversation. Conversational partners invest heavily in maintaining their role in the social aspect of the interaction, even i f i t is at the expense of comprehension and content. As a result, the evidence obtained from the discourse analysis was largely ambiguous with respect to comprehension, despite individual differences in assertiveness and willingness to interrupt the conversational flow for the sake of clarif ication. Concluding Remarks As discussed above, the present study has several implications for cl inical practice. First, i t suggest some ways in which the comprehension measures used here may be adapted for assessment in clinical settings. That i s , the recognition recall task could make use of facts embedded at strategic points in the conversation, such as abrupt topic shifts. Also, the most relevant 118 aspects of the discourse analysis procedure may be selected to provide a prof i le of the c l ien t 's conversational behaviour. These may include the production of new content, the use of repair strategies, and subtopic management. Second, assessment needs to take place in at least two signal-to-noise ratio conditions. A description of the c l ient 's performance in favourable conditions is necessary to establish his or her individual conversational sty le. A second conversation in more adverse conditions seems required to challenge the c l ien t ' s perceptual and cognitive resources and reveal the comprehension di f f icul t ies that might occur in everyday conversations, which frequently take place under less than ideal l istening conditions. Third, cl ients are l ike ly to perform differently with different conversational partners. From a functional point of view, i t may be advantageous to include the c l ien t 's most common conversational partner in the assessment procedure. On the other hand, there is some value in conducting the conversations with an unfamiliar partner (such as a c l in ic ian) , since in their everyday l ives hard-of-hearing cl ients are l i ke ly to encounter persons with whom they do not share the common knowledge base and experiences that frequently fac i l i t a te conversation with friends, spouses, or colleagues. 119 The fact that a client's comprehension is l ikely to vary with the listening condition, the conversational topic, and the conversational partner, as well as the observation that for most participants, transfer of information does not constitute the sole purpose of conversation, leads to an important conclusion with respect to conversation-based therapy. We may assume that in most real-l i f e settings, comprehension is unlikely ever to be perfect. In fact, most normal-hearing listeners are bound to experience comprehension diffculties at least some of the time. Perfect comprehension may therefore not be a realistic goal for therapy, nor may i t be a desirable objective. Clients may choose to sacrifice some of the information value of conversation for the purpose of social interaction. It may be as important to help clients identify when to request clarif ication as i t is to help them identify how to employ repair strategies. Thus, the results of the present study support the concept of hearing handicap as a client-based notion, according to which therapy follows from the needs and problems identified by the hearing-impaired person, with the assistance of the cl inician. 120 REFERENCES Adams, Cynthia (1991). Qualitative age differences in memory for text: A l i f e -span developmental perspective.. Psychology and Aging, 6(3), 323-36. Bi lger, R.C., Neutzel, J .M. , & Rzeczowski, C. (1984). Standardisation of a test of speech perception in noise. Journal of Speech and Hearing Research, 27, 32-48. Brown, G., & Yule, G. (1983). Discourse Analysis. Cambridge: Cambridge University Press. Brown, John (1976). An analysis of recognition and recall and of problems in their comparison. In John Brown (Ed.), Recall and Recognition (pp. 75-102). London: John Wiley and Sons. Bubl i tz , William (1988). Supportive Fellow-Speakers and Cooperative Conversations. Amsterdam & Philadelphia: John Benjamins Publishing Company.-CHABA (Committee on Hearing , Bioacoustics, and Biomechanics). (1988). Speech understanding and aging. Journal of the Acoustical Society of America, 83, 859-95. Cohen, G i l l i an (1989). Memory in the Real World. London, UK and Hi l lsdale, USA: Lawrence Erlbaum Associates. Craik, Fergus I.M. (1983). On the transfer of information from temporary to permanent memory. Ph i l . Trans. R. Soc. Lond., 302, 341-59. Craik, Fergus I.M. (1986). A Functional Account of Age Differences in Memory, In F. Kl ix & H. Hagendorf ( E d s J , Human Memory and Cognitive Capabil i t ies. Mechanisms and Performances (pp. 409-22). s . l . : Elsevier Sciences Publishers. Craik, Fergus I.M., & Lockhart Robert S. (1972). Levels of Processing:' Framework for Memory Research. Journal of Verbal Learning and Verbal Behaviour, 11, 671-84. 121 Daneman, M. & Carpenter P. (1980). Individual differences in working memory and reading. Journal of Verbal Learning and Verbal Behaviour, 19, 450-466. Duncan, Starkey (1972). Some signals and rules for taking speaking turns in conversations. Journal of Personality and Social Psychology, 23(2), 283-92. Erber, Norman P. (1988). Communication therapy for hearing-impaired adults. Abbotsford, Austral ia: Clavis Publishing. Erber, Norman P. & Lind, Christopher (1994). Communication Therapy: Theory and Practice. In Jean-Pierre Gagne and- Nancy Tye-Murray (Eds.), Research in Audiological Rehabil itation: Curent Trends and Future Directions. Journal of the Academy of Rehabilitative Audiology Monograph, 17, 267-90. Erdman, Sue Ann (1994). Self-Assessment: From Research Focus to Research Tool. In Jean-Pierre Gagne and Nancy Tye-Murray, Research in Audiological Rehabilitation: Curent Trends and Future Directions. Journal of the Academy of Rehabilitative Audiology Monograph, 17, 67-92. Frederiksen, Carl H., Brackwell, Robert J . , Breuleux, Ala in, & Renaud, Andre. (1990). The Cognitive Representation and Processing of Discourse: Function and Dysfunction. In Yves Joanette-& Hiram H. Brownell, Discourse ab i l i ty and Brain Damage. Theoretical and Empirical Perspectives (pp. 69-110). New York: Springer Verlag. Giolas, Thomas G. (1990). The Measurement of Hearing Handicap Revisited: A 20-Year Perspective. Ear and Hearing, 11(5, Supplement), 2-5. Giolas, Thomas G., Owens, Elmer, Lamb, Stanford H., & Schubert, Earl D. (1979). Hearing Performance Inventory. Journal of Speech and Hearing Disorders, 44, 169-95. Hess, Thomas M. (1992). Adult Age Differences in Script Content and Structure. In Robin L. West & Jan D. Sinnott (Eds.), Everyday Memory and Aging. Current Research and Methodology (pp. 87-100). New York: Springer Verlag. Johnston, Irvine F., & Strickland, Lloyd H. (1985). Communication mode, affect and reca l l . Canadian Journal of Behavioural Science, 17(3), 226-31. 122 Just, M.A., &, Carpenter, P.A. (1992). A capacity theory of comprehension: Individual differences in working memory. Psychological Review, 99, 122-49. Kintsch, Walter (1981). Memory for Text. In August Flammer & Walter Kintsch, Discourse Processing. Advances in Psychology, 8, 186-296. Kintsch, Walter (1991). The role of knowledge in discourse comprehension: A construction-integration model. In Guy Denhiere & Jean-Pierre Rossi, Text and Text Processing. Advances in Psychology, 79, 107-54. L ine l l , Per (1995). Troubles with mutualities..towards a dialogical theory of misunderstanding and miscommunication. In Ivana Markova, Karl Graumann & Klaus Foppa, Mutualities in Dialogue (pp. 176-213). Cambridge: Cambridge University Press. Lockhart, Robert S. . . Craik, Fergus I.M., & Jacoby, Larry (1976). Depth of Processing, recognition and recal l . In John Brown (Ed.), Recall and Recognition (pp. 75-102). London: Johm Wiley and Sons. Long, Debra L. (1994). The Effects of Pragmatics and Discourse Style on Recognition Memory for Sentences. Discourse Processes, 17, 213-34. Mulrow, Cynthia D., Tuley, Michael R., & Aguilar, Christine (1990). Discriminating and Responsiveness Abi l i t ies of Two Hearing Handicap Scales. Ear and Hearing, 11(3), 176-80. Nabelek, Anna K., & Nabelek, Igor V. (1994). Room acoustics and speech perception. In Jack Katz (Ed.), Handbook of Cl inical Audiology (pp. 624-37). Baltimore: Williams & Wilkins. Neisser, U. (1982). Memory Observed: Remembering in Natural Contexts. San Fransisco: W.H. Freeman. Noble, William (1993). A Historical Perspective on Self-Assessment in Audiology. Seminars in Hearing, 14(4), 299-301. 123 Pichora-Fuller, M. Kathleen, .  & Johnson, Carolyn (submitted). Measuring Comprehension in a Hard-of-Hearing Person who Pretends to Understand Conversation. Part I: Recall . Unpublished manuscript, University of Br i t ish Columbia. Pichora-Fuller, M. Kathleen/ & Johnson, Carolyn (submitted). Measuring Comprehension in a Hard-of-Hearing Person who Pretends to Understand Conversation. Part II: Discourse Analysis. Unpublished manuscript, University of Br i t ish Columbia. Pichora-Fuller, M. Kathleen, & Kirson, Ruth S. (1994). How allocation of Cognitive Resources May Alter Handicap. JSLPA, 18(4), 223-34. Pichora-Fuller, M. Kathleen & Robertson, Laura F. (1994). Hard-of-Hearing Residents in a Home for the Aged. JSLPA, 18(4), 278-88. Pichora-Fuller, M. Kathleen, Schneider, Bruce, & Daneman, Meredith (1995). How young and old adults l is ten to and remember speech in noise. J . Acoust. Soc. Am. 97(1), 593-608. Raven, J .C. (1946). The Mi l l House Vocabulary Scale. London: Lewis. Renkema, Jan (1993). Discourse Studies. An Introductory Textbook. Amsterdam, Philadelphia: John Benjamins Publishinng Co. Strunk Sachs, Jacqueline (1967). Recognition memory for syntactic and semantic aspects of connected discourse. Perception and Psychophysics 2(9), 437-42. Tannen, Deborah (1994). Gender and Discourse. New York: Oxford University Press. Tye-Murray, Nancy (1994). Communication Strategies Training. In Jean-Pierre Gagne and Nancy Tye-Murray, Research in Audiological Rehabil itation: Curent Trends and Future Directions. Journal of the Academy of Rehabilitative Audiology Monograph, 17, 193-208. van Dijk, Teun A. , & Kintsch, Walter (1983). Strategies of Discourse Comprehension. New York: Academic Press. 124 Ventry, I.M. & B.E. Weinstein (1982). The Hearing Handicap Inventory for the Elderly. A new tool . Ear and Hearing, 3(3), 128-134 Wingfield, Arthur & Stine, Elizabeth A.L. (1992). Age differences in perceptual processing and memory for spoken language. In Robin L. West & Jan D. Sinnott (Eds.), Everyday Memory and Aging. Current Research and Methodology (pp. 101-123). New York: .Springer Verlag. World Health Organization (1980). International Classi f icat ion of Impairments, D isab i l i t ies , and Handicaps. (World Heal th Organization: Geneva) Yoder, Carol Y. & Jeffrey W. El i as (1992). A proposed role for affect in everyday memory. In Robin L. West & Jan D. Sinnott (Eds.), Everyday Memory and Aging. Current Research and Methodology (pp. 223-34). New York: Springer Verlag. 125. APPENDIX A Responses to the Hearing Handicap Inventory for the Elderly (HHI/E) Questions Responses EL ZW GK • 1. Does-a hearing problem cause, you to feel embarrassed when meeting new people? no no .' sometimes 2. Does a hearing problem cause you to feel frustrated when talking to members of your family? no yes no 3. Do you have d i f f i cu l ty hearing when someone speaks in a whisper? no yes yes 4. Do you feel handicapped by a hearing .problem? ' no sometimes -no 5. Does a hearing problem cause you d i f f i cu l t y when v is i t ing friends, relatives and neighbours? . • no no sometimes 6. Does a hearing problem cause you to attend religious services less often than you would 1 i ke? -no no no 7. Does a hearing problem cause you to have arguments with family members? .no yes no 8. Does a hearing problem cause you d i f f i cu l ty when l istening to TV or radio? no sometimes sometimes 9. Do you feel that any d i f f i cu l ty with your hearing l imits or hampers your personal or social l i fe? no no no 10. Does.a hearing problem cause you d i f f i cu l t y when in a restaurant with . relatives and friends? no sometimes sometimes 126 APPENDIX B Verbal Protocol A. Post-Conversation Verbal Protocol (Adapted from Erber, 1988) la . Overall, how much of the conversation do you think you understood? (all=10, none=0) lb . Overall, how much of the conversation do you think the experimenter understood? 2a. Overall, how effortful or t i r ing was the conversation? (extreme! y=10, not at all=0) 2b. Overall, how effortful or t i r ing was the conversation for the experimenter? 3. What are the advantages and disadvantages of discussing a familiar/unfamiliar topic? 4. Who talked more during the conversation? Why? . 5. Who asked more questions during the conversation? Why? 6. Who answered more questions during the conversation? Why? 7. What caused problems understanding during the conversation? 8. What parts of the conversation were the easiest? Why? 9. What parts of the conversation were the hardest? Why? 10. Was it-necessary to solve problems in understanding? What solutions were tried? How well did-they work? 11. What was the effect of the noise on you? Were you aware of i t? How distracting was i t? B. Real Estate (EL)' l a . 10 lb. 10 2a. 2 2b. 2 .3. Interest level is higher; also,, you should know something about i t 4. Alternating, 50/50, maybe because of interest or because you were drawing 127 me out 5. Probably you did, maybe to help me along and speed up the conversation 6. Maybe I did, because you asked more questions 7. No 8. No 9. No 10. Negligible; I was a l i t t l e aware of i t at f i r s t , but i t became background noise and was not distract ing; the recall task was harder than the actual conversation; I think I got most of i t , but not necessarily in the right order. C. Holidays (EL) l a . 9 2a. 5 3. No 4. 50/50 5. You, to keep the conversation going 7. Yes, towards the end, because of volume of noise 9. Possibly when talking about your parents' holidays, though maybe I wasn't paying as much attention then as I should have; I possibly missed.the multiple choice question on Eden II 10. No. I just kept going, maybe I should have asked you for repetit ion, sometimes, i f i t ' s a general conversation and you miss something, you just let i t go 11. i t was more d i f f icu l t to concentrate; I was very much aware of i t and would, not have l iked to continue with the conversation beyond 15 minutes; any conversation with an interesting person is fine for however long, but not i f you don't hear well and you have to keep asking for c la r i f i ca t ion . 128 D. Babies (ZW) l a . 9 2a. 9 (would not have requested as many repairs and tried as hard in real l i f e ) 3. It was easy-because i t ' s -interesting, everybody knows about babies. 4. You did. I'd have trouble leading. I wouldn't really know where to start . I responded rather than lead. 5. You did. i t was an opening to the areas that you were going to discuss. 6. Not sure. 7. No, other than not hearing something. I had to have you repeat some of the expressions/medical terms, always ask for repetit ion, though depends on the type of conversation. I'm afraid people might get i r r i ta ted i f I ask too frequently in real l i f e , would ask only i f I thought i t was important, causes strain at home. 8. I don't think so. 11. It was very distract ing. I was very aware of i t . I'm always aware of i t . i t ' s probably a way of l i f e now. this noise was louder and not as subtle as noise in real l i f e , but also easier to differentiate your voice from the noise. E. Cats (GK) l a . 10 lb. 10 2a. 2 2b. 1 3. Less stressful talking about something you know though you may not learn as much from i t . Another disadvantage is that you tend to have your own viewpoints and may 'override' the other person's viewpoint. 4. Me, since you asked the questions and I tend to give very lengthy replies 5. You did because you were the interviewer 6. I did since you asked the questions 7. I don't own a cat; I was depending on knowledge of my daughter's cats 8. Talking about affection and love for animals was easier since my knowledge about these areas is greater than my technical knowledge about raising cats 9. It was hard to talk about the animals suffering or being hurt-- i t was an 129 unpleasant topic 10. There weren't any problems 11. Didn't bother me; was hardly aware of i t , except when there was a pause in the conversation; then I could hear and was aware of the noise. F. Restaurants (GK) l a . 5 lb. 10 2a. 8 2b. 1 3. ' I don't think there were any particular advantages; just possibly re-enforcing my own feeling that I'd rather eat at home than go out and spend money that I can't afford 4. I did; I always talk a lot ! 5. Dr. Ful ler; that's her job 6. I did because that's what I'm here for 7. I had to ask her more than once to repeat what she said 8. When I was talking because I didn't have to l is ten when Dr. Fuller was talking, especially when there was overhead sound; probably understood more when talking about the baby; because I'm part icularly interested in babies because of new grandchild 9. Noise interfered with hearing and was also mentally distracting 10. I think I should be a l i t t l e b i t more aggressive about asking people to repeat--I tend to when people are talking, and what I consider mumbling, I pretend I understand and sometimes I don't; but there's nothing wrong with saying I'm sorry I didn't hear you; but I think we a l l tend to try and hide i t 11. It was annoying; frustration; you want to hear what people say, that's what the world is a l l about--communicating; i f you can't hear what people say i t ' s very frustrating and i t makes you inef f ic ient and ineffectual [ i f there was that amount of noise in the real world I'd] move to a quieter place until the noise passes; would not try to carry on a conversation in that amount of noise. 130 APPENDIX C Definition of Discourse Analysis Categories Clause: a unit of speech that selects independently for mood; that i s , i t can stand by i t se l f as a declarative, imperative, interrogative, or exclamatory structure. It is an independent clause consisting of a single independent clause, along with any subordinate clauses grammatically related to i t . Verbless utterances (e.g. "Yeah," "Oh") are counted as an independent clause i f they have a sentence intonation contour, count as one idea unit, and are not backchannel responses.5 Content clause: a clause contributing new information to the discourse topic. Contentless clause: a clause that does not contribute new information to the discourse topic; contentless clauses were excluded from some analyses because they did not contribute information that could be remembered and include the following categories: * Acknowledgements of previously conveyed content ("yeah", "that's right" etc.) * Statements of uncertainty ( "we l l . . . " , "I don't know" etc.) * Repeated statements of previously conveyed content * Information extraneous to the discourse topic (such as comments on hearing, repair sequences, comments on the logistics of the experiment and "asides" that are not directed at the other speaker) * Information that was misunderstood or not heard. Turn: a turn includes a l l the utterances from when a speaker began to talk following another speaker's turn or following a pause of at least three seconds after his or her own last utterance, when that utterance had utterance-final intonation and was not a f i l l e r or other indicator of intention to continue. Backchannel responses were not counted as turns. de f in i t ion adapted from Pichora-Fuller and Johnson (submitted), with permission 131 Content turn: a turn including at least on content clause. Contentless turn: a turn consisting solely of contentless clauses. Backchannel responses: are similar to acknowledgements but do not occur at transition-relevant points, and.therefore do not constitute a turn. Topic: the subject of the conversation as i t was previously determined by the c l in ic ian based on participants' fami l iar i ty and interest rankings. Subtopic: the. subject of immediate focus and concern; a change in subtopic occurs when the new information in a speaker's turn is discontinuous with the information in the prior speaker's turn but s t i l l relevant to the general topic of the conversation. Smooth subtopic shif t : a subtopic transit ion that is easy to follow because i t is closely related to previous content turns. -Abrupt topic sh i f t : results from a discontinuity in the local centre of focus or interest that may have been hard for the l istener to construct; the new subtopic is highly divergent so that i t may have been d i f f i cu l t for the new speaker to.judge i ts relevance to the interaction. Information Question: questions e l ic i t ing new information from the conversation partner; for the purposes of our analyses these do not include c lar i f i ca t ion questions or repair requests. Repair request (or c lar i f ica t ion question): requests for repetition or c lar i f i ca t ion phrased as a question. Comments about hearing: requests for repetition or .c lar i f icat ion indirect ly phrased as a statement about fai led comprehension (e.g. "I did not get that"). Spontaneous repair: the current speaker provides c la r i f i ca t ion or repetition of her own content in the absence of an exp l ic i t request. 132 Request for ver i f icat ion: the c l in ic ian attempts to determine whether the conversational partner has understood. Request for acknowledgement: c l in ic ian attempts to e l i c i t c la r i f i ca t ion by means of intonational cues (r is ing intonation). Lexical cohesion: speaker 2 repeats a content word in his turn.that occurred in the previous turn by speaker 1; this excludes pronouns, statements of the conversational topic, and replies to questions. Connective: a conjunction or discourse part icle l inking speaker turns (e.g "and," "but," or a phrase such as "I don't know, but . . . " ) . Overlap: occurs when two speakers speak at the same time. Interruption: an overlap that does not occur at transition-relevant points, and is experienced as disruptive in nature. Gap: a pause of 3 or more seconds between turns; uptake of the new turn may be in i t iated by either speaker. 133 •APPENDIX D Lists of Facts A. Real Estate (EL) 1. Our house in Mississauga was a link-house in a new development 2. I'd rather buy.when interest rates are moderately high 3. moving and real estate fees are costly, but at least there are tax deductions 4. I knew nothing about sumps until I bought a house in Vancouver 5. My present house is 50 years old I miss the modern features of our old house, l i ke central vacuuming and air conditioning -7. I wouldn't want a large house because I would accumulate too much stuff 8. An important feature for me is a f ireplace 9. There are no land tax breaks for seniors in Ontario l i ke in BC 10. The idea of land-tax based on market value caused great controversy in Toronto. B. Holidays (EL) 1. At University of Toronto, the second week of February is the "reading week" holiday 2. Back East, Halloween is not celebrated with fireworks 3.. Remembrance Day is not a holiday in Ontario 4. At the hospital where I worked we could take our birthday and anniversary holidays 5. Workers are legal ly enti t led to 4$ vacation pay, about 2 weeks 6. This summer our family, including the 2-year old, w i l l go' to the cottage for a week 7. In 1985, we went to an a l l - inc lus ive resort called Eden II in Jamaica for a week 8. Ever since about 1957, my parents have enjoyed holidays at Cape Cod 9. Travelling across the country, we stayed at a beautiful camping ground with a waterfall on the north shore of Lake Superior 10. Trips to the beach are a problem because my husband sunburns easi ly. 134 C. Family Parties (ZW) 1. After a big family dinner, my father always washes the dishes in a big pot 2. My mother's mincemeat tarts are a favourite at Christmas 3. At Thanksgiving, we al l usually go to my s is te r ' s place in the country 4. As children we opened presents at my grandmother's on Christmas Eve 5. I ins is t in invi t ing my in-laws to our place because my mother-in-law is a terr ib le cook 6. Our family has never bought expensive gi f ts 7. At Christmas, my Australian friend was surprised to see houses decorated with l ights 8. My contribution to large family dinners is usually home-made di11ed carrots and pickles 9. Sometimes we s t i l l get out the home movies that were taken when we were children 10. Nothing bothers me more than having the television on during family parties D. Babies (ZW) 1. The time i t takes for a baby to regain i t s birth weight is a key indicator of i t s early health 2. Jaundice, a minor problem in over half of newborns, is due to high levels of b i l i rubin 3. Babies born as early as 22 weeks gestation age now survive 4. A birth is legal i f the baby is born 28 or more weeks old gestation age 5. A blood test can indicate with certainty that a woman is pregnant even before a menstrual period is missed 6. Hearing can be tested within 48 hours of birth but is-usual ly not tested unti l a baby is about 3 months old 7. Rubella during pregnancy puts a baby at r isk for hearing loss 8. older-mothers may account for the birth rate being higher now than ten years ago 9. Babies weighing more than 10 lbs. are common i f maternal diabetes is uncontrolled 10. Flying in the third trimester is not advised because of the possible need for NICU care. 135 E. Cats (GK) 1. My third-cat was inherited from a friend who moved to Japan 2. One of my cats pulls T-shirts out of the dresser when she is mad •3. My husband says he's afraid i f the cats went outside they'd come home f l a t 4. One of my cats eats prescription food to prevent bladder stones 5. When the new cat arrived, the old cats cornered him in the hall closet for months 6. My husband has a stone cat in his off ice because real cats would get fur in his computer 7. On cat rubs the side of my computer when I am working and now.it 's covered with cat grease 8. The cats have favourite sleeping spots: the l iv ing room armchair, on top of :a stack of papers on my-desk, and the mat in the bathroom 9. The torn cat has a habit of drinking water from the kitchen top • 10. A major act iv i ty for my baby is chasing the cats.' F. Restaurants (GK) 1. By ordering unfamiliar dishes I get ideas for new things to cook at home 2. Ever since my baby was born I haven't eaten out much 3. My Australian friend says that people there eat out much less than here 4. My husband is a big pizza fan but he also enjoys Japanese food 5. My favourite restaurant back east was an Ital ian one called II Molinaro 5. Soft chairs and quiet are important features in choosing a restaurant after a long hard working day 7. I never order dishes that were specialt ies of my grandmother, l i ke beet soup 8. In Montreal, the cost of meals is reduced because people bring their own wine 9. Once in a blue moon I get a Big Mac attack 10. Back east i t is popular to go for Sunday Brunch at inns in the country. APPENDIX E Recognition Recall Test A. Real Estate (EL) 1. Our place in Mississauga was a link-house a townhouse a condominium a bungalow a semi-detached house 2. I'd prefer to buy real estate when interest rates are moderately low interest rates are moderately high interest rates are changing quickly inf lat ion rates are moderately high inf lat ion rates are moderately low 3. The.cost of moving depends mostly on the distance travelled-depends mostly on the time taken does not depend on the value of the goods moved is lower in the summer is tax deductible 4. Since moving to Vancouver I have learned about gardening pi umbi ng sai l ing painting carpentry 5. Our present house was bui l t during WW II before the depression in the late 40's in the 50's after most other houses in the area 6. I wish we had central vacuuming an attached garage a garbage disposal sky 1ights a f ireplace 7. I prefer a small house because • i t ' s easier to resell i t ' s easier to clean i t ' s cheaper to maintain i t ' s easier to find things in i t i t ' s cosier 8. For me, an important sel l ing feature of a house is a f ireplace a finished basement a view gas heating proximity to transportation 9. There are more breaks in BC than in Ontario for landowners who convert to natural gas heating are Canadians are seniors are f i rst- t ime home buyers are low-income 10. There was great controversy in Toronto over foreign land ownership market value taxation purchase value taxation educational tax rates subsidised housing projects . B. Holidays (EL) 1. At University of Toronto, "reading week" holiday the second week of March the second week of February the last week of February the last week of March the week of Easter 2. Back East, Halloween celebrations do NOT include costumes children going door-to-door fireworks candy apples parties 3. In Ontario, there is NO holiday on Remembrance Day Boxing Day Good Friday Canada Day Civic Holiday 4. Where I worked I had holidays on a f loat day on my birthday on the Queen's birthday on Jewish holidays on the hospital 's anniversary day 5. Workers are legally enti t led to 5% vacation pay 4$ vacation pay 6% vacation pay 12 vacation days 15 vacation days 6. This summer we plan to go to the cottage for a week go to the cottage for two weeks go to a lodge for a week go to a lodge for two weeks go to Jamaica 7. We went to Eden II last year 10 years ago in 1983. in 1985 . on our honeymoon 8. For over thir ty years, my parents have enjoyed holidays at the cottage in Ontario in Florida at Hilton Head at Cape Cod 9. we stayed at a beautiful camping ground with a waterfall in Northern Ontario near Niagara Fal ls in the Rockies in Lake Huron in Quebec -10. My husband doesn't enjoy fishing skiing golfing gardening sunbathing C. Family Parties (ZW) 1. After a big family dinner, my father turns on the TV turns on the coffee pot turns on the dishwasher washes the dishes washes the pots 2. Our Christmas favourite is home-made f ru i t cake plum pudding my mother's stuffing mincemeat tarts apple pie 3. My s is ter 's place is • on a lake in the country near my parents' home where we go for summer BBQ's very cozy for family parties 4. On Christmas Eve . we always went to my grandmother's my grandmother came to our house we wrapped presents we opened one present we went to church at midnight 5. I invite my in-laws over because their house is too hot my cooking is better i t ' s easier with the baby my husband insists i t saves them a lot of work 141 6. Our family always gives gi f ts that are small expensive personal practical . funny 7. At Christmas, my friend from Australia was surprised to see real Christmas trees l ights on houses Santa Claus in stores . no snow on the ground mandarin oranges 8. For large family meals, I usually make home-made pie home-made bread pickled carrots j e l l i ed carrots a pickle and cheese tray 9. Occasionally at family parties we play cards play scrabble play computer games watch home movies watch special TV shows 10. The most annoying thing at family parties is when some people arrive late watch TV drink too much smoke at the dinner table leave early 142 D. Babies (ZW) 1. A key indicator of a baby's early health is amount of jaundice birth weight time to regain birth weight time to double birth weight the strength to i ts f i r s t cry 2. Jaundice is caused by high levels of b i l i rub in low levels of b i l i rubin nutrit ional deficiencies poor long function poor heart function 3. Doctors consider premature babies may survive i f they are born as early as 32 weeks gestation 22 weeks gestation 18 weeks gestation 12 weeks gestation 28 weeks gestation 4. A birth is legal when the baby is born 28 or more weeks old gestation- age born al ive registered with the province born at any age cer t i f ied by a doctor 5. A blood test for pregnancy is definite at any time after 6 weeks gestation only when a urine test is definite only for healthy women only for f i rst- t ime mothers 143 6. Today children are at risk for hearing loss usually have a hearing test before going home from the hospital at three months of age at six months of age at one year of age at two years of age 7. A risk factor for hearing loss in infants is rubella during pregnancy Rh-factor incompatibility '• ' . any degree of jaundice diabetes during pregnancy high blood pressure during pregnancy , 8. the birth rate is higher now than ten years ago lower now than ten years ago the same as ten years ago in young mothers the same as ten years ago in teenage mothers-lower now than ten years ago in older mothers 9. uncontrolled maternal diabetes can result in high blood pressure large babies small babies diabetic babies premature babies 10. Flying is not advised during pregnancy because the possible need for intensive care i t may cause a miscarriage pressure changes may damage the baby health insurance is not available for pregnant.women the baby may have to be delivered on the plane E. Cats (GK) •1...Where does the previous owner of my cat l ive? Toronto Thailand Japan Holland Jasper 2.. What does one cat do when she is mad? unravel to i le t paper scratch the back door pull cushions off the couch pull T-shirts out of the dresser knock over ornaments 3. Why doesn't my husband want the cats to go outside? they might come home late they might come home pregnant' they might come home injured they might come home dirty they might come home mad 4. Why does one cat eat special food? weight control bladder stone prevention l iver disorder treatment diabetes control food allergy 5. Where was the new cat cornered? hall closet basement bedroom closet under the stairs furnace room 6. What kind of cat is in my husband's off ice stuffed cat cat poster cat statue plast ic cat cat photo 7. What is on the side of my computer? cat fur cat footprints cat noseprints cat grease cat drool 8. Where do my cats not sleep? bathroom l iv ing room kitchen on my desk in armchairs 9. Where does the new cat drink? bathtub to i le t kitchen sink laundry room sink from a teacup 10. What does the baby l ike to do? pet the cats chase the cats pull the ta i l s of the cats scare the cats pull the whiskers of the cats F. Restaurants (GK) 1. I l i ke ordering new dishes because . I get bored with the same thing al l the time i t ' s a better test of the cook I get new recipe ideas I feel I am getting more for the money what was good once is often disappointing the second time 2. I don't eat out much now because i t ' s too expensive I don't have time having a baby makes i t too d i f f i cu l t I'm new to Vancouver I got t i red of restaurant food 3. My friend told me that people eat out less often in Australia Japan Italy back east Montreal 4. My husband's favourite food is Chinese food pi zza • steak Big Macs pasta 5. The name of my favourite restaurant is Angelo's the Mermaid II Molinam the Wind Surfer the Country Inn 6. If I'm t i red, I ' l l choose a restaurant because the chairs are soft the music is soft service is fast i t ' s close to home the l ighting is good 7. I never order specialt ies of my mother's cabbage ro l ls beet soup pickled herring hamburger 8. One way of making eating out cheaper is to bring your own alcohol to avoid drinking alcohol to bring your own dessert to avoid eating dessert to t ip only i f the service is good 9. Sometimes I get a craving for lobster a Big Mac Kentucky Fried Chicken l iver and onions Chateau Bri and 10. Sunday brunch is popular in large hotels at country inns on the highway in shopping malls at country churches 148 APPENDIX F Results of Free and Recognition Recall Tests Note: In a l l of the following tables: "+C" = Clauses and turns with content " -C" = Contentless clauses and turns 149 CD •— CD ro s- c j Ll_ CD 00 CD C_3 O i — C_) C J <-J + CO CD CD CD CD • CM CD CD CD CD CD CM LO ' CO i—l CO CD -^T LO CD r-- CD ^ - LO i—I CO CO CD " 5 f CD LO CD CD •— CD ra t- o Ll_ CD a; +-> cz CD +-> c o C _ J Z3 + C\] CD CD CD CD CD CD CD CD CD ^1" CO i—I LO CD -3" CD CO CD i — l CD CD CD CO i—l CO CD CD i — CD CD S- CJ Ll_ - CD +-> OO CZ CD CD oo 4-> Z3 C CO O i — C_) C_J C_) + CNJ CD CD CNJ CD LO CD i — I i—I CD C\J CO CO OO CD OO ^3" CD . i—I CD CD CD CD CD. CD CO CNJ CD r-~ CD i— CD ca s- o U _ CD CZ s_ =5 O c_> C_) + CN] CD CD CNJ CO CD CO i — l CD i — l CD CD CD CD •—I ^t" C D CNJ CD CD . CD CO CO LO CD £ ;z £ £ '= S * 3 S1 I i: I 8 S < 00 OO CD CD s_ +J cz Z3 CD ra ra 3 CD +J +-> !_ cn o CJ ra cz cz ra =3 c CD o o CJ +-> +-> CD ro ra cz t - -1— •1— • 1— L O OO cn oo oo cz o +-> +-> o . CD CD ra 00 > CJ CD -1— cn cz cz CJ o t_ cn • i— CD +-> +-> cz f_ CD CD - M i — CD +-> 00 +-> X3 cz ra •1— CD CZ > Q -ja ra +-> !_ 00 c CD • 1— CJ • a •i— CD OO Z3 CD cz o • 1— CD ra o o ro !_ cz CO cc 1—t 2 : s: czr 2 : 1 s: a: CL. •—1 o O 150 CD ro i~ CJ u_ cu OO O C_> CNJ CD CNJ LO CD CD CNJ CO cu LO CD i—l LO CNJ- CNJ CO ro + o o cu CU r -cu in c_> CNJ CD CD CNJ CNJ CD CD co + co CU i — CU ro cu o <_> C_) CD CD CD CD i—I CD CNJ CO rrj C_3 + CD CNJ LT> CNJ OO I— CU •— CU ca S- CJ Ll_ CU c : CO cu 4-> S_ =3 o 1— c_> CNJ CD CD CD CD CD CT CD CD CNJ "3" CNJ C_> + CO rc U CU O > , £ • - £ QJ CU C_> >^  >> >> cu cu s_ q-=3 TO' Q -o Z3 oo c O +J (_) =3 •a cu •a x ra cu X re +-> T3 O C CU C/l cu e o cu X ro ro > CU oo c o CJ T J CU T J X ra cu cj C L cu ra cu <c I— o cu T J > cu T J c ro CU T J cu +-> o CU r— CU ro 5- O Ll_ CU ro CU r— CU ro t- <_) U_ CU ro + SZ oo cu s= +-> S-c: a O r-+ CU r— CU ro t- CJ u _ cu cr: ro cu o CJ> + CU r— CU ro C- CJ Li_ CU c: =5 CJ) + >> - a •7\ cu o CJ cu cu 3 a . o +-> - Q Z3 OO o +-> ro CNJ C D CNJ C D C D C D C D C D C D CNJ L O C O C O C D CNJ C D C D C D C D C D C D C D CNJ CNJ CNJ C D C O r-H CNJ C D C D C D C D C O CNJ CNJ C D r—I 1 CM C D C D C D C D >5 oo ra >~> >> CU ra ro s_ Q TD T J .O •i— >-> CU i — ra cu CJ O Q. ^ <f- c_ c OO sz cu O ro cu cu > s_ OO - c_ 1— O > 3 13 ' ja EE o ro -i— +-> O e O 3 • i— +-> o 1 c n h - ai CJ cu +-> CU CJ ra CJ < c <f- c : c oo s_ cu CJ sz • t— o o cu ro cu 13 •1— a . ra. ra o 1 Z3 1 > a: C_> u_ o o >• • > 1— 151 L O c o CM C D CM CT> -=3" cn co cu .— Ol CD S- CJ Ll_ Ol D i CD C_> Ol i — CD CD t_ CJ Li_ CD CD r— CD CD S- CJ Ll_ CD 00 CD CZ C_ OO CD C co O i — C J c_> C_) + OO c o C_) C_) + CD O i — C_) C_) C_> Ll_ CD t— CD CD C_ CJ C J CD C J + cz ra c j CD 4—> £ CD LO CD CD CD -^j" CD C D . CD CD .—I "=1-CD CNJ CD CD C D . ^ d " CD CD CD ^ j - i — i c o c D ^ j - r - ^ ^ r c o CD t—I CO CO LO CD C\J CD CD CD OO LO CO CD i—l CM CNJ OO CNJ CD CD LO i — ' CD i—I CD co co CNJ CNJ CNJ co CM co CO LO co o Z3 CO c_ o 00 ra +-> CD T3 cz CJ ra CD CO c E 1— ra i — > i O - c_> cz CD o 00 +-> ra > 1— CD >> CD cz cz (— ra Q . +-> +-> •1— CD •J— • o !_ E •i— ra •i— +-> CZ +-> CD £_ o i — •1— CD 1— Z3 ra i Z3 • 1— •i— > t - CD o CZ "O 1 <*- O ra x a • p— •,— > <4- • ,— ra <c i — > E ro +J ro 3 ra cz 1— CD CD ro CJ CD !_ CD c_ ra o CO LL. T CO <c ai z : 1— C J (— 153 CU I — CD ro !_ CJ Ll_ CU or ro C J sz cu CU LO +-> ZS 5Z ro O r — C J C_) C J + r—l C D C O C D C O CNJ C Q co r--C D C D CNJ C D r-H C O i-» co co CNJ C O C O cu CU r— CU ro u_ Oi fZ r y CJ ^ C_> C_) + C D C D C D CNJ C D CNJ r-H C D C O C O C O CNJ CNJ C D C D C D C D " 3 -CNJ CU i — CU ro S- CJ u_ cu =3 ro CU Z5 O r — C J C J C J C J + r-H C D CNJ C O C D CNJ -^ r un CNJ r-H CNJ C D CNJ -=3" CNJ r-H oo co j^- co C D CNJ C D C D C O C O C O CNJ CU r— CU ro S- CJ Ll_ CU O C J C J + CNJ C D C O L O r-H r-H C D C D C D C D r-H C D C D CNJ C D r-H C O CNJ L O C D CNJ r--CNJ E ro I a CJ CJ cu -o 'Zi cu s_ cu CJ CU !_ !_ o C J LO CD C O or CD •(— 4-> - s_ ZS. CD D _ >> L l _ "1 E O) C O 1— L l _ -Q r o £Z 1— o cx o JO 3 CO LO LO LO "a CU cu o • 1 — •p— o LO +-> p— LO M - CU LO c_ • 1 — LO c -CI cu ro LO c ro cu LO LO o • 1— Q - ra ro o •(— +-> • P— CT1 rz • 1— > s— - 1 — ! — "a fZ o CJ LO o >i 4-> +-> ro !_ • i— • 1 — ra • 1 — e LO cu • 1 — !_ ZS c n !_ 4-> LO .a > i • 1 — • (— cn "D +-> O CU ra +-> 3 cu cu <c !_ !_ ra LO > • i— +-! CJ CU ra 1 — £ i — ra SZ ro !_ Z3 ro o ro O • 1 — c cu o o Ll_ C J 1 1— < Li_ C J • 1 Q 1— i n 1— o 1— 1 r a C J c u al c z o •I— +-> •1— c z c n o u cu i—1 ai + J 00 C_ - r a zs. Q _ r - j oo L O CU L l _ • i — _ Q a> r a (— ca - O r a s z r— o CD .— CD "3 C_ CJ Ll_ CD ro C J ro C_> C J + CD CD i — CD ra C_ CJ Ll_ CD fe =3 CO CZ c C J + CD i — CD ra s- cj U _ CD ro C J CD oo Z3 O C J C J + CD CD t— CD ra c o _ U _ CD 5 f V c j C J C J + O > , £ r - £ CD C 2 CD % 81 £ L § C L o -t-> -CI Z3 OO C-o CD CNJ CD i—I CNJ CD CD CD CD CD i—I CD CD CNJ CO CD CD CD CD i—l i—I CTi CTl CO CNJ CD CNJ CD CNJ i—I CD i—I CD i — ' CO CD CD i—l CD i—I i—I CD L O co L O CNJ LO CNJ L O CNJ L O CNJ L O C D L O L O >•> OO +-> CZ OO CD CD +-> C- +-> .CZ X3 CD cn i — >> OO +-> >i • i— • P— o s_ ra 00 CJ OO CD -CZ cz CD c_. a i cz CD 3 CO CJ ro !_ +-> cz ra +-> CD "O CZ Z3 -CZ -CZ • 1— c CD sz 1 •i— cz c n O cn +-> 1— cn j Q +-> <C j n ro CD - Q Z3 c_ - O CD ra c_ r— ra s_ c_ ro ra • p— • 1— s_ • i— • (— O CQ CD Q_ 1 CQ OO a_ CQ 1— ,154 CO CO co CNI co cu •— CD CD C_ CJ U_ CD oo CD 3 CD C CD CD 00 +-> =3 C= CD O i — C J C J C J + CD CD CD CD CO CD CD CO ' LO LO CTl CNJ CD LO ^1-CO LO CD i — CD CD C_ CJ Li_ CD ai OO cz c_ 00 cz o C J CJ + CD CD CD CD CD CO CD CNJ LO CO CN) CNJ CNJ CNJ co CD i — CD co 5- O Ll_ CD CD C J +-> 00 CZ CD CD oo +-> =3 , CZ CD O i — C J C J C J + ro CNJ CO CNJ LO LO CO co CO oo CD r— CD CD C- CJ U_ CD c_ ZJ e CD o C J C J + CD CNJ i—I CD CD LO CO 1— 1 CO C J CD Q i CZ o •I— +-> • i — cz: C D O C J CNJ CU or + J CO C_ CO Q _ CO c o C U L l _ • i — - Q cu r o f— C D _ Q CO cz: 1— o O > , ^ r - ^ CD ' P CD Jr! CD CZ ( . CD -> 8 3 < Q . O +-> ja oo c_ o CJ CJ ra > "O o o . CD S ra o c j CD c j ra •""3 CD •a o CD > ro C_ C-5 CO o C J •a CD CD < I— o CU r— CU ro 5- CJ Ll_ CU CrT ro CU ro C J CD CU r— CU ro + r y CJ ^ - C J CD CU i — CU ro S- CJ U_ CU CJ) + O r— CJ> CJ) C J C J + OO CO CO LO CO CNJ CNJ i—I i—I LO CNJ OO CD CD CD CD CO CNJ CNJ r-H CNJ CO CO CNJ CD CNJ CD CNJ CO CP. LO LO LO LO CU r— CU ro S- CJ Ll_ CU sz cu o C J C J + CNJ CNJ CD CD i—I LO CNJ oo oo CNJ CNJ O >, -o ^ r - ^ QJ c S * 3 CrT >> sz sz >, Q -O o S_ o +-> SZ cu > >1 S_ o S_ S_ CU . 0 ro O . CU cu Q -e o CJ cu CU CD J Z +-) 1 1— >> <c ro JO 1— CU ro o :n CO 1— CD CD CJ + c o 157 CD i — CD CD C CJ U _ CD D i O CJ 3 CD C O C D L O C D C O C D ^ l " C O CD CD LO 3 C J + C D C O i— I C D C D C O " 3 " C N J ^ t - C S J ^ L O C M CD r— CD CD S- CJ Ll_ CD O CJ C N J C D C \ J C D C D C D <—I OO cz c_ C D i—I C D C D CD i — CD CD S- CJ U _ CD O CJ C D t—l i—l C O C D C J + L O L O co CD .— CD ca C_ CJ U - CD C oo CD C +-> c £ = 3 O I — CJ C D i—I C O C D C D c_ 3 L O L O C J + L O co L O £ ^ ^  £ 'E S ^ s c= >, o +-> 3 O O 00 CD O >i >> JD ra ra s_ CQ C J c o o ra CQ 3 O CD CD CD C_ N Q_ CX ra •""3 CD CD CD CD CU i— CU IX! s- o u_ cu ro CD C J + CU i— cu ro s_ o U_ CU L O sz =3 O C_) CD CU r— CU ra S- CJ u_ cu C_) + cu O r— C J C_) + CU ra ^ <_> 3 + O > , T - , +3 |Z 43 £ "E S cu g o £ - I o . o 4-> ja 3 on OO r—I CD CD CD CD r—I CD CD ^t" OO CNJ CD CD CD CD L O CD LO r-H LO LO OO OO CD CD CD OO OO OO LO LO CNJ LO LO CNJ C D C D C D C D C D C D C N J C D ^ " CD CNJ OO CD CNJ CD CD CD CD CD •—I f— CD LO CD CD CD CD CD CD CD r-H CD CNJ CD CNI r—I CD CD CD CD CD r-H *rj- CD CNJ CD CD CD CD CNJ LO LO 4-> cu L O CU L O CJ 1— L O SZ ro CJ ra s_ ra o cu SZ • CJ EE +-> sz <t- o cu +-> +-= 1—1 ra 4-> s_ ra ^£ L O 4— s_ sz CU 2: L O o 4-> ro i — 4-> o - 0 4-> •r— o . L O CU i— s_ SZ s_ ro L O CJ) S_ s_ - i — s_ • 1 — o o o • i — CU • i — JZ •I— o m s: 3 2: 1— Q oo. 00 CJ 158 s_ o CO CO CD CD CD CD CD CD i — CD C- CJ Ll_ CD or CD CD i — CD CD S- CJ Ll_ CD CO CD CD co 4-J CZ CD C= CD O r -C J C J C J + C CD +-J o c_> c_ =3 C_) + CJ CD =3 CO 5= CD CD c/) +-> 3 C= CD O •— C_) C J C J + CD i — CD CD S- CJ Li_ CD Di c =3 O C J C J + r; cp c 01 O CJ CD Di CJ CD t_ C_ o C J c_ CD 3 Q . O X3 =3 OO c_ o CO CNJ L O C\J CTl CD CNJ L O •—< "3" CT> CNJ L O '—1 CO CM i—I CD CD CD CM CD CD CM CO i—I CO CO CD CD CM CD CD CD L O T—I > D 00 O0 TD +-> CD o 00 CD c: • i— o CO > CD +-> <4-CD CD •i— E (— CJ c +-> +-> ro c= - i— ' SZ =3 oo •i— CD 1 > 00 00 CJ CD CJ NJ <c !_ CD TD CD > CD o h-CD C_ • i— X C D . o OO Ll_ U J 1—1 OO Ll_ 1— CD CD CD CD 159 co oo CNJ CD CO CT. CO ro c j CD t_ CD CD C_ c_ 3 •a CD > CD TD CO !_ CD TD CD +-> O 

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:
https://iiif.library.ubc.ca/presentation/dsp.831.1-0099176/manifest

Comment

Related Items