THE EFFECTS OF TIME COMPRESSION UPON THE INTELLIGIBILITY OF THREE SENTENTTAL MESSAGE TESTS M. A .mcwem STATE UNIVERSITY GENE w. BRATT 119.7 5 {wish-5a: attuned .193“; .26-1. 36- 9 *‘I’1g t) a. ,0! ‘. I. . .‘ ,“ . .-y fl 3&7: g N .5 f .: A _/ T... ‘. ‘rr‘. 9‘ vuzb’u; m‘VCJI‘5 a 1‘. «at. .5. . . , I; :1 -.fi '.' ' 0 I: 'r? ‘ . ‘ . ET '. I ,A I v' . , ' 3‘ "v . ‘2": 'T’ A F. ' I. a w”. g-x ~_ . “- -- b L'.."‘q 1- 'a ‘Q" I‘J W ? . a .. I R, u: 0 ~31"? - 31.; . £3:- fi “M5 M4 -~ 1 (J: Al "mmmrafis ABSTRACT THE EFFECTS OF TIME COMPRESSION UPON THE INTELLIGIBILITY OF THREE SENTENTIAL MESSAGE TESTS by Gene w. Bratt A review of the literature suggests that renewed interest is being shown in the sentential speech message. Reports have indicated a high correlation exists between the intelligibility scores obtained with con- tinuous discourse and those obtained with sentential stimuli, simply because the sentential message is thought to more realistically appraise a patient’s discrimination of everyday speech than do the more widely used monosyllabic stimuli. Further, interest in the time domain of the speech signal has warranted investigation of speech stimuli of sufficient duration to permit manipulation of their temporal characteristics. Application of the sentential message to the assessment of central auditory function has been of particular interest. The literature indicates that stimuli employed in the evaluation of central processes must exhibit the stability to withstand peripheral contamination and also sufficient degradation to tax the integrative functions of the central auditory system. ‘Sentence stimuli are characterized by a high degree of homogeniety of intelligibility and of redundancy due to the grammatical and semantic constraints present in the sentential message. Properly degraded, sen- tential stimuli may prove adequate in providing a differential assessment of central auditory functioning. Gene N. Bratt The literature provides both a theoretical and diagnostic basis for the utilization of temporal distortion. particularly time compression, as a means of degrading the speech signal. Normative data have been established for time—compressed monosyllabic word stimuli using a normalehearing population of young adults. Data have also been obtained on the responses to time-compressed monosyllables of individuals with peripheral and central auditory dysfunction. The present study was designed to investigate the responses of normal-hearing young adults to time—compressed sentential stimuli. Three samples of sentential stimuli were administered: five lists (ten sentences per list) each of the Central Institute for the Deaf (CID) Sentence Lists, the Revised CID (RCID) Sentence Lists, and third- order sentential approximations (seven word length) constructed from a pool of words derived from the CID Sentence Lists. Four ratios of time compression (0%, 40%, 60%, and 70%) were employed as were two sensation levels (24 dB SL and 40 dB SL). Ninety-six normal-hearing young adults served as subjects. Each subject was exposed to all 150 sentences at one ratio of time com- pression and at one sensation level. The design resulted in eight experimental conditions, four conditions of time compression by two conditions of sensation level, with twelve subjects in each condition. Within each condition six left ears and six right ears were randomly selected to receive the stimuli. Presentation of the sentence lists was counterbalanced to prevent the occurrence of an order effect. Written responses of each subject were evaluated for the number of words correctly recalled. Scores were converted to percentages and then placed in a four—way analysis of variance with a repeated measures design. Gene N. Bratt The results of this study indicated that the intelligibility of time-compressed sentential stimuli follows the general trend of the intelligibility of time-compressed monosyllabic stimuli, that is, as time compression increases intelligibility decreases, with a dramatic decrease in intelligibility occurring at 70% time compression. However, the CID and RCID mean intelligibility scores were consistently higher than the monosyllabic word mean scores at all ratios of time compression. Sentential approximations were noted to elicit a higher intelligibility than monosyllables at low compression ratios; however, as time compres- sion increased sentential approximation intelligibility decreased more rapidly than did monosyllabic word intelligibility. The results of the present study further supported the concept of a trading relationship between time compression and sensation level, whereby increased sensation level was able to partially offset the negative effects upon intelligibility of increased time compression. Although minimal ear effects were noted in this study, a significant ear effect at 70% time compression in one of the five inter-test analyses suggests the need for further investigations relative to laterality effects at high compression ratios with different types of speech stimuli. The results of the present study support the notion that increased linguistic constraints in sentential messages result in increased intelligibility of the sentence stimuli. This finding implies that sentential stimuli may provide the stimulus stability necessary to assess central dysfunction with a minimum of peripheral contamination of test results. The findings of this study further support the thesis that time compression may ultimately prove to be a means of signal degradation Gene N. Bratt which will permit improved assessment of central auditory processing. Finally, the present findings may provide normative data to encourage investigation of timeecompressed sentential stimuli. THE EFFECTS OF TIME COMPRESSION UPON THE INTELLIGIBILITY OF THREE SENTENTIAL MESSAGE TESTS By (C‘Lh Gene u. Bratt A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Audiology and Speech Sciences 1975 Accepted by the faculty of the Department of Audiology and Speech Sciences, College of Communication Arts, Michigan State University, in partial fulfillment of the requirements for the Master of Arts degree' Co-Director of Thesis Zé/kWLD. Ci, 196%,. ,W Co-Director of Thesis Guidance Committee: Co-Chairman Co—Chairman ACKNOWLEDGMENTS I wish to express my sincere appreciation to Dr. Daniel S. Beasley and to Dr. William F. Rintelmann for their encouragement and , guidance as the Co—Directors of this thesis. I further wish to thank Dr. Steven C. White for his valuable time and assistance in serving as a member of the guidance committee of this thesis. I wish to thank my fellow students, particularly those who served as subjects in this study, for their time and interest invested in this work. A special note of thanks is given to Dan Konkle whose assistance was invaluable in the generation of the stimulus materials. Most of all, to Mary and Lisa, who did without husband and father far too often over the past two years, I express my special and lasting gratitude. iii TABLE OF CONTENTS LIST OF TABLES . . . . . . LIST OF FIGURES . . . . . . LIST OF APPENDICES . . . . . . CHAPTER I. INTRODUCTION AND REVIEW OF THE LITERATURE II. III. IV. Central Auditory Evaluation . Sentence Tests: The Central Institute for the . Deaf Sentence Lists Sentence Tests: Synthetic Sentence Identification . Time Compression . . . . Summary and Statement of the Problem EXPERIMENTAL PROCEDURES Subjects Stimulus Materials Stimulus Generation Presentation Procedures Apparatus and Calibration Analysis RESULTS DISCUSSION Time Compression and Sensation Level Test Differences Ear Differences . . . . . iv Page vi vii viii IO 15 20 23 23 24 24 25 27 28 29 48 48 52 53 TABLE OF CONTENTS (Continued) Implications for Future Research V. SUMMARY APPENDICES LIST OF REFERENCES 55 '58 61 77 LIST OF TABLES TABLE Page 1. Intensity (dB SPL. re: 0.0002 dynes/cmz) data 25 obtained for CID Sentence Lists B,D,E,G, and J, RCID Sentence Lists B,F,G.I, and J, and for five lists of third-order sentential approximations (SA) obtained via spectral analysis of the master recording. 2. Sentential lists included within each analysis 30 of variance. 3. Analysis of Variance Table--Analysis 1: CID List 31 B, RCID List B, S.A. List 1. 4. Analysis of Variance Table-—Analysis 2: C10 List 0, 32 RCID List F, S.A. List 2. ' ‘ 5. Analysis of Variance Table--Analysis 3: CID List E, 33 RCID List G, S.A. List 3. 6. Analysis of Variance Table--Analysis 4: C10 List G, 34 RCID List I, S.A. List 4. 7. Analysis of Variance Table--Analysis 5: CID List J, 35 RCID List J, S.A. List 5. 8. Average articulation mean scores at 24 dB SL and 40 49 dB SL for three time-compressed sentential speech tests (present data) and for time-compressed CNC monosyllables (NU-6) at 24 dB SL (Beasley, Schwimmer, and Rintelmann, 1972b) and 40 dB SL (Beasley, Forman, and Rintelmann, 1972a). vi LIST OF FIGURES FIGURE Page 1. Average articulation mean scores for four levels 36 of time compression collapsed over test type, sen- sation level, and ear mean scores. Figure 1 fur- ther represents average mean scores collapsed over five separate analyses of variance per- formed on the experimental data. 2. Average articulation mean scores for two levels 38 of sensation level collapsed over test type, time compression, and ear mean scores. Figure 2 further represents average mean scores col- lapsed over five separate analyses of variance performed on the experimental data. 3. (a-e) Average articulation mean scores for 40 three tests of sentential stimuli collapsed over time compression, sensation level, and ear mean scores for each of five separate analyses of variance performed on the experimental data. 4. Average articulation mean scores for four levels 42 of time compression plotted by sensation level collapsed over test type and ear mean scores. . Figure 4 further represents average mean scores collapsed over five separate analyses of var- iance performed on the experimental data. 5. (a-e) Average articulation mean scores for four 44 levels of time compression plotted by test type collapsed over sensation level and ear mean scores for five separate analyses of variance performed on the experimental data. vii LIST OF APPENDICES Appendix Page A. CID Sentence Lists B,D,E,G, and J 61 B. RCID Sentence Lists B, F, G, I, and J 64 C. Five Lists of Third-order Sentential Approximations 67 (Seven Word Length) Derived from a Pool of Words Taken from the CID Sentence Lists (Silverman and Hirsh, 1955). 0. Instructions Given to Listeners 70 E. Octave Band Data for Ambient Noise Measured in the 71 Test Booth. F. Cell Mean Intelligibility Scores for CID, RCID, and 72 Sentential Approximations Sentential Stimuli for Four Ratios of Time Compression (0%, 40%, 60%, and 70%) Zresegted at Two Sensation Levels (24 dB SL and 40 B SL . viii CHAPTER I INTRODUCTION AND REVIEW OF THE LITERATURE Renewed interest has been shown in the use of sentential speech messages for the measurement of speech intelligibility (Harris, Haines, Kelsey, and Clack, 1961; Jerger, Speaks, and Trammell, 1968; Duffy and Giolas, 1974) based on the idea that sentences offer a more natural lis- tening task than do the more widely used monosyllabic speech stimuli (Hirsh, 1952). Similarly, because in audiological evaluations an effort is made to appraise a patient's potential for processing 'everyday speech' (Webster, Davis, and Ward, 1965), it has been questioned whether word lists are adequate in fulfilling this purpose. Giolas and Epstein (1963) stated that because continuous discourse is by definition representative of speech encountered in everyday sit- uations, it was the most logical speech message to use in discrimina- tion testing. However, the difficulty in quantifying such stimuli has been a significant deterrent to the use of continuous discourse in the research or clinical setting. As an alternative, Giolas and Epstein sug- gested that stimuli should be used which present the highest intelligibility correlations to continuous discourse while also permitting facility of administration. Giolas and his associates and others (Lehiste and Peterson, 1959; Speaks and Jerger, 1965) have investigated sentential stimuli on the premise that sentences provide a better representation of conversational 1 2 speech than do monosyllabic speech stimuli. It has been argued that factors such as word predictability, stress, accent, intonation, voice quality, and duration normally supply important cues in processing speech, and such factors cannot be weighted in tests employing monosyllabic words. Research has further shown that auditory perceptual processing is temporally biased (Hirsh, 1959, 1967; Aaronson, 1967; Beasley and Shriner, 1973), suggesting time to be a crucial parameter of the speech signal. Speaks and Jerger (1965) suggested that, due to the importance of the time domain, speech material used in auditory assessment should consist of sentences of sufficient length to permit systematic alteration of their temporal characteristics. It is generally recognized that sentential speech stimuli are characterized by a high degree of homogeniety of intelligibility, and that they present an extremely steep performance-intensity function (Speaks, Jerger, and Jerger, 1966; Duffy and Giolas, 1974). Consequently, sentences are considered stable and reliable stimuli. Further, sen- tential messages exhibit a strong resistance to distortion incurred via peripheral auditory disorders. Partly for this reason sentences have been effeCtively employed in the assessment of central auditory function. In the presence of sufficient degradation, such stimuli may challenge the integrating and processing centers of the higher auditory pathways with a minimum of peripheral contamination. Central Auditory Evaluation The strategic problem in constructing measures of central auditory integrity has been to design procedures that will focus stress at precisely the proper level along the auditory pathway. In evaluating 3 the peripheral auditory system the differential audiometric techniques currently employed are successful in isolating the site of the lesion. Despite the presence of occasionally contradictory data within the peripheral test battery, a definitive pattern is generally observed to aid in the clinical diagnosis of pathology. Clinical procedures proven effective in isolating peripheral lesions have n0t been as successful in eliciting results which differen- tiate the pathologically involved central auditory system (Jerger, 1973). The nature of the peripheral battery stimuli has been suspected of contributing to the failure to obtain differential results at the CNS levels. These stimuli consist of essentially simple word items or pure tones. Although such stimuli may tax the analytical functions of the peripheral system, they fail to challenge the integrative function of the central processing centers (Bocca and Calearo, 1963). The excessive redundancy of the elementary stimuli more than compensates for the reduced integrity of the pathologically involved central auditory pathway. Schuknecht (1958) suggested that the lesions which involve the central auditory pathways result in a system which transmits simple signals at threshold intensity, but is incapable of handling the complex signals of speech. Jerger (1973, p.88) summarized four basic fundamentals characteris- tically noted in the presence of a central auditory lesion: 1. The patient will ordinarily show no significant "hearing loss" in the sense of decreased sensitivity for pure tones, on either ear . . . . 2. The patient will ordinarily demonstrate little or no difficulty with pure-tone tasks of the type so useful in the evaluation of peripheral disorders . . . . 3. The primary symptom will be difficulty in understanding speech presented to the ear opposite the affected side of the brain . . . . 4. Speech intelligibility tasks involving binaural, noncoherent S were designed as samples of everyday speech, the specifications of which had been determined by a Working Group of the Armed Forces- National Research Council Committee on Hearing and Bio-Acoustics (Silverman and Hirsh, 1955). An a priori assumption of the Working Group was that the basic sample item of everyday speech was to be the sentence. Controls were established for vocabulary, phonetic structure, and syntactic sentence form. To minimize the effects of memory on intelligibility scores, a maximum length per sentence was limited to twelve words and a minimum limit was designated at two words per sentence. Fifty percent of these sentences were to range between five and nine words in length, while the length of 25 percent of the sentences was to be between two and four words, and the remaining 25 percent were to be between ten and twelve words in length. To increase the validity of such a sample, redundancy was encouraged within each sentence. No test item was to be constructed that would require that every word be heard for correct discrimination of that item (Silverman and Hirsh, 1955). The CID Sentences were subsequently developed and recorded by Silverman and Hirsh (1955) in such a way that the authors felt that they had satisfied the criteria suggested by the Working Group. The recorded version of the CID Sentence Lists was made by ten untrained speakers. Effort was employed to obtain natural, spontaneous samples with normal inflection, tempo, and emphasis. The scoring of the sample was based upon the correct repetition of fifty key words within each of the ten lists. The sentences subsequently were published for use by audiologists (Davis and Silverman, 1960). 6 Harris gt_al. (1961) used the CID Sentences in an investigation of the dependence of intelligibility upon the electroacoustic characteristics of low-fidelity circuitry (hearing aids). Because the CID Sentences were of varied length, the authors revised the original one hundred sentences so that all sentences ranged from five to nine words in length, thereby reducing contamination of intelligibility scores by short-' term memory due to excessive sentence length. The same fifty key words per list were retained; Harris gt_al, called their test the Revised Central Institute for the Deaf (RCID) Sentence Lists. Giolas (1966) investigated the relationship between the intelligi- bility of sentences (CID Sentence Lists) and the intelligibility of a measure of continuous discourse (Ulrich, 1957). The continuous dis- course sample devised by Ulrich (1957) consisted of a fifteen minute lecture followed by a nineteen item test. Giolas (1966) found that as frequency distortion via low-pass filtering increased, errors in dis- crimination increased for both the sentence lists and the continuous discourse. However, Giolas reported that the error curves lay closer together for the sentence lists and continuous discourse than did the error curves for monosyllabic word tests and continuous discourse of a previous study (Giolas and Epstein, 1963). Giolas concluded that the possibility was enhanced for determining a patient's ability to under- stand speech when the CID Sentence Lists were employed as the test stimuli. Furthermore, it was shown that the speech intelligibility scores obtained with these sentences were relatively high compared to those obtained with monosyllables under similar degrees of frequency distortion, due possibly to the higher redundancy of the sentences. A weakness seemingly present in the correlation studies of Giolas 7 and Epstein (1963) and Giolas (1966), where a test such as the Ulrich was used, was the unknown effect of the variable of comprehension. Ideally, in intelligibility testing, an attempt is made to minimize or eliminate the effects of all factors, including comprehension, which may influence the listener's capacity to discriminate speech sounds. A test comprised of continuous discourse which employs the question/an- swer paradigm Would seem to inherently increase the effect of the compre- hension variable. Equivalency of the CID and Revised CID Sentence Lists has been investigated by Giolas (1972) and Giolas and Duffy (1973). Giolas (1972) noted that the revision of the original sentences consisted of adding to or deleting words from some sentences, and consequently the meaning of the sentences changed. Giolas (1972) found that the mean scores for intel- ligibility of the RCID Lists were significantly lower, except for lists H and I, than were the mean scores for the original sentences under the influence of low-pass filtering with a 420 Hz cut-off frequency. Giolas (1972) concluded that, based on the group mean scores for each list, both the CID and the RCID Sentence Lists contained a sufficient number of equivalent sentence lists to be of research utility. Giolas and Duffy (1973) reported low inter-list score correlations and thereby questioned the equivalency of the sentence lists; however, low score correlations do not necessarily preclude an equivalency between lists, but rather suggest that a positive and negative linear relationship between lists was minimized. Giolas gt_gl, (1970) investigated the predictability status of selected CID Sentence Lists as well as a selected list from the Revised CID Sentence Lists (Harris-et al., 1961) and one from the Synthetic 8 Sentence Identification Lists (Jerger gt_al,, 1968). Giolas §t_gl, (1970) proposed that a low predictability status would be obtained if the intelligibility of sentence material was more dependent upon speech discrimination than on the influence of contextual cues. CID Sentence Lists B and D were selected to allow for list equivalency comparisons. List C from the RCID Lists was used to investigate the effect of sentence length on sentence-embedded word predictability. Third-order synthetic sentences were employed to provide a control for list comparisons. Giolas and his associates reasoned that a low predictability score for the synthetic sentences should be obtained because contextual cues in synthetic sentences are minimal. Word predictability was defined by the authors as the property of a sentence which permits prediction of a missing word(s). A modified cloze pro- cedure was used to measure the predictability of words within a sentence structure. Giolas and his associates found that the CID Sentence Lists were highly predictable and similar to each other in intelligibility difficulty. RCID Sentence List C was considerably less predictable than the two CID-Sentence Lists. The synthetic sentences had a negligible predicta- bility value. These findings suggest that the intelligibility of the C10 Sentence Lists was more dependent upon the influence of contextual cues, whereas the synthetic sentences depended more upon speech dis- crimination for intelligibility. The intelligibility of RCID Sentence List C was noted to depend moderately upon both contextual cues and upon discrimination ability. Finally, analysis of the results of the CID and RCID Sentences indicated a wide range of predictability values for key words within each list. 9 Duffy and Giolas (1974) investigated the relationship between the sentential characteristic of word predictability and sentence intelligibility, using the C10 Sentence Lists 8 and D and RCID List C under two conditions of low-pass filtering (420 Hz and 360 Hz cut-off fre- quencies). They found that the intelligibility of easy-to- predict words was consistently higher than that of difficult-to-predict words. These results indicated that the intelligibility of the sentence lists could be affected by controlling the predictability factor of the words in each list. In the C10 Sentences this relationship was found to be strong, due supposedly to the abundance of contextual cues available in these sentences. Duffy and Giolas concluded that, regardless of the scoring method used, when the mean scores for sentence intelligibility were compared to those of word lists (Hirsh, Reynolds, and Joseph, 1954; Epstein, Giolas, and Owens, 1968), it was apparent that sentences were more intelligible under severe low-pass filtering conditions. In summary, the CID Sentence Lists have been shown to exhibit a higher intelligibility correlation to continuous discourse than do monosyllabic word tests. For this reason they may considered a more valid assessment of a person's capacity to process speech. Equivalency studies between the CID Sentence Lists and the RCID Sentence Lists have indicated that the CID Sentences manifest a stronger relationship between word predictability and intelligibility, suggesting increased dependence upon contextual cues for correct recall. The RCID Sentences, with a lessened intelligibility--word predictability relation- ship, were determined to exhibit a greater dependence upon discrimination for correct perception . Based on group mean scores, there was reported enough inter-list equivalency within each test to render each a useful 10 research tool. Sentence Tests: Synthetic Sentence Identification Speaks and Jerger (1965) noted that the generation of the Synthetic Sentence Identification (SSI) material occurred as a result of a number of shortcomings inherent in the conventional monosyllabic word and sentence tests used to test auditory perceptual processing. The sentence tasks described by Fletcher and Steinberg (1929), Hudgins, Hawkins, Karlin, and Stevens (1947), Silverman and Hirsh (1955), and Harris gt_al, (1961), which used real English sentences, were felt to be difficult to administer and score. This was due, in part, to lack of control over the informational content of the test sentences. In- creased redundancy in these sentences allow subjects to accurately discriminate each sentence on the basis of perhaps two or three key words. Speaks and Jerger felt this increased the difficulty of evaluating the results of these test sentences. In addition, they argued that formation of equivalent lists of sentences was extremely difficult due to the effects of word familiarity, word and sentence length, and syntacti- cal structure. The SSI test was constructed as a sentence test whereby each test item possessed sufficient length to permit systematic alteration of its temporal characteristics. In addition, each item was controlled for informational content and length. 'Artificial' sentences were constructed according to predetermined rules on the basis of conditional probabilities of word sequences, as described by Speaks and Jerger (1965). The test procedure employed a closed—set paradigm. The subject was supplied with a list of ten sentences of equal length and order, 11 and was asked to select from that closed set the sentence which was presented to him. Because of the closed-set paradigm, the authors felt that the significance of the linguistic background of the subject was reduced, thus rendering the procedure more objective. It was stated further that this procedure could be easily automated and the effects of learning could be determined. Initial eValuation of the SSI materials (Speaks and Jerger, 1965) with normal-hearing subjects under a low-pass filtered condition (350 Hz cut-off frequency) showed that as constraints on word sequence increased, intelligibility also increased. Under the influence of temporal inter- ruption (1.25 interruptions per second), the same pattern was noted: third-order scores were better than first-order scores, even with 75 percent of the message deleted. Speaks et;al, (1966) investigated the performance-intensity function for first-order approximations, and found it to be 16.1 dB sound pressure level (SPL) at the 50 percent correct identification point, whereas a similar degree of intelligibility was not achieved with spondees until 20.2 dB SPL was reached. The performance region from 20 percent to 80 percent correct for first-order approximations was encompassed within an intensity range of only 5 to 7 dB. It was further determined that presenting the SSI materials under conditions of distortion, such as white noise (Jerger, 1970a) or low-pass filtering (Speaks gt_gl,, 1966) resulted in making the test uniformly more difficult, but did not result in a flattening of the slope of the performance-intensity function. Using third-order approximations, Speaks (1967a) found the point of equal intelligibility to be 725 Hz in the presence of low and high- pass filtering. This was in contrast to the point of equal intelligibility 12 of 1900 Hz for CVC monosyllables (French and Steinberg, 1947). Speaks further stated that under severe filtering conditions, 100 percent discrimination accuracy was achieved with the SSI, but not with CVC monosyllables. Speaks concluded that the SSI material was more redundant than monosyllables, and thus had a greater resistance to degradation due to additional contextual cues available in the SSI stimuli. Speaks, Karmen, and Benitez (1967) determined that the slope of the performance-intensity function for third-order approximations could be considerably flattened by the addition of a competing message presented to either the ipsilateral or contralateral ear, relative to the ear receiving the primary signal. The degree of degradation could be con- trolled by regulating the intensity ratio between the competing message and the primary signal (message to competition ratio, MCR). In com- paring the SSI-MCR task and conventional monosyllabic speech discrimination tasks administered in quiet, Jerger et_al. (1968) employed a competing message of continuous discourse. The competing message was spoken by the same person used to record the primary signal, and was administered at a 0 dB-MCR. The authors reported that the SSI-MCR and phonetically- balanced monosyllables yielded similar information. Generally, the scores of the two tests were similar when the pure-tone audiogram revealed a flat configuration for conductive and cochlear disorders. As the slopes of the audiograms became more severe, the monosyllabic word scores declined more rapidly than did the SSI-MCR scores. This discrepancy was greatest for subjects with good hearing in the low frequencies, coupled with a severe loss in the high frequencies. This result supported the contention that high frequency energy was important to the intelligibility of monosyllabic words, whereas low frequency energy played a major role 13 in the intelligibility of the SSI materials. To make the sentential message difficult enough to challenge the perceptual centers of the central auditory system, it has been necessary to degrade the speech stimuli. Jerger and his associates have reported differentiated responses specific to central auditory lesions by mixing either an ipsilateral or contralateral competing message with a primary message consisting of synthetic sentences (Jerger, Weikers, Sharbrough, and Jerger, 1969; Jerger, 1970 a,b; Jerger, Lovering, and Wertz, 1972; Jerger, 1973). The results of these studies have led Jerger to postulate that patients with unilateral brain-stem lesions exhibit increased difficulty with ipsilateral competing messages, whereas temporal lobe pathology results in difficulty particularly with the contralateral competing message tasks. The sentential speech message has further been degraded by temporally accelerating the speed of delivery of the stimulus. Calearo and Lazzaroni (1957) used this procedure to demonstrate the presence of central dysfunction as a component of presbycusis. Aged subjects were asked to respond to "short significant" sentences presented at three rates of acceleration: 140 words per minute (wpm), 250 wpm, and 350 wpm. The authors reported a dramatic breakdown in intelligibility in presbycusics compared to normal-hearing adults. They reported a similar breakdown in discrimination in ears contralateral to temporal lobe lesions. Calearo and Lazzaroni concluded that in normal-hearing individuals the redundancy of both the speech signal and the neural pathways was sufficient to allow nearly complete neutralization of the negative effects incurred by modestly accelerating the speed of delivery. With further increases in delivery speed, an increase in intensity was 14 needed to insure good discrimination. By contrast, for subjects with altered central pathways, the function of the processing mechanism was affected. This resulted in an increase in the processing time of the auditory pathways and centers, leading to impairment in dis- crimination of accelerated speech. Hirsh (1959) contributed further rationale for temporal alteration of the speech signal to assess the integrative and associative functions of the higher auditory pathways. He discussed time as the dimension within which patterns of speech are articulated for hearing in much the same way vision operates within a dimension of space. Hirsh stated that the essence of auditory perception was change within time. In a series of experiments designed to illustrate the importance of temporal order in auditory perception, Hirsh reported that there seemed to be minimal values of time necessary to allow for discrimination of changes in auditory stimuli. In observing the nature of serial order of stimuli in time, Hirsh (1967) stated that speech consists of elements and element patterns which are in fact temporal patterns. He suggested that the structure of language involves storage, perceptual sketching, and the formation of relationships over periods of time. Aaronson (1967) commented on four temporal factors which are important in auditory perception. The first such factor was the rate at which stimuli are presented. According to Aaronson, most short-term memory studies have shown that fast rates which reduce the time available for perception produce lower accuracy. Another factor was the duration of the stimuli. In experiments involving either visual or auditory stimuli, it has been shown that the amount of time between stimuli may be of at least equal importance in determining recall accuracy as the 15 time the stimuli are physically presented to the subject. The validity of this statement has been questioned by Beasley and Shriner (1973), who reported stimulus duration to be more significant to auditory processing than interstimulus interval. A third factor, post-stimulus events, may be important in that perceptual processes have been strongly suspected to continue after the physical presentation of the stimulus has ceased. Finally, the importance of pre-stimulus events has been established in studies where stimulus recall has been restricted by presenting a series of distractions prior to the presentation of the stimulus. The literature thus provides both a theoretical and a clinical basis for employing temporal degradation as a means of reducing the redundancy of the speech message to permit investigation of the central auditory pathways. 0f the types of temporal alteration available, time compression has been the most systematically and thoroughly investigated. Time Compression The early studies in time-compressed speech used either a person speaking faster than normal or a record speed-changing procedure (Fletcher, 1953). However, the frequency distortion accompanying speed- changing a recording proved undesirable. In addition, the inflectional distortion present when a person speaks faster than normal, and the extent beyond which a person cannot speak faster, limited the use of time-compressed speech. Garvey and Hinneman (1950) described a method to time compress speech stimuli whereby specific segments of the sample were discarded. The remaining segments were abutted in time by manually splicing magnetic 16 tape. This procedure eliminated the unwanted frequency distortion, and was more exact in controlling the information to be discarded. Unfortunately it was an extremely time consuming and cumbersome pro- cedure. Fairbanks, Everitt, and Jaeger (1954) introduced an electromag- netic time compressor which randomly eliminated speech material at fixed intervals. This procedure allowed for independent regulation of duration and interstimulus intervals. Recently, several computer- based procedures have been developed to time compress speech samples. The Lexicon Varispeech I (Lee, 1971) is such a procedure, whereby thirty millisecond (msec) segments are randomly discarded from the speech sample until a desired percentage of time compression has been achieved. This procedure allows for systematic increases in the rate of sentence delivery, introduces minimal pitch distortion, and reduces the amount of noise present in the finished speech sample. Fairbanks and Kodman (1957) employed an electromagnetic sampling technique to obtain a time-compressed sample of speech, consisting of the Harvard PB-50 Word Lists (Egan, 1948). This speech material was ad- ministered to highly trained listeners in an effort to investigate the psychoacoustic nature of the time-compressed speech stimulus. The authors found that, using a maximum intensity of 80 dB sensation level, trained normal-hearing adults were capable of discriminating time- compressed monosyllables until a ratio of 80% time compression was achieved, at which point a marked breakdown in intelligibility occurred. Lutermen, Welsh, and Melrose (1966) presented CID W-22 mono- syllabic word lists, time compressed by 10% and 20% using an electromagnetic sampling procedure, to aged populations and to two groups of younger 17 adults. The authors reported, contrary to the findings of Calearo and Lazzaroni (1957) and de Quiros (1964), that the performance of the aged subjects in their study did not differ significantly from that of young subjects in response to time-compressed speech. However, the findings of Luterman et_al. were limited by the mild ratios of time com- pression used. Sticht and Gray (1969) also presented the CID W-22 word lists to young and aged subjects. Each group was divided into normal-hearing and sensorineural hearing-impaired subjects. Three time compression ratios (36%, 46%, and 59%) and an undistorted control condition were achieved using an electromagnetic procedure for time compression. The authors found no differential responses relative to the nature of the subjects' hearing ability. They did find that the aged subjects did slightly poorer in discriminating time-compressed speech than did the younger subjects. This difficulty in intelligibility was noted to in- crease as the percentage of time compression increased. Bergman (1971) employed time-compressed speech as one test in a battery of degraded speech listening conditions presented to adults in each age decade from 20-89 years of age. The results of his study revealed that, despite relatively normal thresholds for pure tones, the ability to discriminate time-compressed speech decreased as age in- creased. These findings suggested that decreased discrimination ability in the aged was related to reduced perceptual processing capabilities, thereby requiring more time to integrate the complex signal of speech. Beasley, Schwimmer, and Rintelmann (1972b) conducted a study to establish normative data for time-compressed speech in a population of normal-hearing young adults. They administered five conditions of 18 time-compressed speech (30% to 70% at 10% intervals) plus a control condition of 0% time compression. Each of these conditions was presented at sensation levels of 8, 16, 24, and 32 dB. The stimuli presented consisted of the four lists of Form B of the phonetically-balanced Northwestern University Auditory Test No. 6 (NU-6) (Tillman and Car- hart, 1966). The authors reported that as the degree of time compression increased, intelligibility decreased. It was found that this effect could be partially offset by increasing the intensity level of the signal. Ear differences were found to be minimal. Specifically, it was reported that losses in discrimination were minimal until a ratio of 40% time compression was reached. From 40% time compression, break- down in discrimination was gradual until 70% time compression was achieved, at which point a marked breakdown in intelligibility occurred. At a compression ratio of 70%, increased sensation level had little effect, although the maximum intensity administered was only 32 dB sensation level. Beasley, Forman, and Rintelmann (1972a) extended the intensity limits of the Beasley, Schwimmer, and Rintelmann (1972b) study to 40 dB SL. Their findings indicated that, despite the increased intensity, a marked breakdown in intelligibility again occurred at 70% time compression; however, the extent of the intelligibility loss was not as great at the higher (40 dB) sensation level. The Beasley gt_al, studies (1972a,b) provided a basis for comparison between normal-hearing and pathologically involved patients by establishing normative data for time-compressed monosyllables with normal-hearing young adults. Kurdziel (1972) administered time-compressed CNC mono- syllables (NU-6) to carefully selected individuals with noise-induced hearing loss. She found that sensorineural hearing-impaired subjects 19 displayed time compression articulation functions which were essentially parallel to normal functions but with reduced discrimination scores at all sensation levels. Kurdziel and Noffsinger (1973) administered two ratios (40% and 60%) of time-compressed NU-6 monosyllables to ten subjects with central lesions involving the temporal lobe. Three of these subjects had undergone hemispherectomies. All subjects were capable of understanding normal speech and responding reliably; most had pure tone audiograms within normal limits. Under time compression, the scores of the subjects via the ipsilateral ear were near those of normal subjects. Scores obtained in the contralateral ears were considerably depressed. Differences in the mean scores between the ipsilateral and contralateral ears were 13% at 40% time compression and 25% at 60% time compression. By contrast, none of the ears demonstrated differences of greater than 4% with un- diStorted speech. Normative data on the intelligibility of time-compressed monosyllables employing normal-hearing young adults have been reported by Beasley et_al, (1972a,b). Furthermore, normative data for monosyllables (PB-K and WIPI tests) with children have been supplied by Maki, Beasley, and Orchik (1974), and with an aged population by Konkle, Beasley, and Bess (1974). However, no normative data appropriate for audiological mea- surement of intelligibility have been reported for time-compressed sentential stimuli. Such information is needed in light of findings of studies which have investigated the pr0perties of the sentential message. These findings indicate that time-compressed sentential speech stimuli may be effectively employed in the investigation of central auditory dysfunction. 20 Summary and Statement of the Problem In summary, a review of the literature suggests both a theoretical and a diagnostic basis for utilizing time-compressed speech as a tool for isolating lesions of the central auditory system (Calearo and Lazzaroni, 1957; Bocca and Calearo, 1963; Aaronson, 1967; Kurdziel and Noffsinger, 1973). The clinical significance of time compression had been limited earlier (Calearo and Lazzaroni, 1957; Bocca and Calearo, 1963) by a failure to establish normative data in the presence of standardized speech stimuli. Furthermore, these early clinical studies failed to describe adequately the procedure used to temporally alter the speech stimuli. Later studies (Luterman gt_al,, 1966; Sticht and Gray, 1969) employed limited ratios of time compression and failed to use an adequate number of subjects to allow norms to be generated for time-compressed speech. Beasley eggal, (1972a,b) have provided normative data for normal- hearing young adults using CNC monosyllabic stimuli (NU-6). Normative data for aged subjects (age 54-74) have been reported by Konkle gt_al, (1974) using the NU-6 stimuli, and for children, using the PB-K and WIPI tests, by Maki gfl;j{L. (1974). Finally, data using pathological listeners responding to time-compressed monosyllables have been reported by Kurdziel (1972) for sensorineural hearing-impaired subjects and by Kurdziel and Noffsinger (1973) for patients with CNS lesions. No normative data for audiological measurement of intelligibility have been reported in the literature where sentential messages have been employed as the test stimuli. However, sentential speech stimuli have received renewed attention in assessing auditory function in general due to their correlation of intelligibility scores with those 21 of continuous discourse (Giolas and Epstein 1963; Giolas, 1966), thus providing perhaps a more realistic clinical appraisal of hearing impairment (Lehiste and Peterson, 1959; Speaks and Jerger, I965; Duffy and Giolas, 1974). Furthermore, under temporal and competing message degradation, the sentential speech message has been reported to be an effective measure of central auditory pathology (Calearo and Lazzaroni, 1957; Jerger gt_al,, 1969; Jerger, 1970a,b; Jerger gt_al,, 1972; Jerger, 1973). Speaks (1967a) suggested that the sentential message also has a greater resistance to distortion than monosyllabic stimuli. Thus the possibility of peripheral contamination in the assessment of central auditory dysfunction is seemingly lessened by the use of stimuli such as the CID Sentence Lists (Silverman and Hirsh, 1955), the Revised CID Sentence Lists (Harris gt_al,, 1961), and the Synthetic Sentence Identification test (Speaks and Jerger, 1965). These stimuli, properly degraded, may possess a high degree of stability and a listening difficulty sufficient to challenge the integrity of the central audi- tory system. Because time has been shown to be crucial parameter of the speech signal (Hirsh, 1959, 1967; Aaronson, 1967; Beasley and Shriner, 1973), temporal alteration of the speech signal would seemingly provide an effective means of signal degradation. Among the procedures available which provide temporal alteration of the speech signal, time compression has been the most extensively and systematically investi- gated. The purpose of this study was to determine the effect of several ratios of time compression upon the intelligibility of the sentential speech message with normal-hearing listeners. Furthermore, the effect 22 of intensity changes was investigated with respect to each ratio of time compression. Specifically, the following questions were investigated: 1. Does the intelligibility of sentential stimuli (CID Sentences, RCID Sentences, third-order sentential approximations) vary inversely with the ratio of time compression used; i.e., as time compression increases does sentence intelligibility decrease? Does the intelligibility of time—compressed sentential stimuli (CID Sentences, RCID Sentences, third-order sentential approximations) vary directly with the sensation level of the stimulus presentation; i.e., as sensation level increases does time-compressed sentence intelligibility increase? Do the intelligibility curves follow a similar pattern for each of the three time-compressed sentential stimuli (CID Sen- tences, RCID Sentences, third-order sentential approximations) investigated? Does the loss of intelligibility for time-compressed sentences follow a pattern similar to that reported with time—compressed monosyllables using normal-hearing listeners? CHAPTER II EXPERIMENTAL PROCEDURES Ninety-six subjects were assigned to eight experimental conditions of twelve subjects each. Each experimental condition was comprised of one of three ratios of time-compressed speech (40%, 60%, and 70%) or an unaltered control condition (0%), presented at one of two sensation levels. Each group was exposed to three samples of sentential speech stimuli: five lists (ten sentences per list) each of the C10 Sentence Lists, the Revised CID Sentence Lists, and third—order sentential approximations. Subjects The subjects consisted of ninety-six young adults, selected from a university population, whose hearing was assessed to be within normal limits. These subjects were randomly assigned to one of eight ex- perimental conditions. Within each group of twelve subjects, six right ears and six left ears were randomly selected to receive the stimulus presentation. To insure hearing within normal limits, each subject was required to pass a pure-tone sweep frequency screening test admin— istered bilaterally at 25 dB Hearing Threshold Level (HTL) (re: ANSI, 1969) at octave intervals ranging from 250 Hz to 8000 Hz. 23 24 Stimulus Materials Three samples of sentential speech stimuli were used in this study: the CID Sentence Lists (Silverman and Hirsh, 1955), the Revised CID Sentence Lists (Harris gt_al., 1961), and third-order sentential approxi- mations. The five lists (B,D,E,G, and J) of the CID Sentence Lists and the five lists (B,F,G,I, and J) of the Revised CID Sentence Lists were selected for presentation because these lists have the highest inter-list intelligibility equivalency as reported by Giolas and Duffy (1973). Also, five lists of third-order sentential approximations were constructed from a pool of words derived from the CID Sentence Lists. The words which comprise the C10 Sentence Lists were randomized into list form and presented to individuals selected from a university pop- ulation. Third-order sentential approximations (seven word length) - were derived according to the procedure reported by Speaks and Jerger (1965). The sentence lists employed in the present study are shown in Appendices A-C. Stimulus Generation The three sentential speech tests were recorded by a trained male speaker via an Ampex AG-440-B tape deck (frequency response 50-15,000 Hz, :_2 dB) coupled to an Electrovoice 635-A microphone. This tape recording was spectrally analyzed with a Bruel and Kjaer type 2112 spectometer, and the analysis was recorded with a Bruel and Kjaer type 2305 graphic level recorder. The results of this analysis appear in Table 1. The original recording was then electromagnetically duplicated from an Ampex AG-440-B tape deck to a master cassette tape with a 25 Sony TC-180AV tape deck (frequency response 50-15,000 Hz, :_2 dB). This duplication procedure permitted temporal processing using a Lexicon Table 1. Intensity (dB SPL. re: 0.0002 dynes/cmz) data obtained for CID Sentence Lists B,D,E,G, and J, RCID Sentence Lists B,F,G,I, and J, and for five lists of third-order sentential approximations (S.A.) obtained via spectral analysis of the master recording. Intensity range between Mean intensity above Standard peaks of minimum and minimum intensity peak deviation maximum intensity CID 22 16.80 3.97 RCID 19 15.29 3.16 S.A. 22 15.74 3.39 Varispeech I time compressor (Lee, 1971). Each speech sample was pro- cessed at time compression ratios of 0%, 40%, 60%, and 70%. A master c0py of each of these four processed tapes was made by recording from the Lexicon and obtaining a reproduction on an Ampex AG-660-2 tape deck (frequency response 50-15,000 Hz, :_2 dB). Research copies of each tape were produced by recording from the Ampex AG-440-B tape deck to an Ampex AG-660-2 tape deck. Presentation Procedures After the pure-tone screening test, a Speech Reception Threshold (SRT) was obtained using a tape-recorded version of the CID W—l spondaic word lists. The subject was familiarized with the spondee test vocabulary. The SRT was then established using a modification of the procedure suggested by Tillman and Olson (1973). Two words were presented initially at a 30 dB HTL level via earphones. If both words were repeated correctly 26 the signal was attenuated 10 dB. This procedure was continued until the subject missed both words. At that point the level of the signal was increased 8 dB and then attenuated in 2 dB steps, with two words presented at each step. The criterion for starting was that five out of six words were to be correctly repeated. If they were, the descent was continued. If not, the level was increased by 10 dB and a new 2 dB descent was begun. This descent, with two words presented at each level, was continued until five out of six words were missed by the subject. The SRT was defined as the lowest level where the subject repeated both words correctly, minus 1 dB for each word correctly received thereafter. Because the speech audiometer was calibrated in 2 dB steps, in those cases where the SRT was an odd integer, the threshold was increased by 1 dB. Next, each subject was exposed to all three speech stimuli (150 sentences) at one time compression ratio and at one sensation level (SL) re his SRT. The order of presentation of the three stimulus sets was rotated within each group. In addition, the presentation order of the five lists within each set was reversed for half of the subjects to prevent the occurrence of an order effect. Sensation levels were employed of 24 dB and 40 dB. The Ampex AG-660-2 tape deck was used to present the stimuli. All subjects were tested individually in a two-room prefabricated double-walled test chamber (IAC 1200 series), with the experimenter in a single-walled control room (IAC 400 series). The ambient noise level in the test room was sufficiently low (45 dB SPL, C-scale via a Bruel and Kjaer type 2204 sound level meter) so as not to interfere with the test results. The specific octave band data for the ambient noise 27 are reported in Appendix E. Each subject was given standardized instructions (see Appendix D) and was asked to write down the appr0priate response on the forms provided. Fifteen seconds were allowed between each sentence presentation for the response to be written. Apparatus and Calibration The pure-tone screening was accomplished with a commercial audio- meter (Beltone Model 15-C) which supplied the signal to TDH-39 ear- phones housed in MX 41/AR cushions. The calibration of the audiometer was checked every two days during the investigation to assure compliance with the standards specified by ANSI (1969). The speech stimuli were presented via a Grason-Stadler 162 speech audiometer through TDH-39 earphones mounted in MX 41/AR cushions. This audiometer was calibrated so that audiometric zero was defined as being 20 dB above 0.0002 dynes/ cmZ. The level of a 1000 Hz calibration tone recorded on the tape was adjusted to 0 VU for all speech stimuli. Before each subject was tested, the speech audiometer, including earphone and cushion, was calibrated with an artificial ear assembly (Bruel and Kjaer type 4152) using a condensor microphome (Bruel and Kjaer type 4144) and a sound level meter (Bruel and Kjaer type 2204). In order to calibrate the TDH earphone, the earphone was coupled to the condensor microphone of the sound level meter by means of a standard 6 cc artificial ear assembly. The level of the tone at a given attenuator setting was adjusted until it produced a deflection to zero on the speech audiometer VU meter. The resulting acoustic output of the system was measured, and this value was accepted as the average intensity of 28 the sentential stimuli as determined in the spectral analysis previously discussed. Thus, with the speech audiometer attenuation set at 40 dB HTL, the output of the artificial ear was 60 dB SPL, re: 0.0002 dynes/cmz. Analysis The individual word responses of each subject were hand-scored by the experimenter and then converted to percentage correct scores. For each subject, fifteen such percentage scores were obtained, one score for each sentence list administered. The data were then placed into a four-factor analysis of variance (ANOVA) with repeated measures design (Winer, 1962). Within each of the eight cells there were twelve subjects for a total of ninety- six subjects, forty-eight subjects at each sensation level (twenty-four right ears, twenty—four left ears) and twenty-four subjects at each time compression level (twelve right ears, twelve left ears). Each of the eight time compression by sensation level groups received three sets of sentential stimuli (five lists per set) so that fifteen intelligibility scores were recorded for each subject. CHAPTER III RESULTS In the present study the intelligibility of sentential stimuli was investigated with respect to the variables of time compression, sensation level, ear differences, and sentential test type. Analysis of the experimental data revealed that significant main effects were associated with each of the variables studied with the exception of ear differences. Further, there were noted significant interactions between the effects of time compression by sensation level, time compression by test type, and time compression by ear differences. One analysis of variance with repeated measures design was per- formed on each of the five sets of corresponding lists across the three speech tests used in this study. For example, Analysis 1 was performed on the data obtained from the set of three first lists across the three sentence tests. Thus, Analysis 1 was performed on the data obtained from CID List 8, RCID List B, and sentential approximation list 1. A complete schedule of analyses with corresponding sentence lists is shown in Table 2. The five ANOVAS are shown in Tables 3 through 7. The mean data associated with each significant interaction are illustrated in Figures 1 through 5. There were significant effects associated with the time compression factor for each of the five inter-test analyses (see Tables 3 through 7). 29 30 Table 2. Sentential lists included within each analysis of variance. Analysis No. 010 List RCID List S.A. List 1 B B 1 2 0 F 2 3 E G 3 4 e I 4 5 .1 J 5 mooo . o W .33. 31 moodw: moéw. «. awe. _N. m.o~ a H.NNH A x up x N x 4m Hmm. m. e.w N e.om P x up x N one. m. N.¢H N o.mN k x N x 4m «samooo.o N.mm H.¢¢¢N o o.¢mo¢H p x up mNm. H. m.N N N.¢ h x N moH. N.H m.om N m.mHH » x 4m «aamoco.o m.¢mN m.mNmN N m.mmNeH » amm. N. m.m~ m N.ov up x N x 4m Hmfi. m.H m.ONH m w.oom uh x N samoo. o.m N.mHm m N.Hem oh x 4m nmq. N. N.¢m H N.¢m N x 4m «atmooo.o o.NmN o.mm¢¢H m o.NHmm¢ up mmN. H.H “.mm H H.mo N aaNoo. m.oH o.N¢o H o.N¢m Nm .pmpm N No mucmupmwcmvm .xocaa< .pmpm N msascm cam: .$.u .m.m wucmwsm> mo mugaom .Ahv pmw» .ANV saN .Auhv cowmmeQEoo me?» .Ava Fm>w4 cowuwmcmm "msouowm .H pmwN .<.m .m pmwN oHom .m umwN oHu "H mwmxpwc No mvmapmc< .m mpnm» mooo . W «.3. 32 Noon: monw a NNN. . ‘ ,. . e. N.N N N.NN N x NN x N x NN aNN. e. N.N N N.NN N x UN x N NNe. N. m.NN N N.om N x N x NN «esmooo.o N.Na e.omm N N.NNNN N x UN NNe. N. N.NN N N.Nm N x N Nae. N. m.eN N N.NN N x Nm asemooc.o N.NNN o.eNNm N N.NeNN N Nem. N. N.om m c.NN NN x N x Nm ammo. N.m N.omN m e.NNm NN x N «atmooo.o e.oN o.eee N N.NNNN NN x Nm NNN. N.N N.NN N N.NN N x Nm stamooo.o N.NNN o.ONONN m H.0NNNN UN eNN. o.N N.¢N N N.eN N aaNoo. N.NN N.NNN N N.NNN Nm .HmHm “— ..vo mocmuwmwcmwm .XOLQQ< .flwwm “— mhwzdm cam: Nth .m.m wucvam> $0 moo—30m .Ahv paw» .NNV smN .Aopv copmmmggeou mewh .Ava Fw>mN cowucmcmm “msouumN .N NNNN .<.m .N NNNN NNNN .N NNNN NNN "N mNmaNae<--aNNaN aucaNsa> Ne mNmaNae< .e aNNaN 33 mooo .o w 33. moodwt. modws NNN. N.N N.NN N N.NNN N x NN x N x NN owe. a. NAN N 902 N. x up x N mmm. N. 92 N 5R N. x N x ._m sasmoood mém czaNmH N 533 .— x PF So. «N N.Nm N e6: k x N «.3. m. mg N Na“ N. x ._m 3.380 .0 N . mNN flow? N N .82: N. cam. m. 92 m wKe up x N x ._m Nmm. mo. NN m 9N up x N 3.80. N.m méNm m ménm B. x ._m wNm. mo. N.N H N.N N x 4m «.3886 m.mmN NéNNNH m magma up Sm. a. HAN H H.HN N «Loo. N.NH o.NNN N o.NNN ._m .umum N No mucaowmwcmwm .onNN< .papm N mgmacm cam: .N.u .m.m mucmwgm> mo mugaom .Ahv amok .ANV NNN .Auhv co_mmmgasou wave .ANmV Fm>mN cowpmmcmm "msouuoN .m NNNN .<.m .N NNNN NNNN .N emNN NNN ”m mNmaNae< --aNaaN maeaNLa> Na NNNNNN=< .N NNNNN 34 moood Wests Noodw: Nodwt Nmm. N.N N.NN N N.Nmfi .— x o... x N x ._N omm. N. N.N: N N.NN » x up x N 3N. N. m6 N N.N .— x N x ._m «.5386 cam NANN N N.Nmme .F x 8. oi. m. N.N N N.NN N. x N NE. N.N N.Na N m.mm H x ._m «.3885 N.N: NémmN N N.NNNN .— NNN. m. N.NN m N.NN NN. x N x .5. EN. N.N N.NNH m N.NNN NH x N aimoooé N.N N.NNN m N.NomN up x ._N oma. N. N.NN H N.NN N x ._N aimoooé N.NNN N.NNNS m N.NNNNN up Sm. N.N «.63 H N.NNN N .35886 N.NH N.NNNH N N.NNNH ._m .umum N mo wucmuNwwcmwm .onNN< .umpm N wgmzcm com: .N.u .N.N mocwmcm> No wuczom .Ahv umwh .ANV NNN .Auhv commmmsnsou mENP .ANNV Po>mN coppmmcmm "mgouumN .e pmNN .<.N .H pmNN oNNm .N umNN NNN "e mvmapmc mo mwmxpmc< .N mpnmh 35 mooodWa: moodwai modWa NNN.. emu. nmn. «aamooo.o m¢w. vow. *NNmooo.o eHm. umH. {amoo. Nmm. «tamooo.o moo. «amoo. .pwpm N mo mocmowmwcm_m .onNN< H.NON N.N ¢.m .papm N w.oH N.NN n.¢H m.mmmH H.m 0.x m.¢mm¢ N.NHH N.omH ¢.N¢¢ ¢.mm N.¢¢NmH N.Hmm N.wa mgmzcm cum: N.¢N N.NNH ¢.mN H.H¢Nm H.NH o.¢H m.momm v.w¢m m.va MMMNNNNONKOND N.NNmH 0—1 ¢.mw m m.mmmwm H m.Hmm H N.Nmm .m.u .m.m uh x h x N x Am F x 0h x N k x N x 4m h x up H x N h x Nm k up x N x 4m up x N oh x 4m N x 4m up N Nm wucmwgm> No mogzom .Hpv pmwh .HNN NNN .Hohv conmmNNsoo NEH» .HNNV Hm>mN cowpmmcmm .N NNNN .<.m .N NNNN NNNN .a NNNN NNN "N mNmszz No ”mgouumN N.N»Naeq .N NHNNN 36 Figure 1. Average articulation mean scores for four levels of time compression collapsed over test type, sensation level, and ear mean scores. Figure I further represents average mean scores collapsed over five separate analyses of variance performed on the experimental data. 37 IOOF 90- 80- 60- the el, verags 50- PE RCENT CORRECT 30- 20- IO- O I J L l I J l 40 60 70 PERCENT TIME COMPRESSION 38 Figure 2. Average articulation mean scores for two levels of sensation level collapsed over test type, time compression, and ear mean scores. Figure 2 further represents average mean scores collapsed over five separate analyses of variance performed on the experimental data. 39 IOO - 90- 80- ...Ummmou Pzw ummm 20*- lot- 24 SENSATION LEVEL Figure 3 (a-e). 40 Average articulation mean scores for three tests of sentential stimuli collapsed over time compression, sensation level, and ear mean scores for each of five separate analyses of variance performed on the experimental data. PERCENT CORRECT 41 IOO F (94.1) -—i (953)—- 90 - 11.6 7.8 .. 9.9 8° - (5... 7o — 3-0 60 Analysisl 50 , 1 92.7 ' ' 95.4 90 b 12.1 7.8 84.4 3° ‘ 18. 7o - 60 - 3-b 50 Analysis 2 90_ (94.3) F_' (93.6) *—'i 1 1.3 10.9 . 8° " (80.9 21. 7o— 60' 3-c 50 AnalySTs 3 L. 93.0 ‘_* (91.9 E_' 90 (12.3 142 333 80- 20.3 70 .. 60-3-d . 50 AnalySTS 4 _ 91.0 -—~ 93.2 ‘—"‘ 9° (15.5) (11.6 so — (79.8 1 23 7o - 60 -3-e ' Analysis 5 5° c 1 0 RC1 0 TEST TYPE Figure 4. 42 Average articulation mean scores for four levels of time compression plotted by sensation level collapsed over test type and ear mean scores. Figure 4 further represents average mean scores collapsed over five separate analyses of variance performed on the ex- perimental data. PERCENT CORRECT 5O 4O 30 20 IO 43 .......... = 40 dB SL = 24 dB SL 1 1 '1 1'1 40 60 PERCENT TIME COMPRESSION Figure 5 (a-e). 44 Average articulation mean scores for four levels of time compression plotted by test type collapsed over sensation level and ear mean scores for each of five separate analyses of variance performed on the experimental data. PERCENT CORRECT loo 9O :0 7O 60 40. 888' 50 45 _.—.—o—o_0'_.—°-—-o;- ~~ ‘~~. \ \ 5-a . Analysisl \J— * ‘7‘ r-....—--..— .—.—.—.—.—.—..—.—.—_ — _.—... \\ x, x, \O \O _ 5-b AnalysisZ - 5-c \ SO Analysis 3 — 5-e .\. AnalySTslS A l l l l V PERCENT TIME COM PRESSION 46 As can be observed in Figure 1, there was a moderate decrease in intelli- gibility through 60% time compression, and at 70% time compression there was a dramatic decrease in intelligibility scores. There was also a significant main effect associated with sensation level over the five inter-test comparisons. Figure 2 shows that generally the 40 dB SL stimuli resulted in higher intelligibility scores than did the 24 dB SL condition. Figure 3 (a-e) depicts the significant main effect of test type. Minimal differences were noted between the CID and the RCID mean scores for all five inter-test analyses, while the sentential approximation means were significantly lower in each case. The effect of time compression interacted significantly with sensation level and with test type for the five inter-test comparisons. As can be observed in Figure 4, there were minimal differences between the two sensation levels through 60% time compression. However, at 70% time compression, the 40 dB SL score was clearly superior to that of the 24 dB SL condition. The interaction between time compression and the three sentential tests employed, shown in Figure 5 (a—e), illustrated that the trend of a decrease in intelligibility for the four time compression ratios was similar for all three tests. In addition, Figure 5 shows that the sentential approximation scores were generally lower than the C10 and RCID mean intelligibility scores, and that this effect became increasingly marked as time compression increased. Although the differences between the C10 and RCID scores were minimal, Figure 5 shows that the superiority of intelligibility scores between the CID and RCID tests did not remain constant over the five inter-test comparisons. Analyses 1,2, and 5 47 revealed a RCID mean intelligibility score superiority, while Analyses 3 and 4 depict a CID mean intelligibility score superiority. Finally, for Analysis 2, there was a significant ear by time compres- sion interaction. Specifically, at 70% time compression the left ear mean score (74%) was better than that of the right ear (69%). CHAPTER IV DISCUSSION Time Compression and Sensation Level The results of this investigation support the thesis that, for normal-hearing young adults, intelligibility of speech stimuli decreases as the ratio of time compression increases. Further, the loss of intelligibility associated with increased ratios of time compression can be partially, but not completely, offset by an increase in the intensity of the speech stimuli. The nature of the analysis performed on the data obtained in this study particularly strengthened the validity of this thesis (Tables 3 through 7). The five inter-test analyses of variance each pointed to identical significant effects associated both with the main factors of time compression and sensation level, and to the interaction between these two main factors. Previous investigations with normal-hearing young adults (Beasley, Forman, and Rintelmann, 1972a; Beasley, Schwimmer, and Rintelmann, 1972b) using CNC monosyllables have indicated that time compression influences intelligibility in such a way that a moderate decrease in intelligibility is associated with compression ratios of 0% through 60%. However, a marked decrease in intelligibility occurs between 60% and 70% time compression. It has been suggested (Aaronson, 1967; Daniloff, Shriner, and Zemlin, 1968) that at such high compression 48 49 ratios, the individual may lack sufficient time and/or information to process the incoming signal. The results of the studies by Beasley gt_gl, (1972a,b) indicated also that, for each increase in sensation level from 8 to 40 dB SL in 8 dB increments, intelligibility improved. This effect was most evident at 70% time compression. The results of this study, using time-compressed sentential speech stimuli, revealed the effects of the increased redundancy of the time- compressed C10 and RCID stimuli relative to time-compressed monosyllabic stimuli. For all conditions of time compression and sensation level, the scores of the CID and RCID stimuli were superior to those obtained using the CNC monosyllables (Table 8). This observation is illustrated Table 8. Average articulation mean scores at 24 dB SL and 40 dB SL for three time-compressed sentential speech tests (present data) and for time-compressed CNC monosyllables (NU-6) at 24 dB SL (Beasley, Schwimmer, and Rintelmann, 1972b) and 40 dB SL (Beasley, Forman, and Rintelmann, 1972a) Time Compression CID RCID S.A. NU-6 24 dB SL 0% 99.4 99.7 96.9 91.9 40% 99.2 99.3 94.0 88.6 60% 97.5 97.7 84.9 83.6 70% 70.2 73.4 42.7 50.1 40 dB SL 0% , 99.6 99.4 98.3 98.2 40% 98.8 99.2 -94.1 97.2 60% . 97.9 98.3 91.4 94.2 N 70% 81.4 84.8 52.1 80.4 50 by the fact that, at compression ratios of 60% or less, at both sensation levels, the intelligibility scores for real sentential stimuli remained well above 90% correct. By contrast, the CNC monosyllabic word scores were below the 90% correct level at both 40% and 60% time compression in the 24 dB SL condition. The third-order sentential approximations presented quite a different intelligibility pattern relative to monosyllabic word stimuli. Table 8 shows that the approximations elicited higher discrimination scores than did the CNC monosyllables at the lower time compression ratios. However, as time compression increased, the scores for the sentential approximations decreased more rapidly than did the monosyllabic word scores to the point where, at 70% compression, the sentential approximation scores were inferior to those obtained using monosyllabic words. The higher intelligibility of the time-compressed C10 and RCID sentences relative to the time-compressed monosyllabic word scores may be the result of the increased redundancy of these sentential stimuli. These tests are composed of real English sentences and thus the numerous linguistic constraints operative in normal language may be expected to influence the intelligibility of these stimuli in the test situation (Miller and Isard, 1963). Such constraints serve to reduce the number of sentences possible under a given condition, and so limit the possible alternatives available from which words may be chosen to make up the resultant sentences. In this way these constraints reduce the amount of information contained within a given sentence, thus increasing the redundancy of the sentence. As a result, intelligibility is enhanced compared to monosyllabic words when these constraints are present. The sentential approximations maintain a level of linguistic 51 constraints relative to the degree they approximate real sentences. Thirdeorder approximations are more grammatically and semantically constrained than first or second-order sentential approximations, but are less constrained than real sentences. This would explain the reduced intelligibility of third-order approximations at all time com- pression ratios compared to the real sentence stimuli (CID and RCID sentences) as noted in Figures 3 and 5 and in Table 8. As time compression increases, the moderate level of redundancy present at 0% compression decreases as a result of the degradation process. At 70% compression, it would appear that the positive effects of the linguistic constraints associated with third-order approximations have been negated to a large extent by the effects of the temporal alteration of the stimuli. Thus, the listener is confronted with the difficult task of discriminating a string of seven words at a very rapid rate of presentation. Rather than enhancing discrimination by increasing constraints on word selection, the successive words may in fact be introducing concurrently a forward and reverse masking effect (Aaronson, 1967), thus impairing discrimination. This would explain reduced sentential approximation intelligibility relative to that of monosyllabic stimuli at high time compression ratios. Speaks (1967a) reported data that suggested that even under conditions of severe filtering, sentential approximations (synthetic sentences) elicited higher intelligibility than did CVC monosyllabic stimuli. However, Speaks employed a closed-set test paradigm where the subject was asked only to indentify the correct sentence out of ten possible selections. Thus, a sentence was discriminated as either entirely correct or entirely wrong. In the present study a more difficult open-set selection procedure was employed whereby each word 52 within a sentence had to be accurately recalled. Furthermore, considering the temporal bias of the speech message, temporal distortion as employed in the present study may provide a more efficient means of sentential degradation than was obtained by the frequency degradation used by Speaks (1967a). Test Differences Three sentential speech tests were employed in this study: five lists each of the CID Sentence Lists, the RCID Sentence Lists, and third-order sentential approximations. Mean scores for the CID and RCID tests were similar in all five inter-test comparisons in that intelligibility decreased slightly for both tests through 60% time compression, and dramatically decreased between 60% and 70%. Further, there was no consistency in the superiority of the mean scores of one test over those of the other test. In Analyses 1,2, and 5 the RCID mean scores appear to be slightly higher than the mean scores for the CID sentences, whereas with Analyses 3 and 4 the C10 mean scores exceed the RCID mean scores. These findings are not in agreement with those of Giolas (1972) and Giolas and Duffy (1973). Giolas (1972) investigated the equivalency of the CID and RCID tests, and reported that under the influence of 420 Hz low-pass filtering, the RCID scores were consistently lower than the C10 scores. He concluded that the revision process had resulted in an increase in the difficulty of the RCID stimuli. The results of this study indicated that under the degradatory influence of time compression, a small and inconsistent difference existed between these tests. Giolas and his associates did not attempt 53 to explain why the RCID scores were depressed under the influence of 'frequency distortion. However, possibly this unexplained effect, present under the influence of frequency distortion (Giolas, 1972) was negated in the present study, which employed a temporal degradation, by the advantage of sentence length homogeniety associated with the RCID sentences. Also, the discrepency in the findings between the present study and the investigation of Giolas and associates may be due, at least in part, to the differing effects of low-pass filtering versus time compression. The results of the present study suggest, therefore, that homogeniety of test scores is better preserved when time compression is employed as the degradatory influence of these sentence tests. A further observation in regard to the interaction between types of tests employed concerns the consistently lower intelligibility of the sentential approximations, whereby the difference between the third- order sentential approximation scores and those of the other two tests became greater as time compression increased. This trend, observed in all five inter-test analyses, supports the suggestion that intelligibility decreases as linguistic constraints decrease in speech stimuli. Ear Differences The dominance of one hemisphere of the brain over the other for speech processes has been recognized at least since the time of Broca's work. Recent research using dichotic listening tasks has suggested that verbal material is better received in the right ear (Kimura, 1961, 1963; Shankweiler, 1966; Studdert-Kennedy and Shankweiler, 1970; Knox and Kimura, 1970) while non-verbal environmental sounds are more accurately perceived in the left ear (Knox and Kimura, 1970). Because contralateral 54 auditory innervation to the respective temporal lobes is stronger than the ipsilateral pathway (Kimura, 1961), the results of these studies strongly imply a left hemispheric dominance for verbal processing. Although Bakker (1969) and Nagafucci (1970) reported data exhibiting right ear superiority in response to monotic tasks, research has sug- gested that ear asymmetry is much less evident when monotic speech tasks are empl0yed (Corsi, 1967). Dirks (1964) presented filtered phonetically-balanced words both dichotically and monotically to normal- hearing listeners. He reported a significant difference in the scores of the right and left ears only in the presence of the dichotic stimu- lation. Under the monotic condition, the right ear superiority was dramatically lessened, thereby suggesting that a need exists for competition between contralalteral and ipsilateral pathways to the dominant hemisphere to elicit asymmetry in ear performance. The concept of laterality is crucial to the validity of clinical use of speech tasks. Because the same scoring criteria are used in evaluating both the left and right ears under monotic stimulation, it has been necessary to verify absence of significant ear asymmetry in the devel0pment of auditory test procedures. Ear symmetry is particularly important in central auditory assessment where a significant difference between ears in speech discrimination in the absence of pure tone loss is indicative of central dysfunction. Beasley gt_gl, (1972a,b) reported the performance of right and left ears was equal in the presence of monotically presented time- compressed CNC monosyllables. The results of the present study, using time-compressed sentential stimuli, supported the results of the Beasley studies in that all five inter-test comparisons revealed the 55 absence of a significant main effect associated with ear differences. There was, however, a significant ear by time compression interaction revealed in Analysis 2 at 70% time compression. Surprisingly, this interaction revealed a left ear superiority. Because a similar inter- action was not present in the other four analyses, and because the direction of the asymmetry contradicted the weight of existing research, the noted interaction presently remains unexplained. However, further research is warranted regarding ear laterality associated with high degrees of time compression, employing monosyllabic and sentential speech stimuli. Implications for Future Research Jerger (1973) emphasized that stimuli used for central auditory evaluation must provide sufficient signal degradation to focus stress on the processing and integrating functions of the higher auditory pathways. In the presence of such stimuli, the ear contralateral to the side of the lesion may be expected to exhibit impaired performance of sufficient magnitude to allow a differential diagnosis to be made. Katz et_al, (1963) further stated that stimuli employed in central auditory evaluation should possess sufficient stability to enable an assessment to be made of the central fUnctions with a minimum of peripheral auditory contamination. Thus, stimuli should possess con- comitantly a high degree of redundancy and the demanding features of a degraded speech task. Hirsh (1959, 1967), Aaronson (1967), and Beasley and Shriner (1973) have suggested a theoretical rationale for temporally manipulating the speech signal to achieve a significant degree of degradation. 56 Kurdziel and Noffsinger (1973) have provided diagnostic data which indicate that time-compressed speech can effectively challenge the perceptual processing centers and differentiate the centrally impaired auditory system. The results of the present study revealed that time-compressed CID and RCID speech stimuli intelligibility was consistently higher than that of time-compressed monosyllabic stimuli. Thus it appears that these sentential stimuli maintain a greater redundancy and may be expected to exhibit a greater resistance to distortion than monosyllables. The present study has reported normative data for the administra- tion of time-compressed sentential stimuli. An immediate goal for future research lies in the evaluation of subjects exhibiting peripheral and central pathology, employing time-compressed CID and RCID speech stimuli. Because of the lowered intelligibility of third-order sen- tential approximations in the presence of time compression, these stimuli may not provide the redundancy required for effective central auditory assessment. However, it may be noted from the results of this study that sentential approximation intelligibility was sensitive to the effects of time compression, and thus these stimuli may supply differential results in central auditory evaluation. The possibility of laterality effects at high time compression ratios should be further investigated. The results of the Beasley 53:31, studies (1972a,b) with time-compressed monosyllables indicated minimal differences in intelligiblity between ears. This finding was generally supported by the results of the present study. However, there was also evidence to suspect the possibility of a laterality effect at high time compression ratios with speech stimuli of longer 57 duration than a single syllable. Further research is needed concerning ear performance at high time compression ratios employing both single syllable and sentential stimuli. CHAPTER V SUMMARY Because audiometric procedures used to assess peripheral auditory integrity fail to challenge the processing and integrating functions of the higher auditory pathways, a separate test battery must be employed to evaluate central auditory function. The tests which comprise this special battery must provide sufficient difficulty to focus stress at the higher auditory centers. There is evidence in the literature to suggest that degraded speech stimuli may be effectively used in the diagnosis of central auditory pathology (Bocca and Calearo, 1963; Jerger, 1973). In addition to possessing the difficulty of a degraded speech task, such stimuli must further exhibit a high degree of stability to avoid peripheral contamination of test results (Katz gt_al,, 1963). . Hirsh (1959, 1967), Aaronson (1967), Beasley and Shriner (1973), and Kurdziel and Noffsinger (1973) have suggested a theoretical and a diagnostic rationale for using temporal distortion, particularly time compression, as a means of achieving the degradation necessary to analyze central dysfunction. Sentential speech stimuli have been effectively employed in central auditory evaluation (Calearo and Lazzaroni, 1957; Jerger §t_gl,, 1969; Jerger et_al,, 1972), in part because they maintain sufficient length to allow effective manipulation of their temporal characteristics (Speaks and Jerger, 1965), and because 58 59 they have been shown to be stable speech stimuli (Speaks, 1967a). The present study was designed to investigate the effects of time compression and intensity upon the intelligibility of selected sentential stimuli. The present findings indicated that time-compressed sentential stimuli intelligibility followed the general trend of time-compressed CNC monosyllabic word intelligibility, reported by Beasley EEJEL. (1972a,b). These studies found that the intelligibility of time-compressed speech stimuli decreased moderately through 60% time compression, and that a dramatic loss of intelligibility occurred between 60% and 70% time compression. Also, as in the Beasley studies, the present investigation demonstrated the presence of a trading relationship between time com- pression and sensation level; that is, the negative effects on intelligibility incurred through increased time compression ratios were partially negated by increased sensation level. The present findings clearly portray that, at all time compression and sensation levels, the mean intelligibility scores of the CID and RCID sentences were essentially equal and were superior to the CNC monosyllabic word scores of earlier studies by Beasley et_al, (1972a,b). The intelligibility scores of third-order sentential approximations proved to be particularly sensitive to the influence of time com- pression, with intelligibility decreasing more rapidly at the 70% time compression ratio than with either the C10 or RCID stimuli. The results of this study support the thesis that time compression may ultimately prove to be a means of signal degradation which will permit improved assessment of central auditory processing. Further, the present findings suggest that the sentential stimuli employed in 60 this study may display the stimulus redundancy required to assess central auditory activity while minimizing the effects of possible concomitant peripheral pathology. These suggestions await further investigation. Generally, the five inter-test analyses employed in this study indicated minimal differences between the performance of right and left ears; however, a significant interaction, at 70% time compression, was reported in one of the analyses. These results suggest that further investigation is warranted into the laterality of ear performance, particularly at high time compression ratios, employing various types of speech stimuli. APPENDICES 'APPENDIX A CID Sentence Lists B,D,E,G, and J H O \OCDNOSU‘l-bwm H O H O RomNO‘IU'l-wa O—l O 61 List B The water's too cold for swimming. Why should I get up so early in the morning? Here are your shoes. It's raining. Where are you going? Come here when I call you! Don't try to get out of it this time! Should we let little children go to the movies by themselves? There isn't enough paint to finish the room. . Do you want an egg for breakfast? List D It's time to go. If you don't want these old magazines, throw them out. Do you want to wash Up? It's a real dark night so watch your driving. I'll carry the package for you. Did you forget to shut off the water? Fishing in a mountain stream is my idea of a good time. Fathers spend more time with their chderen than they used to. Be careful not to break your glasses. . I'm sorry. O C QmNOfiU'I-bOONH H O 62 List E You can catch the bus across the street. Call her on the phone and tell her the news. I'll catch up with you later. I'll think it over. I don't want to go to the movies tonight. If your tooth hurts that much you ought to see a dentist. Put that cookie back in the box. Stop fooling around! Time's up. How do you spell your name? List G I'll see you right after lunch. See you later. White shoes are awful to keep clean. Stand there and don't move until I tell you! There's a big piece of cake left over from dinner. Wait for me at the corner in front of the drugstore. It's no trouble at all. Hurry up! The morning paper didn't say anything about rain this afternoon or tonight. . The phone call's for you. OGDVOSU'l-waI-J 1—0 O 63 List J Breakfast is ready. I don't know what's wrong with the car, but it won't start. It sure takes a sharp knife to cut this meat. I haven't read a newspaper since we bought a television set. Weeds are spoiling the yard. Call me a little later. 00 you have change for a five—dollar bill? How are you? I'd like some ice cream with my pie. . I don't think I'll have any dessert. APPENDIX B RCID Sentence Lists B,F,G,I, and J O O O Omem-fiwNO-i H O mNmm-bWNi-l 10. 64 List 1—8 The water's too cold for swimming. Why should I get up so early? Shine your own shoes this time. It's raining right here in the room. Where are you going this morning? You should come here when I call. Don't try to get out of it. We let little children go to the movies. There isn't enough paint to finish. . Do you want eggs for breakfast? List 1-F Music always makes me cheer up. My brother's in town for a short while. We live a few miles off the main road. This suit needs to go to the cleaners. They ate enough green apples. Have you been sick all this week? Where have you been working lately? There's not enough table room in the kitchen. It's hard to see where he is. Look out for new business. I-OWVO‘U‘I-hOONI-I I—I O H . moowmm-hwm g...- C 65 List I-G I'll see you right after lunch. I'll see you later this afternoon. White shoes are awful to keep clean. You stand over there until I move. There's a piece of cake left for dinner tonight. Don't wait for me at the front corner. It's no trouble at all to tell. Hurry up with the morning paper. It didn't say anything about a big rain. . That drugstore phone call's for you. List l—I Where can I find a place to park? I like those big red apples. You'll get fat by eating candy. The color show's over in the fall. Why don't they paint their other walls? How come you always get to go first? What are you hiding under your coat? I should always buy new cars. What's wrong with sugar and cream in my coffee? . I'll wait just one minute. 66 List IeJ But we won't be ready to start. I don't know what's wrong with the car. It sure takes a sharp knife to cut meat. I haven't read a newspaper since we got television. The weeds are spoiling this yard. Call me a little later for breakfast. 00 you have change for a five-dollar bill? How are the things we bought? homflmm-thl—l I'd like some ice cream with my pie. 10. I don't think I'll have dessert. APPENDIX C Five Lists of Third- order Sentential Approximations (Seven Word Length) Derived from a Pool of Words Taken from the CID Sentence Lists (Silver- man and Hirsh,1955). O I O OmflwthNi—O I H 53 O i O I mNmm-wab—i 10. 67 List I catch little news here only minutes after spoiling apples fall late tonight if music quiet people think hard so please me breakfast is before them fell first under while a change tonight so you see ate breakfast around people it's only music window is under the main newspaper box after working tonight while they're spoiling music white teeth are clean but it isn't coat isn't dirty with egg dessert would List 2 windows once up at night after you kitchen floors are always open my letter cup is white and with hate here cellar window that's dirty if she calls not here so please don't believe I'd it's much too little help until now only pe0ple clean floors lately change themselves exercise once again they left magazines on corner there's much changing room takes up I'm home by my kitchen floors take 01-9-me WWNO‘U‘I-fiMNO—I 10. 68 List 3 working dentist isn't careful in changing time ready in our house is miles away business is just a bus left early dentist takes much paint fell outside and an egg white is fooling everyone every people take calls every box ought to tonight this town newspaper that's nice that across the park would name everyone after wash their dirty children didn't come across there are windows which lately everyone left List 4 she'll hate working with magazines around the cleaner glasses will make good pie won't a morning will come to my little business time with my package is here just when I'm ready right now he everybody was fishing by water's slow but like my television business is spoiling her everyone should exercise my favorite cup is I'll hurry to find him before any try my brother's television did break everybody o ‘OQNOithONG-l 10. 69 List 5 but everyone can't be careful not until around work everyone because they were spoiling make dessert tonight but before I'm late stream water outside doesn't suit my dentist put it in this house across their coat is gone until dark stay here hate isn't enough room for working until kitchen table will look short until she watch until he needs these apples from believe children that look used lately in -APPENDIX D Instructions Given To Listeners 70 This is a test of listening discrimination. You are going to hear a tape of fifteen lists of sentences. There are ten sentences within each list. You are asked to listen carefully to each sentence and write down in the appropriate space on the paper provided exactly what you heard. After each sentence you will be given fifteen seconds in which to write out your response. Be sure that the number of your response corresponds to the number of the stimulus sentence you are listening to. If you are unsure of a word or a group of words, don't be afraid to guess. I want to emphasize that some of the sentences you will hear may only resemble real English sentences, but actually they may have no meaning at all. Please listen to all sentences very carefully, writing down exactly what you hear, even if you hear only part of a sentence. No sentences will be repeated. After the 50th and 100th sentences you will be given a short break. If at any time during the test procedure you have a question raise your hand and someone will be in to help you. If you have no questions, we'll begin now. APPENDIX E Octave Band Data for Ambient Noise Measured in the Test Booth. 71 Sound Pressure Level per octave band of ambient noise as measured in the listener's room of a two-room prefabricated double-walled test chamber (IAC 1200 Series). Measurement via Bruel and Kjaer type 2204 sound level meter. C Scale: 45 dB SPL Octave Band Center Frequency in Hz dB SPL 31.5 44 63 35 125 35 250 . . 16 500 5 1000 5 2000 3 4000 11 8000 8 16000 5 31,500 8 APPENDIX F Cell Mean Intelligibility Scores for CID, RCID, and Sentential Approxi- mations Sentential Stimuli for Four Ratios of Time Compression (0%, 40%, 60%, and 70%) Presented at Two Sensation Levels (24 dB SL and 40 dB SL). 72 H mHm>NNN44