Psychometrically Equivalent Korean Monosyllabic Speech Discrimination Materials Spoken by Male and Female Talkers Richard W. Harris, Eunoak Kim and Dennis L. Eggett (Brigham Young University, Provo, Utah, USA) Richard W. Harris, Eunoak Kim and Dennis L. Eggett. Psychometrically Equivalent Korean Monosyllabic Speech Discrimination Materials Spoken by Male and Female Talkers. 언어청각장애연구, 2003, 제 권, 제 1 호, 217-243. This investigation was undertaken to develop, digitally record, evaluate, and psychometrically equate Korean monosyllabic word lists for use in measurement of auditory word recognition. Two hundred fifty monosyllabic words were selected from a Korean frequency usage dictionary: Frequency Analysis of Korean Morpheme and Word Usage (Kim & Kang, 2000). These words were digitally recorded by both male and female talkers native to Seoul, Korea. Percent correct word recognition was measured for each word at intensity levels from 5 to 40 dbhl in 5dB increments using 20 normally hearing subjects. Difficulty rankings were calculated for each of the 250 monosyllabic words. The 200 monosyllabic words that were easiest to identify were selected for inclusion in the final compact disc recordings. Four equivalent word lists of 50 words each and eight half-lists of 25 words each were formed from the selected monosyllabic words. A chi-square analysis revealed no statistically significant differences in audibility among the lists or half-lists. In order to increase homogeneity of audibility of the lists and half-lists, the thresholds of the monosyllabic half-lists were adjusted so that the threshold of each list was equal to the midpoint (.1 dbhl) between the mean threshold of the male half-lists and the mean threshold of the female half-lists. No attempt to phonemically balance the 4 lists, or half-lists was made. The concept of phonemic balance is briefly discussed. The psychometrically equivalent Korean monosyllabic word recognition lists, spoken by both male and female talkers, are included on the Brigham Young University Korean Speech Audiometry Materials (Disc 1.0) compact disc. Key Words: word recognition, speech discrimination, Korean, equivalent, compact disc, psychometric function, homogeneity, monosyllabic Ⅰ. Introduction Auditory word recognition tests are an important diagnostic tool used during audiological testing. Routine comprehensive audiological evaluations are generally considered 217
언어청각장애연구 incomplete without measurement of auditory word recognition using speech stimuli. The auditory word recognition score is the percentage of a word list that is correctly repeated at a suprathreshold level. There are a number of factors that have been identified which can influence the word recognition scores (WRS) including word selection and familiarity, usage frequency, and presentation level (Beattie, Svihovec & Edgerton, 175; Campbell, 5; Hood & Poole, 10; Pisoni, 15), talker dialect, and lexical neighborhood (Brandy, 6; Cambron, Wilson & Shanks, 11; Hood & Poole, 10; Kreul, Bell & Nixon, ; Luce, ; Penrod, 17), number of words included in the list (Elpern, 1; Grubb, 3a, 3b; Resnick, 2), method of presentation (Beattie, Svihovec & Edgerton, 175; Brandy, 6; Creston, Gillespie & Krohn, 6); and type of recording (Kamm et al. 10; Ridgway, ). There has also been considerable discussion about the question of phonetic balance (Eldert & Davis, 151; Lehiste & Peterson, 15; Martin, Champlin & Chambers, 1). Recently, Martin, Champlin & Perez (2000) concluded that whether a word list is phonetically balanced or not does not appear to influence word recognition scores for patients with normal hearing or sensorineural hearing impairment. They concluded that the total score based on randomly selected words is not substantially different from total score based on carefully selected, PB word lists (Martin, Champlin & Parez, 2000). It has generally been concluded that recorded presentation of speech audiometry materials is preferable to monitored live voice presentation (American Speech-Languaqge- Hearing Association [ASHA], 1). Recorded presentation of speech materials standardize the composition and presentation of the materials and allow for better control of the presentation intensity of the test stimuli and insure that the speech pattern of the recorded talker will be consistent from one client to the next, and from one clinic to the next. It is important to consider several factors whenever developing speech audiometry materials. Words selected should be both familiar and moderately difficult to identify (Campbell, 5; Comstock & Martin, 14; Weisleder & Hodgson, 1; Zakrzewski et al., 175). Talkers used to make recordings should use the standard or most common dialect of that language (Weisleder & Hodgson, 1). If the word lists are to be recorded, a 21
Psychometrically Equivalent Korean Monosyllabic Speech Discrimination Materials Spoken by Male and Female Talkers digital recording method should be used. Digital recordings offer numerous advantages over tape recordings. These advantages include improved signal-to-noise ratio; increased channel separation, dynamic range, and frequency response; reduced harmonic distortion; elimination of wow and flutter associated with tape playback mechanisms; longer storage life without degradation through use (Kamm et al, 10; Nakamihi, n.d.; Ridgway, ; Sony, 11). Perhaps one of the greatest advantages of a digital recording is that with the use of computers, the digital signal can be modified in a highly efficient and uniform manner including random access to tracks and even randomization of word order in lists using custom software (Harris et al., 2001; Kamm et al., 10; Ridgway, ). To date there are no high quality digital recordings of speech stimuli that can be used to obtain measures of auditory word recognition in individuals whose native language is Korean. The purpose of the present investigation was to develop and evaluate high quality digital recordings of speech stimuli that can be used to measure auditory word recognition in patients whose native language is Korean. Psychometrically equivalent phonemically balanced lists (50 words each) and half-lists (25 words each) will be constructed using male and female Korean talkers. Ⅱ. Method Subjects All subjects participating in this study were natives of Korea. A total of 20 subjects ( male & female), ranging in age from 1 to 2 years (M = 22.7 years), participated in the evaluation of the bisyllabic words. Summary statistics of the subject thresholds are listed in Table 1. Each participant had puretone air-conduction thresholds of dbhl at octave and midoctave frequencies from 5 to 000 Hz and had static acoustic admittance between 0.3 and 1.4 mmhos with peak pressure between 0 and +50 dapa (ASHA, 10; Roup et al., 1). 21
언어청각장애연구 Materials Word lists. Monosyllabic words were selected as the stimuli for auditory word recognition testing. Initially, 260 monosyllabic words were selected for recording from a list of frequently used Korean words (Kim & Kang, 2000). From this list of 260 monosyllabic words that were recorded, 250 monosyllabic words were selected for evaluation in this study. Words that had poor recordings were eliminated from evaluation. In addition, when selecting the monosyllabic words, words that had the same pronunciation but had different meanings and spellings were avoided. Talkers. Initial test recordings were made using seven native Korean-speaking individuals, three males and four females. All talkers were native Koreans, from the Seoul area, and spoke the standard Korean dialect. After the initial recordings were made, a panel of ten Korean judges evaluated the performance of each female talker while a panel of fifteen Korean judges evaluated the performance of each male talker. Both panels of judges were asked to indicate whether the vocal quality and accent of the talker is acceptable or unacceptable and then they were asked to rank order the talkers from best to worst on evaluation forms. The highest ranked talkers (1 male & 1 female) were selected as the talkers for all subsequent recordings. Recordings. All recordings were made in a large anechoic chamber located on the Brigham Young University campus in Provo, Utah, USA. A Larson-Davis model 2541 microphone was positioned at a 0 azimuth and was covered by a 7.62 cm windscreen. The microphone was connected to a Larson-Davis model 00B microphone preamp, which was coupled to a Larson-Davis model 2200Cpreamp power supply. The signal from the preamp power supply was then routed through an Apogee AD-00024-bit analog-todigital converter; the digitized signal was stored on a hard drive for later editing. A 44.1 khz sampling rate with 24-bit quantization was used for all recordings, and every effort was made to utilize the full range of the 24-bit analog-to-digital converter. Ambient noise levels in the anechoic chamber were < 0dBSPL which allowed a signal-to-noise ratio of at least 65 db during recording. During the recording sessions, the talker was asked to pronounce each word several 220
Psychometrically Equivalent Korean Monosyllabic Speech Discrimination Materials Spoken by Male and Female Talkers times. A native Korean judge rated each word for perceived goodness of production, and the best production of each word was then selected for inclusion on the Korean speech audiometry test CD. Any word that had a clipped waveform was discarded. After the rating process, the intensity of each word to be included on the CD was edited using Sadie Disk Editor software (Studio Audio and Video Limited, ) to yield the same intensity as that of the 00 Hz calibration tone contained on the CD. The CD containing the final edited words was produced on a PlexWriter /4/32 CD-R/W drive using a 44.1 khz sampling rate and -bit quantization. The NS high dither option in the Sadie Disk Editor software was used to convert the recordings from 24 to bit quantization. Procedures Custom software was used to control randomization and timing of the presentation of the words. The signal was routed from a computer-controlled CD-ROM drive to the external inputs of a Grason Stadler model 1761 (GSI-61) audiometer. The stimuli were routed from the audiometer to the subject via TDH-50P headphones. Prior to testing each subject, the inputs to the audiometer were calibrated to 0 VU using the 00 Hz calibration tone on track 1 of the Korean CD. All testing was carried out in a sound suite that met American National Standards Institute [ANSI] standards (11) for maximum permissible ambient noise levels for the ears not covered condition. Evaluation of monosyllabic words. The subjects were not familiarized with the monosyllabic words prior to testing. The 250 monosyllabic words were divided into ten lists of 25 words each. Ten presentation levels were selected at which to present the lists: -5 to 40 dbhl in 5dB steps. One list was presented at each of the ten presentation levels. The order of the presentation of the lists and the order of the words within the list were randomized for each subject. Each word was presented an equal number of times at each intensity level across the entire subject population. Prior to the administration of the speech discrimination test, the following instructions were given: (Korean) 다양한음향높이에서일음절 ( 한글자 ) 의단어를들으시겠습니다. 아주작고조용한높이에 221
언어청각장애연구 서는단어를알아듣기힘들것입니다. 단어가주어질때마다불빛으로신호해드리겠습니다. 집중해서잘들어주시고, 들으신단어를준비된응답지에정자로바르게써주십시요. 들으신단어가확실하지않을때는추측해도좋습니다. 추측할단어가없으면빈칸으로남겨주십시요. 질문있습니까? (English) You will hear monosyllabic words (1 syllable) at a number of different loudness levels. At the very soft loudness levels it may be difficult for you to hear the words. An indication light will come on, informing you that a word has been presented. Please listen carefully and print legibly the words you hear in the spaces provided on the response sheets. If you are unsure of the word, you are encouraged to guess. If you have no guess, please draw a blank line in the space provided for that word. Do you have any questions? Calibration The audiometer was calibrated prior to, weekly during, and at the conclusion of data collection. Calibration was performed in accordance with the specifications of the ANSI S3.6 (). No changes in calibration were necessary throughout the course of data collection. Ⅲ. Results After the raw data were compiled, each monosyllabic word received a difficulty ranking. The ranking was based on the number of times each word was correctly identified across all intensity levels and subjects. The more often the word was identified correctly, the higher the ranking. Table 2 lists the difficulty ranking for each monosyllabic word from the male recording and Table 3 lists the difficulty ranking for each monosyllabic word from the female recording. On receiving a difficulty ranking, the 200 words with the highest ranking were divided into four lists of 50 words each. Table 4(male) and Table 5(female) contain the four phonemically balanced lists for each gender. The first four words from the rank-ordered list of 200 words were randomly assigned to one of the four lists. This process was repeated 222
Psychometrically Equivalent Korean Monosyllabic Speech Discrimination Materials Spoken by Male and Female Talkers until each list contained 50 words. Eight half-lists of 25 words each were constructed after the creation of the four phonemically balanced lists. The eight male half-lists are found in Table 6; the eight female half-lists are found in Table 7. Two half-lists were formed from each full list by randomly designating the first word in a list as either an A or a B, designating the second word with the letter that was opposite from that assigned to the first word, and then counterbalancing the assignment of the remaining words. Once all words were assigned a letter, the full list could be divided into two half-lists: half-list A and half-list B. Once the monosyllabic lists and half-lists were created, the raw data were used to create psychometric functions for each of the lists and half-lists for both the male and female talkers. The raw data were reanalyzed using logistic regression to obtain regression slope and regression intercept values for each of the 4 lists and each of the half-lists for both the male and female talker recordings. The values obtained for the regression slope and regression intercept for each list and half-list are presented in Table (male talker) and Table (female talker). The calculated regression slope and regression intercept values were then inserted into a modified logistic regression equation (Equation 1) that was designed to calculate percent correct performance at any specified intensity level. %= ( 1- exp ( a+b db) 1+exp(a+ b db) ) 0 (Equation 1) In Equation 1, a is the regression slope, b is the regression intercept, and db is the presentation intensity level in dbhl. For a more detailed discussion of the derivation of Equation 1 the reader is referred to Harris et al. (2001). By inserting the regression slope, regression intercept, and intensity level into Equation 1, it is possible to predict the percent correct word recognition at any specified intensity level. Percent correct word recognition was predicted for each of the bisyllabic lists and half-lists for a range of - to 40 dbhl in 2dB increments. Psychometric functions were then produced using the predicted percentages. The threshold (presentation intensity 223
언어청각장애연구 required for 50 % word recognition performance), the slope at threshold, and the slope from 20 to 0 % were calculated for the bisyllabic lists and half-lists by inserting the desired proportions into Equation 2. The data for the threshold, slope at threshold, and slope from 20 to 0 % for each psychometric function are presented in Table (male) and Table (female). db = log p 1-p -a b (Equation 2) After the lists and half-lists were compiled, a chi-square (χ²) analysis was performed in order to determine if there were any statistically significant differences among the monosyllabic full=lists or half-lists. No significant differences were found among the four full-lists for the male talker(χ² = 0.72, p = 0.7) or the female talker(χ² = 0.34, p= 0.5). There were also no significant differneces found among the half-lists for male talker(χ² = 2.0, p=0.5) or the female talker(χ²=.23, p=0.31). While there were no statistically significant differences among the lists or half-lists, intensity level adjustments were made digitally using Sadie Disk Editor software (Studio Audio and Video Limited, ) in an attempt to increase the psychometric equivalency of the lists and half-lists. The thresholds of each word in the bisyllabic lists (4 male & 4 female) and the bisyllabic half-lists ( male & female) were digitally adjusted so that the 50 % threshold of each list was equal to the approximate midpoint (.1 dbhl) between the mean threshold of the eight male half-lists and the mean threshold of the eight female half-lists. The intensity adjustments made to each word in the 4 lists and half-lists are presented in Table (male) and Table (female). The adjusted psychometric functions for the adjusted male talker bisyllabic lists and half-lists after intensity adjustment are presented in Figures 1 and 2; the psychometric functions for the female talker bisyllabic lists and half-lists after intensity adjustment are presented in Figures 3 and 4. Figure 5 contains mean psychometric functions for the combined male and combined female bisyllabic lists both before and after intensity adjustment to equate performance. Inspection of Figure 5 224
Psychometrically Equivalent Korean Monosyllabic Speech Discrimination Materials Spoken by Male and Female Talkers indicates that the predicted psychometric functions were identical for the male and female talker lists after making intensity adjustments to equate performance. It should be pointed out that only minor (= 0.5 db) adjustments to the words in the lists were required to equate the lists for both male and female talkers. Ⅳ. Discussion The main purpose of this study was to develop a set of homogeneous Korean monosyllabic word lists for use in measuring auditory word recognition. Inspection of Figures 1-5 indicate that we have been able to develop a set of lists and sub lists which have very homogeneous performance with respect to audibility and psychometric function slope. A chi-square analysis was performed to determine whether there were any statistically significant differences among the monosyllabic lists or half-lists. No significant differences were found among the four lists or the eight half-lists, as can be seen in Tables (male) and (female). Slopes from 20 to 0 % for the monosyllabic lists and half-lists ranged from 4.2 to 4. % / db (M = 4.4 % / db) for the male recordings and from 3. to 5.6 % / db (M = 4.4 % / db) for the female recordings. Others have reported means for English auditory word recognition materials that are very similar to the current Korean recordings. Beattie, Edgerton & Svihovec (177) reported a mean slope of 4.2 % / db for the NU-6 word lists and a mean slope of 4.6 % / db for the CID W-22 word lists. Wilson & Oyler (17) found the following when evaluating the recordings from the Auditec of St. Louis CD: 4.4 % /db(nu-6 word lists) and 4. % / db (CID W-22 word lists). The present investigation made no specific attempt to phonemically balance the 4 speech discrimination lists or the half-lists. Kim & Kang (17) present detailed description of the occurrence of initial consonants, final consonants, and vowels in Korean. Their sample size was very large. Any attempt to phonemically balance speech discrimination word lists containing 25 or 50 words will be hindered because so few words 225
언어청각장애연구 are in each list and the proportion of occurrence of phonemes in a list is difficult, if not impossible, to match to the occurrence of phonemes in the language as a whole. While there have been previous attempts in English to develop phonemically balanced lists, these attempts have, by and large, been unsuccessful due to this and other reasons. In addition, while there are significant coarticulation effects in running speech, these effects are minimal in isolated words. This also would have a dramatic effect upon any attempt to develop a list of 50 words whose proportion of occurrence of phonemes matches the proportion of occurrence of phonemes in the language as a whole. While an initial phoneme of a word in a list is truly an initial phoneme, it is not an initial phoneme in running speech because of coarticulatory influences. A great deal of research remains to be done in the field of Korean speech audiometry materials. Future research could examine the similarities between the mean SRT obtained with the 25 adjusted bisyllabic words from this study and the mean PTA of the test subjects. More research could also be done with the monosyllabic lists. A comparison could be made between the auditory word recognition score obtained when the ten most difficult words of a list are presented and the score obtained when the whole list is presented. If the score obtained when the ten most difficult words are presented is similar to the score obtained when the whole list is presented, test time could be shortened by presenting only the ten most difficult words. Future research could also include examining list and half-list equivalency for hearing impaired individuals with varying types and degrees of hearing impairment. In addition to the research that can be conducted on the current Korean speech audiometry materials, there is also a need to develop additional Korean speech materials for a variety of purposes. For example, speech materials could be created for children on the basis of word familiarity. Sentence materials similar to the HINT test (Nilsson, Soli & Sullivan, 14) could be developed and be of benefit in hearing aid and other types of evaluations. There is also a need to develop high-quality recordings of Korean speech materials used in aural rehabilitation for those with cochlear implants. It takes a great deal of time to identify, record and evaluate new digitally recorded speech materials but 226
Psychometrically Equivalent Korean Monosyllabic Speech Discrimination Materials Spoken by Male and Female Talkers the benefits are considerable in terms of time saved during evaluations and test-retest reliability across different clinical facilities if they all utilize the same recorded materials. Having speech audiometry materials on compact disc also provides the audiologist with the ability to select which words or test stimuli to administer from a longer list of items and allows for varying interstimulus intervals and random orders under computer control. In summary, we have been able to develop digitally recorded monosyllabic Korean auditory word recognition lists and half-lists that are very homogeneous with respect to audibility and psychometric function slope for both male and female talkers. These lists can be used to evaluate auditory word recognition in individuals whose native language is Korean. The monosyllabic lists and half-lists for both the male and female talkers are contained on the CD entitled Brigham Young University Korean Speech Audiometry Materials (Disc 1.0). 227
언어청각장애연구 References 김흥규 강범모 (17). 한글사용빈도의분석. 서울 : 고려대학교민족문화연구소. 김흥규 강범모 (2000). 한국어형태소및어휘사용빈도의분석. 서울 : 고려대학교민족문화연구소. American National Standards Institute (11). American National Standard maximum permissible ambient noise levels for audiometric test rooms (ANSI S3.1-11). New York: ANSI. American National Standards Institute (). American National Standard specifications for audiometers (ANSI S3.6-). New York: ANSI. American Speech-Language-Hearing Association Committee on Audiologic Evaluation (1). Guidelines for determining threshold level for speech. Asha, 30, 5-. American Speech-Language-Hearing Association (10). Guidelines for screening for hearing impairments and middle ear disorders. Asha, 32(Supplement 2), 17-24. Beattie, R. C., Edgerton, B. J. & Svihovec, D. V. (177). A comparison of the Auditec of St. Louis cassette recordings of NU-6 and CID W-22 on a normal-hearing population. Journal of Speech and Hearing Disorders, 42, 60-64. Beattie, R. C., Svihovec, D. V. & Edgerton, B. J. (175). Relative intelligibility of the CID spondees as presented via monitored live voice. Journal of Speech and Hearing Disorders, 40, 4-1. Brandy, W. T. (6). Reliability of voice tests of speech discrimination. Journal of Speech and Hearing Research,, 461-465. Cambron, N. K., Wilson, R. H. & Shanks, J. E. (11). Spondaic word detection and recognition functions for female and male speakers. Ear and Hearing,, 64-70. Campbell, R. A. (5). Discrimination test word difficulty. Journal of Speech and Hearing Research,, -22. Comstock, C. L. & Martin, F. N. (14). A children s Spanish speaking word discrimination test for non-spanish-speaking clinicians. Ear and Hearing, 5, 6-170. Creston, J. E., Gillespie, M. & Krohn, C. (6). Speech audiometry: Taped vs live voice. Archives of Otolaryngology, 3, 40-43. Eldert, E. & Davis, H. (151). The articulation function of patients with conductive deafness. Laryngoscope, 61, 1-0. Elpern, B. S. (1). The relative stability of half-list and full-list discrimination tests. Laryngoscope, 71, 30-35. Grubb, P. (3a). Phoneme analysis of half-list speech discrimination tests. Journal of Speech and Hearing Research, 6, 271-275. Grubb, P. (3b). Considerations in the use of half-list speech discrimination tests. Journal of Speech and Hearing Research, 6, 24-27. Harris, R. W., Goffi, M. V. S., Pedalini, M. E. B., Gygi, M. A. & Merrill, A. (2001). Psychometrically equivalent brazilian portuguese trisyllabic words spoken by male and female talkers. Pro- Fono, (1), 37-53. 22
Psychometrically Equivalent Korean Monosyllabic Speech Discrimination Materials Spoken by Male and Female Talkers Hood, J. D. & Poole, J. P. (10). Influence of the speaker and other factors affecting speech intelligibility. Audiology, 1, 434-455. Kamm, C., Carterette, E. C., Morgan, D. E. & Dirks, D. D. (10). Use of digitized speech materials in audiological research. Journal of Speech and Hearing Research, 23, 70-721. Kreul, E. J., Bell, D. W. & Nixon, J. C. (). Factors affecting speech discrimination test difficulty. Journal of Speech and Hearing Research,, 21-27. Lehiste, I. & Peterson, G. (15). Linguistic considerations and intelligibility. Journal of the Acoustical Society of America, 31, 20-26. Luce, P. A. (). A computational analysis of uniqueness points in auditory word recognition. Perceptual Psychophysiology, 3, 5-. Martin, F. N., Champlin, C. A. & Chambers, J. A. (1). Seventh survey of audiometric practices in the United States. Journal of the American Academy of Audiology,, 5-4. Martin, F. N., Champlin, C. A. & Perez, D. D. (2000). The question of phonetic balance in word recognition testing. Journal of the American Academy of Audiology, (), 4-43. Martin, F. N. & Sides, D. G. (15). Survey of current audiometric practices. Asha, 27, 2-36. Nakamichi. (n.d.) Nakamichi MR-1 discrete head professional cassette deck owner s manual. Japan: Nakamichi Corporation. Nilsson, M. J., Soli, S. D. & Sullivan, J. (14). Development of the Hearing In Noise Test for measurement of speech reception thresholds in quiet and noise. Journal of the Acoustical Society of America, 5(2), 5-. Penrod, J. P. (17). Talker effects on word-discrimination scores of adults with sensorineural hearing impairment. Journal of Speech and Hearing Disorders, 44, 340-34. Pisoni, D. (15). Speech Perception. Journal of the Acoustical Society of America, 7, 31-3. Resnick, D. (2). Reliability of the twenty-five word phonetically balanced lists. Journal of Auditory Research, 2, 5-. Ridgway, J. (). Compact disks - A revolution in the making. Canadian Library Journal, 43(1), 23-2. Roup, C. M., Wiley, T. L., Safady, S. H. & Stoppenbach, D. T. (1). Tympanometric screening norms for adults. American Journal of Audiology, 7, 55-60. Sony. (11). Compact disc player operating instructions. Japan: Sony Corporation. Studio Audio and Video Limited. (). Sadie Disk Editor (Version 3.0) [Computer Software]. Weisleder, P. & Hodgson, W. R. (1). Evaluation of four Spanish word-recognition-ability lists. Ear and Hearing,, 37-33. Wilson, R. H. & Oyler, A. L. (17). Psychometric functions for the CID W-22 and NU Auditory Test No. 6 materials spoken by the same speaker. Ear and Hearing, 1, 430-434. Zakrzewski, A., Jassem, W., Pruszewicz, A. & Obrebowski, A. (175). Identification and discrimination of speech sounds in monosyllabic meaningful words and nonsense words by children. Audiology,, 21-26. 22
언어청각장애연구 0 Percent Correct Recognition (%) 0 0 70 60 50 40 30 20 1 2 3 4 0 - -4 0 4 20 24 2 32 36 40 dbhl Figure 1. Psychometric functions for the four male monosyllabic Korean lists. 0 Percent Correct Recognition (%) 0 0 70 60 50 40 30 20 1A 1B 2A 2B 3A 3B 4A 4B 0 - -4 0 4 20 24 2 32 36 40 dbhl Figure 2. Psychometric functions for the eight male monosyllabic Korean half-lists. 230
Psychometrically Equivalent Korean Monosyllabic Speech Discrimination Materials Spoken by Male and Female Talkers Percent Correct Recognition (%) 0 0 0 70 60 50 40 30 20 1 2 3 4 0 - -4 0 4 20 24 2 32 36 40 dbhl Figure 3. Psychometric functions for the four female monosyllabic Korean lists. 0 Percent Correct Recognition (%) 0 0 70 60 50 40 30 20 1A 1B 2A 2B 3A 3B 4A 4B 0 - -4 0 4 20 24 2 32 36 40 dbhl Figure 4. Psychometric functions for the eight female monosyllabic Korean half-lists. 231
언어청각장애연구 0 0 Percent Correct Recognition (%) 0 70 60 50 40 30 20 male before adjustment female before adjustment male after adjustment female after adjustment 0 - -4 0 4 20 24 2 32 36 40 dbhl Figure 5. Mean psychometric functions for male and female monosyllabic lists: before and after intensity adjustment. Table 1. Age (in years) and Pure Tone Threshold (dbhl) Descriptive Statistics for the 20 Korean Participants M Minimum Maximum SD Age 22.7 1 2 3. 5 Hz 6.3 0 4. 250 Hz 4.3-5 4.4 500 Hz 4.0-5 4.2 750 Hz 4.0-5 4.2 00 Hz 3.5 0 3.7 00 Hz 4. -5 5.0 2000 Hz 5.5 0 4.6 3000 Hz 3. -5 3.6 4000 Hz 2.0-5 4.1 6000 Hz 2.3-5 5.0 000 Hz 3.3-6.3 232
Psychometrically Equivalent Korean Monosyllabic Speech Discrimination Materials Spoken by Male and Female Talkers Table 2. Korean Monosyllabic Male Words for Speech Discrimination Lists in Rank Order from Easiest to Most Difficult Word Rank Word Rank Word Rank Word Rank 말 17 장 꿀 영 차 쪽 끝 뺨 책 주 구 칠 귀 죽 금 침 회 리 활 춤 천 멍 효 땀 달 명 잠 뚝 돌 나 절 길 깨 날 집 흙 꽃 신 전 쥐 공 쌀 중 맛 향 탕 칼 묵 혀 털 키 문 입 왜 코 면 자 왕 맥 눈 종 월 만 풀 줄 원 망 편 너 위 먹 평 팔 약 미 색 피 옆 남 섬 양 안 널 선 용 방 오 시 배 벼 옷 소 밤 창 산 솔 빵 촌 성 술 초 철 싹 탈 이 층 통 탑 개 도 턱 터 꿈 돈 틈 티 끈 등 육 열 그 은 예 알 힘 강 연 백 형 끌 욕 뱀 (table 2 continues) 233
언어청각장애연구 234 Word Rank Word Rank Word Rank Word Rank 발뽕병닭때딸땅동뒤감김곳국겁관핵한흠일잣죄점매밀밑팥품표사서 쉰실속손수숨쌈탓윷뼈벗총답뜰뜸뜻독간골굴군광귤해흉잔콩목물멋 님놈파살설틀억요비칡댁덫갑짐몸뇌늘상식십솥운엿앞반밭봄분벽별 떡둘글굿못새삼솜숲톱복불흰쇠밥벌곰적셋법넷육박뿔묘년씨북 7 7 7 7 6 6 6 6 6 5 5 5 4 4 3 3 3 2
Psychometrically Equivalent Korean Monosyllabic Speech Discrimination Materials Spoken by Male and Female Talkers Table 3. Korean Female Monosyllabic Words for Speech Discrimination Lists in Rank Order from Easiest to Most Difficult Word Rank Word Rank Word Rank Word Rank 말 1 벌 반 육 팔 17 병 방 욕 영 17 총 차 박 빵 층 창 뼈 동 땅 촌 북 구 도 달 법 장 돌 딸 벼 표 이 뒤 벽 편 은 감 칠 서 개 꽃 철 양 강 길 천 예 꿈 곳 땀 열 혀 한 떡 뱀 잠 힘 돈 별 집 일 김 책 전 자 곰 침 콩 잔 공 끈 만 짐 관 해 맛 점 광 향 날 칼 회 절 너 망 흠 목 널 먹 효 명 오 미 형 옷 평 못 흉 피 십 물 입 산 솔 멍 종 쌀 왜 묘 죽 월 위 살 매 원 약 삼 밑 용 연 식 몸 백 요 숨 멋 밥 앞 솥 면 밭 발 왕 남 (table 3 continues) 235
언어청각장애연구 236 Word Rank Word Rank Word Rank Word Rank 팥풀새설선신터틀턱털틈육알안배봄벗댁답둘끝굿국금겁핵흙잣적눈 키리밀문놈파상성시쉰실소술통윷뽕비초춤때뜰뜸독등끌귤묵넷늘억 품사색섬쇠손탈탕불군탑톱뚝덫간글골그굴귀흰활쪽죄중맥수싹씨 엿밤복분칡뜻갑주뇌솜쌈숲티옆꿀코나년탓뺨깨쥐님속운셋닭뿔줄 7 7 7 7 7 6 6 6 6 6 6 5 3 1
Psychometrically Equivalent Korean Monosyllabic Speech Discrimination Materials Spoken by Male and Female Talkers 237 Table 4. Korean Male Monosyllabic Lists in Rank Order from Easiest to Most Difficult List 1 List 2 List 3 List 4 말돌공혀너밤개힘형리날왕원옆창등끌금집칼만남성통예칠땀문눈색시탑백발동뒤핵잣매밑사숨쌈답간광귤멋파억책천깨종줄양빵그장죽신왜월안철도강구전코맥오산턱욕춤흙묵풀평솔터알뽕딸김관일밀표실수탓총독군해목놈틀귀회꽃입피배이끈쪽멍쌀탕약벼층은끝활잠키먹널싹틈연침길맛면섬소탈열닭땅감국한점팥쉰손윷벗뜸굴흉콩님설차달향자팔용초꿈주명나털위방촌돈꿀효절중망미옷육영뺨뚝쥐편선술티뱀병때곳겁흠죄품서속뼈뜰뜻골잔물살요
언어청각장애연구 23 Table 5. Korean Female Monosyllabic Lists in Rank Order from Easiest to Most Difficult List 1 List 2 List 3 List 4 팔구편뱀책향산원밭땅이잠콩맛솔왜앞창감곳자칼못삼숨박법땀김회입밑면새선털알답굿잣적파상소윷비뜸독눈사영장양예침절쌀월병총개혀집날십연요방달꽃힘망미식솥북칠천돈흠형종멋설터역안벗끝흙키밀시통뽕초춤묵품섬말동표별끈목옷용벌도돌꿈만널오약발차딸길일짐물묘왕욕벽철공관효죽몸풀신턱배댁금겁리문쉰실때뜰끌손탈탕빵서열해명피백밥층은강전너평위반촌뒤한잔점먹멍살육뼈벼떡곰광흉매남팥틀틈봄둘국핵줄놈성술등귤넷늘색쇠
Psychometrically Equivalent Korean Monosyllabic Speech Discrimination Materials Spoken by Male and Female Talkers 23 Table 6. Korean Male Monosyllabic Half-lists in Rank Order from Easiest to Most Difficult 1A 1B 2A 2B 3A 3B 4A 4B 돌공밤개형왕옆창끌집만통칠문눈탑발뒤핵매사쌈간귤억말혀너힘리날원등금칼남성예땀색시백동잣밑숨답광멋파책깨줄빵죽왜안철강코오산춤묵평터알김일표실탓군해놈천종양그장신월도구전맥턱욕흙풀솔뽕딸관밀수총독목틀귀입배이멍쌀벼은끝키널틈침맛면탈열감한점손벗뜸흉설회꽃피끈쪽탕약층활잠먹싹연길섬소닭땅국팥쉰윷굴콩님차자팔꿈명나방돈꿀절미육뺨쥐선티뱀곳겁죄서뜰골물요달향용초주털위촌효중망옷영뚝편술병때흠품속뼈뜻잔살
언어청각장애연구 240 Table 7. Korean Female Monosyllabic Half-lists in Rank Order from Easiest to Most Difficult 1A 1B 2A 2B 3A 3B 4A 4B 팔뱀책산밭이맛솔창곳칼삼박법회입새털답굿적상비뜸사구편향원땅잠콩왜앞감자못숨땀김밑면선알잣파소윷독눈장양침쌀총혀집연방꽃망식솥천흠종설터벗끝키시초품섬영예절월병개날십요달힘미북칠돈형멋역안흙밀통뽕춤묵동표끈옷벌돌만약차딸일묘왕벽공효몸신배겁문실뜰끌탕말별목용도꿈널오발길짐물욕철관죽풀턱댁금리쉰때손탈빵열명백층은전위촌한잔멍육떡곰흉팥틀봄핵놈술등넷쇠서해피밥강너평반뒤점먹살뼈벼광매남틈둘국줄성귤늘색
Psychometrically Equivalent Korean Monosyllabic Speech Discrimination Materials Spoken by Male and Female Talkers Table. Mean Performance of Korean Male Monosyllabic Lists and Half-lists Slope from List a a b b Slope at 50 % c 20 to 0 % d Threshold e Change in db f 1 2.22-0.20 4. 4.2.34 0.3 2 2.33-0.20 5.1 4..41 0.4 3 2.20-0.1 4. 4.2.33 0.3 4 2.3-0.21 5.2 4.5.46 0.4 1A 2.2-0.20 5.0 4.3.4 0.4 1B 2. -0.1 4. 4.2.1 0.1 2A 2.25-0.20 4. 4.3.3 0.3 2B 2.42-0.21 5.3 4.6.42 0.4 3A 2.23-0.1 4. 4.2.4 0.4 3B 2.1-0.20 4. 4.2.1 0.1 4A 2.21-0.1 4. 4.2.4 0.4 4B 2.60-0.23 5.7 4..44 0.4 M 2.2-0.20 5.0 4.4.4 0.3 Minimum 2. -0.23 4. 4.2.2 0.1 Maximum 2.60-0.1 5.7 4..5 0.4 Range 0.44 0.04 0. 0.7 0.3 0.3 SD 0. 0.01 0.3 0.2 0.1 0.1 a a = regression intercept. b b = regression slope. c Psychometric function slope ( % / db) at 50 % was calculated from 4. to 50.001 %. d Psychometric function slope ( % / db) from 200 %. e Intensity required for 50 % intelligibility. f Change in intensity required to adjust the threshold of a word to the mean PTA of the subjects 241
언어청각장애연구 Table. Mean Performance of Korean Female Monosyllabic Lists and Half-lists Slope from List a a b b Slope at 50 % c 20 to 0 % d Threshold e Change in db f 1 2.25-0.21 5.2 4.5.7-0.3 2 2.23-0.21 5.2 4.5.7-0.4 3 2. -0.20 5.0 4.3.6-0.5 4 2. -0.1 4. 4.2. -0.1 1A 2.07-0.1 4. 4.2. -0.3 1B 2.4-0.23 5. 5.0.6-0.4 2A 1. -0.1 4.4 3..6-0.4 2B 2.76-0.26 6.5 5.6.7-0.4 3A 2.23-0.21 5.2 4.5.7-0.3 3B 2.03-0.1 4. 4.2.5-0.6 4A 1.6-0.1 4.6 4.0.7-0.3 4B 2.2-0.21 5.1 4.5.1 0.1 M 2.20-0.21 5.1 4.4.7-0.3 Minimum 1. -0.26 4.4 3..5-0.6 Maximum 2.76-0.1 6.5 5.6.1 0.1 Range 0. 0.0 2.0 1. 0.7 0.7 SD 0.24 0.02 0.6 0.5 0.2 0.2 a a = regression intercept. b b = regression slope. c Psychometric function slope ( % / db) at 50 % was calculated from 4. to 50.001 %. d Psychometric function slope ( % / db) from 200 %. e Intensity required for 50 % intelligibility. f Change in intensity required to adjust the threshold of a word to the mean PTA of the subjects 242
Psychometrically Equivalent Korean Monosyllabic Speech Discrimination Materials Spoken by Male and Female Talkers 국문초록 검사어표간의대등한명료도를갖는한국어음명료도검사어표의제작 Richard W. Harris*, Eunoak Kim*, Dennis L. Eggett** (*Dept. of Audiology and Speech Language Pathology, **Dept. of Statistics, Brigham Young University) 본연구의목적은청력검사시어음명료도측정에사용될한국어단음절검사어표를제작하는것이다. 이연구에사용된 250개의단어들은한국어형태소및어휘사용빈도의분석 ( 김흥규 & 강범모, 2000) 이란자료를통해선택되었다. 검사용단어들은한국어표준말을구사하는각각의남녀에의해디지털방식으로녹음되었다. 각단어들을정상청력자 20명에게 5에서 40 dbhl 까지 5dB 간격으로청취하게한후, 개의자극강도에서명료도검사성적을측정하였다. 250개의단어중명료도성적이좋은 200개의단어를최종컴팩트디스크에포함시켰다. 따라서, 한개의표당단어가 50개씩구성되어있는 4개의검사어표 (full-lists) 와각표를다시둘로나누어 25개의단어를포함하게한 개의검사어표 (half-lists) 가제작되었다. 이검사어표들은표간에대등한명료도성적을갖도록단어의강도를조절하였다. Chi-square 분석은이들검사어표간에통계학적으로차이가없음을나타내었다. 한국단음절검사어표는남녀각각에의해브리감영대학교한국어음청력검사어표 (Disc 1.0) 컴팩트디스크에녹음되었다. 핵심어 : 단어재인, 어음변별, 한국어, 대등한, 컴팩트디스크, 심리측정적기능, 동질성 게재신청일 : 2003 년 1 월 24 일 게재확정일 : 2003 년 2 월 1 일 Richard W. Harris ( 제 1 저자 ): Professor Department of Audiology and Speech Language Pathology, Brigham Young University, e-mail: Richard Harris@byu.edu Eunoak Kim ( 교신저자 ): Dept. of Audiology and Speech Language Pathology, Brigham Young University, e-mail: eunoak@hotmail.com Dennis L. Eggett ( 공동저자 ): Professor Department of Statistics, Brigham Young University, e-mail: collings@byu.edu 243