충남방언 화자에 의한 /오/와 /우/의 청지각적 분류

다국어 초록 (Multilingual Abstract)

Previous studies in vowel production and perception on korean vowels revealed that there is considerable overlap between the vowel spaces of korean vowel [o] and [u] (Yang, 1992; Seong, 2004; Moon, 2007, inter alia). The present research aimed at investigating the boundary in perception between these two vowels.
The material consisted of a continuum of eleven stimuli between the vowels [o] and [u]. Praat 5.3.23 was used for manipulation of vowel formants and the glottal pulse was extracted from male speech voice. With this method the first two formants were manipulated while F3 and F4 remained constant. The stimuli were numbered from 0 to 10. F1 and F2 for stimulus0 were 300 Hz and 750 Hz, and for stimulus10 were 400 Hz and 650 Hz respectively.
Native speakers of Korean consist of 16 female and 16 male were recruited for the experiments. All participants were between the ages of 18 and 29. They were undergraduate or graduate students either at Chungnam National University or Kongju National University. For all participants the length of residence in Chungnam and/or Daejeon province was longer than 18 years. They had no known speech or hearing impairment. All participants were paid at the end of the session.
The experiment consisted of three tasks, i.e. (a) task0: /o/ identification task(yes-no), (b) task1: /u/ identification(yes-no), and (c) task2: /o/-/u/ classification. Task2 involved a forced choice between two alternatives, the vowels /o/ and /u/. The tasks were performed in two fixed orders: for the group0 the order of the tasks was task0-task1-task2, and for the group1 task1-task0-task2.
There was no response time limit, yet the participants were encouraged to respond as fast as possible. In each task, each stimulus had to be identified 15 times in a random order. Single stimulus was played on each trial and the participant had to identify it by mouse-clicking on a response field.
The data of first and final blocks of each task were eliminated before the statistic analysis. ROC analysis for stimuli with response as the classification variable showed that the criterion for vowel /o/ was ≤ 5 and for /u/ > 5. For the /o/-/u/ classification task, the cutoff value for /u/ was > 5. F1 and F2 for the stimulus5 were 344 Hz and 694 Hz respectively.
Fisher's exact probability test suggested that no significant effect of gender and group on individual criterion, while the difference was significant only in the 'group' variable of task2[F = 12.615, p = .002].
Spearman's rank correlation coeffcient was used to assess the relationship between the stimuli and the rank of response time. The test revealed that there were significant correlation on the data sets of both male and female participants for task0 and task1. The significance for group0 data set was present only in task0, and for group1 only in task1.

목차 (Table of Contents)

• 1 서론 1
• 1.1 연구의 필요성 및 목적 1
• 1.2 연구문제 2
• 2 이론적 배경 3
• 2.1 모음 포먼트 3
• 2.2 한국어 단모음 /오/와 /우/의 음향적 특성 5
• 2.3 청지각 특성 10
• 2.4 합성음성 12
• 2.5 지각실험설계 13
• 2.6 ROC 분석 15
• 2.6.1 신호탐지이론 15
• 2.6.2 ROC 곡선 18
• 3 연구방법 21
• 3.1 참가자 21
• 3.2 실험자극 22
• 3.3 실험방법 25
• 3.4 결과분석 26
• 4 연구결과 27
• 4.1 자극에 따른 반응의 ROC 분석 27
• 4.2 성별 및 집단에 따른 규준값 교차분석 33
• 4.3 반응시간와 자극의 상관계수 36
• 5 논의 및 결론 38
• 참고문헌 41
• 영문초록 49
• A 부록 51
• A.1 실험참가자 51
• A.2 과제 및 집단별 ROC 곡선 53
• A.3 실험참가자별 Youden index 및 규준값 61
• A.4 성별 및 집단에 따른 규준값 빈도 도표 64
• A.5 반응시간 기술통계량 및 히스토그램 68
• A.6 자극별 반응시간 구간순위 빈도 71