<%BANNER%>

Acoustic Properties of the Japanese Tap by Non-native as Compared to Native Speakers of Japanese

Permanent Link: http://ufdc.ufl.edu/UFE0041678/00001

Material Information

Title: Acoustic Properties of the Japanese Tap by Non-native as Compared to Native Speakers of Japanese
Physical Description: 1 online resource (54 p.)
Language: english
Creator: Rowlings, Bethany
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: acoustic, acquisition, english, japanese, language, native, non, phonetics, second, tap
Linguistics -- Dissertations, Academic -- UF
Genre: Linguistics thesis, M.A.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Extensive research has been done on native Japanese learners of English as a second language; however, very little work exists on native English learners of Japanese as a second language. It is a well-known fact that native Japanese learners of English have difficulty with English /r/ and /l/ phonemes; however, native English learners of Japanese have a similar problem in reverse: acquiring the Japanese tap. This study aims to examine the acoustic properties of the Japanese tap as spoken by native speakers of Japanese and compare the native production to that of advanced learners of Japanese as a second language who are native speakers of English. Acoustic measurements of duration and number of silent interval (the brief moment of relative silence prior to the release of the tap sound) and frequencies of the second and third formants will be examined to determine which cues are salient for proper production of the Japanese tap. Tap production will be compared not only between language groups, but also in a variety of syllable environments to gain a full picture of how the tap is produced by native speakers of Japanese and examine how well advanced learners have acquired this production. Results showed that advanced learners of Japanese are able to closely approximate native speaker production with limited interference from native English sounds.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Bethany Rowlings.
Thesis: Thesis (M.A.)--University of Florida, 2010.
Local: Adviser: Wayland, Ratree.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0041678:00001

Permanent Link: http://ufdc.ufl.edu/UFE0041678/00001

Material Information

Title: Acoustic Properties of the Japanese Tap by Non-native as Compared to Native Speakers of Japanese
Physical Description: 1 online resource (54 p.)
Language: english
Creator: Rowlings, Bethany
Publisher: University of Florida
Place of Publication: Gainesville, Fla.
Publication Date: 2010

Subjects

Subjects / Keywords: acoustic, acquisition, english, japanese, language, native, non, phonetics, second, tap
Linguistics -- Dissertations, Academic -- UF
Genre: Linguistics thesis, M.A.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract: Extensive research has been done on native Japanese learners of English as a second language; however, very little work exists on native English learners of Japanese as a second language. It is a well-known fact that native Japanese learners of English have difficulty with English /r/ and /l/ phonemes; however, native English learners of Japanese have a similar problem in reverse: acquiring the Japanese tap. This study aims to examine the acoustic properties of the Japanese tap as spoken by native speakers of Japanese and compare the native production to that of advanced learners of Japanese as a second language who are native speakers of English. Acoustic measurements of duration and number of silent interval (the brief moment of relative silence prior to the release of the tap sound) and frequencies of the second and third formants will be examined to determine which cues are salient for proper production of the Japanese tap. Tap production will be compared not only between language groups, but also in a variety of syllable environments to gain a full picture of how the tap is produced by native speakers of Japanese and examine how well advanced learners have acquired this production. Results showed that advanced learners of Japanese are able to closely approximate native speaker production with limited interference from native English sounds.
General Note: In the series University of Florida Digital Collections.
General Note: Includes vita.
Bibliography: Includes bibliographical references.
Source of Description: Description based on online resource; title from PDF title page.
Source of Description: This bibliographic record is available under the Creative Commons CC0 public domain dedication. The University of Florida Libraries, as creator of this bibliographic record, has waived all rights to it worldwide under copyright law, including all related and neighboring rights, to the extent allowed by law.
Statement of Responsibility: by Bethany Rowlings.
Thesis: Thesis (M.A.)--University of Florida, 2010.
Local: Adviser: Wayland, Ratree.

Record Information

Source Institution: UFRGP
Rights Management: Applicable rights reserved.
Classification: lcc - LD1780 2010
System ID: UFE0041678:00001


This item has the following downloads:


Full Text

PAGE 1

1 ACOUSTIC PROPERTIES OF THE JAPANESE TAP BY NON-NATIVE AS COMPARED TO NATIVE SPEAKERS OF JAPANESE By BETHANY ROWLINGS A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLOR IDA IN PARTAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ARTS UNIVERSITY OF FLORIDA 2010

PAGE 2

2 2010 Bethany Rowlings

PAGE 3

3 ACKNOWLEDGEMENTS I thank my professors fo r their endless assistance and knowledge on this project, and my friends for their support and help in finding participants for my research.

PAGE 4

4 TABLE OF CONTENTS page ACKNOWLEDG EMENTS ............................................................................................... 3 LIST OF TABLES............................................................................................................ 6 LIST OF FIGURES .......................................................................................................... 7 LIST OF ABBR EVIATIONS ............................................................................................. 8 ABSTRACT..................................................................................................................... 9 CHA PTER 1 INTRODUCTION.................................................................................................... 11 2 BACKGRO UND ...................................................................................................... 13 Japanese Learners of Eng lish ................................................................................ 13 Difficulties for Englis h Learners of Japanese .......................................................... 14 Acoustic Cues for /r/ a nd /l/ per ception................................................................... 15 Acoustic Cues for tap s............................................................................................ 15 The Locus Equation ................................................................................................ 16 Expanding the Loc us Equat ion ............................................................................... 17 3 METHOD................................................................................................................ 20 Participants............................................................................................................. 20 Materi als................................................................................................................. 20 Procedur e............................................................................................................... 22 4 ANALYSIS AND RESULTS .................................................................................... 25 Analys is.................................................................................................................. 25 Results.................................................................................................................... 26 Silent In tervals.................................................................................................. 26 Silent Interv al Duration ..................................................................................... 28 Silent Interval Intensity ..................................................................................... 29 Locus Equation of the 2nd Formant................................................................... 29 3rd Formant R egressi ons.................................................................................. 30 5 DISCUSSION......................................................................................................... 38 The Silent Interval ................................................................................................... 38 Second Fo rmant ..................................................................................................... 42 Third Formant ......................................................................................................... 43

PAGE 5

5 6 CONCLUS IONS ..................................................................................................... 44 APPENDIX A WORD-ONLY CONDIT ION STIM ULI..................................................................... 47 B SENTENCE CONDIT ION STIMULI ........................................................................ 49 C NONSENSE WORDS ............................................................................................. 52 REFERE NCES .............................................................................................................. 53 BIOGRAPHICAL SKETCH ............................................................................................ 54

PAGE 6

6 LIST OF TABLES Table page 3-1 Self-rated Japanese proficien cy by native Eng lish spea kers ......................................... 24 3-2 Native English speakers age of Japanes e acquisition and to tal time studied................ 24 3-3 Self-rated English proficien cy by native Japa nese speak ers ......................................... 24 3-4 Average frequency rating of target word s by English and Japanese speakers ............. 24 4-1 Total number of tokens per subject (d ivided into conditions) compared to total number of silent intervals (SI) appeari ng in d ata per speaker (divided into conditi ons).................................................................................................................... 31 4-2 Possible number of silent interv als and total number of present silent intervals for each language group broken down by tap sylla ble .................................... 31 4-3 Percentage of silent intervals present by syllabl e and language groups ....................... 31 4-4 Percentage of silent intervals pres ent by subject and ov erall for condition .................... 32 4-5 Average duration in milliseconds of silent interv als per speaker ................................... 33 4-6 2nd formant locus equation values per condition and overa ll by s peaker....................... 35 4-7 2nd formant regression values averaged across a ll speakers........................................ 35 4-8 3rd formant regression values per condition and over all by speaker ............................. 37 4-9 3rd formant regression values averaged across all speak ers......................................... 37 A-1 Nonsense Words created, one in each of all possibl e envir onments ............................. 52

PAGE 7

7 LIST OF FIGURES Figure page 2-1 Spectrogram showing Onset and Midpoint areas for frequency measurements .................................................................................................... 194-1 Locus equation of a single native speaker's 2nd formant frequencies (in Hz)...... 334-2 Locus equation of a single non-native speaker's 2nd formant frequencies (in Hz)...................................................................................................................... 344-3 Regression of a single native speaker's 3rd formant frequenc ies (in Hz)............ 364-4 Regression of a single native speaker's 3rd formant frequenc ies (in Hz)............ 36

PAGE 8

8 LIST OF ABBREVIATIONS F2 2nd Formant F3 3rd Formant L1 Native Language L2 Second Language SI Silent Interval SLM Speech Learning Model

PAGE 9

9 Abstract of Thesis Pres ented to the Graduate School of the University of Florida in Partial Fulf illment of the Requirements for the Degr ee of Master of Arts ACOUSTIC PROPERTIES OF THE JAPANESE TAP BY NON-NATIVE AS COMPARED TO NATIVE SPEAKERS OF JAPANESE By Bethany Rowlings May 2010 Chair: Ratree Wayland Major: Linguistics Extensive research has been done on nat ive Japanese learners of English as a second language; however, very little work exists on native English learners of Japanese as a second language. It is a well-k nown fact that native Japanese learners of English have difficulty with English /r / and /l/ phonemes; however, native English learners of Japanese have a sim ilar problem in reverse: acquiring the Japanese tap. This study aims to examine the acoustic pr operties of the Japanese tap as spoken by native speakers of Japanese and compare th e native production to that of advanced learners of Japanese as a se cond language who are nativ e speakers of English. Acoustic measurements of duration and number of silent inte rval (the brief moment of relative silence prior to the release of the tap sound) and frequencies of the second and third formants will be examined to determine which cues are salient for proper production of the Japanese tap. Tap pr oduction will be compared not only between language groups, but also in a variety of syllable environment s to gain a full picture of how the tap is produced by native speak ers of Japanese and examine how well advanced learners have acquired this produc tion. Results showed that advanced

PAGE 10

10 learners of Japanese are able to closely a pproximate native speaker production with limited interference from native English sounds.

PAGE 11

11 CHAPTER 1 INTRODUCTION Research in the field of phonetics and second language acquisition focuses on acoustic properties of sounds and the abilitie s of second language learners to properly acquire the correct characteri stics of their target language sounds to achieve native-like production. Concerning Japanese and English, extensive research has been done on the ability of native speakers of Japanese as well as native speakers of other Asian languages to properly produce English approximants, the /r/ and /l/ phonemes, in a phonetically correct manner. Despite the extensiv e research on learners of English, very little has been done in reverse with a focus on native English learners of Japanese as a second language. The difficulties faced by a native speaker of Japanese learning English are similar to those that will face a native English learner of Japanese. Native English speakers will face difficulties learning the Japanese tap just as native Japanese speakers experie nce difficulties learning the English approximants. In order to examine how well a native Eng lish speaker acquires these sounds in Japanese, the acoustic properties of the t ap phoneme itself must be examined when produced by native speakers. A Japanese learner's production of this phoneme can then be compared acoustically to determine if t he defining characteristics of this sound can be properly acquired by learners. This study examines the acoustic pr operties of the Japanese tap by native speakers of Japanese who are late learners of English and compares to them the acoustic properties of the tap as produced by native English speakers who are advanced, late-learners of Japanese. For the J apanese speakers, it is predicted that the native tap phoneme will hav e similar acoustic characteristics to a stop consonant, just

PAGE 12

12 on a faster time scale. Potential invariant features will also be examined. For the nonnative Japanese speakers, it is predicted t hat, despite a high level of fluency, the English speakers will not have completely mastered the production of the tap phoneme and may experience interference from the nati ve English approximants causing different acoustic characteristics to appear in their data. Overall, however, it is expected that the advanced learners of Japanese will have achieved a high level of proficiency and have acquired a high level of production accuracy by this time in their study.

PAGE 13

13 CHAPTER 2 BACKGROUND Japanese Learners of English While very little research has been done on English learners of Japanese, extensive research has been conducted on the reverse. Native Japanese speakers have difficulty acquiring the English /r/ and /l/ approximants, as the closest phoneme in Japanese is phonetically an apico-alveolar tap (Price, 1981). For Japanese learners, there is a 2 to 1 mapping of the English approximants to the Japanese tap (Takagi, 1993), causing Japanese learners to have diffi culty separating their native /r/ phoneme (the tap) into the English /r/ and /l/. According to the Speech Learning Model (SLM) (Flege, 1995), the larger the phonetic dissimilarity between a sound of a second language (L2) and the closest native language (L1) related sound, the easier it will be for L2 learners to adapt and distinguish the L2 phone from their L1 sounds. However, when they are similar though not identical sounds, as is the case with the Japanese ta p and the English /r/ and /l/, L2 learners will have difficulties distinguishing the L2 phonemes from their L1 sound and have more difficulties acquiring correct pronunciation of the L2 sounds. The SLM accounts for the difficulties that Japanese learners have when acquiring the English /r/ and /l/, though it is complicated for the native Japanese speakers by the prev iously mentioned 2 to 1 mapping, requiring Japanese sp eakers not only to acquire a closely related phoneme, but also to acquire two separate closely related phonemes where the distinction does not exist in t he native language.

PAGE 14

14 Difficulties for English Learners of Japanese English learners of Japanese encounter similar, but reversed, problems to the Japanese learners of English. For the native English sp eakers, the 2 to 1 mapping exists, but for them the difficu lty is in converting two similar sounds (English /r/ and /l/) into a different, but related sound (the Japanese tap). Here, the SLM still applies, making it difficult for English speakers to acquire both the perceptual knowledge to distinguish the tap fr om /r/ and /l/ as well as the production knowledge to accurately produce the Japanese phoneme. Acoustically, the Japanese /r/ bears more si milarities to the intervocalic flapped /t/ or /d/ in English (Price, 1981) rather than the E nglish /r/ or /l/. In English, the flapped /t/ or /d/ occurs in unstre ssed syllables between vowels. Un fortunately, the association of the English flap to the Japanese tap is ra rely made in classrooms to English learners. While associating the Japanese tap with the E nglish flap may, according to the SLM, make it more difficult to acquire, the articulatory motion of the Japanese tap is not emphasized in a language classroom, and knowing a similar sound (i.e., the intervocalic flap) exist in English may actually be advantageous to learners.. Additionally, English learners of Japanese face difficulties from the biasing of the modern orthographical conventions for wr iting Japanese sounds in the English alphabet. In previous studies, the Englis h /l/ has been rated more similar to the Japanese tap than the English /r/ (Sekiyam a & Takhura, 1993), however the modern orthography represents the Japanese tap with the English alphabet 'r'. Learners, however, often are not informed that this sound is not appropriately mapped from the English to the Japanese, leaving learners to initially produce all Japanese tap's as if they were English approximant /r/ as oppos ed to the tap. Native Japanese speakers

PAGE 15

15 will thus interpret the English /r/ as a wors e pronunciation than if it were represented orthographically as the English /l/, thus the addi tional difficulty of English learners of Japanese in acquiring a true native-like pronunciation. Acoustic Cues for /r/ and /l/ perception According to Lisker (1957), the distincti on between /r/ and /l/ for English speakers lies in the 3rd Formant (F3). In order to perceiv e the difference between these two sounds, the listener must be sensitive to this formant. Thus, English speakers, theoretically, are able to hear differences in sounds based on this formant. Japanese speakers, on the other hand, have shown insensitivity to the 3rd formant at the /r/-/l/ boundary, which causes the difficulties for native Japanese speakers acquiring English (Iverson, et al., 2003). This also shows that in order to perceive the Japanese tap, the distinctions in the 3rd F\formant would be unnecessary and possibly detrimental to the production or perception of the tap. We might expect the perception of the 3rd formant by English speakers to cause interference in the proper production of the Japanese tap, especially if the underlying r epresentation that an English speaker has of the Japanese tap is the English approximant /r/. Acoustic Cues for taps Taps involv e a fast and brief closure of the articulators. As a stop consonant requires a complete closure of the articula tors, it might be expe cted that a tap would have similar acoustic cues. Many researchers have identified an interval of relative silence prior to the release of a stop closure as a critical identifier of stop articulation on a spectrogram (Raphael 2005). Additionally, Z ue and Laferriere (1979) found that flaps, which are very similar to taps, have a brief silent interval resemblin g that of a stop, but much shorter. From these observations, we would expect that the Japanese tap be

PAGE 16

16 preceded by a brief silent interval visible on the spectrogram, and that this interval would be shorter than the interval appearing prior to a stop consonant. This silent interval would distinguish t he sound from an approximant, such as the English /r/ and /l/ as in an approximant there is no closure of the articulators and no silent interval should appear. Thus, the salience of the brief silent interval prior to the Japanese tap may prove a cue for perception or production of the tap and could be used to track differences in production between native and non-native speakers of Japanese. The Locus Equation Originally suggested by Lindblom (1963) the Locus Equation has been used by many researchers as an invariant cue (a salient feature that does not change based on context) in perception of st op consonants. The first use of the Locus Equation was with voiced stop CV syllables and has since been expanded to include voiceless stops, nasals, fricatives, and approximants (Sussman, 1994). The locus equation is effectively a best fit regression line of vowel onset and vowel midpoint frequencies of the second fo rmant when plotted on a graph. Figure 2-1 shows the locations on a spectrogram w here the onset and midpoint measurements are taken. The F2 frequency at the onset of the consonant is measured and plotted along the y-axis, while the F2 frequency at the midpoint of the vowel following the consonant is plotted on the x-axis. Once the frequencie s of multiple tokens have been plotted, a best fit, straight light regression is graphed through the plotted frequencies. This results in a straight line with an equation form of y = m x + b where b is the y-intercept of the line where it cross the y-axis, and m is the slope of the line. These two numbers, the yintercept and slope are considered constant fo r each place of articulation of a given manner, and thus are an invariant cue to perception of these sounds.

PAGE 17

17 As stated, Sussman later expanded Li ndblom's locus equation to apply to multiple manners of articulation, not just st op consonants. It logically follows that this invariant cue should also be able to be used as an invariant cue for the perception of the Japanese tap. By invest igating the regressions re sulting from plotting F2 frequencies at the onset of the tap and midpoint of the vowel, the resulting yintercepts and slopes can be used to identify differences in production of the tap by the non-native speakers as compared to the native Japanese speakers, or show that the non-native speakers have successfully acquired the acous tic properties of t he tap pertaining to frequency changes, if indeed the locus equation can be used as an invariant cue to tap production. Expanding the Locus Equation Though the locus equation in its original definition by Lindblom pertains to differences in the 2nd formant, a similar idea of usi ng a regression line to examine similarities or differences in place or manner of articulati on could be expanded to the 3rd formant. By recording the third formant frequencies of the tap at vowel onset and midpoint, and plotting them as with the F2 lo cus equations (with mi dpoint on the x-axis and onset on the y-axis), will the y-intercept and slope values be invariant for F3 as they are for F2 and is this cue used in tap production? As previously stated, t he English /r/ and /l/ distinct ion, or distinction between place of articulation of t he approximants, lies in the 3rd formant, as the research by Iverson and colleagues previously m entioned showed that the Japanese were insensitive to the 3rd Formant at the English /r/-/l / boundary. The question then remains whether native Japanese speakers are wholly ins ensitive to F3, or if they are insensitive only concerning the boundary of non-native s ounds. If native English speakers are

PAGE 18

18 more sensitive than Japanese to the F3 as is necessary to distinguish between the native English approximants, will the English s peakers show more refinement in the F3 than the Japanese speakers? To answer these questions, F3 locus regressions will be measured for the tap in order to investi gate the role of F3 in tap production.

PAGE 19

19 Figure 2-1. Spectrogram showing Onset and Midpoint areas for frequency measurements

PAGE 20

20 CHAPTER 3 METHOD Participants Five native English speakers (1 ma le) and seven native Japanese (3 male) speakers from the University of Florida volu nteered to participate in this study for no compensation. Participants were asked to be between the ages of 20 and 40 and have learned their second language after puberty. Tw o female Japanese speakers were omitted from analysis: one due to learning E nglish prior to puberty, and one due to age. The English speakers were all advanced Japanese learners with average self-rated proficiencies of 6.4 out of 10 in Writing, 6.6 out of 10 in Reading, 6.6 out of 10 in Speaking, and 6.8 out of 10 in Listening (see Table 3-1). All had participated in a Study Abroad programs in Japan, though the length of study varied (see Table 3-2). The native Japanese speakers were all born in Japan and were late learners (after age 12) of English. Their self-rated proficiency in Japanese was a 10 in all categories, and the average self-rating in English wa s 6.8 out of 10 in Writing, 6.8 out of 10 in Reading, 6.2 out of 10 in Speaking, and 6.8 out of 10 in Listening (See Table 3-3). Materials Japanese words learned in the first se mester of Japanese classes at the University of Florida containing one or more examples of the Japanese tap were selected for a total of 25 words. All were native Japanese words as opposed to borrowed words, thus all could be presented in the hiragana syllabary. Of the 25 words, 6 contained a word initial tap with varying following vowel environments (including RA, RI, RU, RE, RO, and RYO) and the rest contained an intervocalic tap also with all possible contexts of vowels preceding the 6 sylla bles uses as word initial taps. A total of

PAGE 21

21 30 Japanese nonsense words (5 word initial t aps) were also creat ed for the experiment consisting of two syllables. For the 25 words with an intervocalic tap, there was one of every possible combination of preceding and following vowel (see Appendix A). The stimuli were arranged into four lists in two conditions: a Word-only Condition and a Sentence Condition. All lists were written using only the hiragana writing system, the syllabary used for native Japanese words. No kanji (characters) or katakana (syllabary used for foreign words) were used in any of the lists. Kanji were not used to ensure that all non-native parti cipants would be able to r ead all stimuli and so that words would not have to be discarded due to potentially unknown Kanji. Katakana was not used as no foreign words were used in any sentences or as target words to ensure that subjects were not bias ed by the original language of the foreign word, if it was known to them. List A consisted of the ta rget words arranged into sentences. Some target words were used in multiple sentences to provide for a greater number of collected tokens. All sentences contained only vocabulary and gramma tical structures learned within the first semester of Japanese classes at the Univer sity of Florida (see Appendix B). List B consisted of just the target words written in hiragana. Fill er words common in the first semester of Japanese that did not contain an instance of the tap token were used sparsely throughout the list to distract s ubjects who might make a connection between the words-only list and a focus on the tap phoneme (see Appendix C). List C consisted of the Japanese nonsense words arranged in a carrier phrase Machi no namae wa _____ which translates to The name of the to wn is ____. The final list, List D,

PAGE 22

22 consisted of the nonsense words alone. Nonsense lists were used as distracters to encourage participants to be more relaxed on the actual experimental lists. In all lists the stimuli were randomi zed by hand and presented on sheets of paper with each word or sentence on an individual line. The lists were double spaced. In all Lists but List C, the critical word s and sentences appeared twice and the first occurrence of the stimuli wa s not analyzed. The order of the stimuli the second time was not the same as the first. In List C the stimuli were not presented twice, as the first participant felt that the list was too lengt hy and, due to it being a repetitive carrier phrase, it was too easy to get lost with the length. As List C and D were not used in analysis, but only as filler lists, it was not cr itical to present these sentences twice. Procedure Participants first signed a consent form and answered questions concerning their education, language background and pr oficiencies. They were then told that they would be reading four lists of word s or sentences, two containing real words and the following two containing made-up words. The partici pants were recorded alone in a sound proof both using a Marantz PMD 660 digital recor der with a sampling rate of 44100 Hz and external Shure SM 10 A headset microphone. The lists were presented to participants one at a time and each list was recorded onto a separate track. Lists A and B were always presented first, though hal f of the participants saw A fi rst and B second and the other half read B first and A second. List C and D were always presented second and always in this order, as the Nonsense Word s-only Condition (List D) made it more obvious that all nonsense words involved t he tap segment than Li st C which presented the words using carrier phrases and, as filler lists, the order was not critical.

PAGE 23

23 Prior to recording, and as a reminder pr ior to each new list, participants were told to pause between each line and not to read the individual words together as if they were a sentence or the sentences together as if they were a paragraph. This was to ensure that target tap tokens occurring sentence or word initially woul d be properly isolated from preceding words. Participants were inst ructed to take as much time as necessary to read the sentences, including reading word s or sentences silently in advance of speaking if necessary, and were told to read the lists in a normal conversational speaking voice with speech rate and volume appropriate to asking a teacher a question in an average size (30-40 student) classroom. Once all four lists had been recorded, participants were asked to fill out two additional forms. The first asked native Japanes e speakers to rate on a scale of 1-5 how often the target words were used in everyday Japanese speech, with '5' corresponding to 'very frequently'. The native English speakers were asked to rate their familiarity with the word based on their current level of Japanese knowledge to get a frequency rating of the target words in both the native and non-native speakers vocabularies (see Table 3-4). The average frequency rating for target words was 4.3 out of 5 for native speakers and 4.09 out of 5 for non-native speakers.

PAGE 24

24 Table 3-1. Self-rated Japanese proficiency by native English speakers Table 3-2. Native English speakers age of Japanese acquisition and total time studied Table 3-3. Self-rated English prof iciency by native Japanese speakers Table 3-4. Average frequency rating of target words by English and Japanese speakers SubjectAgeAge First AcquiredTotal Years of StudyLength of Study Abroad (months) e301211839 e302211839 e303211839 e304221561 e305211839 Subject Native Language Average Frequency Rating e301 English 3.77 e302 English 2.55 e303 English 4.86 e304 English 4.66 e305 English 4.59 j001 Japanese 4.73 j003 Japanese 3.77 j004 Japanese 4.68 j005 Japanese 5 j006 Japanese 3.27 Overall English 4.09 Overall Japanese 4.3 Subject SpeakingListeningReadingWriting e301 8887 e302 7777 e303 7556 e304 6776 e305 5766 Subject SpeakingListeningReadingWriting j001 81097 j003 6677 j004 5555 j005 7768 j006 5677

PAGE 25

25 CHAPTER 4 ANALYSIS AND RESULTS Analysis Recordings were transferred from the digi tal recorder to com puter via USB. The files were then segmented into individual words. For each subject, there were 28 possible tokens in the Word-Only Condition and 47 possible tokens from the Sentence condition for a total of 75 tokens per participant. Though 750 tokens should have been collected across all participants, a total of 748 to kens were collected as two participants accidentally skipped a word in the Words-Only condition. Some tokens had to be rejected due to recording or analysis problems such as the participants making extra noise during reading, difficulties in segmenting the file due to a participants pronunciation, or improperly read words resulting in obscur ed vowel environments; thus, a total of 737 tokens (364 English and 373 Japanese), or 98.5% of collected tokens, were analyzed across all subjects. Acoustic measurements were taken us ing the Praat software (Boersma and Weenink, 2010) and recorded in Excel. For each token, 2nd formant (F2) and 3rd formant (F3) frequencies were averaged over approxim ately 25.6 millisecond intervals at both tap onset and vowel midpoint for use in Locus Equation regressions. In some tokens, the vowel after the tap was t oo short, and a 25.6 millisecond interval was not possible. In these situations, intervals as close to 25.6 as possible were used, but none were less than 20 millisecond intervals. Additionally, the duration of the silent interval prior to t he onset of a tap, indicating stop or flap-like closure, was recorded as well as intensity of these in tervals. Intensity at the onset of the tap was also recorded over the same 25.6 millisecond interval that

PAGE 26

26 onset F2 and F3 frequencies were recorded in order to compare with intensity of the silent intervals. Once the F2 and F3 measurements were recorded, Locus Equation regressions for both the second and third formants were graphed for each subject in the Word-Only Condition, Sentence Condition, and Overall. Midpoint was graphed on the X-axis and Onset on the Y-axis. The slopes, y-intercepts, and R2 values of the regressions were recorded in Excel. Using the R Statistical program (R Development Core Team, 2006), a repeated measures ANOVA wa s run on the slopes and y-intercepts with condition (Sentence or Word-only) as the within s ubject factor and native language (Japanese or English) as the between subject factor. The other recorded data (duration and intensive of silent intervals, vowel onset intensity) was also imported to R for statistics. Repeated measures ANOVA was run on the duration of the silent intervals changes in intensity from silent interval to onset with condition (Word-only or Sentence) and sylla ble environment (RA, RI, RU) as within subject factors and native language as the between subject factor. Additionally, all speakers had some degree (certain number) of missing silent intervals, thus ANOVA's were also run comparing the amount of Silent Intervals of the non-natives to that of the native speakers for both the Words-Only and Sentence conditions as well as overall with syllable environment as an addi tional within-subject factor. Results Silent Intervals Duration of silent intervals was re corded for each participant. There were a substantial number of missing si lent intervals across subjec ts, and this number was also recorded. Table 4-1 shows the total number of silent intervals possible for each subject

PAGE 27

27 overall and a breakdown by cond ition in the first set of columns and then a breakdown of how many silent intervals actually occu rred in their data overall and broken down by the syllable environment of t he tap per subject. Table 4-2 shows the sum of the total possible silent intervals and the sum of the to tal silent intervals present in each syllable environment for each native language group and Table 4-3 shows the percentage of silent intervals present by syllable and l anguage group. Table 4-4 shows the percentage of silent intervals present by subject and condition. While native English speakers had a tendency to have a lower percentage of silent intervals (47.55% compared to nat ive Japanese speakers at 64%), the difference in percent present silent intervals bet ween the two groups was not significant [F(1,8)=3.45, p=.1]. There were, however, significant differences by condition (Wordonly or Sentence) [F(1,8)=12.36, p<.05] indi cating that there was significantly higher number of silent intervals in the word-only condition than in the sentence condition (overall for both groups 58.8% for Word-onl y vs 53% for Sentence). The analysis also revealed that the number of present silent interval varied significantly across syllable environments [F(1,8)=11.04, p<05). In addi tion, the interaction between syllable environment and native language group was significant [F(1,8)=3.44, p<.05]. Furthermore, there was a significant inte raction between syllable environment and condition [F(1,8)=5.11, p<.05] and a signifi cant three way interaction between syllable environment, condition, and native language group [F(1,8)=3.03, p<.05] was also found. Post-hoc analysis using T-Tests show that the percentage of silent intervals in the RU syllables was significantly greater among native Japanese speakers than among native English speakers (59% vs 46%) [T(16) =-2.2, p<.05]. Similarly, there was a

PAGE 28

28 significantly higher percentage of silent intervals among the Japanese than the native English speakers for the RO syllables (51% vs 39%) [T(17)=-2.8, p< .05). Additionally, for English speakers, the difference between the percentages of present silent intervals was significantly higher in RU syllables than in RI syllables (46% vs 30%) [T(17)=2.8, p<.05] and in RE syllables t han in RI syllables (47% vs 30% ) [T(18)=-2.5, p<.05]. For Japanese speakers, the differences were si gnificant between RA and RU (39% vs 59%) [T(16)=-2.8, p<.05] and RI and RU (41% vs 59%) [T(16)=-2.7, p<.05]. Post-hoc analysis also showed a significant difference betw een Word-only and Sentence for RA syllable (47.1% vs 33.3%) [T(9)=-3.1, p<.05] fo r RU syllable environm ents (69.4% vs 45%) [T(17)=-2.9, p<.05] and for RE syllabl es (70% vs 45%) [T(17)=-4.7, p<.05]. Silent Interval Duration The average duration of the silent intervals across all conditions and environments was relatively similar in each language group (See Table 4-4 for average duration of silent intervals for each subject across all conditions). However, there were significant variations based on conditi on and syllable envir onment. The average duration of the silent intervals showed significant variation based on condition [F(1,5)=7.90, p<.05] with the Sentence c ondition showing longer durations (average duration 24.1 ms compared to the Word-only condition av erage duration of 21.5 ms). However, there was no significant main effect of native language group [F(1,5)=.0002, p=.99]. In addition, it was found that duration of silent intervals varied significantly across syllable environments [F(5,35)=9.12, p<.05]. The interaction between syllable environment and language was also significant [F (5,35)=2.95, p<.05]. Additionally, there was a significant interaction between condition, syllable environment, and language [F(4,30)=6.62,p<.05].

PAGE 29

29 Post-hoc T-Test analysis showed t hat for native Japanese speakers the difference between RI and RU syllables was significant [T(15)=2.8, p<.05] and the difference between RI and RO syllables was also significant [T(17)=3.45, p<.05], with RI having a longer duration (at 26.0 ms) than both RU (at 20.5 ms) and RO (at 17.9 ms), suggesting that some aspect of the RI syllable environment causes increased duration of the silent interval. Silent Interval Intensity Intensity of the silent in terval and intensity of the onset of the tap were also recorded in efforts to examine intensity dips as characterist ic of the tap. However, the difference in intensity between these inte rvals was not significant between the two language groups [F(1,8)=4.68, p=.06]. A r epeated measures ANOVA with condition and syllable environment as within subject factors and language as a between subject factor also showed no significant results fo r any main effects or interactions. Locus Equation of the 2nd Formant Locus Equation regressions were taken of the 2nd formant frequencies for each speaker to examine both the Locus Equat ion as an invariant feature and also to examine differences in the Locus for each gr oup. Regressions were plotted (Figure 4-1 and 4-2 for example Locus Equation graphs ) for both the Word-Only Condition and Sentence Condition and Overall (See Table 4-5). As the overall regressions were effectively the aggregate data, an analysis of variance was run on the overall slope and Y-intercepts with language as a between subjects factor. There were no significant effects of native language on slope [F(1,8)=.51, p=.49] or y-intercept [F(1,8)=.85, p=.38] values, suggesting both language groups had similar acoustic properties in the 2nd formant.

PAGE 30

30 3rd Formant Regressions Regressions were al so plotted for the 3rd formant frequencies with midpoint on the x-axis and onset on the yaxis (Figure 4-3 and 4-3 for ex ample F3 Regressions) for the Word-Only Condition, Sentence Condition, and Overall (See Table 4-6). As with the F2 regressions, the goal was to determine if t here is an invariant feat ure such as a locus in the 3rd formant that can be found and also to examine differences between language groups. Again only the Overall (aggregate) r egressions were used in the analysis of variance. Similar to the F2 data, there was no significant variation by language group for the slope [F(1,8)=.45, p=.52] or y-intercept [F(1,8)=.35, p=..57], suggesting no substantial difference in the Loc us of the two language groups.

PAGE 31

31 Table 4-1. Total number of tok ens per subject (divided into conditions) compared to total number of silent intervals (SI) appeari ng in data per speaker (divided into conditions) Table 4-2. Possible number of silent in tervals and total number of present silent intervals for each language group broken down by syllable Table 4-3. Percentage of silent interval s present by syllable and language groups SyllableJapanese Possible SIEnglish Possible SIJapanese Present SIEnglish Present SI RA1751726868 RI1061024331 RU85835038 RE1301266559 RO85854333 RYO252589 SubjectNative LanguageTotal PossibleWord-OnlySentencePresent SIWord-Only SISentence SI j001Japanese752847592138 j003Japanese742846371621 j004Japanese752847492128 j005Japanese752847482028 j006Japanese742846381622 e301English732845391821 e302English742846331122 e303English70244623914 e304English742846431825 e305English732746341321 Syllable Native English SI Percentage PresentNative Japanese SI Percentage Present RA 40 39 RI 30 41 RU 46 59 RE 47 50 RO 39 51 RYO 36 32 Average 40 46

PAGE 32

32 Table 4-4. Percentage of silent intervals pr esent by subject and overall for condition SubjectNative LanguageWord-Only Percent SI PresentSentence Percent SI Present j001Japanese7581 j003Japanese5756 j004Japanese7560 j005 Japanese 71 60 j006 Japanese 57 48 e301 English 64 47 e302 English 39 48 e303 English 38 30 e304 English 64 54 e305 English 48 46 Overall Japanese 67 61 Overall English 50.6 45

PAGE 33

33 Table 4-5. Average duration in milliseco nds of silent intervals per speaker Figure 4-1. Locus equation of a single native speaker's 2nd formant frequencies (in Hz) 50010001500200025003000 0 500 1000 1500 2000 2500 3000 3500f(x) = 0.46x + 1058.16 R = 0.54Midpoint F2Onset F2 Subject Native Language Average Silent Interval Duration (ms) j001 Japanese 18.5 j003 Japanese 23.8 j004 Japanese 23.6 j005 Japanese 19.8 j006 Japanese 22.3 e301 English 20.5 e302 English 18.7 e303 English 26.9 e304 English 20.7 e305 English 18.9 Average Japanese 21.6 Average English 21.1

PAGE 34

34 Figure 4-2. Locus equation of a single non-native speaker's 2nd formant frequencies (in Hz) 100015002000250030003500 0 500 1000 1500 2000 2500 3000f(x) = 0.5x + 977.51 R = 0.48Midpoint F2Onset F2

PAGE 35

35 Table 4-6. 2nd formant locus equation values per condition and overall by speaker Table 4-7. 2nd formant regression values averaged across all speakers SubjectNative LanguageConditionslopeY-intercept e301English Word-Only 0.47 1075.87 0.61 e301English Sentence 0.52 929.13 0.41 e301English Overall 0.5 977.51 0.48 e302English Word-Only 0.55 801 0.61 e302English Sentence 0.57 760.19 0.61 e302English Overall 0.56 776.66 0.61 e303English Word-Only 0.46 1019.61 0.59 e303English Sentence 0.48 974.95 0.61 e303English Overall 0.47 994.23 0.6 e304English Word-Only 0.52 951.94 0.49 e304English Sentence 0.59 757.57 0.7 e304English Overall 0.56 839.82 0.6 e305English Word-Only 0.58 792.22 0.71 e305English Sentence 0.51 887.27 0.34 e305English Overall 0.55 925.92 0.49 j001Japanese Word-Only 0.44 1147.96 0.48 j001Japanese Sentence 0.47 1010.52 0.59 j001Japanese Overall 0.46 1058.16 0.54 j003Japanese Word-Only 0.69 524.71 0.7 j003Japanese Sentence 0.67 552.11 0.74 j003Japanese Overall 0.68 542.77 0.72 j004Japanese Word-Only 0.57 847.18 0.5 j004Japanese Sentence 0.29 1294.11 0.13 j004Japanese Overall 0.39 1121.9 0.24 j005Japanese Word-Only 0.59 761.13 0.62 j005Japanese Sentence 0.72 535.73 0.73 j005Japanese Overall 0.66 631.3 0.68 j006Japanese Word-Only 0.69 511.54 0.71 j006Japanese Sentence 0.67 548.7 0.74 j006Japanese Overall 0.68 535.94 0.73R 2 value Native Language Condition slopeY-intercept English Word-Only 0.52928.13 0.6 English Sentence 0.53861.82 0.53 English Overall 0.53902.83 0.56 Japanese Word-Only 0.6758.5 0.6 Japanese Sentence 0.56788.23 0.59 Japanese Overall 0.57778.01 0.58R 2 value

PAGE 36

36 Figure 4-3. Regression of a single native speaker's 3rd formant frequencies (in Hz) Figure 4-4. Regression of a single native speaker's 3rd formant frequencies (in Hz)

PAGE 37

37 Table 4-8. 3rd formant regression values per condition and overall by speaker Table 4-9. 3rd formant regression values averaged across all speakers SubjectNative LanguageConditionslopeY-intercept e301EnglishWord-Only0.631031.20.71 e301EnglishSentence0.392678.590.28 e301EnglishOverall0.51391.120.45 e302EnglishWord-Only0.77635.270.34 e302EnglishSentence0.431552.70.21 e302EnglishOverall0.531287.820.24 e303EnglishWord-Only1.18-609.510.57 e303EnglishSentence1.13-433.50.58 e303EnglishOverall1.17-546.510.56 e304EnglishWord-Only0.321931.860.27 e304EnglishSentence0.182397.530.06 e304EnglishOverall0.32025.210.21 e305EnglishWord-Only0.281864.680.11 e305EnglishSentence0.361653.390.11 e305EnglishOverall0.331738.070.11 j001JapaneseWord-Only0.491330.220.18 j001JapaneseSentence0.311775.820.09 j001JapaneseOverall0.371657.350.12 j003JapaneseWord-Only0.511226.820.52 j003JapaneseSentence0.491404.790.29 j003JapaneseOverall0.511256.530.38 j004JapaneseWord-Only0.321890.090.14 j004JapaneseSentence0.22193.170.09 j004JapaneseOverall0.242099.160.11 j005JapaneseWord-Only0.9772.080.37 j005JapaneseSentence0.67950.040.23 j005JapaneseOverall0.81551.80.3 j006JapaneseWord-Only0.192057.650.24 j006JapaneseSentence0.491288.860.29 j006JapaneseOverall0.271878.290.23R 2 value Native Language Condition slopeY-intercept English Word-Only 0.64970.7 0.4 English Sentence 0.51569.74 0.25 English Overall 0.571179.14 0.31 Japanese Word-Only 0.51315.37 0.29 Japanese Sentence 0.431522.54 0.2 Japanese Overall 0.441488.63 0.23R 2 value

PAGE 38

38 CHAPTER 5 DISCUSSION The Silent Interval A properly articulated tap should involve a silent interval of some kind (based on the data, an interval with duration around 21 or 22 ms on average). However, this data shows that not all native speakers include a silent interval in their production of the tap phoneme. In fact, only 46% of tokens collected from native Japanese speakers contained a silent interval. When the silent interval is absent, there are no gaps or changes in intensity between the surrounding vowel environments, though the sound itself is still present, suggesting the tap tu rns into or is reduced to more of an approximant by native speakers. There were a larger number of missed silent intervals in the Sentence Condition, wit h only 59% of possible silent intervals actually appearing as opposed to 67% in the Word-Only condi tion, suggesting that in normal speech, a larger number of reduced taps were produced. However, even in more careful speech, as in the Word-Only condition, there was still a substantial number of missing silent intervals even by native speakers. Non-native speakers consist ently had a greater number of missing silent intervals overall. Interestingly, the native Englis h speakers had a higher percentage of silent intervals present in the Word-Only condition (76%) but a much lower number in the Sentence condition (45%). While the av erage number of silent intervals was not significantly different between the two langu age groups, it was found that the amount of silent intervals varied significantly across condition and syllable type within each language group. The variation patterns also differed between the two language groups. It is interesting that the native English speak ers had a higher overall percentage of silent

PAGE 39

39 internals in the Word-only condition than the native Japanese speak ers. This suggests that the native English speakers are perhaps over-exaggerating in the Word-Only condition in an attempt to imitate the nativ e Japanese production, but due to their focus in careful speech, actually achieve a higher degree of 'correct' tokens which is not actually natural or native-like based on every day speech. In the Sentence condition, the native English speakers accuracy in producin g silent intervals drops dramatically (by 31%) suggesting that when not producing care fully articulated speech, the English speakers become less accurate, dropping to a le vel of accuracy lower than the native speaker level (at 59%). These percentages suggest that the Engl ish speakers are possibly experiencing more interference from the native Englis h /r/ and /l/ approximants in the Sentence Condition, as they are not able to monitor their speech as accurately. However, when they are able to monitor their production, they over-compensate by not reducing the same number of sounds to approximants as the native Japanese speakers. The number of silent intervals present also significantly varied by syllable environment and condition across both language groups. This suggests that certain environments are more prone to becoming appr oximants by both native speakers and non-native speakers. Post-hoc analysis supports this, as there were significant differences between RI and RU and RI and RE for the native English speakers and differences between RA and RU and RI and RU fo r the native Japanese speakers. As the RI syllable appears as showing significa nt differences in both language groups, reviewing the percentages of silent interval s present in RI syllables by both native English and Japanese speakers shows that the percentage of RI silent intervals falls

PAGE 40

40 below average for both English speakers (30% compared to 40% overall) and Japanese speakers (41% compared to 46% overall) suggesting that the RI syllable becomes an approximant more often in both language groups. Other syllables have a higher than average per centage of silent intervals present, such as the RU syllable which also showed significant differences in both speaker groups, which contained a silent interval in 46% (as compared to 40% overall) of native English speaker productions and contained a s ilent interval in 59% (as compared to 46% overall) of native J apanese speaker productions. With this highly accurate syllable, the difference between the native language groups was significant, suggesting that the native English speak ers have not yet achieved the performance level that native Japanese speakers reach, despite mirroring the general trend for the RU syllable to have a higher number of present silent intervals. The most accurate syllable from the non-native speakers is the RA syllable, where both native and non-native speakers were within one percent in the number of present silent intervals. Interestingly, the native English speakers did not mirror the si gnificant difference in RA and RU syllables that native speakers showed, likely because the difference in RU syllables between language groups was also significant. Other di fferences were substantial, such as the significant difference between native RO (at 51%) and nonnative RO (at 36%). Overall, this suggests that certain syllables are either mo re difficult to produce as a tap, or due to context, are more likely to be produced as an approximant, and this is generally true of both the native and non-native s peakers, though non-native s peakers had higher overall percentages. This also suggests that ce rtain vowel environments may be easier to acquire than others by the non-native speaker s and that while production of the tap in

PAGE 41

41 some environments (such as the RA syllable) has been relatively well learned, in other environments (such as the RO syllable) the production is not near t hat of the native Japanese speakers. As with the number of silent intervals present, the duration of silent intervals did not differ significantly between the two gr oups. There was, however, a significant variation in silent interval duration ac ross condition and syllable environment and the variation pattern differed significantly between the two language groups. Average silent interval duration was longer in the S entence condition in both language groups, suggesting that in careful speech such as in the Words-only condition the silent interval does not have to be as long in order to be sa lient, whereas in normal speech the silent intervals may be extended. As the duration of the silent interval by native speakers overall had a small range (from average 18. 51 ms to 23.80 ms) and only one native English speaker fell outside of this range (at an average of 26.86 ms), it seems reasonable to assume that when English speakers are able to produce the Japanese tap (as opposed to an approximant /r/), with their production closely mirroring native production in silent interval duration. The difference between RI and RU syllable environments and RI and RO syllable environ ment was significant for native Japanese speakers only. This is more interesting when compared to the percentage of silent intervals present in the RI condition by both native and non-nat ive speakers. As mentioned above, the percentage of present silent interv als was below the average in the RI condition and the difference between the RI syllable and other syllables was significant in both languages. This pattern of results suggests that in the context of a following vowel [i], the Japanese tap production is more likely to vary or become

PAGE 42

42 produced as an approximant. That both the duration of the inte rval and the total present intervals differ suggest that this vowe l environment is unique and has different characteristics from the other vowel envir onments and, due to t he low numbers shown by the English speakers, this suggests that the RI environment would be one of the most difficult environments to master when learning proper pronunciation. Second Formant Using the Locus Equation to look for in variant features for the tap and identify differences based on native language proved in significant. There was no significant difference in either y-intercept or slope values between the two groups. This suggests that overall, native English speakers production of Japanese tap closely approximately that of native speakers production. It is impor tant to note, however, that both slope and y-intercept values varied greatly across i ndividuals within each group, suggesting that there was a large degree of variation in the production of a Japanese tap among both native and non-native speakers. Interestingly, the native English speakers showed less variation in their yintercepts than the native J apanese speakers. Though the differences are not significant, it is curious that the English speakers could be said to be showing a relatively more stable production of the Japanese tap than did the Japanese speakers. This may have been due to the fact that, in Japanese, there are no other phonemes to confuse the tap wit h, and thus its production does not have to be as well defined. The smaller variation in yintercepts by native Englis h speakers, on the other hand, suggest that they consistently a ssociate Japanese tap with a particular phoneme in English (i.e., an English /r/ or /l/, /t/ or /d /) However, further rese arch is needed to test this hypothesis.

PAGE 43

43 Third Formant Regressions of the third formant values similar to the locus equation concept, like the F2 equations, showed no significant di fference between the two language groups. Specifically, there was no significant diffe rence in the slope (average English slope = .57, Japanese slope = .46) or y-intercept (average English y=1239.86, Japanese y=1442.18) or R2 value (English R2 = .32, Japanese R2 =.24) Again, these results suggest that native speakers of English are quite successful in their production of the Japanese tap. The R2 values for correlation for both groups are low, which suggests that their production of various tokens of the tap varied greatly.

PAGE 44

44 CHAPTER 6 CONCLUSIONS Examining the acoustic properties of the Japanese tap and comparing salient features between native Japanese speakers and native English advanced learners of Japanese yielded interesting results. Not all of the features predicted to be useful as an acoustic cue to production were salient, but interesting differences and similarities occurred between the two groups of speakers. These data suggest that the advanced learners of Japanese are able to approximate native Japanese speak er production relative of the Japanese tap relatively well. The silent interval duration is the only acoustic characteristic that showed any significant variation based on language group, which suggests that this is a strong indicator of the properties and accuracy of the tap production. While the English speakers have fewer instances of correctly produced taps (as suggested by the lower amount of silent intervals), when these interv als exist there is minimal difference outside of individual variation between the nat ive English and native Japanese speakers productions. Even variation based on syllable environment is relatively well reflected between the native and nonnative speakers. It does seem that there is some inte rference occurring for the native English speakers from the closely related Englis h /r/ and /l/; however, this may be easily improved with increased practice and exposur e by the English le arners of Japanese. There are no significant variations between native English and native Japanese in either the 2nd or 3rd formant, suggesting that English s peakers are able to approximate the Japanese speaker frequencies accurately, or in the case of the 3rd formant, that distinctions at this formant level are unnecessary for distinct ion, and that distinctions for

PAGE 45

45 /r/ and /l/ specific to the 3rd formant will not interfere with the proper production of the Japanese tap by non-native speakers of Japanese. The most salient characteristic of determining when a properly produced tap occurs in a word or sentence remains the si lent interval closure prior to the release sound, which can be easily identified on a spectrogram. Variations in the 2nd formant suggest that, while the second formant locu s equation may be a valid acoustic cue for other consonants such as stops there may be a wider range of values accepted for the locus of the Japanese /r/, which makes it di fficult to use as a comparison for proper production of the Japanese tap by any group of speakers (that is, if the locus is poorly defined, it is difficult to tell if a poorly defined locus on non-native speakers is the result of proper acquisition of the lo cus, or the result of c onflicting L1 sounds causing a scattered locus). Additionally, though the third formant regressions were interesting, they revealed nothing significantly usef ul as an acoustic cue for production or comparison between native and nonnative speakers. It is po ssible that one or both of the formant regressions could yield more interesting results in a larger data set or, specifically, when able to be compared to other native and non-native sounds between groups. Without a comparison to other sounds the usefulness of the locus comparison is limited. Overall, the silent interval proved to show differences not only between language groups but also within language groups based on the vowel environm ent of the tap. Further research could investigate the diffe rences in environment with more success by polling larger numbers of nativ e speakers on a larger number of tokens. The possibility of L2 English affecting the native Japanese s peakers production of the taps could also

PAGE 46

46 be a point of interest that should be further investigated with Japanese speakers who have limited English knowledge, rather t han the relatively fluent and immersed in English native Japanese speakers used in this project. Additionally data from both the silent intervals and the F2 and F3 regressions showed that the advanced learners of Japanese have acquired a high degree of sens itivity to the Japanese tap; however, improvements are still needed in order to match the native Japanese production.

PAGE 47

47 APPENDIX A WORD-ONLY CONDITION STIMULI Japanese v ocabulary used in the Word-Only condition. In the experiment, it would have been presented in hiragana. Kore this arigatou thank you kirei pretty yoroshiku please treat me kindly tokoro place rainen next year benri convenient hiru afternoon raishuu next week mura town/village kuruma car tsukareru To become tired ringo apple renshuu practice roku six ryouri food atarashii new okiru to wake up hiragana hiragana (Japanese syllabary) karui light

PAGE 48

48 sorekara and from here/after that kawamura Kawamura (Japanese surname) omoshiroi interesting

PAGE 49

49 APPENDIX B SENTENCE CONDITION STIMULI Japanese sentences and translations used in the Sentence Conditions. In the experiment, they would have been presented in hiragana. The words containing the tap token have been made bold for easy visibility, they were not accentuated in any way in the actual experiment. Ano hon wa atarashii desu. The book is new. Eigakan wa benri desu. The movie theater is convenient. Kore ha ringo desu. This is an apple. Ringo ga suki desu. I like apples. Shukudai wa hiragana no renshuu desu. The homework is hiragana practice. Ano hito wa kawamura san desu. That person is Mr. Kawamura. Mura wa shizuka desu. The village is quiet. Domo arigatou gozaimasu. Thank you very much Watashi no ie wa totemo benri desu. My house is very convenient. Raishuu no suiyoubi, eigakan ni ikimasu. Next Wednesday I am going to the movies Renshuu shimashou. Let's practice Kyou no hiru hon o yomimasu. This afternoon I will read a book

PAGE 50

50 Watashi wa rainen oosaka ni ikimasu. Next year I will go to Osaka Watashi ha ringo ga suki desu. I like apples. Hiragana wa muzukashii desu. Hiragana is difficult. Ryouri wa oishii desu. The food is delicious. Nihongo no renshuu o shimasu. I do Japanese practice. Machida san wa yonjuu roku sai desu. Mrs. Machida is 46 years old. Ano hana wa kirei desu. That flower is pretty. Atarashi kuruma o kaimasu. I will buy a new car. Donna ryouri ga suki desu ka. What kind of food do you like. Ano hon wa karui desu. That book is light. Watashi wa raishuu no shuumatsu, benkyou shimasu. This weekend I will study. Sorekara hirugohan o tabemasu. And then I ate lunch. Are ha kawamura san no kuruma desu. That is Mr. Kawamura's car. Roku nin imasu. There are 5 people there. Ashita, omoshiroi eiga o mimasu. Tomorrow, I will watch an interesting movie.

PAGE 51

51 Donna tokoro desu ka. What kind of place is it? Hajimemashite. Dozo yoroshiku. Pleased to meet you. I ask that you tr eat me kindly. (traditional greeting) Soko ni kuruma ga arimasu. There is a car there. Kawamura san no hon ha omoshiroi desu. Mr. Kawamura's book is interesting. Rainen, nihon ni ikimasu. Next year I will go to Japan. Watashi wa tsukaremasu. I am tired. Sorekara, hon o yomimasu. And then I read a book Totemo kireina hito desu ne. That person is very pretty

PAGE 52

52 APPENDIX C NONSENSE WORDS The following table consists of all of the nonsense words created for List C and D in the experiment. This also accurately re flects all of the environments used. They are shown here using English orthography, though in the experiment they would have been presented in hiragana. Table A-1. Nonsense Words created, one in each of all possible environments _RA_RI_RU_RE _ROWord Initialratarimeruni reso rodeAR_taranarimarutare daroIR_nirachirishirunire niroUR_surarunitsurusure meroER_deraserinerudere neroOR_moradoridorumoretoro

PAGE 53

53 REFERENCES Boersma, Paul and Weenink David. (2010). Praat: do ing phonetics by computer (Version 5.1.26) [Computer Software ]. Retrieved Januar y 25, 2010, from http://www.praat.org/ Flege, J. E. (1995). Second language speech l earning: theory, find ings, and problems. In W. Strange (Ed.), Speech perception and linguistics experience: Issues in cross-language research (pp. 233-277). Baltimor e, MD. York Press. Iverson, et al. (2003). A Perceptual Interfer ence Account of Acquisition Difficulties for Non-native Phonemes. Cognition 87, 1, Feb. B47-B57. Lisker, L. (1957). Minimal cues for separating /w, r, l, y/ in intervocalic position. Word 13. 257-67. Price, P. (1981). A cross-linguistic study of flaps in Japanese and in American English. Ph.D. Dissertation, University of Pennsylvania. R Development Core Team. (2006). R: A Language and Environment for Statistical Computing [Computer Software]. R F oundation for Statistical Computing. Vienna,Austria. Retrieved February 1, 2010, from http://www.R-project.org/ Raphael, Lawrence J. (2005). Acoustic Cu es to the Perception of Segmental Phonemes. In David B. Pisoni and Robert E. Remez (Ed.), The Handbook of Speech Perception (pp. 182-206). Oxford, UK. Blackwell Publishing. Sekiyama, K. & Yohkura, Y. (1993). Inter-language differences in the influence of visual cues in speech perception. Journal of Phonetics 21. 427-444. Sussman, Harvey M. (1994). The Phonologi cal Reality of Locus Equations across Manner Class Distinctions: Preliminary Observations. Phonetica 51. 119-131. Takagi, N. (1993). Perception of American E nglish /r/ and /l/ by adult Japanese learners of English: A unified view. Ph.D. Dissertat ion, University of California, Irvine. Zue, V. W., and Laferriere, M. (1979). Acous tic study of medial /t, d/ in American English. Journal of the Acoustical Society of America 66. 1039-1050

PAGE 54

54 BIOGRAPHICAL SKETCH Bethany Rowlings earned her bachelor's degree in linguistics from the University of Florida in Spring 2008 with a minor in Teaching English as a Second Language. She entered graduate school in Fall 2008. S he graduated in May 2010 with her master's degree in linguistics.