Training Japanese L1 Speakers in American English L2 Production of the Phoneme /R/ Using Visual Biofeedback in the Form ...

MISSING IMAGE

Material Information

Title:
Training Japanese L1 Speakers in American English L2 Production of the Phoneme /R/ Using Visual Biofeedback in the Form of a Spectrogram a Preliminary Study
Physical Description:
1 online resource (85 p.)
Language:
english
Creator:
Patten, Iomi B
Publisher:
University of Florida
Place of Publication:
Gainesville, Fla.
Publication Date:

Thesis/Dissertation Information

Degree:
Master's ( M.A.)
Degree Grantor:
University of Florida
Degree Disciplines:
Communication Sciences and Disorders, Speech, Language and Hearing Sciences
Committee Chair:
Edmonds, Lisa Anna
Committee Members:
Wingate, Judith M

Subjects

Subjects / Keywords:
acquisition -- american -- biofeedback -- f3 -- formant -- japanese -- japanese-language-speakers -- language -- native -- perception -- phoneme -- production -- r -- rhotic -- second -- speaker -- spectrogram -- treatment -- visual
Speech, Language and Hearing Sciences -- Dissertations, Academic -- UF
Genre:
Communication Sciences and Disorders thesis, M.A.
bibliography   ( marcgt )
theses   ( marcgt )
government publication (state, provincial, terriorial, dependent)   ( marcgt )
born-digital   ( sobekcm )
Electronic Thesis or Dissertation

Notes

Abstract:
Japanese speakers who are learners of English have difficulty both producing and perceptually identifying the phoneme /r/. This phoneme is produced in variable ways by Native Speakers of English, so there is little concrete articulator movement or placement instruction that would be of help to these learners. However, the identifying feature of /r/ is found in the acoustic signal: the third formant (F3) "dips"unusually low for this phoneme (below 1950 Hz;but productions under 2300 are also perceived as correct). Japanese speakers do not auditorily attend to this formant because it does not hold identifying linguistic information in the Japanese language, rendering auditory perception alone an unreliable cue as well. This study uses spectrographic visual biofeedback to encourage these baseline design was implemented with two participants. Treatment stimuli included words with /r/ in initial position which were presented in hierarchical phases of increased syllable shape complexity. Within each phase, cues were reduced from auditory-visual to visual only to delayed feedback. The treatment protocol required participants to replicate native speaker model sounds of the trained syllables while monitoring a computer screen to try to achieve an F3 below 2300 Hz. Generalization trends were monitored with weekly probes on 71 untrained words containing /r/ in a variety of positions. Effect sizes (d) were also calculated to evaluate magnitude of change between the baseline phase (N = 5 probes) and the final probe for each participant. Potential changes in perceptual distinguishing /r/ and /l/ in minimal pairs was also evaluated. Both participants showed improvement in trained syllables with evidence of generalization to a number of untrained syllable shapes. Effect sizes for trained and untrained items varied from 2 to 4.6, respectively for P1 and from 1 to 6 for P2. Perceptual discrimination of /r/ and /l/ scores rose 6 percentage points for P1 and 15 percentage points for P2. These preliminary results suggest that spectrographic visual biofeedback for /r/ production may be an effective treatment approach for training /r/ in Japanese speaking Learners of English.
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 Iomi B Patten.
Thesis:
Thesis (M.A.)--University of Florida, 2012.
Local:
Adviser: Edmonds, Lisa Anna.
Electronic Access:
RESTRICTED TO UF STUDENTS, STAFF, FACULTY, AND ON-CAMPUS USE UNTIL 2014-05-31

Record Information

Source Institution:
UFRGP
Rights Management:
Applicable rights reserved.
Classification:
lcc - LD1780 2012
System ID:
UFE0044211:00001


This item is only available as the following downloads:


Full Text

PAGE 1

1 TRAINING JAPANESE L1 SPEAKERS IN AMERICAN ENGLISH L2 PRODUCTION OF THE PHONEME /R/ USING VISUAL BIOFEEDBACK IN THE FORM OF A SPECTROGRAM: A PRELIMINARY STUDY By IOMI PATTEN A THESIS PRESENTED TO THE GRADUATE SCHOOL OF THE UNIVERSITY OF FLORIDA IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF ARTS UNIVERSITY OF FLORIDA 2012

PAGE 2

2 2012 Iomi Patten

PAGE 3

3 To every traveler who has ever yearned to say more: it is my wish that you have every opportunity t o share your own words, in your own voice, in a way that all may understand you; for these shared words make the whole world a far richer place. And to my grandmother, Mary Elbert, for leading the way.

PAGE 4

4 ACKNOWLEDGMENTS I am deeply grateful to the Aph asia and Bilingualism laboratory at the University of Florida for the resources, guidance, and encouragement provided; without these, this project would not have been possible. My committee has been integral to this entire process: many thanks to Dr. Judi th Wingate for the spark, and to Dr. Lisa Edmonds for opening the door. I am grateful to all of the faculty and staff in the University of Florida Speech Language and Hearing Sciences department who have supported and encouraged curiosity. Idella King, Cas sie Mobley, and Dalila Johnson were ongoing sources of support and help. Dr. J. Rosenbek, Dr. Jamie Reilly, and Yukari Nakamura Sensei Dr. Mehmet Yavas (FIU) for origina lly introducing me to phonology and phonetics (and particularly second language phonology), and to Dr. Kenneth Logan for demonstrating exactly how phonological acquisition training can be enjoyable for both instructor and learner. Thank you also to Angela LaGambina and Daniel Furnas for crucial support, advice, assistance, and so much more. Neal Musson: thank you for putting together all of the dusty pieces over and over again until it worked. To my friends and family: thank you for your patience and suppo rt while I disappeared from your lives for two years. I appreciate all of the English language learners in my classrooms over the years; you propelled me toward this work, and I thought of you daily. Most of all, I am grateful to Participant 1 and Particip ant 2 you were both such a pleasure to know. Ganbatte!

PAGE 5

5 TABLE OF CONTENTS page ACKNOWLEDGMENTS ................................ ................................ ................................ .. 4 LIST OF TABLES ................................ ................................ ................................ ............ 7 LIST OF FIGURES ................................ ................................ ................................ .......... 8 LIST OF ABBREVIATIONS AND SYMBOLS ................................ ................................ .. 9 ABSTRACT ................................ ................................ ................................ ................... 10 CHAPTER 1 BACKGROUND AND INTRODUCTION ................................ ................................ 12 Overview of English Language Learning ................................ ................................ 12 Japanese /r/ ................................ ................................ ................................ ............ 13 ................................ ................................ ...... 15 From a Dichotomy to a Continuum ................................ ................................ ... 16 Interspeaker and Intraspeaker Variability ................................ ......................... 17 ................................ ................................ ................................ ............. 18 Summary ................................ ................................ ................................ ................ 21 Defining the Task ................................ ................................ ................................ .... 21 F3 Visual Feedback and / /: Previous Training Research ................................ ....... 22 Research Questions ................................ ................................ ............................... 24 2 METHODS ................................ ................................ ................................ .............. 28 Participants ................................ ................................ ................................ ............. 28 JL1 Participants ................................ ................................ ................................ 28 Native Speaker Models ................................ ................................ .................... 29 Study Design ................................ ................................ ................................ .......... 30 Stimuli Development ................................ ................................ ........................ 30 Treatment stimuli ................................ ................................ ....................... 30 Probe stimuli ................................ ................................ .............................. 31 Control stimuli ................................ ................................ ............................ 31 Perception stimuli: Listening t ask for pre and post treatment testing ........ 32 Suggestions list ................................ ................................ .......................... 32 F3 Height Criterion Selection ................................ ................................ ............ 32 Procedure ................................ ................................ ................................ ............... 33 Pre Test ................................ ................................ ................................ ............ 33 Baseline ................................ ................................ ................................ ..... 33 Perception task minimal pairs ................................ ................................ ... 34 Treatment Procedure ................................ ................................ ....................... 35 Task familiarization ................................ ................................ .................... 35

PAGE 6

6 Treatment ................................ ................................ ................................ ... 36 Stimulus Progression ................................ ................................ ....................... 37 Stages: Reduced cueing ................................ ................................ ............ 38 Progression ................................ ................................ ................................ 39 Participant preference/request ................................ ................................ ... 40 Probes ................................ ................................ ................................ ....... 40 Treatment scoring and reliability ................................ ................................ 41 3 RESULTS ................................ ................................ ................................ ............... 45 Participant 1 ................................ ................................ ................................ ............ 45 Participant 2 ................................ ................................ ................................ ............ 47 4 DISCUSSION ................................ ................................ ................................ ......... 56 Research Question 1 ................................ ................................ .............................. 56 Research Question 2 ................................ ................................ .............................. 57 Research Question 3 ................................ ................................ .............................. 58 General Discussion ................................ ................................ ................................ 58 Limitations and Future Directions ................................ ................................ ........... 64 APPENDIX A PARTICIPANT SELF RATINGS ................................ ................................ ............. 66 B TREATMENT STIMULI ................................ ................................ ........................... 67 C UNTRAINED PROBE STIMULI ................................ ................................ .............. 68 D CONTROL STIMULI ................................ ................................ ............................... 69 E PERCEPTION MINIMAL PAIRS ................................ ................................ ............. 70 F PERCEPTION ANSWER SHEET ................................ ................................ ........... 71 G SUGGESTIONS SHEET FOR JL1 PARTICIPANTS ................................ .............. 72 H INSTRUCTIONS FOR NATIVE SPEAKER JUDGES ................................ ............. 73 I INTRODUCTORY TRAINING MATERIALS ................................ ............................ 74 J TREATMENT PROGRESSION AND CUEING H IERARCHY ................................ 80 LIST OF REFERENCES ................................ ................................ ............................... 81 BIOGRAPHICAL SKETCH ................................ ................................ ............................ 85

PAGE 7

7 LIST OF TABLES Table page 3 1 Participant 1 production effect size calculations. ................................ ................ 52 3 2 Participant 1 raw production data. ................................ ................................ ...... 52 3 3 Participant 2 production effect size calculations. ................................ ................ 52 3 4 Participant 2 raw production data. ................................ ................................ ...... 54

PAGE 8

8 LIST OF FIGURES Figure page 1 1 Japanese and American /r/ interaction. ................................ .............................. 26 1 2 ................................ ......................... 27 2 1 Training screen shot A. ................................ ................................ ..................... 43 2 2 Training screen shot B. ................................ ................................ ....................... 43 2 3 Training screen shot C. ................................ ................................ ...................... 44 3 1 Participant 1: trained and untrained item results by percentage correct. ............ 49 3 2 Participant 2, A: trained item r esults by percentage correct. .............................. 50 3 3 Participant 2, B: untrained item and control results by percentage correct. ........ 51

PAGE 9

9 LIST OF ABBREVIATION S AND SYMBOLS C Conso nant JL1 First Language Speaker of Japanese who is also a learner of English L1 First Language L2 Second Language NNS Non Native Speaker of English NS Native Speaker of English r The rhotic North American /r/, how it is speci fically pronounced Alveolar tap or flap (similar to what is produced by American V Vowel

PAGE 10

10 Abstract of Thesis Presented to the Grad uate School of the University of Florida in Partial Fulfillment of the Requirements for the Degree of Master of Arts TRAINING JAPANESE L1 SPEAKERS IN AMERICAN ENGLISH L2 PRODUCTION OF THE PHONEME /R/ USING VISUAL BIOFEEDBACK IN THE FORM OF A SPECTROGRAM: A PRELIMINARY STUDY By Iomi Patten May 2012 Chair: Lisa Edmonds Major: Communication Sciences and Disorders Background: Japanese speakers who are learners of English have difficulty both producing and perceptually identifying the phonem variable ways by native speakers of English, so there is little concrete articulator movement or placement instruction that would be of help to these learners. However, unusually low for this phoneme (below 1950 Hz; but productions under 2300 are also perceived as correct). Japanese speakers do not auditorily attend to this formant because it does not hold identifying linguistic in formation in the Japanese language, rendering auditory perception alone an unreliable cue as well. This study uses spectrographic visual biofeedback to encourage these learners to visually identify the low F3 and to replicate it in their own productions. A single subject multiple baseline design was implemented with two participants. Treatment stimuli included words with / hierarchical phases of increased syllable shape complexity. Within each phase, cues were reduced from auditory visual to visual only to delayed feedback. The treatment

PAGE 11

11 protocol required participants to replica te native speaker model sounds of the trained syllables while monitoring a computer screen to try to achieve an F3 below 2300 Hz. Generalization trends were monitored with weekly probes on 71 untrained words containing /r/ in a variety of positions. Effect sizes ( d ) were also calculated to evaluate magnitude of change between the baseline phase (N = 5 probes) and the final probe for each participant. Potential changes in the perception of distinguishing / minimal pairs was also evaluated. Both participants showed improvement in trained syllables with evidence of generalization to a number of untrained syllable shapes. Effect sizes for trained and untrained items ranged from 2 to 4.6, respecti vely for P1 and from 1 to 6 for P2. production may be an effective treatment approach for traini learners of English.

PAGE 12

12 CHAPTER 1 BACKGROUND AND INTRO DUCTION Overview of English Language Learning It would be very difficult to overstate the importance of second language (L2) spoken English in the world today. Though it is har d to get an accurate count, there are an estimated 1 2 billion people currently learning English as a second (i.e., in an English speaking environment) or international/foreign (i.e., in a non English speaking environment) language worldwide. In fact, non native speakers of English (NNSs) now outnumber native speakers of English (NSs) (Jenkins, 2002). English frequently and lack of English can be a barrier to career advancement (Graddol 20 06; Carlson & McHenry 2006). Increasingly, English proficiency is a requirement for entrance into institutions of higher education (Graddol, 2006). In English Language Learner (ELL) instruction, there is a general trend toward ignoring pronunciation instru ction in favor of vocabulary, grammar, and literacy skills. In fact, pronunciation training has been all but abandoned in many classrooms (See Saito & Lyster, 2010 and Derwing & Munro, 2005 for more details). This is due to a variety of factors: most teac hers are NNSs, frequently with the same L1 influence on pronunciation that the students have; both NS and NNS teachers may feel underprepared or unsure of what to teach; there is a dearth of empirical evidence on how to teach pronunciation in the education al materials and literature and when such research does exist, it is underutilized (Derwing & Munro, 2005; Jenkins, 2002). A recent survey of adult English learners who had been studying for an extended period of

PAGE 13

13 time revealed that only 8 in 100 had been taught anything about pronunciation (Derwing & Munro, 2005). There are a number of components that comprise what we perceive as intelligible speech, including rate, word and syllable stress, intonation, and the individual phonemes of the words. Thus it is possible that any of these elements could be trained to start with phonology because we know that certain segmental errors can cause breakdowns in intelligibilit y (Lambacher, 2010). Frequently, second language phonology is taught by articulator placement or by imitation. Articulator placement training is effective for many consonants, and for some ive the target sound and then hypothesize a set of articulatory gestures to produce that sound. Both methods are through audition or proprioception. However, when articu lator placement target is variable, includes more than one or two points of constriction, or includes articulators for which there is poor tactile and kinesthetic feedback, instruction becomes appreciably s a challenge for many English learners, and most particularly for Japanese speakers who are learners of English (referred to hereafter as JL1s), for whom perception of this phoneme is also unreliable. This is the concept we will explore. Japanese /r/ Ja panese does not contain a phoneme that accurately resembles /r/. Japanese /r/, alveolar tap or flap [ Yamada, & Yamada, 2004; Lambacher, 2010).

PAGE 14

14 This phoneme is produced with the tongue tip in light contact with the alveolar ridge and Jenkins even /l/), and may actually be more similar to NS productions of flapped /t/ and /d/ (the St yle and pace of articulation may play a role. According to Flege, Takagi, and Mann (1995), Japanese /r/ particularly sounds like flapped /d/ when it is produced rapidly, and more like /l/ when it is produced emphatically, since there is lateral airflow. Na tive discussed by Flege et al., 1995). The flap also includes a good deal of allophonic variation (Lotto, Sato, & Diehl, 2004). Japanese /r/ includes allophonic variations th at overlap acoustically with American vowels are typically mapped in F1/F2 spac e) ( Figure 1 1 ). They found that: [T]he flap distribution partly overlaps the optimal boundary between /l/ and /r/. Exemplars of both English liquids fall within the flap distribution in F2 x F3 space. That is, Japanese speakers would categorize some exemp lars al.,2004). This overlap in distribution thus includes part of both phoneme categories (though slightly more in /l/ space than in / as well as further space that does not include either phoneme or the boundary between them The consequences of this ov

PAGE 15

15 /l/ in English there is a poorly defined boundary between the two sounds (Lotto et al., 2004). (which are frequently another English phoneme such as [d], or possibly as an ambiguous or uncategorizable sound in English (Flege et al., 1995). Past research has shown wide variation and many outliers in this perce ptual identification, including clusters such as /gr/, /dl/ and /wl/, or /r/ productions, but occasionally /w/, /d/, /br/, and /bl/. a group, share properties not only with consonants, bu t also with glides. Thus, liquids are not the same as traditional consonants such as stops, but rather share some vowel articulatory contact the way most consonants do. space the way vowels can be mapped in F1xF2 space which is not the case for most consonants (this is discussed more later in this paper, but see Lotto and colleagues for ll known, though for the purposes of McReynolds, 1975). Between the two phonem

PAGE 16

16 exhibits high F3 (Kent & Read, 2002). th ere is narrowing, but not closure, in the palatal region. Though there is some variation in descriptions (and names) of these two positions, they can be broadly defined as: the wered The reasons one person produces retroflex position while another produces bunched are still unclear. Individual differences in vocal tract anatomy and size do not appe ar to be the deciding factor in whether an individual uses bunched or retroflex / Zhou, Espy Wilson, Boyce, Tiede, Holland, and Choe (2008) describe a case of two male subjects who produce remarkably different bunched and retroflex shapes despite sharing very similar vocal tract shape, size, and overall anatomy. Importantly, it has been demonstrated that listeners do not perceptually identify (Guenther, Espy Wilson, Boyc e, Matthies, Zandipour, & Perkell, 1999; Zhou et al., 2008). The main acoustic/spectrographic difference between them is seen only in higher purposes they are the same 2008). From a Dichotomy to a Continuum Figure 1 2 ). However, a body of recent

PAGE 17

17 literature has revealed that these two points are not dichotomous but rather are merely general endpoints of what is actually a continuum of articulator movement in the ashi, & Lindstrom 1998; Espy Wilson, Boyce, Jackson, Narayanan, & Alwan, 2000). As noted above, this continuum follows the same acoustic pattern throughout the characteristic low F3 dip. Such extreme articulator variability may at first seem counterintui tive how can so many different lingual movements create the same sound? Articulation of / limited, however, to lingual movement and resultant palatal constriction. The primary constrictions: labial (lip rounding), lingual/palatal (raising the tongue), an d pharyngeal (narrowing of the pharynx) (Bradlow, 2008). Locations, shapes, and degrees of these constrictions vary widely and interact with one another dynamically (Zhou et al. 2008 ; Espy Wilson et al., 2000). Each of these movements individually contribu tes to the lowering of F3, but all three are used by native speakers interactively to create / pharyngeal constriction and/or lip rounding in order to maintain sufficiently low F3 height Further complica ting the matter is the fact that while these three main constrictions are relatively well documented in the literature, additional vocal tract areas have been hypothesized to contribute to the lowering of F3, such as sublingual space and the pyriform sinus es (Espy Wilson et al, 2000). Interspeaker and Intraspeaker Variability Westbury and colleagues (1998), found, in regards to lingual placement for [

PAGE 18

18 [T] here are as many kinds of /r/ as there are speaker by word same, though many were closely alike. Guenther and colleagues (1999), using an electromagnetic midsagittal articulometer (imaging of transducers adhered to the tongue) to measure articulator placement in 7 NS subjects, demonstrated that articulator position varied widely across subjects. All 7 participants exhibited strong trading relations between tongue back height and ton gue front horizontal position changes in one part of the tongue position (e.g. tongue back height) caused changes in another part of the tongue position (e.g. tongue front horizontal position). Importantly, wide variation was also seen within certain sub jects, in different utterances. For example, some speakers used exclusively a different configurations depending on word phonetic context or prosodic variation (Espy shown et al., 1999). All of these variations continued to show the same acoustic cue: lowered ake it so. Despite the clear articulatory differences, JL1s often perceptually identify same is true for /l/) (Aoyama et al., 2004). Additionally, it has been demonstrated JL1s

PAGE 19

19 1). How can this be so? What are they listening to for identifying information? It has been shown that in the speech stream, Japanese speakers are attending most to F2, which holds high salience in liquid, from the glide /w/. (Iverson, Kuhl, Akahane Yamada, Diesch, Tohkura, Ketterman, & Siebert, 2003; Yamada & Tohkura,1992; Lotto et al., 2004). Studies (e.g. Iverson et al., 2003) have clearly shown that JL1s simply do not attend to F3 because it doe s not hold salience i n Japanese for distinguishing one phoneme from the next (in just the same way that native speakers do not attend to F4 Miyawaki et al. (1975) demonstrated that Japanese spe akers could identify and distinguish differences in F3 equally as well as native speakers do when F3 was synthetically removed from the speech stream. Thus, this is not an issue of whether or o it because it does not hold any phonological information in Japanese. are taught anything at all. This position is gesturally more similar to /l/ than to bunched this English /l/ or /d/. Thus if a JL1 is creating muscle tension by e xcessively attending to

PAGE 20

20 tongue front shape and position, it would logically follow that further ambiguities and production of these erroneous substitutions could arise. create s obvious difficulty with the concept of articulator placement training. If each differently for each context, how might we go about training a nonnative speaker in the positions (Guenther et al. 1999; Westbury et al., 1998), how may we lead a learner to The very nature of the articulators and movements may provide further challenge. First, while lip movement may be very easy to visualize and describe, the constriction tongue within (Lambacher, 2010). Next, the types of tongue positions used by native speakers vary widely, and even if one of the two endpoint positions is selected (e.g. retroflex or bunched), it does not then become an easy task: neither position includes true points of contact for tactile feedback, and placement cueing may even be detrimental, creating exaggerated movements (Schuster, Ruscello, & Toth, 1995). Then, pharyngeal constriction is all but impossible to describe, let alone to accurately monitor through proprioception and/or to intentionally control (and if the pyriform sinuses and sublingual these three primary movements at the same time and in just the ri is yet a further challenge.

PAGE 21

21 consistent across all native speaker articulatory approaches: the lowered F3. Summary ing of F3. It is a highly variable and intricate sound produced by native speakers in a wide range of manners, both between and within individuals, creating a highly complex articulatory movement target for ELLs. The articulators and movements involved are not simple, easily definable, nor even completely understood. This is further complicated by the fact that each NS articulator movement interacts with other movements in a compound interdependent dance in order to maintain the most important acoustically Thus, articulator placement cueing is a poor option at best. attendance to F3 in the speech stream. JL1s do not have a reliable metric by which to discern what is and is not an /r/ in either perception or production. This, along with the which JL1s are essentially grasping at sounds in an attempt to meet the target, and as a result, their comprehensibility is compromised. Self perception alone is also, therefore, not an appropriate or reliable cue. Defining the Task trying to assemble a mysterious multi dimensional puzzle with your hands concealed behind a curtain. How is one to know what the puzzle must look like? How does one know if one is moving closer to, or farther from, the goal? What is clearly needed here is a well defined target, and reliable feedback on progress.

PAGE 22

22 If the idea is that we need to create a reliable / something concrete to reference. That thing is clearly F3. Saito and Lyster (2010) found important factor is low F3 height. But Because a spectrogram displays a clear visual depiction of formant shapes and heights, it allows learners to visualize and understand their own productions of formants without relying on perception (which we k now is problematic for JL1s) or articulator position (for which there is no clear, single target). Furthermore, the spectrogram provides an opportunity for learners to compare their own productions to those of native speakers in a single glance. Spectrog raphic measurement of formants has been used in many studies as a excellent example of this). Over time, efficacy and ease of use of this technology have combined wit h lowered cost to create a climate in which it is now a relatively reasonable tool for clinical use. F3 Visual Feedback and / /: Previous Training Research F3 training has been used previously in speech therapy for adult and pediatric populations. Schuste r, Ruscello, and Smith (1992) describe a case study of a young speech therapy treatment. With a combination of traditional articulator placement therapy and spectr ographic biofeedback, the participant was able to accurately produce rated by two experienced judges as correct.

PAGE 23

23 In a follow up study, Schuster and colleagues (1995) provi ded similar treatment to responsive to the biofeedback, was sessions, when he was transferred back to traditional therapy methods and continued to level by the end of the study. More recently, McAl lister and Hitchcock (In Press) produced a similar study with eleven children. Of the eleven, eight demonstrated test, as identified by independent judges and F3 formant height measurements. Of these eight, four demonstrated generalization Spectrograms have also been used as visual biofeedback with language learners, primarily for vowel instruction. Brett (2004) described the in progress development of a program to train Italian L1s on Standard British English vowels through formant interfaces with PRAAT Motohashi Saigo and Hardison (2009) used visual biofeedback in the form of waveform displays to trai n American students of the Japanese Language in perception of geminates (long sounds which are contrastive in Japanese but not in English). The results showed auditory visual feedback training significantly improved perception for words (from 65% correct t o 88% correct) and

PAGE 24

24 sentences (from 54% to 87% correct). This was found to be more effective than auditory only feedback. Furthermore, this auditory visual perception training showed some generalization to production. Additionally, Lambacher (2010) used spe ctrogram training instructionally in the classroom for JL1s in Japan, for production of all phonemes in all positions (including / and reports that students using spectrogram training to improve product ions (by better matching of spectrographic patterns to a teacher model), he did not provide a systematic analysis of the effect of the treatment on trained and/or untrained stimuli. Thus, there is preliminary evidence that spectrogram training holds promi se for improving individual phonemes in monolingual children with developmental articulation have been no experimentally controlled studies evaluating the effect of / F3 feedback for JL1s, which is the aim of the current study. Research Questions Research Question 1 : Does spectrographic visual biofeedback of F3 height Research Question 2 : Does spectrographic visual biofeedback of F3 height positions? speaker judges in t he University of Florida Aphasia and Bilingualism Lab. Our

PAGE 25

25 generalization t o untrained words and positions in JL1s. However, if the extensive is conceivable that improvement to untrained words in the same positions would improve. Whether ge neralization will occur to untrained positions is less certain. Research Question 3 : Does spectrographic visual biofeedback of F3 height Previous literature is somewhat unclear on this topic, however, there is evidence that perception can improve in adult JL1 speakers. For example, a length of residence effect has been demonstrated such that JL1s who have resided in the US for an extended period (mean=21 years) have exhibited imp roved perception of the distinction between / l position (which was trained) increased by 18%, but with little generalization to other positions. Whether F3 visual feedback (coupled with perception of their own production) hat it is possible.

PAGE 26

26 Figure 1 frequencies obtained from productions of English /l/ and /r/ and the Japanese flap. a) Native English productions of /l/ and /r/; b) English produ ctions of /l/ and /r/ with distribution of native Japanese productions of the flap; c) Native Japanese productions of /l/ and /r/; d) Native productions of flap with L2 Diehl (2 004). Reprinted with permission.

PAGE 27

27 Figure 1 formant height (Hz) in the X axis. The three formants are clearly visible, as the beginning of the word opening up at transition to a visibly higher F3 in the vowel. Note: this syllable shape.

PAGE 28

28 CHAPTER 2 METHODS Participants JL1 Participants Two parti cipants were enrolled in the training program. Both participants are native speakers of Japanese that were living near the University of Florida. Both self reported normal vision, hearing, and learning capabilities. Both were initially exposed to English in middle school in Japan but reported no significant increases in fluency until arrival in the US. Both participants were enrolled in English as a Second Language classes in the local community concurrent to treatment. Both participants travel to Japan a nnually or biannually for approximately one month, during which they exclusively speak Japanese. Please see information below and Appendix A for more detailed language information about each participant. Participant 1: This participant, age 33 years, sta rted learning English in middle school with exclusively JL1 instructors, but she was not consistently exposed to English until her arrival in the US four years prior to enrolling in this study. After 6 months in the US, she returned to Japan for approximat ely 1 year, after which she returned to the US and has resided here for the last 2.5 years. Since her arrival she has been inconsistently enrolled in local conversation classes sometimes taught by NSs. She has also worked with a language partner (a NS with whom she converses 50% in Japanese and 50% in English for to provide language practice for both individuals). Her current education involves <3 hours weekly of English Language conversational instruction in a group setting with a volunteer instructor. She reports that none of her English courses

PAGE 29

29 have included a significant pronunciation component. She currently speaks Japanese 100% of the time at home, and 50% Japanese 50% English in social situations. Participant 2: Participant 2, age 43, also started l earning English in middle school with exclusively JL1 instructors and was not consistently exposed to English until her arrival in the US 6.5 years ago. Since her arrival she has also been inconsistently enrolled in English Language courses at adult school s in group settings, and reports that there has been no specific focus on pronunciation in any of her courses. Her current educational situation includes concurrent enrollment in twice weekly general English Language group classes. She reports that she exc lusively speaks Japanese at home, though it should be noted that some members of her household are bilingual with native like fluency in American English, and these individuals frequently speak to her in English, though she exclusively responds in Japanese Outside of the home, she typically speaks English. Four additional participants were consented into the study, but they were not performance on the baselines was too high to allow for improvement and one was fluent in a third language. Native Speaker Models Two participants, one female (NS1 Production Model) and one male (NS2 Perception Model) were recruited by IRB approved flyers from the University of Florida campus. NS1 was an undergraduate student, and NS2 was a graduate student. Both participants are monolingual native speakers of American English standard dialects, and neither has worked, lived, or studied abroad for more than a few weeks. Neither

PAGE 30

30 has received more than two years of college level instruction in a foreign language, nor spent extensive time with English language learners. Study Design A multiple baseline single subject design was used (e.g., Kearns, 1986). Five baseline sessions were conducted. There after, the effect of treatment to the production of various trained and untrained syllable shapes (described in Treatment Stimuli) was evaluated throughout the treatment phases (described in Treatment Procedure). Stimuli Development Treatment stimuli A hie rarchical treatment list of 15 words was developed. In an attempt to maximize utility of treatment and realism of items, real words or highly word like nonwords were selected. The words developed for Phase 1 of treatment consisted of 5 CV syllables with i Phase 1 vowels included a variety of positions, from the low back lax vowel /a/ (e.g. included, Appendix B for full list of stimuli). These stimuli follow the recommendations of word st atus. However, while participants were encouraged to match each model vowel, they were never penalized for inaccurate pronunciation of vowels (or incomplete diphthongs) the focus Phase 2 words consist of five r initial CVC words and nonwords utilizing the same

PAGE 31

31 was chosen for this phase in order to add complexity with a phoneme that is both generally unmarked and has no lingual component, as wel l as being both common to and early learned in both languages (Preston & Seki, 2011). Phase 3 words follow a similar pattern to Phase 2, with a /b/ replacing the /m/ in ected due to its similarity to /m/ and its commonality to both languages (Preston & Seki, 2011). All training stimuli were recorded by the female native speaker participant. She was instructed to read each word individually and clearly in as natural a way as possible. Probe stimuli Seventy one untrained real (initial, medial, and final along with voiced and voiceless consonant clusters of varying complexity) were developed to measure any generalization of treatment to production of Appendix C for complete list. Control stimuli Fifteen control stimuli included the voiced interdental fricative // in similar positions to those included in treatment stimuli. This phoneme was selected based on its relative difficulty and its absence in the Japanese language (Thompson, 2001). Complete matching of controls to treatment items was not possible due to the limited occurrence of // in common English words, but // was included i n word initial, word medial, and word final positions ( Appendix D for complete list).

PAGE 32

32 Perception stimuli: Listening task for pre and post treatment testing Sixty simple r l minimal pairs were developed in four positions (initial, medial, final, and clus ter), for a total of 120 additional words. None of these words was included in included /l/. Each selected word (minimal pair half) was recorded by the male native speaker (perception model) for a total of 60 items. Suggestions list A series of general suggestions was developed in order to aid participants in attempting to match the spectrog ram. These included broad general suggestions such well, but in such a way that no specific recommendation or bias was introduced, (e.g. b Appendix G for the complete list. F3 Height Criterion Selection describe F3 to be under 1950 Hz for most NSs, regardless of gender, age, or vocal tract size (Westbury et al. 1998; Saito & Lyster, In Press). It is important to note, however, the 2200 range. Though these measurements may first appear to be outliers, there is (typically the measures include males). These F3 measurements, though outside the typical range reported by others, still demonstrated t a pattern of relative difference between the F3 of /r/ and the F3 of the adjacent vowel

PAGE 33

33 2300 r NS judges. With these results in mind, we selected 2300 as a reasonable limit of Procedure Pre Test Baseline Participants completed a total of 5 baseline sessions each over a two week period. During each session, all 101 stimuli were presented on index cards in randomized order (15 treatment stimuli + 71 probe stimuli+15 control stimuli =101). All subjects were tested individually by a single examiner in a quiet roo m in the Aphasia and Bilingualism Lab at UF. Recordings were made using the PRAAT program, downloaded from http://www.fon.hum.uva.nl/praat/ on agreement with the Visi Participants were seated at a normal desk a comfortable distance from the microphone (Shur model SM 78) and asked to read each in dex card. No instructions were given as to pronunciation of the elicited words and sounds. Participants were given breaks periodically throughout recording and were informed that they could request a break at any time. Recording the 101 stimuli took approx imately 10 25 minutes. Each participant was asked (and agreed) not to practice or attempt to learn any of the words in any way throughout the duration of the study.

PAGE 34

34 During baseline collection, no feedback related to pronunciation accuracy of any kind was provided. In the early sessions, positive feedback of task performance was early sessions that they did not know how to say a word, they were instructed in a frie ndly manner that they should try their best and that this would meet the requirements of the task to satisfaction. Perception task minimal p airs The perception task was performed after the baselines were complete during session for Baselines 3 and 4. Be fore beginning the listening task, participants were provided with a short 4 item familiarization task that matched the target task, but with a different (female) speaker. The 60 minimally paired words developed for this study were presented in two equal halves during consecutive baseline recording sessions. Participants were presented with a recording of one asked to write what they heard. In order to reduce non target cognitive load and any unrel ated variation in vowel perception or spelling errors, stimuli were presented with an accompanying answer sheet upon which were typed cloze type (gap fill) written spell ings of the real answer, and to minimize any effect of word spelling on letter choice (e.g. the minimal n as to how many sounds or which sounds might be written into the spaces provided, and the two target phonemes were never mentioned explicitly (i.e. participants were told rced

PAGE 35

35 (Appendices E and F for full list of minimal pairs and corresponding answer sheet). Treatment Procedure Task familiarization Treatment was introduced in treatment session 1 phoneme was given verbally, without any specific articulatory instruction. Participants were informed that native speakers produce the /r/ sound in a variety of ways (with a variety of articulatory gestures), but that a ll of these methods produced a certain spectrographic image. The Visi Time Spectrogram module was then participant was asked to repeat the utterance into t he bottom screen. A very brief orientation followed, in which the beginning and ending of the word was pointed out, and it was noted that each utterance may look a little bit different even though it is still correct. Subsequently, participants were famili arized with the target spectrogram through a series of printed images highlighting a) the target formant (F3) b) the treatment target depicting high verbal feedback until the participant was familiar with the task ( Appendix I ). with continuous live spectrographic feedback. Periodically, the investigator stopped the live spectrographic stream to point out formants and formant shapes. This period lasted a few minutes until

PAGE 36

36 the participant indicated that she understood the relations hip between making sounds and seeing a spectrographic image. Treatment Participants were seated as above with the same microphone, computer, and speaker. For treatment, the Real 3950 was used for live spectr ographic display. This module provides a horizontal split screen, allowing for display of a target in one half of the screen and trials in the other half. Throughout treatment, participants were provided feedback exclusively related to the visual feedba ck. The native speaker spectrogram model was displayed in the top half of the screen as a visual feedback guide, which participants attempted to match in the bottom half of the screen through repeated trials. Improvement was defined by a low F3 (under, ove rlapping, or directly contacting the 2000 Hz mark, visually approximating Throughout the entire treatment, the experimenter was seated next to the participant and provided an monitor Figure 2 1 for a visual depiction). At no time was explicit articulatory instruction given. If partici pants requested suggestions sheet (Appendix G) and reminded that, because native speaker /r/ is

PAGE 37

37 to reach the In addition, at no time was the phoneme /l/ mentioned in any way by the experimenter, and no contrast or mention was made when participants produced /l/ d Stimulus Progression The treatment protocol was hierarchical with 3 Phases and 3 stages of reduced cueing within each phase. Each phase included 5 hierarchically ordered stimuli, with i ncremental increases in task difficulty. For example, Phase 1 began with the stimulus height and tension associated with the primary substitution errors (specificall y, /l/ and the voiced alveolar tap). Following this logic, Phase 1 stimuli presentation subsequently progressed upward through high back, rounded vowels, then low front vowels, and ( Appendix B ). It is important to note that because it took more time to meet criteria at the beginning of treatment, the phases did not take an equal amount of time. Phase 1, Stage 1 lasted considerably longer than subsequent stages and phases (4 weeks/12 13 sessions of treatment were dedicated solely to Phase 1, Stage 1, compared to around three sessions for subsequent phases). This provided ample opportunity for practice of the initial 5 stimuli, which were repeated in varying order throughout the first phase. Subsequent stage 1s included an initial group of 40 trials of each word for practice (whether or not criteria was met, to ensure adequate practice of newly introduced words) followed by a second group of trials with the same words in which the parti cipant attempted to meet criteria.

PAGE 38

38 Each phase added one incremental increase in task complexity. Phase 2 added the continuant /m/ in final position to the same CV syllables used in Phase 1. Phase 3 followed the same pattern, but with /b/ replacing /m/ in the final position. ( Appendix J for complete list and order of stimuli.) Stages: Reduced cueing In order to move toward more independent production, each phase consisted of three stages of reduced cueing conditions, each of which repeated the same set of 5 stimuli. Stage 1: stimuli (in order) in three ways: orthographically (index cards), auditorily (NS example recording), and visually (corresponding spectrogram of the NS example recording as well as real stimulus was presented orthographically on an index card. Then, the auditory native speaker model recording was opened and played in the top half of the window, providing a simultaneous spectrographic visual example across the screen. Participants were then asked to attempt to match the picture above by recording and viewing their own attempt in the bottom half of the screen. ( Figure 2 1 for examples). The nativ e speaker recording was played once before every participant attempt. When criteria were met for four of the five words, the participant moved to the next stage. Stage 2: Visual Only Condition. Stage 2 presented the same stimuli in the same order as in s tage 1, but this time without the auditory example (orthographic and visual cues only). Stage 2 presented the same stimuli orthographically and visually in the same order as in stage 1. However, the spectrographic example image was static and the native

PAGE 39

39 s peaker model recording was not played. Participants continued with trials in order, and criteria were the same. Stage 3: Delayed Feedback Condition. This final stage presented only an orthographic cue and delayed visual feedback. In stage 3, the same 5 words were presented orthographically only, and recorded on PRAAT After recording, the investigator asked the participant to give a binary up/down judgment on correctness (i.e. investigator, and the participant wa s able to view only her own corresponding spectrogram and results (without any NS example spectrogram for comparison). Progression For Stage 1s of each phase, participants were first provided at least 40 repetitions of each word in the set, to familiarize themselves with new words as they were introduced. After this had been completed, the word set was re started and attempts were made toward reaching criteria. For stages 1 and 2, participants were given 40 opportunities to meet criteria (a consecutive str ing of 8/10 correct) on each individual word. If criteria were met within 40 trials, the participant was moved to the next consecutive word in the set. This process continued until all words in the set had been attempted, or until the session time had elap sed. For Stage 3, participants were given 4 attempts at each word. Criteria to move on to the next phase an F3 measurement of 2300 or less during /r/ production of 8/10 total for all 5 words, with at least one correct answer from each word category. It is important to note that during the very first stage of treatment (Phase 1, Stage 1) participants were encouraged to meet the stricter criteria of an F3 under 2000 Hz if they could, though the criteria for stage progression were the same (under 2300 Hz).

PAGE 40

40 Th is stage lasted considerably longer than the other stages (approximately 1 month compared to a day or two) as the participants learned how to form the phoneme and attempted to meet criteria. During this first stage, the same 5 initial words were repeated i n the auditory visual feedback condition until criteria were met, as described above. Participant preference/request In order to reduce potential frustration when a word was difficult or repeated many times, participants were provided with three options: Move, Stay, or Free Practice. Any of these options were permitted at any time throughout treatment. ed participants to continue further practice on the same word. Participants typically selected this when incremental improvements were being seen but criteria had not yet been met (e.g. if the F3 shape was moving gradually lower in successive trials, but h ad not yet met criteria). production with live spectrographic visual feedback (i.e. without the customary pause required for measurement, data collection, and replaying of the a uditory example). During Free Practice data was not collected, and participants were encouraged to experiment freely and watch the F3 in order to experiment with their own articulatory hypotheses. Probes Once per week (or every third session) a recording was made of the full set of treatment, non treatment, and control words (101 total items). Probes were always administered before the treatment began. Just as in baseline testing, no feedback was given.

PAGE 41

41 Treatment scoring and reliability Trained native spe aker judgments were used for scoring baseline and probe items. Listener judgments are considered the gold standard for the evaluation of L2 pronunciation (Derwing & Munro, 2009; Saito & Lyster, In Press). After a period of working together to define rules and ensure good inter rater reliability, three trained assistant) listened to individual recordings of each baseline and probe utterance and assigned one of three judgments: 1) correct, in (another sound such a tap, /l/, or other), or 3) ambiguous, impossible to judge (infrequently assigned). See Appendix G for details on the scoring guidelines the raters used. Composite scores for each word were provided based on a) unanimous agreement or b) 2/3 agreement. Any scores with a three way disagreement (correct, incorrect, ambiguous) were re judged in a separate meeting of all three judges, who listened simultaneously and judged the items again. The raters did not know their original rating when they re scored these words. The new scores were combined in the same way as above unless a tie persisted, in which case the ambiguous rating was automatically assigned. Scoring reliability averaged 96% for P1 (unanimous plus 2/3) with an average of 62% unanimous ratings (4% of ratings resulted in a 3 way tie). For P2, reliability averaged 98% with 56% unanimous ratings (2% of ratings resulted a in 3 way tie). A trained research assistant observed o ne third of the training sessions live and recorded whether the clinician was adhering to the prescribed protocol with special attention on cueing to check for the use of any specific cues other than looking at the

PAGE 42

42 spectrogram (such as articulatory cues). Reliability was 99.2% for adhering to the training protocol for item presentation (one word presented out of order once), and 100% for avoidance of articulatory cueing or instruction.

PAGE 43

43 Figure 2 1. Training screen shot A. NS example is visible in the top half of each screen, orrect participant production (counts toward criteria). Figure 2 count toward criteria).

PAGE 44

44 Figure 2 indicating a tap or flap is visible, does not count toward criteria).

PAGE 45

45 CHAPTER 3 RESULTS To evaluate Research Question1 and Research Question 2 (magnitude of treatment phase for both participants. Effect sizes were calculated by de termining the difference between the final data probe point and the average of the baseline points (N knowledge, there are virtually no published studies with effec t size benchmarks on /r/ training results in Japanese English Language Learners, so the findings in the current study will provide preliminary benchmarks. Saito & Lyster (in press) calculated effect sizes for their production treatment study using F3 heigh t measurements, which is which is most similar to our task they had an effect size of 0.59). Also, results can be compared within participants across stimuli in order to gauge relative improvement, particularly with trained and untrained stimuli. The results for Research Question 1, 2, and 3 are described below. See Figures 3 1 3 2 and 3 3 for visual representation of the data and Tables 3 1 and 3 2 for the raw data. Partici pant 1 Research Question 1 : incorrect, and stable (i.e., non ascending) baselines were established for all syllables. She then participated in seven weeks of treatment with three sessions per week ( total of 13 sessions including introductory session). Before treatment began on the last day of

PAGE 46

46 treatment for each week, P1 read the same words she had read for the baseline sessions in order to gauge potential improvement. It was intended for P1 to contin ue through the entire protocol; however, she had to discontinue the training due to travel requirements. Consequently, P1 was trained on five r initial syllables ending in different vowels (Phase 1 stimuli), so the remaining 10 intended treatment items (fr om Phase 2 and 3) were added to initial position generalization measures for a total of 81 probe items (Research Question 2). The average baseline for Participant 1 was 2.0, and Probe 4 was 4.0. Thus, the calculated effect size between the baseline results and the last treatment phase probe (Probe 4) was 2.0. Research Question 2 : Research Question 2 investigated generalization to untrained initial words as well as to untrained words in untrained syllable shapes with initial baseline average was 10.80 (N = 27), and Probe 4 was 26.0 with an effect size of 4.65. Similar to the r initial syllable shape is the medial shape where the baseline average was 0.8 (N = 12) with a Probe 4 average of 5 resulting in an effect size of 5.02. The cluster shape baseline average was 4.4 (N = 19) with a Probe 4 of 8 resulting in an effect size of 3.16. Final position st arted with an average of 4.5 (N = 15) in baseline with a Probe 4 of 5, resulting in an effect size of 0.22. Finally, the control condition () had the lowest effect size (0.06) of all conditions, indicating that experimental control was exhibited. Researc h Question 3: In the pre treatment perception task P1 scored 33/60 for 55% correct. In post treatment (when Probe 4 was collected), she scored 37/60 for 61.67%.

PAGE 47

47 Participant 2 Research Question 1 : wa s rated as correct or incorrect. She then participated in 3 phases of treatment. The phases were kept in baseline in order to evaluate the potential effect of treatment on (Probe 5), the total was 4, resulting in an effect size of 1.6. At the end of this phase, the criteria within one week, and her probe was maintained at 100%. For phase 3 stimuli production returned to 80%, with an effect size of 1.5 over baseline. T he average baseline for Phase 2 stimuli was 3 and probe 7 was 4 for an effect size of 1. The average baseline for Phase 3 stimuli was 1.8 and probe 7 was 4 for an effect size of 1.5. Research Question 2 : Research Question 2 investigated generalization to u ntrained r initial words as well as to untrained words in untrained syllable shapes with / all r initial syllables, so generalization will look at the entire span of probes for each participant. The untrained r initial baseline average was 13 (N = 19), and P robe 7 was 18 with an effect size of 1.71. Similar to the r initial syllable shape is the medial shape where

PAGE 48

48 the baseline average was 4.8 (N = 11) with a Probe 7 score of 9 resulting in an effect size of 1.6. The cluster shape baseline average was 2.8 (N = 16) with a Probe 7 of 5 resulting in an effect size of 1.5. Final position started with an average of 4.2 (N = 13) in baseline with a Probe 7 of 15, resulting in an effect size of 6, and performance at ceiling. Finally, the control condition () had no ef fect size at all (0.0), indicating that experimental control was exhibited. Research Question 3: In pre treatment, the perception task was administered to P2 with an score of 37/60 for 62% correct. In mid treatment (collected before treatment during sessi on 15), she scored 46/60 for 77%.

PAGE 49

49 Figure 3 1. Participant 1 : trained and untrained item results by percentage correct

PAGE 50

50 Figure 3 2. Participant 2, A: trained item results by percentage correct

PAGE 51

51 Figure 3 3. Participant 2, B: untrained item and control results by percentage correct

PAGE 52

52 Table 3 1. Participant 1 production effect size calculations. P1 Probe 4 Baseline Average Baseline Standard Deviation d Trained Phase 1 4 2 1 2 Untrained UT Initial 26 10.8 3.27 4.6 5 Clusters 8 4.4 1.14 3.1 6 Medial 5 0.8 0.8 4 5.0 2 Final 5 4.5 2.30 0.2 2 Control 4 3.4 10.11 0.0 6 Table 3 2. Participant 1 raw production data. P1 Training Generali zation Control Phase 1 UT Initial UT Clusters UT Medial UT Final // N 5 27 19 12 15 15 Baseline 1 Raw 2 16 4 2 9 6 Percent 40% 59% 21% 17% 60% 40% Baseline 2 Raw 1 7 5 0 6 3 Percent 20% 26% 26% 0% 40% 20% Baseline 3 Raw 3 11 6 0 5 3 Percent 60% 41% 3 2% 0% 33% 20% Baseline 4 Raw 3 10 4 1 3 3 Percent 60% 37% 21% 8% 20% 20% Baseline 5 Raw 1 10 3 1 4 2 Percent 20% 37% 16% 8% 25% 13% Training Probe 1 Raw 1 21 5 5 2 3 Percent 20% 78% 26% 42% 13% 20% Probe 2 Raw 4 21 4 5 8 5 Percent 80% 78% 21% 42% 53% 33% Probe 3 Raw 5 25 7 5 6 4 Percent 100% 93% 37% 42% 40% 27% Probe 4 Raw 4 26 8 5 5 4 Percent 80% 96% 42% 42% 33% 27% Three words were removed from UT Initial N because of lost recordings from a probe or baseline: read rude and rusk ** Five words were removed from UT Final N because they were frequently pronounced as disyllabic dire ire mere spare and spire Note that no pronunciation instruction whatsoever was provided for u ntrained words.

PAGE 53

53 Table 3 3. Participant 2 production effect size calculations. P2 Probe 7 Baseline Average Baseline Standard Deviation d Trained Phase 1 4 2.2 1. 1 1.64 Phase 2 4 3 1 1 Phase 3 4 1.8 1.48 1.48 Untrained UT Initial 18 13 2.9 2 1.71 Clusters 5 2.8 1.48 1.48 Medial 9 4.8 2.5 9 1.62 Final 15 4.2 1.7 9 6. 03 Control 3 3 2 0

PAGE 54

54 Table 3 4. Participant 2 raw production data. P2 Training Generalization Control Phase 1 Phase 2 Phase 3 UT Initial UT Cluster UT Medial UT Final // N 5 5 5 16 19 11 15 13 Baseline 1 Raw 2 2 1 8 1 2 3 5 Percent 40% 40% 20% 42% 6% 18% 20% 38% Baseline 2 Raw 4 3 2 15 2 9 3 3 Percent 80% 60% 40% 79% 13% 82% 20% 23% Baseline 3 Raw 2 4 4 15 5 4 5 1 Percent 40% 80% 80% 79% 31% 36% 33% 8% Baseline 4 Raw 2 4 2 14 3 4 3 5 Percent 40% 80% 40% 74% 19% 36% 20% 38% Baseline 5 Raw 1 2 0 13 3 5 7 1 Correct 20% 40% 0% 68% 19% 45% 47% 8% Training Probe 1 Raw 3 5 3 12 5 5 9 3 Correct 60% 100% 60% 63% 31% 45% 60% 23% Probe 2 Raw 2 2 3 13 4 3 8 3 Correct 4 0% 40% 60% 68% 25% 27% 53% 23% Probe 3 Raw 2 3 3 16 5 3 13 4 Correct 40% 60% 60% 84% 31% 27% 87% 31%

PAGE 55

55 Table 3 4. Continued P2 Training Generalization Control Phase 1 Phase 2 Phase 3 UT Initial UT Cluster UT Medial UT Final // N 5 5 5 16 19 11 15 13 Probe 4 Raw 3 3 5 16 5 4 9 4 Correct 60% 60% 100% 84% 31% 36% 60% 31% Probe 5 Raw 4 5 4 18 7 6 15 6 Correct 80% 100% 80% 95% 44% 55% 100% 46% Probe 6 Raw 4 5 2 15 7 4 13 4 Correct 80% 100% 40% 79% 44% 36% 87% 31% Probe 7 Raw 4 4 4 18 5 9 15 3 Correct 80% 80% 80% 95% 31% 82% 100% 23% 7 words were removed from various position Ns because of lost recordings of a probe or baseline: deary, rusk, spar, drink, tray, teething, and them. ** Five words were removed from U T Final N because they were frequently pronounced as disyllabic with medial / /: dire, ire, mire, and spire. Note that no pronunciation instruction whatsoever was provided for untrained words.

PAGE 56

56 CHAPTER 4 DISCUSSION The purpose of this study was to eva luate the effect of spectrographic visual suggest an overall positive effect of the training on the production and perception of /r/ by JL1s. Research Question 1 Research Question 1 asked whether treatment improved production of / / in trained the words. P1 only participated in one phase of treatment on V syllable shapes (N=5). The trained items were variable (but did not rise) in baseline, suggesting an unstable represen tation of / /. The first probe did not reflect improvement, but the three final probes demonstrate more accuracy and increased stability. Based on visual inspection and the effect size of change, it appears that the trained items improved. However, one syl phase. This syllable shape was predicted a priori to be the most difficult because of the high front tongue position and tension involved serve to interfere with disambiguation from /l/ and flap production. Participant 2 engaged in three phases of treatment and thus was trained in on one syllable was evaluated on the other syllables. P2 also showed high variability in trained items during baseline (similar to P1). For probe 5, when only Phase 1 stimuli ( V) had been introduced, Phase 1 gains were d =1.6 (N=5). There is a linear trend pattern as P1. At Probe 6, for which Phase 2 stimuli ( Vm) had been introduc ed, Phase

PAGE 57

57 2 gains were d =2 (N=5). Phase 3 stimuli ( Vb) were introduced before Probe 7, and results demonstrate Phase 3 gain of d =1.5 (N=5) There was no improvement to control items, indicating maintenance of experimental control (i.e. observed improveme nt was not a result of stimulating the entire language system but was likely due to the treatment itself). Research Question 2 Research Question 2 asked whether this treatment demonstrated generalization from trained items ( V) to untrained items in initi al, medial, and final and cluster positions. Both participants demonstrated a trend of improvement from trained items to untrained items. in the generalization measures The items most similar to the treated items initial position demonstrated immediate increases and remained high (over 78% correct) throughout the remainder of treatment. Medial position began with a low baseline and immediately rose to 42% correct, whi ch remained stable throughout the remainder of treatment, for the largest effect size that P1 achieved (5.0). P2 also demonstrated generalization to untrained items, with large effect sizes for each untrained position. Initial position (which was most sim ilar to trained items) demonstrated an effect of 1.7. Dissimilar untreated items demonstrated gains as well, with final position being the largest ( d =6), followed by medial position ( d =1.6) and clusters ( d =1.4). Furthermore, for P2 during probes 1 5, Phas e 2 and 3 stimuli may be viewed as generalization measures because they had not been introduced yet. During that time increases were seen in Phase 2 stimuli ( d =2) and Phase 3 stimuli ( d =1.5).

PAGE 58

58 ores did rise on the final baseline measure, so results should be interpreted cautiously. Research Question 3 An overall increase was demonstrated in the discrimination abilities of both participants. It should be noted that because NS Model 2 (male) was instructed to produce the items in as natural a way as possible, the utterances were challenging (the rate was rapid and no attempt was made toward particular articulatory clarity). However, one of the excluded participants (who had lived in the US for an extended period) was able to judge them with 100% accuracy, thus the task is possible for a highly phoneme identification task (not forced choice); evidenced by the fact that the participants mark indicating that they had heard nothing or only the adjacent sound. improvement (15 percentage points), these results suggest an overall positive influence of production training on perceptual discrimination abilities. General Discussio n Both participants demonstrated a general linear upward trend in all trained and most untrained items over the course of treatment, suggesting that visual feedback of F3 may be a promising approach to training / / in JL1s Control was maintained, indicati biofeedback for JL1s. Thus there are few appropriate benchmarks (besides possibly

PAGE 59

59 Saito & and percentage increases suggest improvement. treatment, suggesting an overall solidifica tion of the phoneme from a volatile pre treatment state. This follows the concept that JL1s may be hypothesizing about the treatment, sometimes hitting the mark and sometimes missing it entirely, but with little ability to self mon itor their own productions for accuracy (following Goto, 1971). This treatment appeared to assist the participants in formulating a more generated articulatory hypotheses with reliable feedback. Thi s treatment appears to hold promise as an alternative to articulator placement instruction, as participants were free to hypothesize about and choose any combination of constrictions that created the effect they saw onscreen, with immediate knowledge of th eir results. While it was not possible to know which lingual position or positions were being used by the participants or how they may have changed over time as their approximations became spectrographically and acousticall y closer to native productions during later stages of treatment. P1 noted during her exit interview that was of help in achieving correct productions. Around Probe 5, P2 began to intermittently make a new type of substitution error in treatment trials: the back consonant /g/ began to appear occasionally in place of t substitutions before). Taken together,

PAGE 60

60 these suggest the possibility of the participants beginning to utilize a more back or bunched type lingual position for / / as treatment progressed. Anecdotally, all three judges independently noted qualitative im provements in category (i.e. a distorted / / rather than an entirely different phoneme) during probes though judged as in category for both conditions typically sounded more native like during probes. While both participants produced gains in generalization measures, t hey did so in markedly different ways. P1 improved most in medial position, with large gains also seen in UT initial position. P2, on the other hand, improved most in final position, with relatively small gains in medial position. This pattern suggests tha t treatment affected each participant differently generalized most to r gains in medial position, which could support this idea, as medial posi colored vowel generalization gains could be largely attributed to r colored vowel improvement. This lophone, with different levels of generalization to different allophones varying by participant with no particular pattern noted.

PAGE 61

61 Pre ings of overlapping category boundaries in production ( Figure 1 1 ). Both reported that before treatment, NS listeners would rticularly salient for P1, whose name 2000 Hz in the introductory sessi reproduce these results for another few sessions. Both reported that before treatment treatment with clarifying the situatio before it got better this may be due to a reorganization of the mental representation of the phoneme, occurring gradually word by word or sound by sound, and some of this variability can be seen clearly in the line graphs. Some words or categories appeared to improve incrementally, and then as a new challenge was introduced, dip down before finally rising. Affective and educational differences may have played a role in the variation between the results of P 1 and P2. P1 reported that over the course of her English language education she had never been taught any particular articulatory positioning for / hand, was instructed in strict retroflex articulator positioning, and informed by instructors that this phoneme would always be very difficult or impossible for JL1 speak ers. She

PAGE 62

62 muscle tension and a great deal of attention to articulator placement during every attempted production (at one point, she even asked Investigator 1 about surgical course, vehemently discouraged by the investigator). These past educational and affective differences also may have played a role in how P1 and P2 differed i n their tactile realization of what was causing a (alveolar tap or flap) during their / / productions As noted elsewhere, during treatment no articulatory cueing was provided. This meant that when a tap, flap, or other stop was produced, the only feedback the participants received was a blank space in the also mentioned that position after treatmen t) P2 struggled with this for a longer time, exhibiting frustration treatment, possibly because she had received articulator placement instruction for a tense retroflex posi tion in production of / taps or flaps during treatment sessions began to decline more precipitously than before.

PAGE 63

63 There is at least some indicatio n that through the course of treatment, both participants experienced a reorganizing or reshaping of their concept of the phoneme. Notably, P2 indicated in the early weeks that when the spectrogram showed a correct in treatment productions were becoming markedly more consistent and correct, she had nothing to do with langua ge. Between Probes 4 and 5, she mentioned that they supports the idea that F3 is not i findings about lack of self perception of correct tokens. These subjective comments provide fascinating insight int o the participant experience of gaining a new phoneme parameter, and are echoes of remarks made by all three of the young NS participants produced correctly (Schuster et al. 1992; Schuster et al., 1995). higher scores in the perception task may have been reflective of the one additional session of treatment that she received before testing Alternately, there may have been influence of the fact that she receives more NS like auditory input in her daily life, and thus potentially had more opportunity to generalize perception than P1 did. These (1982) finding that production can

PAGE 64

64 precede perception, as well as that individual results may vary (also see Catford & Pisoni 1970). Limitations and Future Directions One limitation was that only two females participated; investigation with more male and f emale participants is warranted due to some differences in F3 across gender. Further participant demographic diversity in areas such as age, length of study, country of residence, and even L1 would provide additional insight. Second, a more task specific d esign for the computer program could provide participants more control of the practice with a user friendly interface, or color coded target prompts (e.g. a red F3 center formant frequency, a blue line for the 2300 Hz target). Automatic data collection co uld provide a richer data set and fine grained analysis, such as F3 formant heights, response time measures in the no feedback condition, session data tracking, etc. Third, while we feel confident with the target F3 threshold of 2300 Hz, it would have been indi cation of how productions are perceived by nave NS listeners or NNSs; further exploration into any corresponding increases in overall speech intelligibility with these listener populations is warranted in order to most fully provide for the types of audie nces L2 speakers of English encounter in their daily lives. Fifth, detailed error analys e s of incorrect productions would likely show more detailed changes in changes over time. For example, substitutions of /l/, while relatively frequent in baseline, prec ipitously reduced once treatment commenced; by the final probe for each participant utterances

PAGE 65

65 distortions rather than substitutions. Variations in treatment design may also reveal more specific info rmation on how to most effectively provide this type of treatment. There may well be a way to train this sound in a shorter amount of time, and with more efficiency, for example if erroneous tap or /l/ production were explicitly described, with fewer stimu li, or with variations to frequency and intensity of treatment. Semi exploration is warranted. While Elbert & McReynolds (1975) describ e high efficacy for into stimuli selection and progression order are indicated. Furthermore, a large group study of spectrographic biofeedback with the addition o f articulator movement measures (such as with a midsagittal articulometer) could provide knowledge of which articulator positions are being used effectively by the learners and subsequently be extrapolated into more effective articulator placement instruct ion models for new learners in classroom and clinical environments. Overall, we believe that this application shows promising results and could have great potential for instructional use.

PAGE 66

66 APPENDIX A PARTICIPANT SELF RATINGS Participant 1 Please indicat e your own level of proficiency in English: Nonfluent Native Fluency Reading 1 2 3 4 5 6 7 Writing 1 2 3 4 5 6 7 Listening 1 2 3 4 5 6 7 Speaking 1 2 3 4 5 6 7 In your opinion, how much of a foreign accent do you have in English? Heavily acc ented Native like 1 2 3 4 5 6 7 Please rate how often others identify you as a non native speaker based on your accent in English: Always Never 1 2 3 4 5 6 7 Participant 2 Please indicate your own level of proficiency in English: Non fluent Native Fluency Reading 1 2 3 4 5 6 7 Writing 1 2 3 4 5 6 7 Listening 1 2 3 4 5 6 7 Speaking 1 2 3 4 5 6 7 In your opinion, how much of a foreign accent do you have in English? Heavily accented Native like 1 2 3 4 5 6 7 Please r ate how often others identify you as a non native speaker based on your accent in English Always Never 1 2 3 4 5 6 7

PAGE 67

67 APPENDIX B TREATMENT STIMULI Phase Stimuli 1. 2. 3. 4. 5. Phase 2 1. 2. 3. 4. 5. Phase 3 1. 2. 3. 4. 5.

PAGE 68

68 APPENDIX C UNTRAINED PROBE STIMULI UT Initial (20) UT Medial (12) UT Cluster (19) UT Final (20) Roy airy bray air rye Ari Bree are raid arrow br ew ear ram eerie bro ire read era draw or rep deary drew mar rhyme marry drink mare rib Perry gray mere rim sorry grew mire ripe tarrow grow more road sparrow pro bar rod starry tray bear rub tree beer rude true bore rum try dire rasp ab rade spar risk abridge spare roast agree spear roost aground spire rusk spore

PAGE 69

69 APPENDIX D CONTROL STIMULI CONTROL (15) they though thy them then these thine those thou bathe smooth teethe bathing seething teething

PAGE 70

70 APPENDIX E PERCEPTION MINIMAL PAIRS Perception Minimal Pairs (60) /l/ /l/ 1 care kale 31 rate late 2 race lace 32 core coal 3 rude lewd 33 pour pole 4 rake lake 34 rag lag 5 star stall 35 raced laced 6 tire tile 36 car call 7 raid laid 37 rode load 8 reek leek 38 near Neil 9 Ross loss 39 rot lot 10 rite light 40 tour tool 11 rice lice 41 dire dial 12 sear seal 42 sore sole 13 root loot 43 deer deal 14 dare dale 44 pre* plea 15 stare stale 45 cray clay 16 ran lan 46 rad lad 17 mart malt 47 tear* tale 18 tar tall 48 Ron lawn 19 pyre* pile 49 rock lock 20 craw claw 50 rope lope 21 pry ply 51 read lead 22 pray play 52 store stole 23 pair/pear pail/pale 53 rest lest 24 tier* teal 54 red led 25 peer peal 55 par pall 26 door dole 56 gore goal 27 crew clue 57 tore toll 28 rug lug 58 rain lane 29 rid lid 59 rust lust 30 reap leap 60 steer steal

PAGE 71

71 APPENDIX F PERCEPTION ANSWER SHEET Perception Answer Sheet 1 ka__e 31 __ate 2 __ace 32 co__e 3 __ude 33 po__e 4 __ake 34 __ag 5 sta__ 35 __aced 6 ti__e 36 ca__ 7 __aid 37 __oad 8 __eek 38 nea__ 9 __oss 39 __ot 10 __ight 40 too__e 11 __ice 41 di__e 12 sea__ 42 so__e 13 __oot 43 dea__ 14 da__e 44 p__ee 15 sta__e 45 c__ay 16 __an 46 __ad 17 ma__t 47 ta__e 18 ta__ 48 __o n 19 pi__e 49 __ock 20 c__aw 50 __ope 21 p__y 51 __ead 22 p__ay 52 sto__e 23 pa__e 53 __est 24 tea__ 54 __ed 25 pee__ 55 pa__ 26 do__e 56 go__e 27 c__ew 57 to__ 28 __ug 58 __ane 29 __id 59 __ust 30 __eap 60 stee__

PAGE 72

72 APPENDIX G SUGGESTIONS SHEET FOR JL1 PARTICIPANTS Ideas about what to try: try a lot of things and then look at the machine to see what happens. We can try as long as we want to. The purpos again. Here are some ideas: 1. Go slow. 2. Experiment. 3. Change one thing about your mouth. 4. Change two things about your mouth. 5. Try small changes. 6. Try big changes. 7. Change everything try something new. 8. Relax your tongue. 9. Relax your jaw. 10. Change the front of your tongue big or small, up or down, or just differently. 11. Change the back of your tongue big or small, up or down, or just diff erently. 12. Change the shape of your tongue (more flat, more bent, or just different). 13. Move your jaw up or down or just differently. 14. Move your throat open or closed or just differently (maybe this is hard to feel if 15. If you can t hink of new things to try, please try them! 16. 17. Have fun!

PAGE 73

73 APPENDIX H INSTRUCTIONS FOR NATIVE SPEAKER JUDGES R 3 category listening judgments: 1 Diphthongs that are glided so strongly as to b 0 Touch (For tap/touch of tongue on the palate this category includes voiced alveolar taps of all varieties, including abbreviated trills, flaps, taps, etc.). L D W ) Two (Two sounds or some extra consonant like sound in there: not just because they added a vowel or syllable elsewhere in the word, but the actual R location has two sounds in it.) Unfinished (The R might be started, but never finished or just trails off. Also use this for a deleted R or a vowelized R.) Epenthesis: If an /r/ etc), mark it incorrect even if the /r/ sounds right it is not being produced correctly in the called for position. Other just try to give a little note on what you heard/what sounds incorrect if you why. ? or otherwise unclear. Not a definite yes, but not a definite no. Give this a question mark in column B. If you have a comment on why it seems unclear, write it in column C. CONTROL Listening Judgment Incorrect: Voiceless Z S D Deletion Anything else tha t is NOT a voiced interdental fricative As above, mispronunciation of the word/vowels is allowed if it does not affect the position of the target phoneme.

PAGE 74

74 APPENDIX I INTRODUCTORY TRAINING MATERIALS

PAGE 75

75

PAGE 76

76

PAGE 77

77

PAGE 78

78

PAGE 79

79

PAGE 80

80 APPENDIX J TREATMENT PROGRESSION AND CUEING HIERARCHY

PAGE 81

81 LIST OF REFERENCES Aoyama, K., Flege, J. E., Guion, S. G., Akahane Yamada, R., & Yamada, T. (2004). Perceived phonetic dissimilarity and l2 speech learning: the case of Japanese /r/ and English /l/ and /r/. Journal of Phonetics, 32, 233 250. B est, C. T., & Strange, W. (1992). Effects of phonological and phonetic factors on cross language perception of approximants. Journal of Phonetics, 20, 305 330. Beeson, P. M., & Robey, R. R. (2006). Evaluating single subject treatment research: Lessons lear ned from the aphasia literature. Neuropsychology Review, 16 (4), 161 169. Bradlow, A. R. (2008). Training non native language sound patterns: Lessons from training Japanese adults on the English /r/ /l/ contrast (pp. 287 308). Amsterdam, The Netherlands: J ohn Benjamins Publishing Company. Bradlow, A.R., Pisoni, D.B., Yamada, R.A. & Tohkura, Y. (1997 ). Training Japanese listeners to identify English /r/ and /l/. Journal of the Acoustical Society of America, 101 (4), 2299 23. Brett, D. W. (2004). Computer gene rated feedback on vowel production by learners of English as a second language. ReCALL, 1, 103 113. Catford, J. C., & Pisoni, D. B. (1970). Auditory vs. articulatory training in exotic sounds. The Modern Language Journal, 54 (7), 477 481. Derwing, T. M., & Munro, M. J. (2005). Second language accent and pronunciation teaching: A research based approach. TESOL Quarterly, 39 (3), 379 397. Derwing, T. M., & Munro, M. J. (2009). Putting accent in its place: Rethinking obstacles to communication. Language Teaching 42 476 490. Elbert, M., & McReynolds, L. V. (1975). Transfer of /r/ across contexts. Journal of Speech and Hearing Disorders, 40 380 387. Espy Wilson, C. Y., Boyce, S. E., Jackson, M., Narayanan, S., & Alwan, A. (2000). Acoustic modeling of American En glish /r/. Journal of the Acoustical Society of America, 108 (1), 343 356. Flege, J. E., Takagi, N., & Mann, V. (1995). Japanese adults can learn to produce Language and Speech, 38( 1), 25 55. Gick, B., Bernhardt, B, P. Bacsfa lvi P., & I. Wilson, I. (2008) Ultrasound imaging applications in second language acquisition. In J. Hansen & M. Zampini (eds.) Phonology and Second Language Acquisition. Ch. 11, pp. 309 322. Amsterdam: John Benjamins.

PAGE 82

82 Goto, H. (1971). Auditory perception by normal Japanese adults of the sounds "l" and "r". Neuropsychologia, 9 317 323. Graddol, D. (2006). English next. London: British Council. Guenther, F. H., Espy Wilson, C.Y., Boyce, S.E., Matthies, M.L., Zandipour, M., & Perkell, J. (1999). Articulator y tradeoffs reduce acoustic variability during American English /r/ production. Journal of the Acoustical Society of America, 105 (5), 2854 2865. Hagiwara, Robert. (1995). WPP, No. 90: Acoustic Realizations of American /r/ as Produced by Women and Men. UC L os Angeles: Department of Linguistics, UCLA. Retrieved from: http://escholarship.org/uc/item/8779b7gq Hardison, D. M. (2003). Acquisition of second language speech: Effects of visual cues, context, and talker variability. Applied Psycholinguistics, 24, 495 522. Iverson, P., Kuhl, P. K., Akahane Yamada, R., Diesch, E., Tohkura, Y., Ketterman, A., & Siebert, C. (2003). A perceptual interference account of acquisition difficulties for non native phonemes. Cognition, 87, B47 B57. Iverson, P., Hazan, V., & Bann ister, K. (2005). Phonetic training with acoustic cue manipulations: A comparison of methods for teaching English /r/ /l/ to Japanese adults. Journal of the Acoustical Society of America, 118 (5), 3267 3278. Kearns, K.P. (1986). Flexibility of single subjec t designs. Part II. Design selection and arrangement of experimental phases. Journal of Speech and Hearing Disorders, 51, 204 213. Kent, R. D., & Read, C. (2002). Acoustic Analysis of Speech (2nd ed., pp. 179 183). Albany, NY: Singular Thomson Learning. Kn ock, T. R., Ballard, K. J., Robin, D.A., & Schmidt, R.A. (2000): Influence of order of stimulus presentation on speech motor learning: A principled approach to treatment for apraxia of speech, Aphasiology, 14 (5 6), 653 668. Lambacher, S. (2010): A CALL To ol for Improving Second Language Acquisition of English Consonants by Japanese Learners, Computer Assisted Language Learning, 12 (2),137 156. Liberman, A. M., Cooper, F. S., Harris, K. S., & MacNeilage, P. F. (1962). A motor theory of speech perception. Pro ceedings of the Speech Communication Seminar, Stockholm. Liberman, A.M., Mattingly, I.G, (1985). The motor theory of speech perception revised. Cognition, 21 1 36.

PAGE 83

83 Logan, J. S., Lively, S. E., & Pisoni, D. B. (1991). Training Japanese listeners to identif y English /r/ and /l/: A first report. Journal of the Acoustical Society of America, 89 (2), 874 886. Lively, S. E., Pisoni, D. B., Yamada, R., Tohkura, Y., & Yamada, T. (1994). Training Japanese listeners to identify English /r/ and /l./iii. long term rete ntion of new phonetic categories. Journal of the Acoustical Society of America, 96 (4), 2076 2087. Lotto, A.J., Sato, M., & Diehl, R.L. (2004). Mapping the task for the second language learner: The case of Japanese acquisition of /r/ and /l/. J. Slifka J., Manuel, S., & Matthies, M. (Eds.), From Sound to Sense: 50+ Years of Discoveries in Speech Communication. Electronic conference proceedings. Marian, V., Blumenfield, H. K., & Kaushanskaya, M. (2007). The language experience and proficiency questionnaire ( LEAP Q): Assessing language profiles in bilinguals and multilinguals. Journal of Speech, Language, and Hearing Research, 50, 940 967. McAllister Byun, T., & Hitchcock, E. (In press). Investigating the use of traditional and spectral biofeedback approaches to intervention for /r/ misarticulation. To appear in American Journal of Speech Language Pathology. Miyawaki, K., Strange, W., Verbrugge, R., Liberman, A. M., Jenkins, J. J., & Fujimura, O. (1975). An effect of linguistic experience: The discrimination of [r] and[l] by native speakers of Japanese and English. Perception & Psychophysics, 18 (5), 331 340. Motohashi Saigo, M., & Hardison, D. M. (2009). Acquisition of L2 Japanese geminates: Training with waveform displays Language Learning & Technology, 13 (2), 29 47. Munro, M. J., & Derwing, T. M. (1995). Processing time, accent, and comprehensibility in the perception of native and foreign accented speech. Language and Speech, 38 (3), 289 306. Preston, J. L., & Seki, A. (2011). Identifying residual speech sound disorders in bilingual children: a Japanese English case study. American Journal of Speech Language Pathology, 20 73 85. Saito, K., & Lyster, R. (In Press). Effects of form focused instruction and corrective feedback on l2 pronunciation d To appear in Language Learning. Schuster, L. I., Ruscello, D. M., & Smith, K. D. (1992). Evoking [r] using visual biofeedback. American Journal of Speech Language Pathology, 1 29 34. Schuster, L. I., Rusc ello, D. M., & Toth, A. R. (1995). The use of visual feedback to elicit correct /r/. American Journal of Speech Language Pathology, 4 37 44.

PAGE 84

84 Sheldon, A., & Strange, W. (1982). The acquisition of /r/ and /l/ by Japanese learners of English: Evidence that s peech production can precede speech perception. Applied Psycholinguistics, 3, 243 261. Strange, W., & Dittman, S. (1984). Effects of discrimination training on the perception of /r l/ by Japanese adults learning English. Perception & Psychophysics, 36 (2), 131 145. Westbury, J. R., Hashi, M., & Lindstrom, M. J. (1998). Differences among speakers in Speech Communication, 26 203 226. Yamada, R. A., & Tohkura, Y. (1992). The effects of experimental variables on the perception of American English /r/ and /l/ by Japanese listeners. Perception & Psychophysics, 52 (8), 1992. Zhang, Y., Kuhl, P. K., Imada, T., Kotani, M. L., & Tohkura, Y. (2005). Effects of language experience: Neural commitment to language specific audit ory patterns. NuroImage, 26, 703 720. Zhou, X., Espy Wilson, C. Y., Boyce, S., Tiede, M., Holland, C., & Choe, A. (2008). A magnetic resonance imaging based articulatory and acoustic study of "retroflex" and "bunched" American English /r/. Journal of the A coustical Society of America, 123(6), 4466 4481.

PAGE 85

85 BIOGRAPHICAL SKETCH Iomi Patten is from the San Francisco Bay Area. She received a Bachelor of Arts degree from the Friends World Program at Long Island University, including three years of study abroa d. She taught English to adult speakers of other languages for several years before earning a Master of Arts in Speech Language Pathology in 2012.