Abstract
Shadowing, a practice of repeating what one hears as simultaneously and accurately as possible, has been researched in the Teaching English to Speakers of Other Languages (TESOL) field for years. The research findings have shown that shadowing contributes to English as a Foreign Language (EFL) learners’ bottom-up listening skills, which leads to their overall listening comprehension skills. However, the accumulated research findings have not uncovered what aspects of bottom-up skills shadowing precisely contributes to. Thus, this study attempts to examine the aspects of bottom-up skills to which shadowing contributes and proposes a new shadowing procedure to compensate for the limitation of the current shadowing procedure. To this end, a preliminary study and a primary study were conducted. In the preliminary study, the bottom-up skill development through shadowing practice was precisely examined, using a 112-item bottom-up listening test. Thirty-six Japanese university students participated in the experiment and engaged in shadowing practice in eight lessons for a month. The result showed that shadowing practice was effective for developing the skill of identifying prominence in a speech, and word recognition skills but not effective for enhancing phonemic discrimination skills. In the primary study, to overcome the limitation of the shadowing procedure, a new shadowing procedure including three components of attention to output, corrective feedback, and explicit instruction was proposed. Twelve Japanese university students participated and engaged in the new shadowing procedure for three months. Their progress was assessed by a 32-item phonemic discrimination test, and the result showed that the new output-based shadowing procedure with explicit instruction and corrective feedback improved phonemic discrimination skills for intermediate level Japanese EFL learners.
Introduction
The practice of shadowing, ‘a paced, auditory tracking task which involves the immediate vocalization of auditorily presented stimuli’ (Lambert, 1992: 266), has been adapted from training simultaneous translators to English teaching. In recent years, shadowing has been widely researched internationally, including studies in Iran (Amoli and Ghanbari, 2013; Babapour et al., 2018), Puerto Rico (Guerrero and Commander, 2013), Canada (Foote and McDonough, 2017), Taiwan (Hsieh et al., 2013; Lin, 1998), and Malaysia (Omar and Umehara, 2010). Thorough and detailed academic books have also been recently published to summarize its theory and practice (Hamada, 2017; Kadota, 2019). Shadowing was first introduced to the Japanese EFL teaching context by Tamai (1992, 1997) and Murphey (2001). Since then, Japanese research has focussed primarily on the effect of shadowing on listening comprehension skills in EFL classroom teaching (e.g. Hamada, 2016; Kato, 2009; Mochizuki, 2006; Nakayama and Suzuki, 2012), the mechanisms of shadowing (Kadota, 2007; Kadota, 2019; Miyake, 2009; Oki, 2011; 2012; Onaha, 2003; Shiki et al., 2010), and learner psychology (Hamada, 2011).
The accumulated research has shown that shadowing contributes to language learners’ bottom-up listening process, which then impacts listening comprehension skills in their entirety. However, research on shadowing has reached a plateau and few new impactful findings have been presented recently. To advance the research and practice of shadowing, the function of shadowing needs to be more precisely examined, and where there are limitations, an adapted model of shadowing may need to be developed for the EFL context. To this end, the preliminary study presented herein will attempt to identify and examine the aspects of bottom-up skills to which the standard shadowing practice contributes. Subsequently, the primary study will attempt to propose a new shadowing procedure for language learning to compensate for limitations in the shadowing model.
Brief Review of L2 Shadowing Research
Shadowing, originally a technique to train simultaneous interpreters (see Lambert, 1991; Lambert, 1992), has evolved into a theoretically and empirically supported second language (L2) teaching technique for L2 learners (Hamada, 2016; 2017; Kadota, 2019). Shadowing is known for effectively helping L2 learners overcome their challenges in listening. Listeners often have difficulty with a fast speech rate and identifying individual words (Graham, 2006); put simply, ‘most L2 learners are not adept at bottom-up processing’ (Kissling, 2018: 653). When L2 learners are shadowing, most of the learners’ attention is directed to the sounds they are listening to, unlike when listening for communication, in which their attention is directed at both recognizing the sounds, as well as comprehension (Kadota, 2019). Therefore, as learners practise shadowing, their bottom-up processes in listening improve; they can recognize more words, which consequently smooths the listening process and leads to better comprehension (Hamada, 2017; Kadota, 2019). This is supported by empirical case studies conducted in the L2 classroom (e.g. Hamada, 2016; Kato, 2009; Mochizuki, 2006; Tamai, 1992; Tamai, 1997).
What is Missing in L2 Shadowing Research?
Despite the merits of shadowing, the recognition and uptake of this practice remains limited and it is only popular in several regions, particularly in Asia. The lack of popularity of shadowing is partly because current approaches to teaching listening hold that by making the best of their top-down processing and cognitive and metacognitive strategies (e.g. Field, 2008; Rost, 2011; Vandergrift and Goh, 2012), L2 listeners can compensate for their lack of bottom-up processing skills and still manage to communicate. Although it has long been debated whether L2 learners use bottom-up or top-down listening processes when listening (e.g. Vandergrift, 2004; Field, 2008), there is no doubt that the improvement of the bottom-up listening process strengthens L2 learners’ listening. In fact, research has proven that there is a fairly strong relationship between aural vocabulary and listening comprehension (e.g. Masrai, 2019). Matthews and Cheng (2015: 10) claim ‘the role of lower order listening skills in listening comprehension instruction should not be undervalued’, finding a strong relationship between word recognition and listening comprehension. In other words, for better listening comprehension, learners should be able to recognize words aurally, which occurs in the bottom-up listening process.
Although bottom-up listening skills are important, research has only found that shadowing improves bottom-up listening processing in general, leaving identification of the specific processing features that show improvement unclear. The identification of the specific improvement brought about by shadowing practice will generate more practical and concrete teaching implications, and if current shadowing procedure has weak areas, it needs to be developed.
To examine learners’ bottom-up listening skills, several aspects need to be tested, as summarized by Yeldham (2017). In fact, most case studies on shadowing have only shown that participants improved their listening for comprehension questions (e.g. Kato, 2009; Mochizuki, 2006; Tamai, 1997). Hamada (2016) indicated an improvement in learners’ phoneme perception skills using a 20-item partial dictation task; however, this is still not precise enough. Therefore, to advance shadowing as a teaching technique, research should analyse more details of L2 learners’ bottom-up listening skills development through shadowing practice and explore a potentially new shadowing procedure that would compensate for the existing limitations of the practice.
Purpose of the Study
The primary objective of this research is to develop a new procedure to compensate for any deficits of standard shadowing and provide a bridge between past and future studies. To achieve the ultimate objective, the research first aims to explore precisely which aspects of bottom-up listening skills are improved by shadowing for L2 learners and identify the scope and limitations of the standard shadowing procedure. It then aims to develop a new procedure to compensate for any deficits of standard shadowing.
Preliminary Study
The preliminary study’s research question looks at which aspects of bottom-up listening processing will be improved by shadowing practice.
Methods
Participants
A total of 36 Japanese university students (24 females, 12 males) participated in the experiment. Participants were non-English majors and had intermediate English proficiency level (estimated CEFR level: A2-B1). A contrast group was not established for the following reasons. First, given that the effectiveness of shadowing on bottom-up listening skill development has been proved (e.g. Hamada, 2016; Kadota, 2019; Kato, 2009; Mochizuki, 2006; Tamai, 1997), it is ethically questionable to set a contrast group that would not receive effective instruction for the duration of the study. Still, to control the other potential confounding effects as much as possible, a group of students that were non-English majors were chosen and no other English classes were given during the experiment period, thus it is likely that the study class was the participants’ only exposure to English. Prior to the experiment, written consent forms from participants for data use and collection were procured by the author.
Materials
For the study material, eight passages were chosen from the textbook, Reading Fluency 2 (Nation et al., 2018). Among the four levels of the Reading for Speed and Fluency series, Level 2 (CEFR A2+) was chosen, where each story is written in 250 words, using a vocabulary of 700 words, designed for speed reading (Nation et al., 2018: 4). This textbook was chosen to control the vocabulary size within an easy range and passage length as per the participants’ level.
In order to assess the participants’ bottom-up listening skills, a 112-item listening test that consisted of five sections was developed, by referring to the summative review of bottom-up skill assessment (Yeldham, 2017). As Field (2008) proposes, bottom-up processing requires the phonological skills to distinguish and identify the meanings of sounds, as well as word recognition skills. Along with this, Yeldham (2017) outlines the common assessment techniques as follows. In phonological skills, common assessment techniques include standard discrimination tasks, matched discrimination tasks, and underlining tasks; in word recognition skills, common assessment techniques include partial dictation, gap-fill sentences task, and paused transcription. As the primary function of shadowing is to direct learners’ attention to the sounds they are listening to, enhancing the lower process of the bottom-up skills, four tasks were selected among the choices; standard discrimination tasks and underlining tasks for the phonological skills, and partial dictation and gap-fill sentences task for the word recognition skills (see below). All tasks were played in North American English. The details of each assessment follow.
Standard Discrimination Task (Section I)
Phonemic discrimination tasks (Buck, 2001) were chosen to assess the participants’ phoneme perception skills of difficult phonemes for Japanese English learners (See Appendix). Minimal pairs are appropriate if participants share a first language and common problems of distinguishing certain types of sounds (Buck, 2001). Regarding the lists to be tested, the ‘five crucial segmentals’ (Saito, 2014: 265), /l, ɹ, ð, θ, v/, were selected. In Japanese English learners, the substitution of /l/ for /r/* and /b/ for /v/ are typically observed (Ohata, 2004). /ð and z/, /θ and s/, and / ð and d/ are other typically difficult contrasts because Japanese language does not have /ð/ and /θ/. The phonemic discrimination task makes it possible to examine whether participants can categorize the difficult phonemes and distinguish them from the close sounds in their L1 category. The participants heard only one of the words in each pair, and were asked to mark the word they had heard. For example, in item 1, participants heard the word ‘collect’, so they marked collect, rather than correct (See Appendix). Participants heard a total of 32 items only once. 1
Underlining Tasks (Sections II and III)
In underlining tasks, learners underline the stressed syllables on a transcript while listening to the speech (Celce-Murcia et al., 2010; Yeldham, 2017). The measurement of stress and intonation has three options: identifying stressed syllables in a word, words in a sentence, or intonation patterns (e.g. Harding, 2015). For this study, based on their English proficiency, participants were asked to identify the most stressed word in a sentence; the intonation pattern identification would have been too difficult while stressed syllable identification would have been too easy.
This task consisted of two sections: one, idenfitying the stress only in a sentence, and the other, identifying the stress in a conversation. In the former section, 15 sentences were taken from Harvard sentences (IEEE, 1969) and Gilbert (2012). As these were all independent sentences, it was difficult for listeners to predict the most stressed word from the context. Most words in the 15 sentences are in the 2000 words General Service List (GSL) vocabulary coverage, checked in Cobb (2002) and Heatley et al. (2002) to make sure the vocabulary was within easy range of the participants’ knowledge. The conversation transcript for Section III was taken from Celce-Murcia et al. (2010: 272), and the listeners were asked to identity the stress in 10 sentences. There was a pause between each sentence so that the listener had enough time to mark the answer in Section II, thus it was played once. In Section III, the entire passage was read at a relatively fast rate, approximately 161 words per minute (WPM), and no pause was given between the sentences, thus it was played twice.
Word Recognition Skills (Sections IV and V)
To assess word recognition skills, two types of tests were used: gap-fill sentences and a partial dictation test. The gap-fill sentences task, in which learners fill in the blank by listening to each sentence, was adopted to assess reduced speech (Yeldham, 2017). The gap-fill sentence task was also used in Gilbert (1993; Gilbert, 2005), although Gilbert’s tests did not specifically intend to assess reduced forms. Ito (2001) used reduced forms only, but they were grammatically predictable. Given these, in this study, Harvard sentences (IEEE, 1969) were chosen because they were designed for speech quality measurements; the gap was made up of function words to prevent learners from guessing through context. All the vocabulary in the 20 sentences was within easy range of the 2000 GSL level. The audio was played once.
Lastly, a partial dictation was used to decode stretches of reduced speech as in Yeldham (2017). In contrast to the gap-fill task, participants listened to the entire passage of 308 words, filling in the blanks while listening. The original passage was taken from Nation and Malarcher (2007: 23). Following Yeldham (2016), more blanks were made for function words, with 27 function words out of a total of 40; the remaining 13 were content words. The passage was read at approximately 168 WPM; thus, the audio was played twice as in Section III.
Procedure
The same pre-and post-tests were conducted before and after the experiment. Following this, the participants engaged in shadowing practice in eight lessons, twice a week, for a month.
The procedure in each class was adapted from Hamada (2016), which already showed positive results in a similar context to that of this study. Eight lessons were given in total. In each practice, the participants shadowed the audio twice without reading a script, and then shadowed while reading the script. After a few minutes of individual script check, participants shadowed once again. Finally, participants shadowed while recording their shadowing performance with an IC-recorder and reviewed their performance afterward, referring to Hamada (2015) and Nakayama and Suzuki (2011).
This shadowing practise was given for approximately 20 minutes in a 90-minute lesson. In the remainder of each class, the participants read two other passages, working on speed reading, as well as summary writing, and some discussion on the topic they had read. No listening practise using the audio of the passages was given.
Analysis
In the analysis, the descriptive statistics of the five sections were first calculated. Then, because Sections II and III examine the participants’ skill of finding the prominence, and Sections IV and V examine the learners’ word recognition skills, the data of Sections II and III, and IV and V were tallied. A paired-sample t-test was conducted on each.
Results
The descriptive statistics were first calculated (Table 1). In Section I (phonemic discrimination task), the group showed little progress from 22.36 to 22.94. In Section II and III (identification of prominence), the group increased from 16.72 to 18.36. In Section IV and V (word recognition skill), the group showed an increase from 30.47 to 36.08.
Descriptive Statistics and the Results of t-tests.
effect size: d = .60 (Small), 1.00 (Medium), 1.4 (Large) (Plonsky & Oswald, 2014).
A paired-sample t-test was performed for each (Table 1). In the phonemic discrimination task (Section I), no statistically significant difference was found with little effect size (d = 0.16). In underlining tasks of identifying the prominence (Section II and III), a statistically significant difference was found with a small effect size (d = 0.44). In word recognition skill (Sections IV and V), a statistically significant difference was found with a small effect size (d=0.63). In short, their scores on identifying the prominence and word recognition skill improved, but the phonemic discrimination skill did not; the standard shadowing procedure is not sufficiently effective at helping learners enhance their phonemic discrimination skills.
Primary Study
The Rationale of the Primary Study
The preliminary study found that shadowing is effective for enhancing the skill of finding the prominence and word recognition. However, limitations were found in that the standard shadowing procedure cannot help learners distinguish problematic phonemes of English. To advance shadowing as a more effective technique, the standard procedure needs to be revised to compensate for the deficit of lack of participant skill to distinguish similar phonemes in their bottom-up process.
Considering the nature of shadowing, involving both listening and speaking, the development of a new shadowing procedure lends its theoretical background from Speech Learning Model (SLM) (Flege, 1995) and research on pronunciation instruction (Kissling, 2018; Sakai and Moorman, 2018). First, as SLM (Flege, 1995) posits, accurate perception is necessary for accurate production and if pronunciation improves, it means perception also improves. Sakai and Moorman (2018) hold that speech perception and production modalities are connected, and empirical data shows that pronunciation instruction improves learners’ bottom-up processing for listening (Kissling, 2018). Combined, integration of production training to perceptual training will lead to perception skill development. Therefore, the primary study attempts to develop a new shadowing procedure that adapts pronunciation instruction to shadowing to develop solid learners’ phoneme perception skills.
With the above in mind, by analysing the results of the preliminary study and reviewing the related past studies, three proposals are given. The first proposal is to activate more attention to output. Kissling (2018) suggests that pronunciation instruction should include both perception and production practise, finding a crossover effect of pronunciation instruction on perception. The focus of the current shadowing procedure is mainly on listening, and thus the primary study aims to develop shadowing for both input and output. Additionally, the second proposal is to include corrective feedback, which was missing in the standard shadowing procedure. Borrowing the idea from pronunciation instruction, including corrective feedback into form-focussed instruction was more effective than form-focussed instruction alone for the acquisition of /r/, as in Saito and Lyster (2012). The third proposal is to integrate explicit instruction and shadowing to activate additional attention to the target phonemes. Saito (2013) found inclusion of explicit instruction to form-focussed instruction was more effective than form-focussed instruction alone for acquiring a difficult sound, /ɹ/, and additionally promoted noticing and awareness. In sum, the three proposals suggest a new pronunciation-based shadowing procedure, including corrective feedback and explicit instruction.
Taken together, the primary study thus aims to explore a new shadowing procedure to improve learners’ phonemic discrimination skills. To this end, the new procedure will include the three key components, namely, attention to output, corrective feedback, and explicit instruction, so that the participants can promote noticing and awareness of the difficult phoneme features. The research question is whether the new shadowing procedure improves phonemic discrimination skills for Japanese EFL learners.
Methods
Participants
Twelve university students (4 males and 8 females), aged from 18 to 20 and majoring in education, participated in the study. A small class size was chosen, as in the successful case study of Foote and McDonough (2017), which had 16 participants to conduct production-based shadowing. The English proficiency of participants was estimated at CEFR A2-B1 level. Admittedly, because this study was a longitudinal and exploratory study lasting for three months, it was impossible to set a contrast group to control all the variables. Also, the primary study required deep interactions (i.e. corrective feedback) between the instructor and the participants, along with occasional pair and group work under the instructor’s observation. For these reasons, a small class size, despite being composed of different participants to the preliminary study, was thought to be beneficial.
Materials
A 238-word passage, 90 seconds long, from a famous commencement speech by Steve Jobs at Stanford University in 2005, was selected for shadowing because it is authentic material and contains all the difficult phonemes for Japanese speakers tested in Section I (/f, v, θ, ð, l, ɹ/). The pre-and the post- tests from Section I in the preliminary study were used, as phonemic discrimination was the limitation of the preliminary study and the focus of the primary study. To learn how to pronounce the difficult phonemes, a website that explains the details of each phoneme was used (Eigonokai, 2019). The website explains the International Phonetic Alphabet (IPA) symbols and each consonant and vowel in Japanese, occasionally showing a picture of the mouth shape and tongue position.
Procedure
The procedure was developed based on the principle of including explicit instruction, attention to output, and corrective feedback, with integrations of other multiple shadowing-related activities. The overview of the procedure is summarized in Table 2.
Procedure of the Shadowing Project.
The goal for the participants was to be able to shadow the Steve Jobs speech, and they followed a build-up process. First of all, they transcribed the speech in IPA, as learning IPA is a salient process for pronunciation acquisition (Szpyra-Kozlowska, 2015). Then each week, they reviewed explicitly how the difficult phonemes for Japanese are pronounced; /l, ɹ/ /f, v/ /θ, ð/ /æ// ʌ //ə/ /ɪ/ / sɪ // ʃɪ // tɪ //w/ were selected. Because the experiment was conducted in the regular curriculum, in addition to the five crucial segmentals (/l, ɹ, ð, θ, v/) (Saito, 2014), other important phonemes were also taught. In parallel, participants worked on text-presented shadowing (Kuramoto et al., 2007), in which they shadowed with the help of the script written in Roman letters, to keep up with the speed and imitate the intonation of the target speech. Participants were also engaged in phonemic shadowing, in which they focussed mainly on the phonemes. Finally, participants presented their shadowing of the speech in class.
In more detail, on the first day in the first week, the pre-test was conducted. By Class #2 in the second week, the participants were required to transcribe the script in IPA, and in class, the scripts were checked against the model answer. Classes from #3 to #9 contained the explicit instruction and corrective feedback. Each week, a pair of designated students presented on how the target phoneme should be pronounced (e.g. /l/ and /r/ in Class #3), and participants practiced together, using examples from the Steve Jobs speech. Then, the instructor gave more detailed explicit instruction to make up for any gaps in the student presentations. After the participants understood the mechanism of the target phoneme, they practised the phoneme and the words that contained the phoneme (e.g. dropped, proportionally) individually or in pairs. The instructor checked each individual’s pronunciation and gave corrective feedback to the individual or to the entire class. Additionally, each week, participants were asked to practise shadowing with a special focus on the target sounds they learned in each lesson and submit a recorded performance of shadowing each week. For example, they reviewed /l/ and /r/ in Class #3, and by Class #4 they practised shadowing at home with a special focus on /l/ and /r/ and submitted their recording online. When practising, they could read the scripts in the letters or IPA. In classes #9 to #11, participants checked their shadowing performance in pairs and groups, using their scripts. Finally, on the presentation day, each participant shadowed the target speech in front of the audience.
Analysis
First, the descriptive statistics of the pre- and post- tests are shown. The 32 items are divided into four categories (/l/ and /ɹ/, /θ/, /ð/, and /v /and /b/), each containing 8 items. Then, to examine if the participants improved, a two-way ANOVA was implemented, with category being a between-factor and time being a within-factor.
Results
The descriptive statistics are shown in Table 3. In general, at the post-test, the scores were higher than at the pre-test. In the /r/ /l/ category, the participants showed progress from 6.00 to 6.67; in the /θ/ category, from 6.67 to 7.67; in the /ð/ category, from 5.67 to 6.08; and in the /b/ /v/ category, from 5.42 to 6.42. The two-way ANOVA found a statistically significant difference for time, F(1, 44) = 19.14, p<.01, ηp2 = .30 and category, F (3, 44) = 4.38, p<.05, ηp2 = .23, but not for an interaction, F (3, 44) = .65, p> .05, ηp2 = .04. This means that the group improved their phonemic discrimination skills. Among the four categories, the students appear to have improved distinguishing /θ/ most, judging from the mean score and the medium effect size (1.06).
Statistics of Each Category in the Phonemic Discrimination.
effect size: d = .60 (Small), 1.00 (Medium), 1.4 (Large) (Plonsky and Oswald, 2014).
Discussion
In answering the research question, the new shadowing procedure improved phonemic discrimination skills for the intermediate level of Japanese EFL learners. By incorporating the three principles of attention to output, explicit instruction, and corrective feedback, the pronunciation instruction (shadowing in this study) developed learners’ bottom-up listening skills, as reported in other studies (e.g. Sakai and Moorman, 2018).
It is assumed that the learners’ attention was more on pronunciation in addition to listening in the newly-proposed procedure, while exclusively on listening in the standard shadowing procedure. When practising shadowing only for the development of bottom-up listening skill in the preliminary study, learners did not need to worry too much about the accuracy of their reproduction, but when practising to simulate the target model in this experiment, they were required to copy as accurately as possible, so inevitably they attended to the phonological features more precisely.
In addition to this, the inclusion of explicit instructions and the occasional corrective feedback on the target phonemes may have promoted their noticing of the phonetic differences of the difficult target sounds. This study borrowed the ideas of integration of explicit instruction and corrective feedback from pronunciation research (Saito, 2013; Saito and Lyster, 2012), and these appear to fit well. During the process, especially in the feedback, the instructor carefully checked the six segmentals (/f, v, θ, ð, l, ɹ/). Typically, the instructor found the participants’ friction was too weak for /f/, sounding like /h/; quite often /θ/ was substituted with /s/; sometimes an overcorrection was observed (carrigraphy for calligraphy). Whenever the instructor found these, corrective feedback was given. With the combination of the explicit instruction and occasional feedback, it is presumed their conscious awareness of each phoneme was raised and their distinguishing skills improved.
The theoretical perspective adds support to the positive result, by comparing the standard procedure and the newly-proposed procedure. When learners practise following the standard procedure, they mainly try to recognize the sounds, and thus while they could recognize the difficult sounds, they may not have yet been able to distinguish the two similar sounds. The Perceptual Assimilation Model (PAM) holds that L2 listeners assimilate the sound that is close to their L1 category (Best and Tyler, 2007). For example, when they hear /r/, they can recognize the sound by using their L1 category, but when they are asked to choose one from the two similar minimal pairs by listening, they become confused because /r/ and /l/ fall in the same L1 category. In order to clearly distinguish two similar phonemes, they need to create a new phonetic category (e.g. Flege, 1995). However, the standard implicit shadowing procedure did not help the participants with this process. In contrast, the new procedure appeared to have overcome this limitation by generating more attention to output with corrective feedback and explicit instruction. By attending not only to the input but also the output, the learners needed to not only recognize the incoming sounds but also distinguish them clearly in order to reproduce them. Still, acquiring new sounds missing from their L1 without explicit instruction (i.e. how to pronounce) is challenging, so the explicit instruction and corrective feedback were necessary and were considered to be helpful. Combined, these may have led to creating a new phonetic category in their representational system. In short, the new output-based shadowing procedure, the explicit instruction, and corrective feedback compensated for the deficit of the standard shadowing procedure used in the preliminary study.
Limitations and Future Research Direction
Although the study demonstrates valuable findings, four issues may limit interpretation and need to be further developed to strengthen the findings in the future. First, the tests used in the experiments have room for further development. In this study, the tests used were created by referring to multiple previous studies, but more robust tests for bottom-up listening skills will hopefully be developed to better enable results among studies to be comparable. Second, neither the preliminary nor the primary studies had a control group. The preliminary study was compromised because the findings of past studies had already assured that bottom-up listening skills improved through shadowing; in addition, the primary study was longitudinal and exploratory, so it was difficult to set a control group with similar traits using a similar curriculum. Admittedly, setting a control group would be ideal. Additionally, it must be acknowledged that different groups of learners participated in the preliminary study and the primary study. Despite the merits of a small class in the primary study, the question remains as to whether the same result would have been obtained if the preliminary and the primary studies had used the same participants. Using the same group in both the preliminary and primary studies could strengthen the findings. Third, the number of the participants in the primary study was small. The small class size may have been one of the factors that contributed to the success of the study; a larger sample group and more data are needed to support the findings. Fourth, the primary study used output-based shadowing, but did not examine the pronunciation skill of participants as this was beyond the scope of the study. It is the author’s hope that, as this study has overcome the limitations of the current shadowing procedure, the research will shift to shadowing for pronunciation development, given that perception and pronunciation practice has been a hotly-researched area (Sakai and Moorman, 2018).
Conclusion
The preliminary study examined precisely what features of the bottom-up process can be improved by shadowing, and the primary study attempted to develop a new shadowing procedure to compensate for the limited function of standard shadowing. Taken together, the first valuable contribution of this research is that it shows the limitation of the standard shadowing procedure, in terms of being unable to help with learners’ phonemic discrimination skills. Shadowing research up until now has vaguely posited that shadowing improves learners’ bottom-up listening skills, and thus identifying the strengths and weaknesses of the traditional shadowing procedure was valuable. The new output-based shadowing procedure with explicit instruction and corrective feedback is found to help Japanese EFL learners of an intermediate level to better perceive difficult phonemes. Research shows practising difficult sounds and intensive exposure to them improve learners’ perception (e.g. Thomson, 2012), but shadowing can be more efficient as a teaching technique because it can deal with a wide range of target sounds in natural speech. The author hopes that this study triggers further research into using shadowing.
Footnotes
Appendix
Word list tested in the Phonemic discrimination task (Section I).
| 1. collect/ correct | 12. bow/ vow | 23. vest/ best |
| 2. clothing/ closing | 13. boat/ vote | 24. seem/ theme |
| 3. sing/ thing | 14. fly/ fry | 25. day/ they |
| 4. law/ raw | 15. ban/ van | 26. Dan/ than |
| 5. berry/ very | 16. breathe/ breeze | 27. alive/ arrive |
| 6. sick/ thick | 17. both/ boss | 28. lock/ rock |
| 7. very/ bury | 18. leader/ reader | 29. bowl/ vole |
| 8. clothe/ close | 19. long/ wrong | 30. bent/ vent |
| 9. then/ zen | 20. face/ faith | 31. worse/ worth |
| 10. right/ light | 21. bathe/ bays | 32. sin/ thin |
| 11. teething/ teasing | 22. mouse/ mouth |
Acknowledgements
I would like to thank everybody who helped me with this research, including Dr. Satoko Suzuki, Editor/Dr. Marie Yeo, and the reviewers for their helpful comments.
Funding
The author disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This research was supported by a JSPS KAKENHI Grant [18K00733].
