JLTA Journal, vol. 22: pp. 3–22, 2019 Copyright© 2019 Japan Language Testing Association DOI: 10.20622/jltajournal.22.0_3 Print ISSN 2189-5341 Online ISSN 2189-9746

Relationship Between L2 Vocabulary Learning Strategies and Semi-Contextualized Word Meaning Test Scores

Yusuke HASEGAWA

Joetsu University of Education

Abstract This study aims to reveal whether any specific type of vocabulary learning strategy (VLS) leads to higher scores on semi-contextualized word meaning tests—a multiple-choice gap-filling format in which short written contexts are provided. A total of 132 first-year university students learning English as a foreign language completed a VLS questionnaire and a semi-contextualized word meaning test. The relationship between these two variables was examined using Pearson’s correlation analysis, confirmatory factor analysis, and exploratory factor analysis. The results demonstrated that the relationships between VLS use and test scores were very weak (rs < .20), regardless of the strategy type. The smaller correlations compared to those reported in previous studies using vocabulary size tests may be caused by the more complicated constructs involved in the semi-contextualized word meaning test, which requires not only receptive knowledge about word meanings, but also reading comprehension skills and knowledge about word forms and usage in a sentence. However, imagery strategies, such as creating a mental image of word forms, had a very weak but significant positive correlation with the test scores. Based on these results, this study further discusses how Japanese high school students who will take examinations that employ the semi-contextualized word meaning test format should learn vocabulary.

Keywords: vocabulary learning strategy, semi-contextualized word meaning test, factor analysis

Researchers in the area of second language acquisition have recognized the importance of effective vocabulary learning strategies (VLSs) in the successful learning of a second language (L2). However, although lexical development is central to L2 learning, there is insufficient classroom time for students to learn the thousands of words they need to know to understand most spoken and written discourse. While many students can learn new L2 vocabulary efficiently by themselves, many fail to make such progress (Webb & Nation, 2017). Previous studies have demonstrated that the appropriate use of VLSs contributes to success in vocabulary learning and higher proficiency of English as a foreign language (EFL; e.g., Kojic-Sabo & Lightbown, 1999; Maeda, Tagashira, & Miura, 2003; Mizumoto &

－ 3 －

Takeuchi, 2009). According to Maeda et al. (2003), the frequency of VLS use differs between EFL students with higher and lower English proficiency. In their survey, they divided high-school students into three proficiency groups according to their scores on a cloze test that required participants to fill in the blanks in a short English text. Their correlation analysis demonstrated that high- and middle-proficiency learners were likely to use more VLSs, while high-proficiency learners tended to use so-called imagery strategies, an example of which, provided in their study, is to study an English word multiple times until a mental picture of that word can be evoked. Their results suggested that the effective use of VLSs leads to better performance in contextualized word production tasks, in which participants must write (i.e., produce) the most suitable English word in a given context. On the other hand, many English proficiency tests taken by Japanese EFL learners, such as high-school students, adopt a semi-contextualized test1 with a receptive—as opposed to productive—format, as a tool of lexical proficiency assessment.2 In contrast to the productive test, some receptive tests ask learners to select the best-fitting word to complete a short English text from multiple (often four) options. For example, the English proficiency survey for Japanese third-year high-school students by the Ministry of Education, Culture, Sports, Science and Technology (MEXT) in 2018 included the following question item:

Tyler said that many teachers had had a positive impact on his life, and of those, Mr. Smith, his English teacher, had ( ) him the most. [A] influenced, [B] mentioned, [C] proposed, [D] refused

(MEXT, 2018a, p. 20) In the item above, a point was given for selecting [A] as the most suitable word. A similar type of multiple-choice gap-fill test is used in the National Center Test for University Admissions (in Japanese, Daigaku Nyushi Center Shiken), a standardized test utilized by many public and private universities in Japan. The semi-contextualized word meaning test also comprises the first part of the Eiken test, one of the most widely recognized English language assessments in Japan, developed and conducted by the Eiken Foundation of Japan (formerly the Society for Testing English Proficiency [STEP]). Using the Eiken test and other types of lexical and reading comprehension tests, Morimoto (2006) demonstrated that semi-contextualized word meaning test scores reflect not only the test-takers’ lexical knowledge but also their reading ability. Considering that semi-contextualized word meaning tests have two aspects (i.e., reading and vocabulary), this may lead to the assumption that vocabulary learning methods involving reading contexts, such as example sentences, should lead to higher scores on semi-contextualized word meaning tests. Moreover, given the impact of the National Center Test for University Admissions and the Eiken test on Japanese high-school students, there is a great deal of concern over how to achieve higher scores on the semi-contextualized word meaning test.

－ 4 －

Based on the above, the present study aims to reveal whether any specific type of VLS can effectively improve contextualized lexical processes. Previous studies investigating the relationship between learners’ VLS use and lexical proficiency have mainly used vocabulary size tests without context. For instance, Akase and Uenishi (2011), who conducted a VLS questionnaire and vocabulary size test, found that participants with a larger vocabulary size used the imagery strategy (e.g., imagining the meanings of a target word) more often than those with a smaller vocabulary size. As in Maeda et al.’s (2003) study, the imagery strategy was found to be influential because the effect of the mental image associated to words led to the long-term retention of the learned words (see also Hasegawa, 2016). However, little has been reported on the relationship between VLSs and semi-contextualized word meaning test scores. This study anticipated that participants’ selection of VLSs would affect their scores in the semi-contextualized word meaning test. The following two research questions (RQs) were addressed: RQ1: Do any specific VLSs that Japanese university students used when they were high

school students have significant correlations with their scores on a semi-contextualized word meaning test?

RQ2: Do any VLS factors have significant correlations with the scores on a semi-contextualized word meaning test?

To explore these RQs, this study administered a semi-contextualized word meaning test for Japanese first-year undergraduates, which was created based on past materials of the Eiken test. This was coupled with a questionnaire that asked the students about the strategies they had used when they were high school students. This study adopted a questionnaire developed by Mizumoto and Takeuchi (2009), comprising 25 items that inquired about the frequency of VLS use. However, since this questionnaire did not include items that directly asked about dependency on translations or contexts, two questions regarding how often the learners had used translation- and context-based VLSs were added.

Method Participants In this study, 132 first-year university students (aged 18–19 years) majoring in Education responded to the VLS questionnaire and completed a semi-contextualized word meaning test. However, as five students were unable to participate in all parts of the study, their data were excluded from the analysis and the data obtained from 127 students were ultimately analyzed. The participants were informed that no data obtained from the study would affect their course grades. All participants were Japanese EFL learners. Before entering university, they had studied English for at least six years in Japanese junior and senior high schools. They had

－ 5 －

also participated in Foreign Language Activities (in Japanese, Gaikokugo Katsudo) in grades five and six. Their general English proficiency was estimated to be approximately Grade Pre-2 (CEFR A2) and Grade 2 (CEFR B1) of the Eiken test. It was assumed that all participants were familiar with the format of the semi-contextualized word meaning test because they had all taken the National Center Test for University Admissions before entering university; furthermore, all participants had taken another English lexical proficiency test with the same format prior to this research. Materials The VLS questionnaire. The VLS questionnaire was created based on Mizumoto and Takeuchi’s (2009) material (see Appendix A). Their original questionnaire comprised 25 items using a 5-point Likert scale ranging from 1 (not at all true of me) to 5 (very true of me). The questionnaire included seven Self-management strategies (e.g., “I regularly review the vocabulary I have learned to check if I remember it”), four Input-Seeking strategies (e.g., “I try to make use of media such as TV, radio, Internet, mobile phone, and movies to learn vocabulary”), five Imagery strategies (e.g., “When I try to remember vocabulary, I link my personal experiences to it”), three Writing Rehearsal strategies (e.g., “When I try to remember vocabulary, I write it repeatedly”), three Oral Rehearsal strategies (e.g., “When I try to remember vocabulary, I say it aloud repeatedly”), and three Association strategies (e.g., “When I try to remember vocabulary, I associate it with the synonyms or antonyms I already know”). All statements were translated from English to Japanese by the author of this article. Two items were added at the end of the questionnaire sheet. The first item asked about translation-based vocabulary learning: “When I was a high-school student, I placed priority on using Japanese translation while learning English words.” The second item asked about context-based vocabulary learning: “When I was a high-school student, I placed priority on using contexts while learning English words.” In the original statement, the term contexts was expressed as bun-myaku in Japanese, which may refer to the example sentences in a dictionary, texts in a course book, passages used in an exam, and any other types of written or spoken discourse. The semi-contextualized word meaning test. The semi-contextualized word meaning test was developed based on past materials used in Eiken Grade Pre-1 and Grade 2 in 2017, with written permission from the Eiken Foundation of Japan. First, 135 semi-contextualized word meaning test items were obtained from the past materials. All items had four options. An example is provided below: Q23. Before visitors enter the museum, their bags are ( ) by security officers. The

officers check to make sure visitors are not carrying any dangerous objects. (a) translated, (b) inspected (correct), (c) illustrated, (d) transformed

－ 6 －

According to Oxford Advanced Learner’s Dictionary (Hornby, 2015), the verb inspect has two meanings: “to look closely at something/somebody, especially to check that everything is as it should be” and “to officially visit a school, factory, etc. in order to check that the rules are being obeyed and that standards are acceptable.” In the above example, inspect is used in the first meaning. Most items of the vocabulary part of Eiken Grade Pre-1 and Grade 2 are designed to test examinees’ knowledge of the predominant meanings of the target words. Among the 135 items, those that assessed grammatical knowledge rather than vocabulary were excluded from the present study. This study also excluded items whose correct options were as easy as Levels 1 or 2 in the English vocabulary list called the New JACET List of 8000 Basic Words (Basic Word List Revision Special Committee, 2016). If very easy target words had been included, the test would have comprised a reading test rather than a vocabulary test because accurate comprehension of the context would be much more important than understanding the meanings of the option words. For the same reason, items whose context included words that were more difficult than the correct option were excluded. In the above example (i.e., Q23), all words in the context section (i.e., Before visitors ... objects) were regarded as more basic than inspect with reference to the JACET list. Consequently, the final version of the test comprised 50 items that satisfied the conditions required for this study. In the first 10 items, the correct answers comprised words belonging to Level 3 of the JACET list. In the same manner, the correct answers in the second, third, and fourth 10 items belonged to Levels 4, 5, and 6 of the JACET list, respectively. The answers to the fifth 10 items were not included in the JACET list. In the above example (i.e., Q23), inspect is the 4,411th word in the JACET list. Because the 1,000 words from 4,001st to 5,000th in the list belong to Level 5, this study regarded Q23 as a Level 5 item. However, since the participants’ English proficiency level was around Eiken Grade Pre-2 and Grade 2, test items based on the past material of Eiken Pre-1 were found to be too difficult.3 Therefore, only the first half of the semi-contextualized word meaning test was used for this study (i.e., 25 items belonging to Levels 3, 4, and 5 of the JACET list).4 Procedure The participants completed the semi-contextualized word meaning test (originally 50 items) within approximately 30 minutes. A month later, they answered the VLS questionnaire (27 items in total) within approximately 15 minutes. Before providing the questionnaire sheet, the author briefly explained the purpose of this investigation. The VLS questionnaire asked participants to recall the methods of vocabulary learning they had used in high school, without specifying the purpose or situation of vocabulary learning such as studying English for the university entrance examination. All explanations were provided in Japanese.

－ 7 －

Analyses Data treatment before the main analyses. This study mainly focuses on the relationship between participants’ VLS use and their test scores. However, before the main three analyses, the study examined the test reliability of the semi-contextualized word meaning test. The internal consistency of the 25 items was Cronbach’s α = .575. To improve the test reliability, some items whose reliability coefficients were lower than others were excluded from the main analyses because the test scores should reflect the participants’ ability as accurately as possible. After removing five items, the reliability coefficient became Cronbach’s α = .618. The scores of the remaining 20 items were used for the main analyses hereafter. The three main analyses. To examine RQ1, Pearson’s correlation coefficients were calculated between the semi-contextualized word meaning test scores and the 27 questionnaire items consisting of the 25 VLS items from Mizumoto and Takeuchi (2009) and two additional items inquiring about the participants’ dependency on translation and context while learning vocabulary (see the left part of Figure 1). The alpha level was set at .05 for all statistical analyses in this study.

Figure 1. Design of this study. The first analysis (left) examined RQ1 using Pearson’s correlation coefficients based on the 27 specific VLS items. The second analysis (middle) examined RQ2 using confirmatory factor analysis based on Mizumoto and Takeuchi’s six VLS factors. In addition, the third analysis (right) was carried out using exploratory factor analysis. In the third analysis, factor scores were calculated and RQ2 was reexamined using Pearson’s correlation coefficients. Regarding RQ2, a confirmatory factor analysis using structural equation modeling software was performed based on the 25 answers to the VLS questionnaire adopted from Mizumoto and Takeuchi’s (2009) study (see the middle part of Figure 1). In addition, the structure of factors in this study was different from those in Mizumoto and Takeuchi’s (2009) study; therefore, data of the 25 items were further subjected to an

－ 8 －

exploratory factor analysis, using the maximum likelihood method and oblique (i.e., promax) rotations. The numbers of factors were determined based on Mizumoto and Takeuchi’s six VLS categories. Any items that did not fit Mizumoto and Takeuchi’s model were excluded from the analysis. The results of factor analysis were further used to calculate factor estimates for each participant (i.e., factor scores). Pearson’s correlation coefficients were calculated between the factor scores and the semi-contextualized word meaning test scores (see the right part of Figure 1). Finally, the goodness of fit statistics were calculated again using the new confirmatory factor model. Two models are displayed in the Results section (i.e., Figures 3 and 4).

Results Results of the Semi-Contextualized Word Meaning Test The results of the semi-contextualized word meaning test are presented in Table 1. After removing the items that were not suitable for this study (see Materials and Analyses sections), the average correct rate of the 20 items was 51.77%. Table 1 Results of the Semi-contextualized Word Meaning Test

Frequency level No. of items M SD Range Level 3 10 6.36 (63.62%) 2.03 1–10 Level 4 7 2.70 (38.58%) 1.47 0–6 Level 5 3 1.29 (43.04%) 0.89 0–3 Total 20 10.35 (51.77%) 3.28 3–18

Results of the VLS Questionnaire Descriptive statistics. The mean values (Ms) and standard deviations (SDs) of the 27 questionnaire items are summarized in Appendix A. The items that demonstrated the highest and second highest mean values were Item 2 (i.e., “I keep a vocabulary book or word list to check the vocabulary anytime I wish”) and Item 20 (i.e., “When I try to remember vocabulary, I say it aloud repeatedly”), respectively. The least popular VLS was Item 6 (i.e., “I consciously set aside time to study vocabulary in order to prepare for tests such as TOEIC, TOEFL, or Eiken”). Results of confirmatory factor analysis. The model was constructed based on Mizumoto and Takeuchi (2009). However, the goodness-of-fit indexes were not sufficient: χ2 (260) = 425.51, p < .001, GFI = .80, AGFI = .75, RMR = .14, CFI = .84, TLI = .81, RMSEA (90%CI = .06, .08) = .07. Among the 25 items, Item 6 did not have a significant relationship with the latent factor (i.e., Self-management), p = .368; the other 24 items had significant relationships with the latent factors (ps < .05). 5

A further model was constructed where the six VLS factors would affect the semi-contextualized word meaning test score. However, the goodness-of-fit indexes were not

－ 9 －

sufficient: χ2 (279) = 422.22, p < .001, GFI = .82, AGFI = .77, RMR = .13 CFI = .88, TLI = .85, RMSEA (90%CI = .05, .08) = .06. As a result, none of the six factors, namely, Self-management (p = .806), Input-Seeking (p = .152), Imagery (p = .435), Writing Rehearsal (p = .152), Oral Rehearsal (p = .269), or Association (p = .298) had a significant effect on the score. This result may be due to the misfit between the data and the model. Therefore, in order to test whether any of the factors had a significant relationship with the test score, factor scores were calculated based on a new explorative factor analysis with the items that showed consistency with Mizumoto and Takeuchi’s (2009) assumption. Results of additional exploratory factor analysis. In the additional analysis, six factors were extracted again through the exploratory factor analysis. Because the model did not fit the current data well (χ2 (165) = 218.92, p = .003), this study excluded four items from analysis whose factor loadings were highest for factors that were different from Mizumoto and Takeuchi’s assumptions. The four items removed from analysis and their highest loading factors were Item 6/Factor 4 (originally Self-management), Item 10/Factor 5 (originally Input-Seeking), Item 11/Factor 4 (originally Input-Seeking), and Item 16/Factor 6 (originally Imagery). The remaining 21 items were all consistent with their original labels, resulting in a better fit, χ2 (99) = 121.18, p = .064. The factor loadings of each strategy item on each factor are presented in Appendix B. Factor loadings higher than .40 are in bold. After rotation, the six factors accounted for 54.24% of total variance. The six factors were labeled as follows: Writing Rehearsal (Factor 1), Self-management (Factor 2), Association (Factor 3), Imagery (Factor 4), Oral Rehearsal (Factor 5), and Input-Seeking (Factor 6). Based on this categorization, factor scores were calculated for each participant. The correlation coefficients between the factor scores and the test scores are displayed in the second subsection of the next section (shown as Table 2). Relationship Between VLS Use and Test Scores Results based on the 27 specific items (related to RQ1). Appendix C shows the correlation coefficients between the test scores and 5-point self-ratings for the 25 VLS items, and the two additional items regarding use of translations and contexts. Most importantly, Items 9 and 14 had a significant correlation with the test score, although both were very weak, rs = .18 and .18; ps = .045 and 040, respectively. Additionally, it was found that the use of translations had a very weak but significant positive correlation with the Association strategies (i.e., Items 23, 24, and 25) and that the use of context was related to the use of a variety of strategies such as Items 8 (Input-Seeking) and 22 (Oral Rehearsal). Results based on the new factor models (additional; related to RQ2). Table 2 summarizes the correlation coefficients between the test scores and the six factor scores. As a result, only Imagery showed a significant relationship with the semi-contextualized word meaning test score (r = .19, p = .030). However, the positive correlation between Imagery and the test scores was very weak, as visualized in Figure 2. The other factors did not show a significant relationship with the test score (ps > .05).

－ 10 －

Table 2 Correlation Between VLS Factor Scores and the Semi-Contextualized Word Meaning Test Score

VLS Test score F1 F2 F3 F4 F5 F1: Writing rehearsal −.07 ― F2: Self-management .09 .15 ― F3: Association .14 .00 .38** ― F4: Imagery .19* .06 .34** .31** ― F5: Oral rehearsal .07 −.04 .32** .46** .28** ― F6: Input-seeking .17 .28** .27** .25** .00 .43** Note. A single asterisk (*) refers to p < .05; double asterisks (**) refer to p < .01.

Figure 2. Scatterplot with a regression line showing the relationship between the factor scores for the Imagery strategy (the horizontal axis) and the semi-contextualized word meaning scores (the vertical axis). Finally, the new model constructed based on the data from 22 items shown in Appendix B was examined using confirmatory factor analysis. The model is displayed in Figure 3. Although the goodness-of-fit indexes were still insufficient, all indexes indicated a better fitting than those calculated based on the original 25 items: χ2 (174) = 262.11, p < .001, GFI = .84, AGFI = .79, RMR = .11 CFI = .90, TLI = .88, RMSEA (90%CI = .05, .08) = .06. All VLS items had significant relationships with the latent factors (ps < .01). Using this model, the relationship between VLS use and test scores was reanalyzed, as shown in Figure 4.

－ 11 －

Figure 3. The confirmatory factor analysis employed in this study. Data from the 22 questionnaire items (see Appendix B) were used for indexes of the six latent variables. Rectangles refer to observed variables; circles refer to latent variables. Circles labelled e refer to the errors of measurement. One-sided arrows indicate a standardized direct effect; two-sided arrows show correlation coefficients.

－ 12 －

Figure 4. Results of the confirmatory factor analysis with the six factors affecting the contextualized word meaning test score (see also Figure 3).

－ 13 －

In the final model (i.e., Figure 4), it was assumed that the six VLS factors would affect the semi-contextualized word meaning test score. Although the goodness-of-fit indexes could not be regarded as sufficient, the statistics indicated a better fit than the previous models: χ2 (189) = 274.794, p < .001, GFI = .84, AGFI = .79, RMR = .13 CFI = .91, TLI = .89, RMSEA (90%CI = .04, .17) = .06. However, none of the six factors, namely, Writing Rehearsal (p = .153), Self-management (p = .864), Association (p = .265), Imagery (p = .307), Oral Rehearsal (p = .298), or Input-Seeking (p = .123) had a significant effect on the score.

Discussion This study sought to answer two RQs in order to determine the relationship, if any, between participants’ success in a semi-contextualized word meaning test and their use of VLS. First, regarding RQ1, this study conducted correlation analyses using the 27 individual items of the VLS questionnaire. As a result, 25 of the items had no significant relationship with the test score. This may be due to the complicated constructs involved in the contextualized word meaning test, where both lexical knowledge and reading comprehension skills are required (Morimoto, 2006). The results suggest that scores of contextualized vocabulary tests, including the semi-contextualized word meaning test used in this study, can be predicted with a combination of VLS use and other variables such as L2 reading proficiency. Considering that this type of test measures a more flexible application of lexical knowledge than the usual vocabulary size tests, asking for mere combinations of word forms and their translations or synonyms, it is more difficult for learners to gain more scores by simply repeating some specific methods of vocabulary learning. For example, some teachers may consider that writing out new words multiple times enhances vocabulary retention, while others may suppose that the best way for learners to obtain higher scores on a semi-contextualized word meaning test is to increase the amount of context-based learning methods. Another presumption may be that vocabulary learning methods relying on translations would have a negative influence on students’ scores on a semi-contextualized word meaning test. However, the data collected in this study did not support either of these assumptions. Although there is no doubt that the increased input and output are essential for effective language learning, no clear relationship was identified in this study between learner preference for most VLS types and their test scores. Nevertheless, two VLS items had a very weak but significant correlation with the semi-contextualized word meaning test score. One was to try to manage the learning environment in which the learner is exposed to English vocabulary (Item 9) and the other was to create an image of the spellings or orthographic forms while learning new words (Item 14). Considering the incremental nature of vocabulary acquisition, it is apodictic that, compared to other variables, the amount of learning time has a closer relationship with success in vocabulary growth. In fact, Kojic-Sabo and Lightbown (1999) showed that the time spent on vocabulary learning had a moderate positive correlation with the scores of both

－ 14 －

decontextualized and contextualized vocabulary tests. The results suggest that the frequency at which a learner studies vocabulary and the number of words actually learned are more important than the type of strategies that have been used. Second, in order to address RQ2, further correlation analyses were conducted using the factor scores calculated through the exploratory factor analysis. The results showed that only the Imagery factor, comprising Item 13 (linking vocabulary with personal experiences; factor load = .75), Item 14 (creating a visual image of word forms; factor load = .67), Item 12 (making a mental picture related to word meanings; factor load = .62), and Item 15 (using keyword mnemonic technique; factor load = .36), had a very weak but significant correlation with the test scores. Past studies have found that this type of VLS may lead to a larger vocabulary size (Akase & Uenishi, 2011) and higher proficiency, characterized by higher scores on a contextualized word production test (i.e., a cloze test; Maeda et al, 2003). The current data suggest that the imagery strategies also have some influence on the semi-contextualized word meaning test scores. Since the first step in learning a new word is to gain an understanding of a single meaning that is used in a specific context and to remember the first few letters of the word or its broad structural outline (Schmitt, 2010), the visual image of English words has a bearing on the process of recognizing a written word that the learner has seen somewhere before but has not yet mastered. Hasegawa (2013) suggested that an instruction that encourages learners to imagine the situation described by the context (referred to as imagery instruction) may lead to a better retention of learned vocabulary. In his experiment, however, this effect was found only when a lower proficiency group was presented with meaningful, or imageable, example sentences that evoked imagery more easily than other example sentences. Therefore, even if the imagery strategies have a positive effect on vocabulary retention, their influence may vary across different learning materials, which may have caused the very weak but significant correlation found in this study. In summary, the study data suggest that there is no single method that dramatically enhances test scores. The number of words that have been learned and the time and effort invested in them are more important. The use of translation may have an advantage in terms of the quick understanding of the meanings of new words, while the use of context for vocabulary learning may have other advantages. For example, as suggested by Oyama (2009), learners may understand how to use vocabulary items correctly through context-based learning.6 Nevertheless, it is advisable for high school students to pay more attention to the visual image of English words because it has a weak but significant positive relationship with success in recognizing a written word in a multiple-choice test. One limitation of the present study is that the semi-contextualized word meaning test did not correctly assess the ability to use basic words in context; future research should include target words that are easier than Level 3 and investigate the relationship between VLS use and contextualized word production skills. In addition, further research with a wider range of participants—in terms of both sample size and more diverse characteristics—is

－ 15 －

necessary because this study only captured a group of undergraduates from a single university. The participants in the present study were majoring in Education and their purpose for learning English was different than that of students majoring in other subjects; most of the participants hope to become elementary school teachers in the future. Also, future research should examine whether any significant difference exists between more advanced learners and beginners.

Notes 1. The term semi-contextualized is used to describe a task providing short written contexts,

rather than a vocabulary task embedded in a longer passage (Golonka et al., 2015). 2. Other types of receptive vocabulary tests may include word recognition tests. Although

the term recognition (as opposed to production) often refers to the cognitive process involving recognizing written strings of letters as words, the present study regards recognition as a broader concept that includes the process of matching word forms and their relevant meanings. In Webb’s (2008) study, a multiple-choice test asking participants to identify the synonyms of learned words from four options was assumed to measure meaning recognition, whereas a L2–L1 translation test (without options) was assumed to reflect meaning recall.

3. The average numbers of correct answers for 10 questions from Level 6 and another 10 questions from Over Level 6 were 3.27 (32.7%) and 2.44 (24.4%), respectively, which were close to the chance level of the multiple-choice questions with four options (25%).

4. The guidelines of school education in Japan, or the Course of Study, indicate how many English words should be taught in elementary, junior high, and high schools. According to the current version of the Course of Study, students graduating from high schools should have mastered 3,000 basic words (MEXT, 2009); when the upcoming version is put in effect, the target will have been raised to 4,000–5,000 words (MEXT, 2018b).

5. Even if Item 6 was removed from the model, the goodness-of-fit indexes could not be regarded as sufficient: χ2 (237) = 359.50, p < .001, GFI = .82, AGFI = .77, RMR = .13, CFI = .88, TLI = .86, RMSEA (90%CI = .05, .08) = .06.

6. According to Oyama (2009), translation-based VLS (originally called paired-association vocabulary learning) may be related to the occurrence of English grammatical and usage errors. For example, high-school students who were unable to identify a grammatical error in My mother told me to come home until 10 o’clock (correct word: by) were likely to use more translation-based VLSs and fewer context-based VLSs.

Acknowledgements

This research was supported by JSPS KAKENHI Grant Number 17K13496. I would like to thank the two anonymous reviewers for their helpful comments.

－ 16 －

References Akase, M., & Uenishi, K. (2011). Kokosei no eigo goi gakushu horyaku ni kansuru kenkyu:

Gakushu-sha no goi level ni shoten o atete [A study on high school students’ vocabulary learning strategies and lexical proficiency]. CASELE Research Bulletin, 41, 1–10. https:// doi.org/10.18983/casele.41.0_1

Basic Word List Revision Special Committee (Ed.). (2016). The new JACET list of 8000 basic words. Tokyo, Japan: Kirihara Shoten.

Golonka, E., Bowles, A., Silbert, N., Kramasz, D., Blake, C., & Buckwalter, T. (2015). The role of context and cognitive effort in vocabulary learning: A study of intermediate-level learners of Arabic. The Modern Language Journal, 99, 19–39. https://doi.org/10.1111/ modl.12191

Hasegawa, Y. (2013). Contextualized vocabulary learning by two proficiency groups: Focusing on imagery instruction and context imageability. Annual Review of English Language Education in Japan, 24, 1–16. https://doi.org/10.20581/arele.24.0_1

Hasegawa, Y. (2016). L2 learners’ retrieval of pre-learned word meanings while reading in a new context. Annual Review of English Language Education in Japan, 27, 233–248. https://doi.org/10.20581/arele.27.0_233

Hornby, A. S. (2015). Oxford advanced learner’s dictionary of current English (9th ed.). Oxford University Press.

Kojic-Sabo, I., & Lightbown, P. M. (1999). Students’ approaches to vocabulary learning and their relationship to success. The Modern Language Journal, 83, 176–192. https://doi.org/ 10.1111/0026-7902.00014

Maeda, H., Tagashira, K., & Miura, H. (2003). Kokosei eigo gakushu-sha no goi gakushu horyaku shiyo to gakushu seika [Vocabulary learning strategy use and learning achievement by Japanese high school EFL learners]. Japanese Journal of Educational Psychology, 51, 273–280. https://doi.org/10.5926/jjep1953.51.3_273

Ministry of Education, Culture, Sports, Science and Technology [MEXT]. (2009). Koto-gakko gakushu shido yoryo [The Course of Study for upper secondary school]. Retrieved from http://www.mext.go.jp/a_menu/shotou/new-cs/youryou/1304427.htm

Ministry of Education, Culture, Sports, Science and Technology [MEXT]. (2018a). Heisei 29 nendo eigo-ryoku chosa kekka (koko 3 nen-sei) no gaiyo [Summary of the English proficiency survey for third-year high school students in 2018]. Retrieved from http://www.mext.go.jp/a_menu/kokusai/gaikokugo/1403470.htm

Ministry of Education, Culture, Sports, Science and Technology [MEXT]. (2018b). Koto-gakko gakushu shido yoryo [The Course of Study for upper secondary school]. Retrieved from http://www.mext.go.jp/a_menu/shotou/new-cs/1384661.htm

Mizumoto, A., & Takeuchi, O. (2009). Examining the effectiveness of explicit instruction of vocabulary learning strategies with Japanese EFL university students. Language Teaching Research, 13, 425–449. https://doi.org/10.1177/1362168809341511

－ 17 －

Morimoto, Y. (2006). Goi test no keishiki ga goi chishiki to dokkai noryoku no sokutei ni oyobosu eikyo [Effects of vocabulary test formats on measuring vocabulary knowledge and reading ability]. STEP Bulletin, 18, 77–91. Retrieved from https://www.eiken.or.jp/ center_for_research/pdf/bulletin/vol18/vol_18_p77-p91.pdf

Oyama, Y. (2009). Eitango gakushu horyaku ga eigo no bunpo goho-jo no error seiki ni ataeru eikyo no kento [Strategies used for learning English vocabulary and occurrence of errors in English]. Japanese Journal of Educational Psychology, 57, 73–85. https://doi.org/ 10.5926/jjep.57.73

Webb, S., & Nation, P. (2017). How vocabulary is learned. Oxford University Press.

－ 18 －

Appendices

Appendix A Results of the VLS Questionnaire No. Statement M SD

1 I regularly review the vocabulary I have learned to check if I remember it.

4.07 1.02

2 I keep a vocabulary book or word list to check the vocabulary anytime I wish.

4.26 1.12

3 I try to make it a rule to memorize a certain number of words in a specific time period (e.g., “I will memorize 10 words a day”).

3.56 1.34

4 I try to learn extra vocabulary in addition to what I am taught in class. 3.56 1.255 I try to take time for vocabulary learning. 3.86 1.256 I consciously set aside time to study vocabulary in order to prepare for

tests (e.g., TOEIC, TOEFL, or Eiken). 2.83 1.42

7 I use my own methods for remembering, checking, or reviewing vocabulary.

3.80 1.07

8 I try to expose myself to English vocabulary by reading or listening a lot.

3.29 1.31

9 I try to manage the learning environment so as to expose myself to English vocabulary.

3.28 1.26

10 I try to make use of the media (e.g., TV, radio, internet, mobile phone, or movies) to learn vocabulary.

2.98 1.37

11 I study vocabulary with the intention of using it. 3.61 1.1512 When I try to remember vocabulary, I make a mental picture of what

can be associated with a word’s meaning. 3.74 1.15

13 When I try to remember vocabulary, I link my personal experiences to it.

3.03 1.28

14 When I try to remember vocabulary, I create an image of the spellings or orthographic forms.

3.80 1.20

15 When I try to remember vocabulary, I use the keyword method (i.e., keyword mnemonic technique).

3.10 1.41

(continued)

－ 19 －

Appendix A (continued) 16 When I try to remember vocabulary, I imagine whether the meaning of

the word is negative or positive. 3.60 1.16

17 When I try to remember vocabulary, I write it repeatedly. 3.57 1.3818 When I try to remember vocabulary, I write it on a note or a card. 3.46 1.4119 When I try to remember vocabulary, I remember not only the meaning

but also the spelling of the word by writing it. 3.62 1.44

20 When I try to remember vocabulary, I say it aloud repeatedly. 4.09 1.0921 When I try to remember vocabulary, I vocalize it to remember not only

the meaning but also the pronunciation of the word. 3.92 1.23

22 When I try to remember vocabulary, I say the sample sentence aloud. 3.25 1.3023 When I try to remember vocabulary, I associate it with the synonyms

(e.g., begin and start) or antonyms (e.g., positive and negative) I already know.

3.77 1.09

24 When I try to remember vocabulary, I also memorize the synonyms or antonyms of the word.

3.46 1.15

25 When I try to remember vocabulary, I memorize words similar to it (in meaning, sound, or shape) or the related words in a group.

3.40 1.14

26 When I was a high-school student, I placed priority on the use of Japanese translation while learning English words.

3.94 0.91

27 When I was a high-school student, I placed priority on the use of contexts while learning English words.

3.69 1.08

Note. Items 1 to 25 were adopted from Mizumoto and Takeuchi (2009, pp. 448–449). Items 26 and 27 were created in this study to capture the translation-and-context-based vocabulary learning tendency. All responses were provided on a 5-point Likert scale ranging from 1 (not at all true of me) to 5 (very true of me).

－ 20 －

Appendix B Summary of Factor Loadings in the Exploratory Factor Analysis Factor

Item No. 1 2 3 4 5 6 18 .96 .03 .02 -.01 .03 -.03 17 .83 .02 -.11 .14 .00 -.03 19 .79 -.07 .05 -.11 .00 .09 1 -.10 .73 .04 -.01 .05 -.19 3 -.19 .64 -.02 .03 -.03 .15 5 .11 .62 .04 -.10 -.03 .06 2 .08 .62 -.09 .01 .17 -.02 4 .06 .37 .06 -.02 .09 .08 7 .12 .27 .18 .19 -.18 .06 24 .01 .03 .86 -.09 .07 -.02 23 -.07 -.01 .84 .07 .02 -.05 25 .02 .04 .66 .12 -.06 .08 13 -.03 -.18 .04 .75 .10 -.06 14 .11 .11 .08 .67 -.05 -.12 12 -.09 -.07 -.04 .62 .15 .15 15 .00 .15 -.02 .36 -.18 .09 21 .09 .03 -.06 .07 .85 .06 20 -.05 .20 .01 .00 .76 -.07 22 -.04 -.15 .29 -.09 .43 .10 9 -.05 .08 -.08 .06 -.01 .91 8 .12 -.05 .14 -.08 .07 .54

Note. Factor loadings higher than .40 are in bold. The six factors were labeled as follows: Writing Rehearsal (Factor 1), Self-management (Factor 2), Association (Factor 3), Imagery (Factor 4), Oral Rehearsal (Factor 5), and Input-Seeking (Factor 6). Four items were excluded because their factor loadings were highest for factors that were different from Mizumoto and Takeuchi’s (2009) assumption; the four items and their highest loading factors were as follows: Item 6/Factor 4 (originally Self-management), Item 10/Factor 5 (originally Input-Seeking), Item 11/Factor 4 (originally Input-Seeking), and Item 16/Factor 6 (originally Imagery).

－ 21 －

Appendix C Results of the 27 Specific Items of the VLS Questionnaire (Related to RQ1)

Translation Context Test score Item r p r p r p

Item 1 .08 .361 .06 .480 −.03 .728 Item 2 .07 .428 .09 .327 .04 .695 Item 3 .12 .175 .06 .486 .09 .304 Item 4 .02 .847 .20 .028* .06 .534 Item 5 .13 .159 −.04 .661 .12 .197 Item 6 .06 .505 .28 .001** .13 .155 Item 7 .13 .157 .10 .246 .14 .129 Item 8 −.09 .307 .47 .000** .06 .472 Item 9 −.08 .401 .28 .001** .18 .045* Item 10 −.13 .149 .27 .002** .08 .374 Item 11 −.10 .264 .15 .091 .09 .342 Item 12 −.14 .122 .35 .000** .14 .110 Item 13 .03 .746 .26 .003** .12 .187 Item 14 .11 .208 .18 .038* .18 .040* Item 15 .10 .273 .05 .558 −.04 .638 Item 16 .02 .814 .25 .004** .04 .673 Item 17 .06 .500 .01 .874 −.01 .884 Item 18 .04 .642 .02 .836 −.08 .350 Item 19 .04 .684 .05 .612 −.07 .430 Item 20 .00 .983 .24 .007** .07 .448 Item 21 −.02 .835 .33 .000** .05 .546 Item 22 −.03 .709 .44 .000** −.02 .847 Item 23 .19 .037* .36 .000** .14 .128 Item 24 .23 .011* .36 .000** .12 .162 Item 25 .18 .044* .37 .000** .13 .146 Translation −.03 .750 .12 .180 Context −.06 .535 Note. A single asterisk (*) refers to p < .05; double asterisks (**) refer to p < .01.

－ 22 －