Evaluating lists of high-frequency words: Teachers’ and learners’ perspectives

Second language (L2) learners, especially those studying in English as a foreign language (EFL), have less exposure to the target language and less learning time than children learning their first language (Muñoz, 2008; Webb & Nation, 2017). Identifying which words that L2 learners should learn first is particularly important, because it helps them to get the best return for their learning effort (Nation, 2013). There are a small number of high-frequency words (around 2,000 items) (e.g. think, alright, important) that cover from 70% to 90% of the words in different kinds of texts (e.g. newspapers, general conversation, TV programs, and academic texts) (Coxhead, 2000; Dang & Webb, 2014; Nation, 2004). Knowledge of high-frequency words is important because it may allow learners to recognize a large proportion of words in different spoken and written texts. Such knowledge provides a solid foundation for learners to acquire words at lower frequency levels and achieve a high and stable degree of comprehension. For this reason, high-frequency words have been widely accepted as the crucial starting point for L2 vocabulary learning (Nation, 2013; Schmitt, 2010). Several corpus-based lists of high-frequency words have been developed such as General Service List (West, 1953), CELEX lists (Dutch Centre of Lexical Information; Baayen, Piepenbrock, & van Rijn, 1995), BNC2000 (British National Corpus 2000; Nation, 2006), COCA lists (Corpus of Contemporary American English list; Davies & Gardner, 2010), BNC/COCA2000 (Nation, 2012), SUBTLEX lists (Subtitles-based word frequencies; Brysbaert & New, 2009; van Heuven et al., 2014), New General Service List (Browne, 2014), and New General Service List (Brezina & Gablasova, 2015). These lists are useful for setting learning goals, designing learning materials and activities, and developing tests (Nation, 2016). To the best of our knowledge, little research has been conducted to examine the effectiveness of implementing these corpus-based high-frequency word lists in language classrooms. However, empirical studies with Coxhead’s (2000) Academic Word List (AWL) have indicated that this list is a useful tool for learners (e.g. Banister, 2016; Lesaux et al., 2010; Townsend & Collins, 2009). As the AWL is a corpus-based word list, these findings highlight the value of corpus-based word lists for language learning and teaching.

Given that there are several different high-frequency word lists, one question that arises is which list is the most useful for L2 learners. Five studies have been conducted to address this question (Brezina & Gablasova, 2015, Browne, 2014; Dang & Webb, 2016; Gilner & Morales, 2008; Nation, 2004). Each study used lexical coverage as the sole criterion for determining which list is best. Lexical coverage refers to the percentage of words covered by items from a particular word list in a corpus (Nation & Waring, 1997). Dang and Webb’s (2016) study was the most comprehensive because it compared the coverage of a larger number of word lists in a larger number of corpora, and these corpora had a great degree of diversity in types of texts, sizes, and varieties of English. Their comparison showed that Nation’s (2012) BNC/COCA2000 accounted for the largest coverage while Brezina and Gablasova’s (2015) New-GSL included the largest number of frequent items. This suggested that if one of the lists was used as a whole, the BNC/COCA2000 may provide the greatest value for learners; however, if only a proportion of the list was used, the New-GSL might have the greatest value.

Although lexical coverage is an important criterion to evaluate corpus-based word lists, to make these lists more relevant to L2 learning and teaching, list evaluation should involve their end-users— learners and teachers. Unfortunately, no studies have involved these agents in the evaluation of high-frequency word lists. To address this gap, the present study used learner vocabulary knowledge and teacher perceptions of word usefulness to further compare the BNC/COCA2000 and the New-GSL. The findings should indicate which high-frequency word list is more useful for L2 learners.

The present study is the first attempt to use information from teachers and learners to supplement corpus-based information in the evaluation of high-frequency word lists. Expanding on Dang and Webb’s (2016) study, the present study further compared Nation’s (2012) BNC/COCA2000 and Brezina and Gablasova’s (2015) New-GSL by using two criteria: (a) L2 learner vocabulary knowledge and (b) teacher perceptions of the usefulness of the words for basic functions in English. The research involved the participation of 135 L2 learners and 78 experienced teachers of English as a second language / English as a foreign language (ESL/EFL). The learners were from different proficiency levels and the instructors had experienced teaching in a wide range of EFL/ESL contexts. Therefore, this study is expected to provide an assessment of corpus-based lists of high-frequency words from the perspectives of learners and teachers. The study should shed light on the value of including the perspectives of teachers and learners in corpus-based word list validation, and thus, bring corpus-linguistics research together with other research strands—Second Language Acquisition and teacher cognition. Moreover, the findings should indicate which high-frequency word list is the most suitable for L2 learners. This in turn should help teachers and materials writers to select words for materials, activities, and tests for L2 learners.

The following research questions are addressed:

The BNC/COCA2000 consistently made up a significantly larger proportion of words perceived as useful by teacher participants than the New-GSL. Table 6 presents the results of the analysis with all 78 teachers. When the teachers’ ratings of 973 items (545 BNC/COCA2000 words and the 428 New-GSL words) were examined, depending on the criteria of usefulness, the BNC/COCA words accounted for 60.40% to 79.0% of the useful words while the New-GSL only accounted for 21% to 39.60%. Similarly, when the teachers’ ratings of the most frequent 428 items from each set of words were compared, depending on the criteria of usefulness, the range of percentage of words from each list among the useful words rated by teachers was 61%—80% (BNC/COCA2000) and 20%—39% (New-GSL). The results of the Z-tests indicated that the differences were significant at p < 0.05 in all cases. Interestingly, as shown in columns 4 and 7, as the criterion of usefulness got stricter (from top 500 to top 100), the differences between the percentage of the BNC/COCA2000 and the New-GSL words among the most frequent words rated by the teachers become larger. Appendix 5 in the Supplementary data presents information about the percentage of the BNC/COCA2000 and New-GSL words in each 100-most useful word band rated by all teachers. In most cases, the BNC/COCA accounted for larger proportion of the useful words.

The same pattern was found with each group of teachers (EFL/ESL teachers who were native speakers of English, Vietnamese EFL teachers, and EFL teachers from varying countries). The BNC/COCA2000 always comprised a larger proportion of words rated as useful by the teachers (59–80%) than the New-GSL (20–41%) (for more details, see Appendices 6–8 in the Supplementary data). The results of the Z-tests revealed that except for the case of the top 100, 200, and 500 useful words rated by the EFL teachers from varying countries, the differences were always significant at p < 0.05. Similar to the case of all teachers, for each group of teachers, as the criterion of usefulness got stricter, the differences between the proportion of the BNC/COCA2000 and the New-GSL among the most useful words rated by the teachers became larger (see Appendix 9 in the Supplementary data). It would be interesting to see which target words were consistently rated as the most useful by the teachers. Therefore, for each target word, we also counted the number of teachers indicating it as the most useful (having score of 5). Then, we ranked the 973 target words based on the number of teachers who indicated the word as most useful and identified the top 103² words. The results show 81.55% of these words (84 out of 103 words) are from the BNC/COCA2000 while only 18.45% (19 out of 103 words) are from the New-GSL. The results of the Z test indicated that the difference was significant at p < 0.05. Taken together, the results of the teacher surveys indicated that from the teachers’ perspectives, the BNC/COCA2000 is a more useful high-frequency word list for L2 learners than the New-GSL. It is important to note that there were strong correlations between the ratings of each group of teacher: r_{English L1 teachers-Vietnamese L1 teachers} = .81, r_{English L1 teachers-Various L1 teachers} = .83, r_{Vietnamese L1-Various L1 teachers} = .85, p < 0.001 . This indicates that the ratings of the teacher are very consistent.

The results from the learner data are less transparent than those from the teachers. Columns 2 and 3 of Table 7 present the results of the analysis of all 112 learners. No matter whether the full sets of words or only the most frequent 428 items from each set were compared, the BNC/COCA2000 always accounted for a larger percentage of known words (61.95%, 58.46%) than the New-GSL (38.05%, 41.54%). The differences were always significant at p < 0.05

Analysis with the data of intermediate and advanced learners revealed that there was no significant difference in the proportion of the BNC/COCA words and New-GSL words among the words known by these learners (see the last four columns of Table 7). This finding suggests that it is unclear which list is better known by intermediate and advanced learners. In contrast, as shown in columns 4 and 5 of Table 7, the BNC/COCA2000 always accounted for a significantly larger proportion of words known by pre-intermediate learners (67.36%, 65.44%) than the New-GSL (32.64%, 34.56%), and the differences were always significant at p < 0.05. As L2 learners tend to know more high frequency words than lower frequency words(e.g. Dang et al., under review) and tend to learn high frequency words first (Ellis, 2002), the results indicate that the BNC/COCA words seem to be learned before the New-GSL words.

When the learner data and teacher data were compared, there were 146 words known by at least 90% of the learners and indicated as being useful by the teachers (mean scores of 4 or above) (see Appendix 10 in the Supplementary data). Of these items, 108 words were from the BNC/COCA2000 and 38 words were from the New-GSL. The results of the Z test indicated that the difference was significant at p < 0.05. It is important to note that 7 out of 38 New-GSL words which are indicated as useful and known by most learners (bathroom, bedroom, website, weekend, birthday, classroom, CD) did not appear in the BNC/COCA2000 because the BNC/COCA lists include a separate list of transparent compounds and abbreviations.

Together, the data from teacher surveys and learner yes/no tests indicate that the BNC/COCA2000 is likely to be perceived as more useful by teachers and the items are likely to be learned earlier by L2 learners than the New-GSL. This suggests that the BNC/COCA2000 may be the more useful resource at least for EFL learners in Vietnam. There are two possible reasons for the superiority of the BNC/COCA over the New-GSL in terms of teacher perceptions and learner vocabulary knowledge. The first reason may be the result of the principles under which the two lists were developed. The New-GSL was created with a purely quantitative approach; that is, using the average reduced frequency (Hlavácǒvá, 2006; Savický & Hlavácǒvá, 2002), which takes into account both the absolute frequency of a word and its distribution in the corpus, as the selection criterion. In contrast, apart from these quantitative corpus-based criteria, the development of the BNC/COCA2000 also involved adding to the list the lexical items that did not meet these criteria but may be suitable for L2 learning and teaching purposes. A word list that is solely based on the information from corpora may miss items that have low frequency in corpora but are useful for L2 learning (Nation, 2016). For example, BNC/COCA words such as alright, ok, exam, hello, goodbye, grade, pronounce, schedule, silence were not included in the New-GSL but were known by more than 90% of the learners and were rated as useful by teachers. The greater focus on L2 learning may explain why the BNC/COCA2000 had a larger number of words known by learners and perceived as being useful by teachers than the New-GSL. The second reason may be the corpora used to develop these lists. The BNC/COCA2000 was created from a corpus with a better balance of spoken texts (60%) and written texts (40%) and represents different varieties of English (British English, American English, and New Zealand English). In contrast, the New-GSL may be biased towards British, written English. Three out of the four corpora (LOB, BNC, BE06) on which the New-GSL were based, represented British-English, and three out of the four corpora (LOB, BE06, EnTenTen12) were made up of written discourse. In the only corpus which included spoken English (BNC), spoken samples accounted for only 10%. Given that the BNC/COCA2000 was developed from a corpus which represents a range of spoken and written discourses and varieties of English, it is understandable why the BNC/COCA2000 is likely to be perceived as more useful by teachers from various contexts and to be learned earlier by L2 learners than the New-GSL. By examining high-frequency word lists from the perspectives of corpus linguistics, teachers, and learners, the present study provides a useful methodological innovation that could be implemented in future research.

One interesting finding of this study is related to the teachers’ ratings. The common assumption is that words perceived as being useful for L2 learners may vary greatly between teaching contexts. However, the teacher ratings were less diverse than expected. The teachers in this study came from different L1 backgrounds, had experienced teaching in different EFL/ESL contexts, and varied in years of teaching experience. Yet their ratings were relatively consistent regardless of the criteria of word usefulness and groups of teachers being examined. This finding suggests that L2 teacher perceptions of high-frequency words for L2 learners may be similar across a wide range of contexts. No earlier studies have explored this issue from the perspective of teacher cognition.

This study has an innovative, cross-disciplinary approach towards evaluating high-frequency word lists. It brings together corpus linguistics, Second Language Acquisition, and teacher cognition research under the umbrella of word list studies. While all earlier studies (Brezina & Gablasova, 2015; Gilner & Morales, 2008; Nation, 2004; Nation & Hwang, 1995) used lexical coverage from corpora as the sole criterion, the present study used the information from teachers and learners to supplement corpus-based information in word list evaluation. This approach takes advantage of statistical information to identify and prioritize items that are likely to be encountered by learners. Meanwhile, it ensures that the final entries are appropriate and relevant for L2 learning and teaching. Information from teachers and learners takes into account the contextual and circumstantial realities of a language classroom and provides indicators of the extent to which corpus-based word lists would filter their way to L2 classrooms. As shown in this study, if only corpus information were used in the comparison, it would be challenging to determine whether the BNC/COCA2000 or the New-GSL is a more appropriate for EFL learners in Vietnam. However, when teacher perceptions of word usefulness and learner vocabulary knowledge were used to support corpus-based information, the results provided evidence that the BNC/COCA2000 is the most appropriate list. Additionally, the present study involved the participation of a large number of EFL/ESL teachers and L2 learners. The teachers varied in terms L1 backgrounds, teaching contexts, and years of teaching experience. The learners represented different proficiency levels, university levels, and majors. Furthermore, the information from each source (teachers and learners) was analysed from different angles. A range of criteria were used as indicators of word usefulness and vocabulary knowledge, and the lists were compared as a whole, as well as using the most frequent items in each list. The use of teachers and learners, the large number and great diversity of the participants, and the in-depth analysis of the data provide a thorough assessment of the lists. This provides future studies with a useful model of how to evaluate corpus-based word lists for L2 learning and teaching.

The present study indicated that Nation’s (2012) BNC/COCA2000 appears to be the most suitable high-frequency word list for EFL learners in Vietnam. This list might also be a useful vocabulary resource for learners in many other EFL contexts. Foreign language learning/teaching in Vietnam shares features of typical foreign language teaching/learning situations described in previous studies (e.g. Webb & Nation, 2017; Muñoz, 2008). In such contexts, learners study English in their home country where English is not the first or significant language and have relatively limited contact with English outside classroom. The time allocated for learning English at school is also limited and the exposure to English during the class periods may be limited in source, quantity and quality. Consequently, the majority of learners in various EFL contexts have insufficient knowledge of high-frequency words despite many years of studying English (e.g. Dang, 2019; Akbarian, 2010; Henriksen & Danelund, 2015; Matthews & Cheng, 2015; Webb & Nation, 2017).

Given the importance of high-frequency words and EFL learners’ insufficient knowledge of these words, it is essential for teachers and course designers to ensure that learners have mastered these words before moving on to words at lower frequency levels. At the beginning of a learning program, students’ knowledge of the high frequency words should be assessed. Webb, Sasao and Ballance (2017) Updated Vocabulary Levels Test (developed from the first five 1,000 word lists of the BNC/COCA lists) can be used to measure learners’ knowledge of the BNC/COCA2000. If the test scores indicate that students have not yet mastered the BNC/COCA2000, teachers should ensure that their classroom programs include a clear focus on learning this vocabulary (for suggestions for school wide programs, see Webb et al., 2017).

There are several important issues teachers should consider when helping students learn the BNC/COCA2000 words. First, there are different aspects involved in knowing a word such as form-meaning relationship, word parts, collocations, and associations (Nation, 2013). Teachers should help learners to gain knowledge of the form-meaning relationship first because it is the most important aspect of vocabulary knowledge, which creates a foundation for the acquisition of other aspects. Once learners have mastered the form-meaning relationship of a word, teachers should create opportunities for them to consolidate and expand knowledge of words.

Second, although this study indicated that the BNC/COCA2000 had more words known by learners and indicated as being useful by teachers, 38 New-GSL words were known by at least 90% of the learners and indicated as being useful by the teachers. Together with the BNC/COCA2000 words, these 38 New-GSL words should be the set as the learning goal for L2 learners. As vocabulary learning is an incremental process (Schmitt, 2010), the 2,038 words from the BNC/COCA2000 and New-GSL should be considered as a long-term vocabulary learning goal rather than a short-term goal. In its original format, the BNC/COCA2000 is divided into two sub-lists, each of which consists of 1,000 headwords. These sub-lists may be too large to fit in a single course. Therefore, to better assist teachers and learners in setting learning goals, we also rank the BNC/COCA2000 word family headwords according to their mean coverage in nine spoken and nine written corpora (see Appendices 11 and 12 in the Supplementary data). The information about the mean coverage of the 38 New-GSL word family words is already presented in Appendix 10. Teachers and learners can use the information in these appendices to set the short-term vocabulary learning for their courses. One possible way is focusing on words with higher mean coverage before moving on to those with lower mean coverage. Sequencing vocabulary learning in this way would better scaffold learners’ vocabulary development, because knowledge of known words would facilitate the acquisition of new words while learning new words would consolidate and expand knowledge of known items (Dang & Webb, 2016; Webb & Nation, 2017).

Third, to create opportunities for learners to learn new words and expanding knowledge of known words, teachers should follow Nation’s (2007) Four Strands of meaning-focused input, meaning-focused output, language-focused learning, and fluency development. The four strands provides a framework that includes opportunities to encounter and use words in different contexts. Also, to ensure that students encounter high-frequency words more often, teachers can use vocabulary analysis programs like Lextutor (Cobb, n.d.) or AntwordProfiler (Anthony, n.d.) to check the proportion of BNC/COCA2000 words in the texts and adapt the vocabulary in the texts accordingly.

This study has several limitations which provide avenues for future research. First, only the vocabulary knowledge of Vietnamese EFL learners was examined in the present study. Although these learners share features of learners in many EFL contexts and the data of teachers from different L1 backgrounds with teaching experience in different EFL/ESL (English as a second language) contexts were used to triangulate the information from the learners, bias towards Vietnamese EFL learners is inevitable. Therefore, while the present study provides useful information about the vocabulary knowledge of a specific learner population, further research with L2 learners in other contexts may provide further insight into knowledge of the BNC/COCA2000 and New-GSL possessed by learners from different L1 populations. Second, this study only measured learners’ knowledge of words. It would be useful for future research investigating which words students have learned on their own and which words they have learned from classroom instruction. Third, when measuring the learner participants’ vocabulary knowledge, this study does not consider French/English loanwords in Vietnamese. Fourth, it is unlikely that the teacher participants were aware of the two lists during the data collection (between December 2014 and May 2015) because these lists were fairly recent (both of them were available online in 2013) and the name of the word lists were not mentioned to the participants during the recruitment to avoid biasing them toward a certain word list. However, a rigorous follow-up study with all teacher participants after the study completed would provide more solid evidence about teachers’ awareness of the BNC/COCA2000 and the New-GSL list. Fifth, the main purpose of the current study is to compare the BNC/COCA2000 and the New-GSL; therefore, the BNC/COCA2000 words that also appear in the New-GSL were not used as the target words in the teacher surveys and learner vocabulary tests. It would be useful for future research to explore learner knowledge and teacher perception of the usefulness of these items. Such research would provide further insight into the value of the BNC/COCA2000 words for L2 learners. Finally, this study used the word family version of the BNC/COCA2000 for the comparison, and to deal with the limitation that the sub-lists in the original version of the BNC/COCA2000 is too big (1,000 words) to fit in a single course, this study provided list users with the ranking of the BNC/COCA words in terms of the mean coverage of word family headwords. However, given the current trend in word list studies; that is, offering word lists in different formats using different units of counting (Dang, Coxhead, & Webb, 2017; Gardner & Davies, 2014; Nation, 2016), future research on validating high-frequency word lists should be carried out to validate different versions of the lists using different units of counting and rank the items in the lists according to these units of counting.

Expanding on Dang and Webb’s (2016) study, the present research found that the BNC/COCA2000 had more words known by L2 learners and perceived as being useful by EFL/ESL teachers than the New-GSL. These results suggest that the BNC/COCA2000 is the more suitable high-frequency word list for L2 learners, at least in the Vietnamese EFL context. This study is the first attempt to use information from both teachers and learners to supplement corpus-based information in the evaluation of lists of high-frequency words.