The lexical coverage of popular songs in English language teaching

System xxx (2017) 1e12

Contents lists available at ScienceDirect

System journal homepage: www.elsevier.com/locate/system

The lexical coverage of popular songs in English language teaching Dr Friederike Tegge Massey University, School of Humanities, Private Bag 11 222, Palmerston North, New Zealand

a r t i c l e i n f o

a b s t r a c t

Article history: Received 1 July 2016 Received in revised form 27 April 2017 Accepted 29 April 2017 Available online xxx

Songs are popular among language learners and a text genre that is yet to be fully exploited in language teaching. Questions arise regarding their lexical demand and vocabulary-learning opportunities they afford. Two pop song corpora were analyzed to determine the vocabulary size necessary to comprehend 95% and 98% of words in pop songs. The first corpus comprised 408 songs listed in recent US billboard charts. The second corpus consisted of 635 songs selected by teachers for language-teaching purposes. Results of an analysis using RANGE and 20 BNC word-frequency lists showed that the lexical demand of charts songs is overall clearly lower compared to other written genres but similar to spoken genres, as the most frequent 3000 word families plus proper nouns provided 95.1% coverage of tokens, and knowledge of 6000 word families plus proper nouns was necessary to reach 98.2% coverage. Teacher-selected songs have a lower lexical demand: Knowledge of the most frequent 2000 word families plus proper nouns was necessary to reach 95.5% coverage, while a vocabulary size of 4000 word families plus proper nouns provided coverage of 98.2% of words in the pedagogical corpus. Implications for the use of songs in ESL and EFL classrooms are discussed. © 2017 Elsevier Ltd. All rights reserved.

Keywords: Vocabulary Corpus Song TESOL Teaching materials Text comprehension Lexical load Coverage Word frequency Lyrics

1. Introduction In Western societies, many people listen to music for several hours each day. 15- to 18-year-old US-American adolescents, for example, listen to an average of 3 h and 3 min of music on a typical day (Rideout, Foehr, & Roberts, 2010). Among the many options available to listeners, chart songs rank highly in popularity (North, Hargreaves, & Hargreaves, 2004). And the charts are typically dominated by English-language songs, even in countries where English is not the first language. In Germany, for example, eight out of the top ten chart songs in the week of July 4, 2015 were sung in English (http://www.billboard.com/biz/ charts/international). When considering the popularity and availability of pop music in the light of English teaching and learning, pop songs can, thus, be assumed to provide a large amount of verbal input for learners on a daily basis e both in ESL and EFL settings. Such high interest in and exposure to popular music can also be exploited inside the language classroom. In fact, working with songs is frequently favored by language learners. Green (1993), for example, found that 263 intermediate EFL learners at a University in Puerto Rico ranked song-based tasks highest in terms of enjoyableness compared to other communicative and non-communicative activities. Many language teachers equally express positive views regarding the use of pop songs as a tool to foster language acquisition in the classroom (Tegge, 2015).

E-mail address: [email protected]. http://dx.doi.org/10.1016/j.system.2017.04.016 0346-251X/© 2017 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Tegge, F., The lexical coverage of popular songs in English language teaching, System (2017), http://dx.doi.org/10.1016/j.system.2017.04.016

2

F. Tegge / System xxx (2017) 1e12

The use of songs for language teaching raises the question of how many words learners need to know to understand authentic pop songs in- and outside the classroom. While the usefulness of songs in language teaching is affected by various factors, the present study focuses on their vocabulary load, as unknown vocabulary is understood to be an important obstacle to both reading and listening comprehension (see Hirsh & Nation, 1992; Laufer & Ravenhorst-Kalovski, 2010; Laufer, 1989; Stæhr, 2008). In addition, teachers frequently report using songs to teach vocabulary (Tegge, 2015). Consequently, it is of interest to understand how many words learners need to know to engage with pop songs and what vocabulary-learning opportunities they encounter. Apart from songs, teachers can choose from a variety of authentic text genres to support their students’ language development. Consequently, the aim of the present corpus study is to determine the lexical demand of English pop songs compared to other materials. More specifically, it provides an indication of the vocabulary size needed for comfortable comprehension of pop lyrics. Finally, the present study investigated whether songs selected by language-teaching professionals for pedagogical purposes differ from pop songs found in the recent charts in terms of their lexical demand. 2. Literature review 2.1. Lexical coverage and text comprehension Within vocabulary research, the issue of lexical demand is frequently addressed by asking how many words in a text need to be understood for adequate or reasonable comprehension (Nation, 2006; Stæhr, 2008; Webb & Rodgers, 2009b) which does not require learners to resort to “compensatory strategies” (Laufer, 2013, p. 868) and for incidental vocabulary learning to occur (Webb & Rodgers, 2009a). Coverage of around 95%e98% of running words or tokens in a target text has been suggested (Hu & Nation, 2000; Laufer, 1989; Schmitt, Jiang, & Grabe, 2011). In this context, the term coverage refers to the percentage of known words in the text. 98% coverage is widely accepted to be the optimal threshold (Laufer, 2013) for adequate comprehension of unsimplified written texts. However, while it has been repeatedly demonstrated that 98% coverage is required for optimal reading comprehension (Hu & Nation, 2000; Laufer & Ravenhorst-Kalovski, 2010), the same threshold cannot simply be applied to listening comprehension. Listening differs from reading in many ways, most obviously in the temporary nature of aural texts and the challenge of parallel reception and decoding. Aural texts in contrast to written input do not provide opportunities for perusal and repetition (Lund, 1991). However, spoken discourse affords extra-linguistic support to understanding. Frequently, non-verbal clues are provided, such as gestures, facial expressions and lip movements, which aid listening comprehension and make up for deficient lexical knowledge (van Zeeland & Schmitt, 2013). Compared to readers, listeners also tend to rely more on extralinguistic information and knowledge, including world knowledge, topic familiarity and metacognitive processes of listening comprehension. Consequently, van Zeeland and Schmitt (2013) argued that coverage necessary to comprehend written and spoken texts might differ. Based on an experimental study of listening comprehension in native and non-native speakers, they proposed 95% as an appropriate coverage target for listening comprehension of informal spoken narratives. Bonk (2000) investigated EFL-learners’ comprehension of four audio-recordings of varying levels of lexical difficulty and concluded that coverage of less than 95% of tokens might still result in adequate comprehension if listeners made use of effective listening strategies. In contrast, Stæhr (2008) found that 98% coverage seemed to be a reasonable threshold. 2.2. Vocabulary knowledge and word-frequency lists Another issue that must be addressed when assessing the lexical demand of a text or text genre is the question of how many words learners need to know to reach the threshold required for adequate or comfortable comprehension. This is often done by assessing the coverage of a text provided by word-frequency lists. In this context, coverage refers to the percentage of words accounted for by such word lists (Nation & Kyongho, 1995; Nation, 2004). Frequency lists sort words according to their frequency in general language use, from most to least frequent. Nation’s (2004, 2006) BNC frequency lists, for example, rank English words according to their frequency, range and dispersion in the British National Corpus (BNC). Using such frequency lists to assess the vocabulary knowledge required to understand various text genres is based on the assumption that language learners acquire common words earlier than less common vocabulary. Research has shown that this is indeed the case (Nation, 2006). The present study makes use of 20 BNC wordlists (Nation, 2004, 2006) to assess the lexical demand of song lyrics. These frequency lists, used in a number of studies on lexical coverage, consist of word families rather than individual words. That is, the lists contain headwords along with a number of family members. A vocabulary size of 3000 word families, consequently, refers to knowledge of more than 3000 individual words, as each word family can comprise several members. For the BNC wordlists, a word family is defined on the basis of the level-6 classification described in Bauer and Nation (1993), which includes inflected and derived forms. The use of the word family to measure word knowledge is based on the assumption that “inflected and regularly derived forms of a known base word can also be considered as known words if the learners are familiar with the affixes” (Hirsh & Nation, 1992, p. 692). In the present study, proper nouns are also seen as having such a small learning burden as to be counted as known (Kyongho and Nation, 1989; Hirsh & Nation, 1992). In addition, they are typically clearly recognizable due to a capitalized first letter. So-called marginal words, including interjections, exclamations and hesitation markers, are also counted as known due to their low learning burden (Nation, 2006). Finally, Nation has added Please cite this article in press as: Tegge, F., The lexical coverage of popular songs in English language teaching, System (2017), http://dx.doi.org/10.1016/j.system.2017.04.016

F. Tegge / System xxx (2017) 1e12

3

transparent compounds as a separate category to the word frequency lists, as they can also be considered known by means of knowing their high-frequency parts, for example hometown or sailboat. 2.3. The lexical demand of various text genres No study to date has attempted to establish the lexical demand of pop song lyrics based on the coverage provided by wordfrequency lists. In contrast, various studies have explored the lexical demand of other text genres. Figures differ when it comes to written, unscripted spoken and scripted spoken text genres. Nation (2006) investigated the vocabulary demand of various types of written discourse. He found that vocabulary knowledge of 8000 to 9000 word families is needed to read novels. This is a significantly higher demand than the 5000 word families first proposed by Hirsh and Nation (1992). One reason for this difference could be the use of different wordlists: Hirsh and Nation based their analysis on West’s (1953) General Service List (GSL) and on wordlists adapted from Thorndike and Lorge’s (1944) vocabulary workbook, whereas Nation (2006) used fourteen wordlists developed from the BNC. Secondly, whereas Hirsh and Nation analyzed three short novels written for adolescents, Nation (2006) investigated the vocabulary demand of five novels written for adult readers. These five novels not only constituted a larger corpus from a greater variety of sources but they were also written for a more sophisticated audience. With a required vocabulary size of 10,000 words for 98% coverage, Webb and Macalister (2013) observed an even higher lexical demand of literature targeted at adult readers. This difference might be explained by the significantly higher number of 138 texts in their corpus, comprising a variety of genres, including fiction as well as newspaper articles. In addition, Webb and Macalister observed a similarly high lexical demand, that is, 10,000 word families, in children's literature of various genres written for young L1 readers in New Zealand. Unscripted spoken discourse, that is, spontaneous, unplanned speech, has been found to be less demanding in terms of vocabulary load than written text genres. Nation (2006) analyzed two parts from the spoken section of the Wellington Corpus of Spoken New Zealand English (WSC) of around 100,000 words each and found that knowledge of 6000 to 7000 word families plus proper nouns were necessary to understand 98% of words of friendly conversations and talk-back radio or interviews. Adolphs and Schmitt (2003) investigated general spoken discourse by analyzing the 5-millionword Cambridge and Nottingham corpus of discourse in English (CANCODE), which consists of speech recorded in Great Britain. They found that 3000 word families provided almost 96% coverage of the corpus. For their analysis, Adolphs and Schmitt utilized frequency lists based on the CANCODE itself. That is, coverage was assessed by the frequencies of the words in the corpus itself, rather than by more generalized frequency lists based on both written and spoken texts from a variety of contexts. Another form of spoken language is scripted spoken discourse, which includes lectures, speeches and language produced in movies and TV programs. It differs from unscripted spontaneous language in that it follows a written script prepared prior to speaking. Webb and Rodgers (2009a) examined the lexical demand of British and American movies, while Webb and Rodgers (2009b) investigated the vocabulary demand of British and US-American TV programs. Both the movie and the TV corpus required knowledge of 3000 word families plus proper nouns and marginal words to reach 95% coverage. For 98% coverage of the movie corpus, knowledge of 6000 word families plus proper nouns and marginal words was necessary, while the TV corpus required knowledge of 7000 word families. 2.4. A lexical threshold for the comprehension of song lyrics Murphey (1990) conducted the only other existing song-corpus study with a language-pedagogical focus. For that purpose, Murphey compiled a corpus of 50 songs from the Music & Media Hot 100 Charts from September 1987, comprising 13,161 running words. Based on his lexical analysis, Murphey concluded that pop songs as a genre are short, lexically simple and highly repetitive. Furthermore, he applied Flesch’s (1974) readability formula and found that lyrics can be considered “very easy” in terms of their readability, comparable to texts requiring a reading level of a child after five years of schooling. While Murphey conducted both a content analysis and a lexical analysis of pop songs, he did not investigate lexical coverage. Consequently, no coverage threshold has currently been established for adequate comprehension of song lyrics. However, research on other written and aural materials can give some indication of an appropriate threshold for the comprehension of songs, as lyrics display characteristics of both written and spoken text. Kreyer and Mukherjee (2007) argued that song lyrics “sit somewhat uneasily on the boundary between writing and speech” (p. 37). They compiled the Giessen-Bonn Corpus of Popular Music (GBoP) containing 176,000 words from 442 songs on 27 top albums from the 2003 US Album Charts. They analyzed the corpus regarding a number of lexical, lexico-grammatical and thematical aspects and compared them to the written and spoken sections of the International Corpus of (British) English (ICE-GB). Results indicated that song lyrics display characteristics of both written and spoken text. The short average word-length and the high use of the personal pronouns I and you support the view of song lyrics as resembling spoken discourse. However, the comparatively high standardized type-token ratio speaks for the categorization of lyrics as written text. In addition, Kreyer and Mukherjee observed that song lyrics in the corpus display a low use of the discourse marker you know. In interactional spoken discourse, this expression is frequently used to manage information or to keep the turn. As Kreyer and Mukherjee explained, using such fillers to avoid losing the turn is “not relevant to pop song lyrics as they resemble written texts with regard to the clearly offPlease cite this article in press as: Tegge, F., The lexical coverage of popular songs in English language teaching, System (2017), http://dx.doi.org/10.1016/j.system.2017.04.016

4

F. Tegge / System xxx (2017) 1e12

line text production” (p. 46). Furthermore, the authors observed that published written lyrics often display creative and unconventional ways of spelling words, as in Avril Lavigne's Sk8er boi. Such intentionally deviant spelling emphasizes the relevance of the written text. Overall, Kreyer and Mukherjee identified song lyrics as “a special case of written-to-be-spoken (or, rather, written-to-besung) genre” (p.37). And in fact, when used as teaching material in the classroom, songs are frequently both listened to and read (see Tegge, 2015). It, therefore, seems safe to assume that lexical coverage between 95% and 98%, applied to written and spoken language, is a reasonable estimate for comprehension of song lyrics. 3. The present study The present study addressed the following research questions: 1. How many words do L2-learners of English need to know to gain 95% and 98% coverage of pop song lyrics from the USAmerican charts? 2. How many words do L2-learners of English need to know to gain 95% and 98% coverage of authentic, unsimplified songs chosen by teachers or material designers for in-class use? 3. How does the lexical profile of songs differ from that of other authentic text genres? 4. What are the lexical learning opportunities afforded by authentic, unsimplified English songs chosen by teachers or material designers? 3.1. Data collection For this study, two song corpora were compiled. Both are text corpora and do not include any audio material. The first corpus, the Wellington Corpus of Popular Songs (WOP), comprises 408 English pop songs from the top 100 end-of-year US billboard charts (www.billboard.com) from the years 2014, 2012, 2010 and 2008. In order to avoid any overlap, songs from every other year were selected. The second corpus, the Wellington Corpus of Popular Songs in English Teaching (WOPET), contains the lyrics of 635 authentic, unsimplified English songs recommended by teachers or material designers for use in the English classroom. 314 songs were recommended by 180 ESL/EFL teachers responding to an international online questionnaire (Tegge, 2015). These informants were invited through professional networking and mailing lists of language teacher associations. The 180 respondents were located in 30 countries, with a majority being located in New Zealand (38), Japan (26), Malaysia (17), Canada (14) the USA (13) and the United Arab Emirates (12). They taught at a range of institutions but predominantly at tertiary institutions (79), secondary schools (40) and public or private language schools (38). As Tegge's survey revealed, informants used songs with clear meaning- and language-focused goals in mind and in the context of a directed and diverse teaching unit. Frequently reported teaching purposes included the teaching of authentic language and culture, practicing listening comprehension, new and familiar vocabulary and multi-word items, pronunciation and prosody, grammar and speaking fluency. 62 songs were procured from 26 ESL textbooks (from 11 different series). Only textbooks published in or after the year 2000 were selected to ensure that they were still likely to be used in English classrooms. A further 257 songs were sourced from six websites, for example www.busyteacher.org, which are used by English teachers to share song recommendations and even detailed lesson plans. The songs in the WOPET were further sub-categorized based on the language proficiency of the intended learners. 243 songs in the corpus were seen by teachers as appropriate for beginners (complete and continuing beginners); 356 songs were recommended for intermediate learners (low- and high-intermediate); and only 36 were intended for low- and high-advanced learners. Care was taken that both corpora were of similar size. The WOP comprises 180,892 running words, while the pedagogical WOPET contains 177,384 tokens, about two-thirds of the size of Webb and Rodgers’ (2009b) corpus of TV programs of 264,384 tokens. These two corpora are to date the largest pedagogical song corpora compiled and analyzed for the purpose of establishing a lexical profile of songs used in English teaching. While the above-mentioned GiessenBonn Corpus of Popular Music (Kreyer & Mukherjee, 2007) is, in fact, larger, it was not compiled and analyzed with language teaching in mind. It needs to be mentioned that the two corpora differ regarding the recency of their songs. While the WOP comprises songs from the charts from 2008 to 2014, the WOPET contains songs that were published before 1900 (e.g., national anthems) and as recently as 2011. However, 92% of songs in the WOPET were published after 1959, and 36.4% were published after 2000. Still, the overlap of songs in the two corpora is small. This is, of course, partially due to the fact that the WOP contains songs published in 2014 and 2012, while the data collection for the WOPET finished in 2011. Only 24 songs can be found in both corpora, including Alicia Key's Empire State of Mind, Eminem's I love the way you lie, Michael Buble's I haven't met you yet, Justin Biber's Baby, Taylor Swift's Mine, You belong with me and Love story, Orianthi's According to you, and Daniel Powter's Bad Day. A look at the performers represented in the corpora also reveals differences. Performers contributing more than six songs to the WOP are Rihanna (12), Katy Perry (10), Taylor Swift and Chris Brown (8), Drake and Usher (7). Performers represented with more than six songs in the WOPET are The Beatles (24), Madonna (10), Michael Jackson (8), ABBA, Celine Dion, and Queen (7). Please cite this article in press as: Tegge, F., The lexical coverage of popular songs in English language teaching, System (2017), http://dx.doi.org/10.1016/j.system.2017.04.016

F. Tegge / System xxx (2017) 1e12

5

3.2. Data preparation For both the WOP and the WOPET, the majority of lyrics was obtained from lyrics websites and was checked manually for errors. In addition, the lyrics contained a number of genre-specific features that needed to be addressed in a principled manner. Firstly, so-called fade-outs, that is, the repetition of (parts of) the chorus or verses at the end of the song at decreasing volume, were removed. Secondly, song lyrics often contain a high frequency of words such as ooh, hmm, doobee, shoobee, na. Such words are referred to as non-lexical vocables, as they do not possess lexical content but instead are pronounced for the sake of vocalization itself, to carry the melody and rhythm and to sing without expressing semantic meaning (Chambers, 1980). Given their high frequency, they might greatly influence the lexical profile of the corpus. However, as they have little or no referential or social meaning, they can be assumed to pose only a small or no learning burden. In addition, many are easily recognizable and consistently used across the genre. A certain familiarity with songs among students will, thus, most likely result in familiarity with many of these words. I, therefore, decided to exclude them from the analysis. Clearly identifiable marginal words such as shh, oops, tada or wow, on the other hand, remained part of the analyzed corpus. 3.3. Data analysis The computer program used to analyze the lexical profile of the song corpora was RANGE (Heatley, Nation, & Coxhead, 2002). RANGE counts the number of times a word occurs in a corpus. Results are presented as tokens, types and word families. In addition, RANGE lists the word families according to their frequency in wordlists used during the analysis. For this purpose, Nation's (2004, 2006) twenty BNC wordlists were used, which allowed for the assessment of the overall vocabulary load of the song corpora. When analyzing songs, it needs to be kept in mind that written lyrics are often intended to reflect characteristics of spoken language and frequently contain a high number of contractions (e.g. I've, can't), connected speech (e.g. shoulda, woulda, coulda) and apostrophized abbreviations (e.g. lovin’). The RANGE program (Heatley et al., 2002) automatically separates contractions. For example, can't is counted as two words, that is, as can and not. Connected speech, on the other hand, is counted as one word. For example, coulda is counted as a family member of can. Apostrophized abbreviations were manually added to the frequency lists used in the analysis. For example, lovin’ was subsumed under the lemma love. Quasitranscriptions of spoken variations of words such as ya (you) or cuz (because) where added to their respective word families. 4. Results 4.1. Overview WOP The average number of words in a song in the charts corpus WOP is 435 (median ¼ 399.5). As the last row in Table 1 shows, the 408 songs in the corpus comprise a total of 180,892 tokens, 6486 types and 3942 word families from the BNC frequency lists. The first 1000 most frequent word families make up 160,925 tokens or 88.96% of running words in the corpus. The second 1000 word families account for 4.22% of running words, while the third most frequent 1000 word families make up 1.95% and the fourth 1000 word families comprise 1.19% of tokens. All wordlists beyond the 4000 most frequent word families account each for less than 1% of the tokens in the corpus with a decline over the mid-frequency levels to coverage of less than 0.1% beyond the 11th word frequency band. Proper nouns account for 0.69% of running words. While this result shows the relative importance of recognizing and knowing proper nouns, this percentage is much lower than in many other genres, such as movies with 2.67% (Webb & Rodgers, 2009a), TV programs with 2.96% (Webb & Rodgers, 2009b), novels with 1.53% (Nation, 2006), graded readers with 4.02% (Webb & Macalister, 2013), and even unscripted spoken discourse with 1.03% (Nation, 2006). The low number of proper nouns in the present charts corpus supports Murphey’s (1990) claim that songs are short and vague regarding personal referents and places referred to in the lyrics. The 20 BNC wordlists used for the analysis did not account for 0.68% of tokens in the corpus. These words seem to be reflective of particular sociolects or can be considered genre-specific, for example homie (12 tokens), tatted for tattooed (10 tokens) or illest (7 tokens); they tend to be fairly recently coined or popularized, for example dougie (41 tokens), twerk (14 tokens) or purp (8 tokens); or they display poetic or creative use or play with language, for example disturbia (15 tokens), beaching (4 tokens) or slizzard (3 tokens). The most frequent unlisted words are shawty, shorty or shortie, comprising 144 tokens, and hitters (91 tokens). Nine words were classified as marginal words and produced a total of 37 or 0.02% of running words: gosh, wow, yay, er, ugh, boohoo, oops, shh and sh. The most frequent transparent compounds were bartender (8), payphone (6) and streetlight (4). A separate RANGE analysis assessing the number of academic words in songs revealed that words from the Academic Word List (Coxhead, 2000) account for 589 or 0.33% of tokens and 162 word families in the WOP. 4.2. The lexical demand of charts songs The first research question asked how many words English learners needed to know to gain 95% and 98% coverage of charts song lyrics. Table 2 shows that knowledge of 3000 word families plus proper nouns, transparent compounds and marginal words is necessary to reach 95% coverage of the charts corpus WOP, while 6000 word families plus proper nouns, Please cite this article in press as: Tegge, F., The lexical coverage of popular songs in English language teaching, System (2017), http://dx.doi.org/10.1016/j.system.2017.04.016

6

F. Tegge / System xxx (2017) 1e12

Table 1 Coverage of the charts corpus WOP provided by 20 BNC wordlists, proper nouns, marginal words and transparent compounds in tokens, types and families. Word list

Tokens

1000 2000 3000 4000 5000 6000 7000 8000 9000 10,000 11,000 12,000 13,000 14,000 15,000 16,000 17,000 18,000 19,000 20,000 Proper nouns Trans. compounds Marginal words Not in the list Total

Types

Families

N

%

N

%

N

160,925 7632 3524 2160 1499 490 676 370 183 138 226 123 63 58 50 43 44 31 39 38 1242 67 37 1234 180,892

88.96 4.22 1.95 1.19 0.83 0.27 0.37 0.20 0.10 0.08 0.12 0.07 0.03 0.03 0.03 0.02 0.02 0.02 0.02 0.02 0.69 0.04 0.02 0.68

2015 1102 720 413 324 190 142 100 94 76 61 45 37 23 24 19 15 17 20 10 535 30 9 465 6486

31.07 16.99 11.10 6.37 4.99 2.93 2.19 1.54 1.45 1.17 0.94 0.69 0.57 0.35 0.37 0.29 0.23 0.26 0.31 0.15 8.25 0.46 0.14 7.17

863 654 495 318 260 159 125 86 88 69 57 43 34 22 22 18 14 17 18 10 532 29 9 ? 3942

transparent compounds and marginal words need to be known to reach 98% coverage. If proper nouns (and transparent compounds and marginal words) are not assumed to be known, knowledge of the most frequent 3000 word families is still sufficient to reach 95% coverage of chart songs. However, without proper nouns, knowledge of 8000 word families is required to gain 98% coverage. Assuming that 95% coverage of charts songs is sufficient for adequate comprehension of songs, results suggest that the lexical demand of charts song lyrics is located within the range of English learners’ vocabulary knowledge overall. Laufer (1998), for example, found that High School graduates in Israel have a vocabulary size of 3500 word families, and Laufer

Table 2 Cumulative coverage in percent of the charts corpus WOP, excluding and including proper nouns (PN), transparent compounds (TC) and marginal words (MW). (95%- and 98%-thresholds in bold.) Wordlist 1000 2000 3000 4000 5000 6000 7000 8000 9000 10,000 11,000 12,000 13,000 14,000 15,000 16,000 17,000 18,000 19,000 20,000 Proper nouns Transp. compounds Marginal words

Coverage without PN, TC, MW

Coverage with PN, TC, MW

88.96 93.18 95.13 96.32 97.15 97.42 97.79 97.99 98.09 98.17 98.29 98.36 98.39 98.42 98.45 98.47 98.49 98.51 98.53 98.55 0.62 0.06 0.02

89.70 93.92 95.87 97.06 97.89 98.16 98.53 98.73 98.83 98.91 99.03 99.10 99.13 99.16 99.19 99.21 99.23 99.25 99.27 99.29

Please cite this article in press as: Tegge, F., The lexical coverage of popular songs in English language teaching, System (2017), http://dx.doi.org/10.1016/j.system.2017.04.016

F. Tegge / System xxx (2017) 1e12

7

(2001) showed that English university majors in China have a vocabulary size of 4000 word families. However, if 98% coverage is required for unassisted comprehension of song lyrics, learners need to know the 6000 most frequent word families, a vocabulary size likely to be found in only fairly advanced learners. These results suggest that songs listed in the charts can be seen as lexically too demanding for many ESL and EFL learners, if unassisted text comprehension is the goal. This finding begs the question whether songs which are actually recommended for use in English classes by language-teaching professionals display a similar lexical demand. 4.3. Overview WOPET The average number of words in a song in the pedagogical WOPET is 289 (median ¼ 269). Table 3 shows that the 635 songs in the WOPET use a total of 177,384 tokens, 6496 types and 4018 word families from the BNC frequency lists. The first 1000 most frequent word families make up 159,169 tokens or 89.7% of words in the corpus. The second set of 1000 word families accounts for 5.1%, and the third most frequent 1000 word families make up 1.8% of tokens in the pedagogical song corpus. All wordlists beyond the 3000 most frequent word families account each for less than 1% of the tokens in the corpus, with a rapid decline over the mid-frequency levels to coverage of less than 0.2% beyond the 7th word frequency band. Proper nouns account for 0.62% of running words. The 20 BNC wordlists used in the analysis do not account for 0.18% of tokens in the corpus. The most frequent unlisted words are boogie woogie, shawty, biddy and hearties. Eight words were classified as marginal words and produced a total of 40 or 0.02% of running words: shh, neh, er, yay, boohoo, oops, tada and wow. The most frequent transparent compounds were hometown (19), sandman (10), sunshiny (9), cowgirl, schoolyard, and superhighway (6). A separate RANGE analysis revealed that words from the Academic Word List (Coxhead, 2000) account for 584 or 0.33% of tokens and 182 academic word families in the WOPET. 4.4. The lexical demand of teacher-selected songs The second research question asked how many words English learners needed to know to gain 95% and 98% coverage of song lyrics chosen by teachers or material designers for in-class use. Table 4 shows that knowledge of 2000 word families plus proper nouns, transparent compounds and marginal words is necessary to reach 95% coverage of the pedagogical song corpus WOPET, while 4000 word families plus proper nouns, transparent compounds and marginal words need to be known to reach 98% coverage. The lexical demand is slightly higher if proper nouns (and transparent compounds and marginal words) are not assumed to be known. That is, without proper nouns knowledge of 3000 and 5000 word families is required for 95% and 98% coverage of tokens in the corpus. Since the pedagogical WOPET consists of three subcorpora containing songs recommended for use with learners of different proficiency levels, it is appropriate to investigate potential differences in coverage between the three subcorpora. As Table 5 shows, the subcorpora can be ranked in terms of lexical coverage, with the advanced subcorpus displaying the highest and the beginner subcorpus displaying the lowest coverage provided by high frequency words. However, differences are small and a vocabulary size of 4000 word families plus proper nouns, transparent compounds and marginal words is required to gain 98% coverage of all three subcorpora. Given the division into three subcorpora and the high number of very short texts in the corpus, it was further investigated whether the required vocabulary knowledge of 4000 words to reach 98% coverage is a good representation of the lexical demand of the majority of texts in the corpus. In order to ascertain that the overall lexical demand of 4000 word families was indeed representative, the 95%-confidence interval for a subset of 108 songs was calculated. The 108 songs consisted of 36 songs from each subcorpus. Songs from the beginner and intermediate subcorpora were selected randomly. However, the advanced subcorpus only comprised 36 songs in total, and a random selection was not possible. As Table 6 shows, all three corpora contain a similar number of songs recommended by teachers in the survey or on teacher websites. Calculating the 95%-confidence interval can provide an indication whether the average lexical demand of 4000 word families does indeed pertain to the majority of songs in the corpus. In order to calculate the 95%-confidence interval, the BNCfrequency bands were understood to be categorical data ordered from 1 to 20. Results show that the statistical average is 4.27. The standard deviation is 3.14, and the measurement error is 0.592274. The lower bound of the 95% confidence interval is 3.676245, whereas the upper bound is 4.860793, indicating that we can be 95% confident that the mean is within the range of 3.7 and 4.9. A required vocabulary size of 4000 word families is, thus, a good estimate of the lexical demand of the 108 songs and also of the entire WOPET. 4.5. Comparing the WOP and the WOPET A comparison of the WOP and the pedagogical WOPET shows that language professionals tend to choose lyrics that are overall more learner-appropriate in terms of their lexical demand: Teachers select lyrics that are clearly less demanding than the genre as a whole and that make greater use of the 2000 most frequent words. While charts songs require knowledge of 6000 word families plus proper nouns, teacher-selected songs only require knowledge of 4000 word families plus proper nouns to reach 98% coverage. In addition, songs actually used for language-teaching purposes are overall shorter, that is, they contain on average one third fewer words than pop songs in general. Please cite this article in press as: Tegge, F., The lexical coverage of popular songs in English language teaching, System (2017), http://dx.doi.org/10.1016/j.system.2017.04.016

8

F. Tegge / System xxx (2017) 1e12

Table 3 Coverage of the pedagogical WOPET provided by 20 BNC wordlists, proper nouns, marginal words and transparent compounds in tokens, types and families. Wordlist

Tokens

1000 2000 3000 4000 5000 6000 7000 8000 9000 10,000 11,000 12,000 13,000 14,000 15,000 16,000 17,000 18,000 19,000 20,000 Proper nouns Trans. compounds Marginal words Not in the list Total

Types

Families

N

%

N

%

N

159,169 9017 3136 1570 930 585 290 161 255 186 131 75 74 22 62 61 28 21 20 38 1094 103 40 316 177,384

89.73 5.08 1.77 0.89 0.52 0.33 0.16 0.09 0.14 0.10 0.07 0.04 0.04 0.01 0.03 0.03 0.02 0.01 0.01 0.02 0.62 0.06 0.02 0.18

2181 1209 800 508 337 206 156 107 106 78 70 48 37 15 11 25 13 9 14 15 356 43 8 144 6496

33.57 18.61 12.32 7.82 5.19 3.17 2.40 1.65 1.63 1.20 1.08 0.74 0.57 0.23 0.17 0.38 0.20 0.14 0.22 0.23 5.48 0.66 0.12 2.22

882 685 532 388 281 174 145 98 97 73 65 46 36 15 10 23 13 9 14 14 356 43 8 ? 4018

Table 4 Cumulative coverage in percent of the pedagogical WOPET, excluding and including proper nouns (PN), transparent compounds (TC) and marginal words (MW). (95%- and 98%-thresholds in bold.) Wordlist 1000 2000 3000 4000 5000 6000 7000 8000 9000 10,000 11,000 12,000 13,000 14,000 15,000 16,000 17,000 18,000 19,000 20,000 Proper nouns Transp. compounds Marginal words

Coverage without PN, TC, MW

Coverage with PN, TC, MW

89.73 94.81 96.58 97.47 97.99 98.32 98.48 98.57 98.71 98.81 98.88 98.92 98.96 98.97 99.00 99.03 99.05 99.06 99.07 99.09 0.62 0.06 0.02

90.43 95.51 97.28 98.17 98.69 99.02 99.18 99.27 99.41 99.51 99.58 99.62 99.66 99.67 99.7 99.73 99.75 99.76 99.77 99.79

Table 5 Cumulative coverage in percent of the complete WOPET and its subcorpora including proper nouns (PN), transparent compounds (TC) and marginal words (MW). (95%- and 98%-thresholds in bold.) Wordlist 1000 2000 3000 4000 5000

þ þ þ þ þ

PN, PN, PN, PN, PN,

TC, TC, TC, TC, TC,

MW MW MW MW MW

Complete WOPET

Beginner

Intermediate

Advanced

90.43 95.51 97.28 98.17 98.69

91.16 96.06 97.57 98.45 98.92

90.23 95.26 97.17 98.02 98.58

88.38 94.88 96.77 97.94 98.43

Please cite this article in press as: Tegge, F., The lexical coverage of popular songs in English language teaching, System (2017), http://dx.doi.org/10.1016/j.system.2017.04.016

F. Tegge / System xxx (2017) 1e12

9

Table 6 The sources of songs in the three subcorpora. Subcorpus (N)

Source

N

%

Beginner (36)

Survey Web Textbook

13 19 4

36.1 52.8 11.1

Intermediate (36)

Survey Web Textbook

21 12 3

58.3 33.3 8.3

Advanced (36)

Survey Web Textbook

30 1 5

83.3 2.8 13.9

The fact that teachers and materials designers select songs containing more high-frequency words than pop songs in general indicates that they make e at least to a certain degree e good judgments of word frequency. This seems to contradict McCrostie’s (2007) and Alderson’s (2007) findings that language teachers and other language professionals like corpus linguists do not display particularly good judgement when it comes to the objective frequency of words in general language use. However, we do not know how these teachers arrived at their judgment and selection. While studies on frequency judgments rely on the participants' intuition, the teachers and materials designers contributing to the WOPET might have relied on outside sources including frequency lists and materials based on such lists. It seems likely that they selected songs that corresponded with the vocabulary needs of the target learners as defined by the curriculum and other core teaching materials. Still, these speculations do not detract from the fact that these language professionals made well-informed choices regarding the lexical profile of the selected songs. Furthermore, the pedagogical song corpus contains clearly fewer words that are not included in any of the wordlists (0.18% in the WOPET compared to 0.68% in the WOP), which indicates that teachers try to avoid lyrics containing language reflective of a particular sociolect, genre-specific words and lyrics displaying creative language play. Given the survey respondents’ overall sensitivity to the lexical demand of songs, it is surprising, however, that there is only a relatively small difference in the vocabulary size necessary to reach 98% coverage of the three subcorpora in the WOPET, as we might expect the selected lyrics to more clearly display lower lexical demand when used with lower-level learners. Webb and Macalister (2013) found a similar lack of clear differences between four stages of graded readers in the Oxford Bookworm Series, which all required a vocabulary size of 3000 words for 98% coverage. The fairly small difference in lexical demand between the subcorpora in the pedagogical song corpus might be explained by the short length of the songs with only 289 words, resulting in a relatively small number of unknown words per song. In addition, song lyrics tend to be highly repetitive, which can result in a high number of unknown tokens but still a low number of unknown types and word families. For example, the word submarine occurs 25 times in only one song, “Yellow Submarine” by The Beatles, and thus accounts for 14% of the tokens in that song. And while we have seen above that 4000 word families is a good estimate of the overall lexical demand of the texts in the WOPET, there is also some variability between songs, indicating that teachers find and use songs with a lower vocabulary load, including lyrics requiring a vocabulary size of only 1000 and 2000 word families to gain 98% coverage. The fairly high vocabulary load in songs intended for beginners and intermediate learners might also reflect their use in the classroom not for extensive but intensive listening and reading, and for intentional rather than incidental vocabulary learning. The threshold of 98% coverage for adequate text comprehension has been established for unassisted, extensive reading, for reading for pleasure (Hirsh & Nation, 1992) and for incidental vocabulary learning (Webb & Rodgers, 2009a). A high vocabulary load can still be considered appropriate, if intensive reading or listening and intentional vocabulary learning is the goal. Tegge (2015) showed that songs used in the classroom are frequently accompanied by a number of activities that assist comprehension and, thus, allow for a higher vocabulary load in the target text. Tegge's teacher questionnaire showed that a song used during a language lesson is typically repeated two to three times and frequently used in conjunction with one or several meaning- and language-focused activities including gap-fill activities, content discussions and a sing-along. 4.6. The lexical profiles of various genres The third research question was concerned with the vocabulary load of pop song lyrics in comparison with other text genres. As Table 7 illustrates, songs found in the charts are similar to other scripted and unscripted spoken genres regarding their lexical demand. Teacher-selected songs, however, require knowledge of comparatively fewer words than both authentic written and spoken genres to reach 95% and 98% coverage. The difference is most marked e up to 4000 word families e when comparing teacher-selected songs and written texts. A comparison with the Oxford Bookworm Series on the other hand reveals that teacher-selected songs and graded readers both require knowledge of 2000 word families plus proper nouns and marginal words to gain 95% coverage. When 98% coverage is the target, graded readers are somewhat less demanding, requiring a vocabulary size of only 3000 word families, while teacher-selected songs require knowledge of 4000 word families. This comparison of various text genres suggests that Please cite this article in press as: Tegge, F., The lexical coverage of popular songs in English language teaching, System (2017), http://dx.doi.org/10.1016/j.system.2017.04.016

10

F. Tegge / System xxx (2017) 1e12

Table 7 Lexical knowledge required for 95% and 98% coverage of different text genres (PN proper nouns, MW ¼ marginal words, TC ¼ transparent compounds). Text Genre Pop songs (charts) Teacher-selected songs Spoken discourse Spoken discourse Movies TV programs Novels (adult) Novels (adolescent) Newspaper articles Children's literature Literature (adults) Graded readers (stage 4 of 4)

95%

98%

Study

3000 þ PN, MW, TC 2000 þ PN, MW, TC 3000 þ PN 3000 3000 þ PN, MW 3000 þ PN, MW 4-5000 þ PN e 4000 þ PN 3000 þ PN, MW 3000 þ PN, MW 2000 þ PN, MW

6000 þ PN, MW, TC 4000 þ PN, MW, TC 7000 þ PN e 6000 þ PN, MW 7000 þ PN, MW

this study (WOP)

Nation, 2006 Adolphs & Schmitt, 2003 Webb & Rodgers, 2009b Webb & Rodgers, 2009a

8-9000 þ PN 5000 8-9000 þ PN 10,000 þ PN, MW 10,000 þ PN, MW 3000 þ PN, MW

Nation, 2006 Hirsh & Nation, 1992 Nation, 2006 Webb & Macalister, 2013 Webb & Macalister, 2013 Webb & Macalister, 2013

this study (WOPET)

teacher-selected lyrics are located between graded readers and authentic spoken and written materials and can, thus, be seen as entry-level authentic reading material or material to be read while listening. 4.7. Lexical learning opportunities The fourth research question addressed the vocabulary learning potential afforded by songs used in English classrooms. As the analysis results show, teacher-selected songs display comparatively high use of high-frequency words from the first three frequency bands. In addition, words from the 10th to the 20th frequency bands account for only 0.56% of tokens compared, for example, to 1.32% in TV programs and 1.25% in movies. This finding further supports the hypothesis that songs are located between graded readers and other authentic text genres that can be used to teach or rehearse mid-frequency words from the 4th to 9th frequency bands. Furthermore, a separate analysis showed that both the WOPET and the WOP contained few academic words, suggesting that popular songs might not be the ideal genre for teaching and learning academic vocabulary. However, such a general statement does not apply to each individual song, as careful assessment and selection of individual texts can, of course, ensure a lexical profile appropriate for the intended purpose. Sting's Russians, for example, contains several words from Coxhead’s (2000) Academic Word List (ideology, ignorant, logical, precedent, respond). 5. Discussion Pop songs can be understood as a written-to-be-spoken (or sung) genre on the performers’ side, while they are intended to be listened to and also listened-to-while-read (Kreyer & Mukherjee, 2007). The present analysis of a chart song corpus (WOP) and a pedagogical song corpus consisting of songs used in English teaching (WOPET) allows us to draw certain conclusions. Firstly, pop songs in the charts appear to be similar to authentic scripted and unscripted spoken discourse in their overall lexical demand while displaying a clearly lower vocabulary load than written genres. Pop lyrics can thus serve as entry-level authentic reading or reading-while-listening material. Secondly, when teachers use songs to teach English, they select songs which are not only shorter but also display a clearly lower vocabulary load than the average song found in the billboard charts. Lexical load is thus, not surprisingly, of concern to teachers when choosing songs for languagepedagogical purposes. However, despite their comparably lower lexical demand, pop lyrics can still be considered outside the range of lexical knowledge of many English learners, if intended for unassisted reading and listening and incidental vocabulary learning. Even teacher-selected songs tend to be lexically too demanding for many learners, in particular beginners. For high-advanced learners, on the other hand, they might offer few encounters with unknown vocabulary. 5.1. Pedagogical implications One way of addressing this problem is to take a tiered approach to song use and differentiate regarding the purpose of the activity and the level of assistance provided by the teacher and the material. As Bowler and Parminter (2002) put it: “Text level of challenge þ task level of support ¼ student success” (p. 59). In other words, in beginning and intermediate classes it seems advisable to use songs for intensive reading or listening and intentional vocabulary learning and to provide a greater amount of assistance for comprehension. Advanced learners, on the other hand, can cope with less support. In addition, as the number of unknown words in pop lyrics is likely to be limited for advanced learners, pop songs seem more useful to consolidate and entrench already familiar words and to support the acquisition of deep word knowledge beyond a first-form-meaning connection, including collocations and syntagmatic relationships. Please cite this article in press as: Tegge, F., The lexical coverage of popular songs in English language teaching, System (2017), http://dx.doi.org/10.1016/j.system.2017.04.016

F. Tegge / System xxx (2017) 1e12

11

Finally, it needs to be acknowledged that there is some variability within the corpora, and that it is possible to find songs that display a very low or very high vocabulary load. In other words, findings pertaining to the whole genre are not necessarily applicable to individual texts and teachers can find songs that suit their individual purposes and learners. In this regard, the relative brevity of the lyrics e with the average song in the charts corpus WOP comprising 435 running words e can be seen as advantageous. The short length of pop songs compared to other text genres can be considered an advantage for several reasons: Firstly, teachers do not have to spend an excessive amount of time assessing a song's lexis. Webb and Rodgers (2009b) pointed out the variability of vocabulary demand in different episodes of the same TV series and suggest that “teachers may need to evaluate a number of programs before they are able to find one that is appropriate for learners” (p. 357). Considering that the average drama in their corpus had a word count of 6,741, songs are faster and easier to assess in terms of their vocabulary load. Another advantage is the possibility to use songs in their entirety during a language lesson. In addition, the short length renders songs suitable for repeated listening and singing during a lesson, a practice that can support the entrenchment of word knowledge. And finally, the relative brevity of songs compared to other scripted aural texts allows more readily for the implementation of one or several additional activities around the song. 5.2. Limitations The present corpus study has limitations: Most importantly, the frequent repetition of individual words in one or only few of the very short texts in both corpora might have resulted in a somewhat inflated estimation of vocabulary knowledge necessary to reach 95% and 98% coverage of the corpora as well as of individual songs. Further research involving larger corpora is needed which also takes into account the range and dispersion of words in the corpus. Furthermore, it can be argued that wordlists with a stronger focus on American English, such as Cobb's BNC-COCA lists, might have been more appropriate. However, no such lists were available when the WOPET was analyzed. In addition, the use of Nation's BNC wordlists allows for a comparison with other studies utilizing the same methodological approach. Another limitation of the present study might be its narrow focus on vocabulary. When investigating the comprehensibility and readability of lyrics, assessing the vocabulary load is a good starting point. As Laufer and Ravenhorst-Kalovski (2010) pointed out, vocabulary is “a good predictor of reading, if not the best” (p. 16). However, assessing the lexical load of lyrics based on word frequencies might not be sufficient to determine whether they are suited for learners. For one, such an analysis , recommended for use with does not capture idiomatic language use. For example, the song “If I were a boy” by Beyonce intermediate learners, appears to be lexically fairly undemanding, requiring knowledge of only 2000 word families for 98% coverage. However, the song makes frequent use of idiomatic language: Expressions like chase after girls, kick it with someone and stick up for someone consist entirely of high-frequency words but can create greater barriers to comprehension than lowfrequency words. Other language- and discourse-features such as sentence structure and text cohesion affect readability as well. Further research is necessary, and such research needs to go beyond word length and frequency and sentence length, which tend to form the core of many readability formulas. 6. Conclusion The present study provides insight into the vocabulary load of pop songs overall as well as teacher-selected songs in particular. It shows that pop songs can function as entry-level authentic texts but also highlights the need for a tiered approach when using songs with learners at different proficiency levels. At the same time, the findings have implications regarding the informative value of assessing the lexical coverage of a whole genre for language learning and teaching: The study highlights the possible contrast between the lexical profile of a text genre overall and that of a pre-selected sample of the same genre, indicating that careful assessment of individual texts can ensure a lexical profile that is more appropriate for the intended learners and use. Acknowledgements The author wishes to thank the reviewers for their insightful and constructive feedback. Furthermore, she would like to express her sincerest gratitude to her doctoral supervisors Dr Frank Boers and Dr Averil Coxhead for their guidance, insight, and encouragement. This research was supported by Victoria University of Wellington and Massey University. References Adolphs, S., & Schmitt, N. (2003). Lexical coverage of spoken discourse. Applied Linguistics, 24(4), 425e438. http://dx.doi.org/10.1093/applin/24.4.425. Alderson, J. C. (2007). Judging the frequency of English words. Applied Linguistics, 28(3), 383e409. http://dx.doi.org/10.1093/applin/amm024. Bauer, L., & Nation, P. (1993). Word families. International Journal of Lexicography, 6(4), 253e279. http://dx.doi.org/10.1093/ijl/6.4.253. Bonk, W. (2000). Second language lexical knowledge and listening comprehension. International Journal of Listening, 14(1), 14e31. http://dx.doi.org/10.1080/ 10904018.2000.10499033. Bowler, B., & Parminter, S. (2002). Mixed-level teaching: Tiered tasks and bias tasks. In J. C. Richards, & W. A. Renandya (Eds.), Methodology in language teaching: An anthology of current practice (p. 59). Cambridge: Cambridge University Press. Chambers, C. K. (1980). Non-lexical vocables in Scottish traditional music. Unpublished doctoral dissertation. United Kingdom: University of Edinburgh. Coxhead, A. (2000). A new academic word list. TESOL Quarterly, 34(2), 213e238. http://dx.doi.org/10.2307/3587951.

Please cite this article in press as: Tegge, F., The lexical coverage of popular songs in English language teaching, System (2017), http://dx.doi.org/10.1016/j.system.2017.04.016

12

F. Tegge / System xxx (2017) 1e12

Flesch, R. (1974). The art of readable writing. New York: Harper & Row. Green, J. M. (1993). Student attitudes toward communicative and non-communicative activities: Do enjoyment and effectiveness go together? The Modern Language Journal, 77(1), 1e10. http://dx.doi.org/10.1111/j.1540-4781.1993.tb01938.x. Heatley, A., Nation, P., & Coxhead, A. (2002). Range and frequency programs. Available at: http://www.victoria.ac.nz/lals/staff/paul-nation.aspx. Hirsh, D., & Nation, P. (1992). What vocabulary size is needed to read unsimplified texts for pleasure? Reading in a Foreign Language, 8(2), 689e696. Hu, M., & Nation, P. (2000). Unknown vocabulary density and reading comprehension. Reading in a Foreign Language, 13(1), 403e430. Kreyer, R., & Mukherjee, J. (2007). The style of pop song lyrics: A corpus-linguistic pilot study. Anglia - Zeitschrift für englische Philologie, 125(1), 31e58. http://dx.doi.org/10.1515/ANGL.2007.31. Kyongho, H., & Nation, P. (1989). Reducing the vocabulary load and encouraging vocabulary learning through reading newspapers. Reading in a Foreign Language, 6(1), 323e335. Laufer, B. (1989). What percentage of text-lexis is essential for comprehension? In C. Lauren, & M. Nordman (Eds.), Special language: From humans thinking to thinking machines (pp. 316e323). Bristol, UK: Multilingual Matters. Laufer, B. (1998). The development of passive and active vocabulary in a second language: Same or different? Applied Linguistics, 19(2), 255e271. http://dx. doi.org/10.1093/applin/19.2.255. Laufer, B. (2001). Quantitative evaluation of vocabulary: How it can be done and what it is good for. In C. Elder, A. Brown, E. Grove, K. Hill, N. Iwashita, T. Lumley, et al. (Eds.), Experimenting with uncertainty: Essays in honour of alan davies (pp. 241e250). Cambridge: Cambridge University Press. Laufer, B. (2013). Lexical thresholds for reading comprehension: What they are and how they can be used for teaching purposes. TESOL Quarterly, 47(4), 867e872. http://dx.doi.org/10.1002/tesq.140. Laufer, B., & Ravenhorst-Kalovski, G. C. (2010). Lexical threshold revisited: Lexical text coverage, learners' vocabulary size and reading comprehension. Reading in a Foreign Language, 22(1), 15e30. Lund, R. J. (1991). A comparison of second language listening and reading comprehension. The Modern Language Journal, 75(2), 196e204. http://dx.doi.org/ 10.1111/j.1540-4781.1991.tb05350.x. McCrostie, J. (2007). Investigating the accuracy of teachers' word frequency intuitions. RELC Journal, 38(1), 53e66. http://dx.doi.org/10.177/ 0033688206076158. Murphey, T. (1990). Song and music in language learning. European University Studies: Education e Peter Lang. Nation, P. (2004). A study of the most frequent word families in the British National Corpus. In P. Bogaards, & B. Laufer-Dvorkin (Eds.), Vocabulary in a second language: Selection, acquisition, and testing (pp. 3e13). John Benjamin Publishers. http://dx.doi.org/10.1075/lllt.10. Nation, I. S. P. (2006). How large a vocabulary is needed for reading and listening? Canadian Modern Language Review/La revue canadienne des langues vivantes, 63(1), 59e82. http://dx.doi.org/10.3138/cmlr.63.1.59. Nation, P., & Kyongho, H. (1995). Where would general service vocabulary stop and special purposes vocabulary begin? System, 23(1), 35e41. North, A. C., Hargreaves, D. J., & Hargreaves, J. J. (2004). Use of music in everyday life. Music Perception, 22(1), 41e77. http://dx.doi.org/10.1525/mp.2004.22.1. 41. Rideout, V., Foehr, U., & Roberts, D. (2010). Generation M2 e media in the lives of 8- to 18-year-olds (a kaiser family foundation study). Retrieved 31/08/ 2015, from http://files.eric.ed.gov/fulltext/ED527859.pdf. Schmitt, N., Jiang, X., & Grabe, W. (2011). The percentage of words known in a text and reading comprehension. The Modern Language Journal, 95(1), 26e43. http://dx.doi.org/10.1111/j.1540-4781.2011.01146.x. Stæhr, L. (2008). Vocabulary size and the skills of listening, reading and writing. The Language Learning Journal, 36(2), 139e152. http://dx.doi.org/10.1080/ 09571730802389975. Tegge, F. (2015). Investigating song-based language teaching and its effect on lexical learning. Unpublished doctoral dissertation. New Zealand: Victoria University of Wellington. Thorndike, E., & Lorge, I. (1944). The Teacher's word book of 30,000 words. New York: Teachers College, Columbia University. Webb, S., & Macalister, J. (2013). Is text written for children useful for L2 extensive reading? TESOL Quarterly, 47(2), 300e322. http://dx.doi.org/10.1002/tesq. 70. Webb, S., & Rodgers, M. P. H. (2009a). The lexical coverage of movies. Applied Linguistics, 30(3), 407e427. http://dx.doi.org/10.1093/applin/amp010. Webb, S., & Rodgers, M. P. H. (2009b). Vocabulary demands of television programs. Language Learning, 59(2), 335e366. http://dx.doi.org/10.1111/j.14679922.2009.00509.x. West, M. (1953). A general service list of English words: With semantic frequencies and a supplementary word-list for the writing of popular science and technology. London: Longman. van Zeeland, H., & Schmitt, N. (2013). Lexical coverage in L1 and L2 listening comprehension: The same or different from reading comprehension? Applied Linguistics, 34(4), 457e479. http://dx.doi.org/10.1093/applin/ams074. Friederike Tegge is lecturer in linguistics at Massey University in Palmerston North, New Zealand. Her research interests include vocabulary studies, classroom-based research and songs in language teaching. In 2015, she finished her PhD thesis entitled “Investigating song-based language teaching and its effect on lexical learning” at Victoria University of Wellington under the supervision of Dr Frank Boers and Dr Averil Coxhead.

Please cite this article in press as: Tegge, F., The lexical coverage of popular songs in English language teaching, System (2017), http://dx.doi.org/10.1016/j.system.2017.04.016