Revisiting the notion of ESL: A corpus-based analysis of English textbook instructional language

Learning English as a second language in the inner circle (ESL) is usually contrasted with learning it as a foreign language in the expanding circle (EFL). The notional landscape of ESL conjures up many positive images about ‘inner-circle pedagogy’, whereas, in fact, our knowledge about ESL activities and tasks is often unspoken, unanalysed and limited. In this paper, the researchers revisit this age-old problem by conducting bottom-up searches of a 154,681-word textbook corpus consisting solely of textbook instructional language. These sections declare task purposes and actions to be done, and are suitable for understanding ESL teaching practices. The study was conducted from an etic perspective of two researchers from the expanding circle. We retrace how we encountered a major impasse during this process, which then prompted a revision of our view of ESL. The Results section presents our corpus findings: There was an imbalance between the four skills, a strong emphasis on grammar, a large number of group/pair work requests and a narrowly-defined approach to literacy training. These findings add details to existing qualitatively-based ESL textbook studies. From expectation to realisation, this research calls for a preparedness in regard to enlarging our current understanding of what counts as ESL

Learning English as a second language in the inner circle (ESL) is usually contrasted with learning it as a foreign language in the expanding circle (EFL). The notional landscape of ESL conjures up many positive images about 'inner-circle pedagogy', whereas, in fact, our knowledge about ESL activities and tasks is often unspoken, unanalysed and limited. In this paper, the researchers revisit this age-old problem by conducting bottom-up searches of a 154,681-word textbook corpus consisting solely of textbook instructional language. These sections declare task purposes and actions to be done, and are suitable for understanding ESL teaching practices. The study was conducted from an etic perspective of two researchers from the expanding circle. We retrace how we encountered a major impasse during this process, which then prompted a revision of our view of ESL. The Results section presents our corpus findings: There was an imbalance between the four skills, a strong emphasis on grammar, a large number of group/pair work requests and a narrowly-defined approach to literacy training. These findings add details to existing qualitatively-based ESL textbook studies. From expectation to realisation, this research calls for a preparedness in regard to enlarging our current understanding of what counts as ESL pedagogy.

Conceptual framework
An issue that is particularly relevant to those of us educated in non-English-dominant contexts is the positive images prompted by the promise of learning English in English-dominant countries. Does it imply an endocentric model of teaching practices? What are common activities in ESL classrooms? There is no single reason that can explain our interest. From a macro-societal perspective, 'Western teaching practices are better than others', 'those from developed countries are better' and 'Western TESOL methodology dominates EFL classrooms' are some common sentiments within the teaching profession noted by World Englishes scholars (Holliday, 2005;Pennycook, 1994;Phillipson, 1992). Hong Kong's education system, for example, constantly looks to English-dominant countries for innovative pedagogical approaches in educational reforms (Morris and Adamson, 2010). Howatt and Widdowson (2004), in their monograph recounting historically important milestones pertaining to the spread of ELT worldwide, describe ESL as 'teaching English in the commonwealth', the 'quasi-mother-tongue', a 'hard necessity' and being 'more orthodox than EFL'. Learning English as a foreign language (EFL), on the other hand, brings to mind 'learning English in non-colonial countries', 'around the world' and being 'an interesting option for schools' (i.e., not a necessity). ESL contexts therefore suggest a closer affinity with native speaker ideals than EFL contexts (McArthur, 2006). In applied linguistics parlance and writing, researchers commonly draw on the ESL and EFL distinction, such as in 'ESL vs. EFL learners' and 'ESL vs. EFL classrooms', outlining an image that studying English in different contexts entails different learning outcomes. Paradoxically, there is also a tendency for the ELT literature to pay close attention to EFL practices, to the extent that practices in ESL settings are generally little discussed and are little known. ESL is in many ways envisaged to be different from EFL, conditionally, linguistically and in terms of teaching practice. This article gives micro-views of tasks and activities in ESL textbooks through corpus-linguistic evidence, in an effort to zero in on this concept. In this study, corpus linguistics provides a 'neutral' search method (McArthur, Lam-McArthur and Fontaine, 2018) through which to understand the pedagogies and routines emphasised in ESL textbooks.
Our major incentive for analysing English textbook activities stems from a simple curiosity: Do Anglo-centric materials imply an Anglocentric pedagogy? Termed global English textbooks or ESOL textbooks in this paper (Cooke and Simpson, 2008;Fletcher and Barr, 2009;Gray, 2010;Masuhara et al., 2008;McGrath, 2013;Tomlinson, 2008), this genre of materials is used in ESL contexts for the teaching of English. Tomlinson and Masuhara (2008, pp. 159-160) remark that 'most EFL learners who come to England have already achieved at least a pre-intermediate level and most of them are using materials published by the major British publishers. Often these materials take the form of global coursebooks. ' McGrath (2013, p. 9) further remarks that '[g]lobal coursebooks […] derive from an Anglo-centric view of the world […] native-speaker norms predominate.' These criticisms set a rather bleak backdrop for the current study. They reflect an emic viewpoint (Pike, 1967) from curriculum analysts, who are concerned about the exportation of Western-based ideologies to learners from other places. Despite these cautions, the phenomenon of 'native-speakerism' continues to be willingly accepted by learners and teachers in the expanding circle (Gong and Holliday, 2013). Nowadays, English is increasingly learned early in China and is offered via strong forms of locally-based immersion programmes (Feng and Adamson, 2019). In the city of Hong Kong, parents volitionally choose English-medium schools for their children over Chinese-medium schools (Li, 2002). Examples such as these in Asian cities (see, e.g., Lin and Man, 2009) suggest that provisions are being made available in the expanding circle to confer conditions that try to match, in spirit and in practice, learning English in English-dominant settings, nursing 'the-earlier-the-better' psyche of learners. These positive expectations in regard to ESL (simply, learning English in an English-immersive environment) are not totally square with the complex character of ESL education, to be discussed below. Adopting an etic perspective (Pike, 1967), the current researchers are curious about what ESL materials have to offer, whatever they may be. We stand in the expanding circle, peering inward. For this very reason, English as a Lingua Franca movement is not seen as relevant to the current discussion of teaching practices, as an etic position implies our interest in others' offerings and an acceptance of 'partisan' practices in others' cultures.
Seeing lexical patterns in instructional texts as non-random choices to repeat certain teaching actions, this study uses such patterns to infer ESL pedagogies. If Western-based textbooks prioritise certain teaching practices, these practices can be studied lexically and textually. The use of corpus data to investigate the interface between lexis and disciplinary ethos, while growing in genre and discourse studies (Hyland, 2012(Hyland, , 2015, is still uncommon in traditional textbook analyses. Much research in this area focuses on textbooks' content and linguistic exposure received by users (Chen, 2016;Hajiyeva, 2015;Walkov� a, 2020;Wood and Appel, 2014). These studies often ask such questions as: What is the lexical quality of English textbooks? How are collocational patterns taught in these textbooks? Is the teaching method based on corpus-informed insights? This group of studies is different from the present one, which uses collocational patterns to infer common teaching routines. As a part of the notion of a 'task workplan' (Breen, 1987;Ellis, 2003;Samuda and Bygate, 2008), instructional texts are purpose-declaring plans outlining how activities can be done (e.g., listen and check; work with a partner; use the words and phrases to …). This section of the textbooks circumscribes linguistic and non-linguistic outcomes to be achieved, and are written to engage learners in the real performance of tasks. The simplicity of these textual plans makes them suitable for grounding the present lexical analyses.
The current paper is structured as follows: The researchers use ESL in England as an example to discuss ESL provision in the inner circle. 1 English language education for those who do not speak English as their first language is generally offered in two ways: via mainstream education (EAL) and pull-out classes (ESOL). We describe how we encountered a major impasse during this study: Discovering that ESL is a nonobligatory part of present-day curricula compelled us to revise our expectations for ESL. We will explain how the proposition of this paper can be pursued by studying a class of Anglo-centric ESOL textbooksa selection that is not ideal but offers a snapshot of ESL. In the methodology section, we demonstrate how the use of a 40-item collocation list can effectively capture some consistent points in textbook writers' advice. Starting as a textbook study, we re-learned modern ESL provisions in English-dominant countries. Re-engaging with ESL from an etic perspective brings with it an unexpected advantage: It allows us, without much concern about research boundaries and domain specialisms, to consider EAL and ESOL as ways of practising ESL. The information summarised below, although probably not entirely new to specialists, is worth being repeated to general readers. From idealistic expectations to progressive realisations, the authors of this article argue for a greater understanding of the notion and scope of ESL nowadaysthe first objective of this studyand present and evaluate common ESL activities through a textbook language corpusthe second objective of this study.

EAL in mainstream education
Generalists who have worked and spent a large part of their education in the expanding circle may be unaware of the fact that ESL (in the form of separate English classes for learners) has been practically nonexistent for 30 years, 2 since the implementation of the statutory National Curriculum in England in 1991 (Costley and Leung, 2013;Leung, 1993). What was traditionally known as ESL is now called the EAL curriculum (Costley, 2014, p. 290;Leung, 2016, p. 159), in which English is seen as an additional language in the repertoire of children who are at the same time exposed to another language at home (DfE, 2019, p. 9). The decision to remove separate English training (traditional ESL) from the classroom means that all children follow the National English Curriculum and receive content-based instruction (i.e., learning English through English literature content, as well as through other content subjects) (Mohan et al., 2001). This move was largely in response to a mix of factors and requests: a vision to provide equal opportunities for all; an aversion to segregated teaching, which works against the spirit of social cohesion; and the belief that only 'Standard English' should be taught in state-funded education (see Costley, 2014;Leung, 2010Leung, , 2016. Furthermore, in reality, providing ESL training to all EAL learners is unnecessary. Not all EAL learners need English to the same degree; many learners were brought up in England and are in fact fluently bilingual (Leung, 2018). 3 In 2019, 1,563,294 learners in state-funded primary and secondary schools were estimated to be EAL learners (DfE, 2019).
Given this situation, there is little incentive for publishers to produce textbooks and learning materials for EAL learners. Leung (2018, p. 234) notes that there is 'no curricular visibility for EAL'; instead, it is 'deeply and invisibly enmeshed in classroom communication' (Leung, 2016, p. 164). Costley (2014, p. 288) notes that EAL is 'not a tangible or recognised curriculum in England.' This policy of integration has been 1 The choice of England as a reference frame for understanding ESL provisions is based on several reasons. There is a large amount of ESL research in this area and its model can also be seen in Scotland, Australia, Canada and New Zealand (e.g., Foley et al., 2013;Leung, 2007). Furthermore, as researchers from Hong Kong, our interest is in part explained by our curiosity regarding the English system. 2 Throughout this paper, we use ESL as a generic label to refer to learning English in classroom contexts in English-dominant countries (e.g., the UK). Although ESL can also mean learning the language at home and in the streets, we focus on classroom-based ESL practices.
3 EAL is a policy-makers' label. It does not provide information on proficiency. The traditional notion of ESL, which is suggestive of proficiency, as in 'ESL norms' and 'ESL usage', is not an entirely equivalent concept to EAL. This explains why ESL provision is non-obligatory, as some EAL learners obviously do not need it.
similarly implemented in both Australia and Canada (Leung, 2007;Mohan et al., 2001). From an etic standpoint, a generalist should now understand ESL as a form of assistive teaching, whereby specialist EAL staff (or teaching assistants) work alongside subject teachers to support a learner's acquisition of content knowledge; during the process, the learner 'picks up' English at the same time. In applied linguistics literature, scholars may use the term 'content-based ESL instruction' or 'content-based approach to L2 learning' (DeKeyser, 2007;Doughty and Williams, 1998) to refer to something akin to EAL, although exactly how they are equivalent is unknown. Use of terminology can contribute to confusion over the current status of ESL (or EAL).
Because it is assumed that English can be most successfully learned through total immersion in school activities, there is no explicit language pedagogy dedicated to EAL teaching (Costley and Leung, 2013). The literature advocates some universal teaching principles for EAL students, including 'student-oriented pedagogy', 'inclusive educational practices', 'opportunities for speaking and listening', 'collaborative learning', 'vocabulary development', 'enquiry-based activities', a 'hands-on approach', 'pupil-sensitive' practice, and 'grammatical input' (see, e.g., Demie and Lewis, 2018;Leung and Franson, 2001). One direct consequence of this policy is that there is an apparent difficulty in sourcing EAL materials: Published and publicly-endorsed materials do not exist. Conceptually, this presents a notional challenge to the researchers, as this had not been our impression of ESL until the chance of revisiting it arose. While it may be a debateable point whether or not ESL, in the form of EAL provision, truly exists, it certainly is different in nature from EFL practicesa point of which generalist researchers may not be fully aware.

ESOL in non-mainstream contexts
Besides EAL for school-aged ESL learners, another form of ESL, English for speakers of other languages (ESOL), is provided for those aged 16 years and above. 4 For many scholars working in the expanding circle, ESOL does not immediately spring to mind as a form of ESL, as it is nonmainstream, non-obligatory and is perceived as being 'non-standard' in the minds of many. In applied linguistics literature, scholars call this group of learners 'adult ESL learners' (Ellis, 2003;Gass, 2013;Lightbown and Spada, 2013;Nunan, 1988); this terminology does not immediately bring up ESOL as an equivalent term. In many English-dominant countries, the provision of ESOL opens up a viable route for people who want to acquire English language proficiency for the purposes of securing work and settling. We should be prepared to broaden our analytical lens to consider the pedagogical practices that can be detected within ESOL materials. In England, almost 20 years have passed since the launch of the official ELT curriculum for adults (Adult ESOL Core Curriculum) in 2001. This curriculum aimed to improve literacy levels in young people and adults (DfES, 2001). Before that, ESOL provision for this group had been informal in nature (Foster and Bolton, 2018). Structurally, ESOL classes take place in language schools and community facilities. In terms of numbers, the ESOL curriculum served 114,000 learners in 2016-2017 (Foster and Bolton, 2018), a fraction of those served by EAL.
ESOL provision in England is similar to arrangements in other English-speaking countries (e.g., the US and Australia). One of its aims is to link ESOL attainment levels to work and labour, employability and citizenship policies (Burns, 2006;Cooke and Simpson, 2008;Roberts et al., 2007;Simpson, 2015). In England and in the UK in general, the demand for this sort of ESL provision has arisen partly in response to the requirement, introduced in 2001, that those wishing to immigrate need to achieve a minimum level of CEFR B1 5 in an English language test (Simpson, 2015). This requirement was reinforced in a recent document published in (2018) by the UK Home Office, Knowledge of Language and Life in the UK (Version 21.0), in which it is stated that those arriving from outside 16 English-speaking countries need to satisfy this language requirement (p. 18). ESOL courses give focused teaching in regard to the four skills and support students in achieving the required English standard through external assessments (e.g., IELTS). ESOL practitioners typically adopt ESOL textbooks for use in their teaching (Cooke and Simpson, 2008;Roberts and Cooke, 2009). McGrath (2013, p. 9) criticises these books as being derived 'from an Anglo-centric view of the world and cultural realities. ' Gray (2002' Gray ( , 2010 considers these textbooks to be carriers of white-dominant ideologies in the form of ideas, images, icons and ways of talking. These critical viewpoints intensify our interest in this set of textbooks; they are native-based and ethno-centred. Our ingrained interest is in their ESL pedagogy. Seen from this perspective, the infiltration of white-dominant ideologies is less of a concern for practitioners like us and life-long learners of English. In terms of ESOL pedagogy, Masuhara et al. (2008, p. 297) note that '[t]he emphasis in most courses is on explicit teaching of declarative knowledge followed by controlled or guided practice.' In terms of the four skills, 'most of the courses give much greater prominence to listening and speaking than they do to reading and writing' (Masuhara et al., 2008, p. 299). Tomlinson, Dat, Masuhara, and Rubdy (2001) have described these books as 'UK-centred' (p. 86), 'teaching-centred' (p. 87) and 'grammar-centred' (p. 97). Tomlinson (2008, p. 3) claimed that 'many ELT materials […] focus on the teaching of linguistic items rather than on the provision of opportunities for acquisition and development.' These qualitative comments mostly concern the pedagogy and cultural content of these materials. The present article does not aim to repeat these efforts, but instead uses WordSmith Tools to identify phraseology in ESOL textbooks that pertains to teaching and learning. This paper provides micro-views of the activities, acts and routines that are found in these ESL materials.

Through the looking glass: A corpus-based analysis of ESL English textbooks
For those who have spent their lives in the expanding circle, learning English in an English-majority country is an enticing notion. 6 Unfortunately, scholars, teachers and parents are often unaware what ESL means in terms of content and method. Given the categories of the English curriculum outlined above (either threadbare, as with EAL, or else narrowly defined in work-or skill-based terms, as with ESOL), either one of these will be incomplete in describing ESL. The present study asks whether or not these inner-circle materials endorse certain ESL pedagogies and, if so, what they are. 7 In the pursuit of an answer to these questions, we are mindful of the risk of generalising ESL norms using a specific set of ESOL textbooks. Given the current impasse, there is a prima facie case to argue that these ESOL textbooks do offer a snapshot of ESL, if ESL is to be understood broadly, inclusive of general ESL teaching conditions, mainstream and non-mainstream. Our proposal to consider EAL and ESOL as forms of ESL can be seen as a function of our etic viewpoint as spectators from the expanding circle; a flexibility is afforded by this position, which is different from the rather secluded research area each of them is. This revision of ESL as a notion (beyond mainstream education) may cause unease among readers. However, a degree of awareness is beneficial for us to continue to read the teaching literature with realistic expectations (for example, when a reader encounters the term 'adult ESL learners' in the literature). The current study proceeds with these provisos.
All 14 books under investigation were benchmarked against the CEFR and corresponded to B1 and B2 levels (i.e., intermediate). The researchers sourced ESOL textbooks from a variety of publishers to avoid simplistic generalisations. Our focus is on one particular sectionthe textbooks' instructionswhich we consider to be purpose-declaring texts that can illuminate authorial preferences in lexical choices for the framing, introduction and delineation of pedagogical tasks.

Research questions
1. What are the 40 most frequent collocations (or collocation frames) in the instructional texts of 14 textbooks that may be used by ESOL learners? 2. How do these collocations reflect ESOL textbook pedagogies?

Building the specialised corpus
The current corpus is made up of the instructional texts of 14 contemporary textbooks from major British publishing houses, including Cambridge Empower, Cutting Edge, English File, Navigate, New Language Leader, Open Mind and Outcomes. The decision to build this corpus departs from the traditional use of corpus data to investigate word semantics, toward decoding teaching practices via textbooks' commands. This genre of ESOL textbooks is quite easy to identify, based on their blurbs, their titles, and the fact that they are all CEFRbenchmarked at B1 or B2 levels. They were all sourced from a specialist bookshop in the UK (https://www.eflbooks.co.uk/). The blurbs of these books show that these series seek balance in the teaching of the skills involved (e.g., Cutting Edge). Several series mention that they are suitable for adult learners (e.g., Navigate and Open Mind). Some of the textbooks pride themselves on the teaching of English in a globalised world (Language Leader), for the 21st century (Open Mind) and with global content (Outcomes). With our teaching experience from the expanding circle, it is evident that these materials are more suited to ESL contexts, rather than EFL contexts (see Appendix B for a list of the 14 textbooks).

Instructional texts
Corpus linguistics comprise a group of neutral techniques that can provide interesting insights when used with special texts. In our experience, instructional texts have almost been forgotten as essential textual plans carrying textbook writers' advice and actionable steps for learners. This function is missing in other parts of textbooks; for example, speaking tasks themselves do not tell the learner what he or she is supposed to do to carry out the tasks. Instructional texts are short and terse. The compactness of these texts means that the writer's meanings are often concatenated within a short space. A single instruction, such as the following, consists of smaller steps and the challenge presented to a qualitative researcher in this context concerns being able to accurately extract important information through 'bare eyes'.
Collocations are preferred meaning patterns. Searching for collocations (instead of single words) means that it is the cohesive links in the writers' advice (read the information) that will signal entire pedagogic actions. Such information, when repeated in a high frequency, arouses a learner's attention and implies importance for action. In constructing the corpus, the researchers took care to exclude short headings (e.g., Grammar Focus, Reading and Speaking, Follow Up: Writing, and so on), in order to prevent these headings from skewing the data. The corpus therefore only includes complete sentences.

Searching for the top 40 collocations
Lexical analysis software, WordSmith Tools 7.0 (Scott, 2017), was used to search for salient lexical collocations in the instructions (see the examples below). Two of the most prominent association measures currently in use, the Mutual Information (MI) score and the t-score (Gablasova, Brezina and McEnery, 2017), were used to indicate the relative strength between two collocates. In general, t-scores are good at indicating prominent occurrences of word combinations. MI scores, on the other hand, are suitable for determining special partnerships in a corpus; for example, in the ESOL textbooks, the word 'change' occurred more frequently with 'role' (MI 9.02) and 'partner' (MI 9.66) than other words (Hunston, 2002;Schmitt, 2010). These association measures are built into the WordSmith software. They were adopted in the present study in a preliminary screening stage.
[1] Read two more stories about mishaps. Complete the gaps with the best verb forms (there may be more than one possibility).
[2] Look at Audio script 1.3 on page 168 and underline examples of the present perfect simple, the past perfect simple and the past simple.
Examples such as these demonstrate that seemingly simple instructions often consist of micro steps that need to be undertaken (example [1] shows that the task involves both reading and vocabulary learning). The availability of statistical data allows us to determine with a greater degree of certainty whether or not certain information is repeated across the textbooks (i.e., the words in bold type). Note that commands may be expressed in different ways. The three types of lexical collocations under focus consist of bigrams (e.g., listen again [MI 5.44;t 24.25]), denoting a contiguous word sequence; three-word skipgrams (e.g., work (in) pairs [MI 6.08; t 32.18]), denoting a collocation with a gap in the middle; and four-word skipgrams (e.g., look (at) (the) photo [MI 5.89;t 10.26]). The notion of lexical bundles, consisting of contiguous words that form a fixed structure (e.g., the meaning of the and in the box below (Wood, 2010)), was not used in the current method, as these bundles are too rigid to allow for flexibility in textbook writing. While it would have been possible to further increase the window span, these three patterns already generated a sizeable amount of data for discussion.

Cleansing the corpus
The first step in constructing the corpus involved the manual scanning of ESOL textbooks and the conversion of these texts into a readable form using the optical character recognition (OCR) function in a scanner. During this process, the second author, along with research assistants, manually checked unusual characters and removed short headings that preceded the tasks. To thoroughly check for errors and typographic inconsistences between textbooks, the text files were then uploaded to Wordsketch.com, which identified further formatting issues, including inconsistent spellings (e.g., 'centre' and 'center', 'role-play' and 'roleplay'), inconsistent punctuation (such as 'e.g.' and 'eg.', 'words/phrases' and 'words or phrases') and other formatting issues (e.g., 'Choose T (true) or F (false) … ' can be simplified to 'Choose true or false … ').

Screening by frequency measures
The raw text files were uploaded to WordSmith Tools 7.0. The authors used a lemma list (which lemmatises related word-forms) and a stop list (which excludes unwanted words, such as function words and pronouns) to delimit the searches. The first stage of the search involved creating WordList index files. According to Scott (2017, p. 301), these 'record the positions of all the words in [a] text file.' The Relationship function further tightens the search criteria (see Fig. 1). In the current study, a collocation has to have both an MI score and a t-score of four or above (Durrant and Doherty, 2010), it must occur over 20 times per million words (Biber, Johansson, Leech, Conrad and Finegan, 1999) and it must have appeared in a minimum of four books. Applying these restrictions enabled us to downsize to a selection of 294 collocations for further screening (see Table 1).

Qualitative screening
In the next step, qualitative screening was essential to screening out combinations that are opaque or fragmented, as well as those that are unhelpful in informing ESL teaching practices. This process involved checking whether or not combinations evince clear semantic sense. Some of the collocations that were considered opaque and uninformative for the present purpose were do þ think (MI 5.59;t 20.42; as in 'Do you think men and women … '), again þ repeat (MI 5.77;t 8.15; as in 'Listen again and repeat the sentences'), people þ who (MI 5.57; t 5.79; as in 'Find people who do the same things as you'), work þ small (MI 4.76;t 11.79; as in 'Work in small groups … ') and then þ practise (MI 4.92; t 6.04).
Despite the removal of words such as 'work' and 'practise', these words were recovered in other collocations. For example, the word 'work', although lost in work small, is counted in the four-word skipgram work __ __ group [MI 4.37; t 7.31; as in 'Work in small groups' and 'Work with another group']. The collocation then practise, although removed due to the word 'then' (it is not a meaning-contributing word), is counted in practise saying (MI 9.11;t 11.81; as in 'Then practise saying the phrases below'). The second author read the concordance lines and the original textbook pages before making an initial decision; any potentially removable items were checked by the first author. Care was taken to not excessively expunge words. After screening, the number of eligible collocations was reduced to 118 (see Table 1).

Regrouping
Several extremely similar collocations were listed separately in the original list. For instance, listen check (531 instances) and listen repeat (185 instances) basically request the same act from ESL learners, as do work pair (1067 instances) and work group (494 instances). These separate listings were unhelpful in showing the extent to which a particular act is called for. Regrouping the collocations into 'frames', such as listen __ (second place in the list) and read __ (third place), offers aggregate views of pedagogical emphases. After similar kinds of collocations (read article, read text, read information and so on) are combined, it becomes possible to work out how common an activity is (639 requests in this case). For the sake of clarity, structurally different collocations were not combined, including read again (a bigram) and read __ article (a three-word skipgram). After regrouping, there were 80 collocations in the list (third column), arranged into 40 entries (second column).

Classroom learning acts
To answer research question two, the collocations were further coded into 'classroom learning acts' in order to estimate the pedagogies in these textbooks. For example, read article (third place; as in 'Read your article again in more detail … ' from Cutting Edge B2, p. 19) and listen again (fourth place) clearly call for reading and listening, and were coded as 'R' and 'L', respectively (in the eighth column). For the less transparent collocations, a decision was made after reading the concordance lines. We therefore decided that the collocation correct form (sixth; as in 'Complete the sentences with the correct form of the verbs in the box' from Cambridge Empower B2, p. 105) indicates a focus on lexical-grammatical accuracy (coded as 'LG'). The collocations write three (14th) was confirmed as referring to writing acts ('W'; as in 'Work alone and write three new predictions' from Cambridge Empower B1, p. 51). The collocation watch video (24th; as in 'You are going to watch a video about Churchill's life' from Navigate B2, p. 124) involves reading an animated film and listening to the narration at the same time; it was  double-coded as 'R' and 'L'. Although resembling writing, the collocation underline example (33rd) was agreed to be 'R' by the authors (as in 'Underline examples of the following stylistic devices' from New Language Leader B2, p. 11). Initial agreement between the authors reached 87.5%. Often, there was only a mild rather than a radical difference in opinion (for example, is asking a teacher's opinion (ask teacher, ranked 39th place) a kind of meaning-making constructivist act?). Each problematic case was resolved by revisiting the original texts and sometimes by involving a further research assistant. In cases of ambiguity, the collocations were classified as 'others' ('/'). For example, the collocations look photo (as in 'Look at the photos of 'yarn-bombing' and discuss … ' from Navigate B2, p. 40), check answer (as in 'Check the answers on page 128' from Navigate B1, p. 69), find out (as in 'Find out where they decide to go and why' from Outcome B1, p. 34), correct option, and correct answer (emphasising correctness is not a call for concrete action) were coded as 'others', because these collocations may not suggest actual reading or writing. The codes are displayed in the penultimate column of the 40item list in Appendix A. The rationale behind this procedure was to give a rough estimation of common ESL teaching practices, something that is hard to achieve through qualitative analyses.

Results: The top 40-item list 8
This paper calls for a broadening of the conceptual terrain of what counts as ESL pedagogy. By investigating the ways in which textbook writers carefully choose words in ESOL textbook commands, this method provides an angle into ESL activities and routines. The steps described above led us to the discovery of 40 frequent collocations (collocation frames). These collocations and the frequency of data facilitate listing, ranking, grouping and comparison. Owing to space limitations, it is not possible to explain each item in depth. The following sections aim to describe the most obvious features in the ESOL textbooks.

Reading the list
The collocations/collocation frames are ranked in the fifth column of the 40-item list ('Frequency'). Topping the list is the work __ frame (1561 instances), comprising the three-word skipgrams work pairs (1067 instances) and work group (494 instances). Since structurally different collocations were not combined, a related four-word skipgram, work __ __, takes the fifth place (468 instances). This latter frame comprises two collocations, work partner and work group, respectively (as in 'Work with a partner. Look at the photos of Paris. What do you recognise in them?' from Navigation B2, p. 44). These high numbers indicate how frequently learners are advised to work with others by these textbooks.
Second in the list is the three-word skipgram Listen __ frame (859 instances). This frame comprises listen check (531 instances), listen repeat (185 instances) and listen part (143 instances; as in 'You are going to listen to part of a lecture on group dynamics … ' from New Language Leader B2, p. 96). As explained above, the coding system attempted to truthfully capture the meanings implied by a collocation. In the case of listen repeat, the word 'repeat' signals a request for oral repetition (as in 'Listen and repeat the sentences. Copy the rhythm and try to get the/ə/sound right', from English File B1, p. 37). The collocation listen repeat was thus doubly coded as both 'L' (listening) and 'S' (speaking) (see the eighth column). Another structurally different collocation, listen again, takes the fourth place (as in 'listen again and tick the phrases you hear' from Cutting Edge B1, p. 66), and yet another frame, including listen people and listen conversation, can be found in the ninth place.
The third most frequent collocation is the read __ frame (639 instances). Reading 'articles' (320 instances) is most common, while reading 'extracts' (25 instances) is least common (as in 'Read the extracts from Bill Bryson's A Short History of Nearly Everything' from New Language Leader B2, p. 20). Further down in sixth place, collocations including correct option, correct answer, correct order and correct place (e.g., 'Write them in the correct place in the chart', from English File B2, p. 29) do not by themselves suggest any major ESL acts. They are contrasted with correct form and correct verb, which ask ESL learners to pay attention to grammatical accuracy (hence, coded as 'LG'). The decision to assign a particular code was therefore made on a case-by-case basis, involving reading and thinking about the original data. The rest of the list can be read in this way.

Estimation of the four skills
The second question asks about the types of pedagogies and learning acts (reading, writing, lexical-grammatical acts, etc.) in ESOL textbooks. A novel way to estimate teaching focuses, as proposed by this study, is to use the frequency data (fifth column) and the codes (eighth column) in the 40-item list. The table below represents this information by rearranging the related collocations under learning acts.
Based on the rearranged data, it is evident that the 14 ESOL textbooks contain many constructivist meaning-making activities (second column of Table 2), followed by a great deal of listening, grammar and, finally, traditional reading and writing activities. This sequence confirms observations made by prior qualitative research , which has noted the prominent emphasis in these books on listening and speaking. The current study adds finer details to these observations, in terms of constituting activities (i.e., 'listening' includes many sub-activities) and an activity's implementation (i.e., whether an activity involves pair work or group work).

Pair work and group work
Among the 2241 instances of 'constructivist' collocations (second column of Table 2), the most common ones include work pair (first;  (129) Look at the photos of (182) Listen and find out (125) a These numbers are calculated from the fifth column of the top 40-item list. In some rare cases, we found that two collocations with a similar function occurred in a command (e.g., 'Work in pairs. Take turns to say a word' from Open Mind B1, p. 47). The total number of 2241 collocations is a conservative estimation, in that it excludes duplications (i.e., 'take turns' was excluded as it only elaborates on 'work in pairs').
repeated more than a thousand times), work group (first), work partner (fifth), other students (23rd; as in 'Ask other students your questions' from Outcome B2, p. 81), another pair (32nd; as in 'Exchange papers with another pair and read the conversations' from Navigate B1, p. 84), change roles (37th) and change partner (37th). On further examination of the textbook pages, we sometimes queried the practicality and genuineness of such invites in regard to whether or not they promote real interactional authenticity and social engagement (Ellis, 2017;Philp and Duchesne, 2016) (i.e., positive encouragement among learners and active listening from others). As seen in the following examples, many of these 'pair/group work' commands ask learners to attend to grammatical issues, instead of engaging learners in meaningful exchanges. The frequencies of such phrases and statistical data may tell one story, but researchers need to actually read the actual texts to determine the focus of these tasks.
'Work with a partner. Look at the highlighted words and phrases in the article' (Cambridge Empower B1, p. 78).
'Work in pairs. Take turns to ask and answer the questions using the phrases in the box' (Cutting Edge B1, p. 10).
'Work in pairs. Read the conversation. Which two words should be stressed? Underline them, then listen and check. Practise the conversation together' (Open Mind B2, p. 13).
'Write sentences for and against these questions, using expressions from 3d. Compare your sentences with other students' (Empower B2, p. 65).

Listening and speaking
Prior studies by Masuhara, Tomlinson and colleagues Tomlinson et al., 2001) have found that ESOL textbooks prioritise listening and speaking. The current phraseological method discovered numerous listening commands (1,797) in the ESOL textbook instructions but only a few for speaking (364). We do not have a theory as to why listening is so greatly emphasised in ESOL classrooms. The UK Home Office document devotes sections (e.g., on pages 10, 11, 13, and 15) to describing the importance of reaching the 'speaking and listening' requirements for immigration applications. This finding is certainly in stark contrast to our findings regarding the Hong Kong and People's Republic of China textbooks (Chan, 2020). These EFL textbooks tend to prioritise reading and writing much more than listening and speaking. This research led us to believe that there may be fundamental differences between inner-and expanding-circle textbooks in regard to their ethos, style and pedagogical focuses. Below are some examples of listening tasks from ESOL textbooks.
'You are going to listen to a conversation between a tourist and a guide at the Jim Thompson House. First put the conversation in the right order' (Navigate B1, p. 22).
'Listen to a radio programme about the trends mentioned in exercise 2. Decide if these statements are true or false' (Navigate B2, p. 58).
'Listen to part of a seminar on identity. What does Sean think about expressing personal identity?' (Open Mind B2, p. 10).
Many of the commands calling for 'speaking' promote a narrowlydefined, practice-orientated kind of speaking. The researchers are uncertain why a greater number of speaking commands was not found, as this has also been noted in foregoing studies. Of course, if pair/group work is also considered as involving speaking, then 'speaking' acts can be greatly expanded. Below are some examples of speaking tasks.
'Listen and repeat the Free Time questions. Copy the rhythm' (English File B1,p. 4).
'Look at the words in the box and decide how the underlined letter e is pronounced. Add the words to the table, then listen and check. Practise saying the words' (Cambridge Empower B2, p. 10).

Reading and writing
What about traditional literacy skills in ESOL textbooks? Along with previous studies, which have criticised the lack of reading and writing activities in these books , the present researchers also reached a similar conclusion (see Table 2). Again, we should go deeper to take a look at how commands mediate reading and writing.
Here are a few examples: 'Read Anita's email. Why is she sending the email? Tick the correct reasons' (Cambridge Empower B1, p. 35).
'Work in pairs. Read the story. Which of the titles below would you choose for it? Which other titles are suitable and why?' (Cutting Edge B2,p. 35).
'Look at the extract again. Which tense do we use when we want to suggest a connection between these things?' (New Language Leader B2, p. 10).
The greater purpose of reading and writing should not be narrowly defined through skill-training activities. Hadfield (2014) and Timmis (2014), in a monograph on textbook writing, point out that textbook tasks should aim to engage learners in the cognitive, affective and social dimensions of operating a text. The key is to provide a motivating vision for learners, enabling them to see what reading and writing are for. The current ESOL commands, however, are neutral and remote from learners' perspectives. Nine out of 10 instructions do not frame tasks as 'events', 'situations' or 'figured worlds'. Rather than helping learners to anticipate real-life challenges, events and practices in English-dominant societies, 9 these reading and writing tasks are akin to small exercises that sometimes end with rather offbeat linguistic questions.
'Think of an event or experience that you want to write about. In your notebook, make notes about the event using the guidelines below' (Open Mind B1, p. 15).
'Write a paragraph of about 120 words. Describe your idea and explain how it would improve the world. Use at least one sentence with if þ past tense þ would' (Navigate B1, p. 107).
'Write an email to a friend explaining that you haven't been well, and saying what you've been doing recently' (English File B2, p. 15).
Similarly, these writing instructions predispose learners to treat texts as objects. We would prefer to see writing tasks designed in a way that is relevant to living, studying and coping in the inner circle. Note that contemporary literacy training lends importance to the cultivation of the writer's voice and identity, as well as working with real-life artefacts Rowsell, 2010, 2012). We use this analytical lens to comment on these tasks. 10

Lexical-grammatical focus
We found a large number of collocations (1,162) calling attention to 9 For example, mock tasks that help learners to tackle local government queries or emails could be situated literacy activities. 10 The fact that we can make these comments is also due to our familiarity with Hong Kong textbooks. More discussions about the framing of tasks can be found in Author (Year). the lexical-grammatical aspects of the language. Such patterns are for the first time quantified by corpus tools. They include check meaning (eight place; as in 'Check the meaning of the bold words in the headlines' from Navigate B2, p. 108), correct form (sixth), correct verb (sixth), correct word (sixth), word phase (10th), complete rule (11th), word box (16th), present simple (18th), present perfect (18th), present continuous (18th), past simple (19th), past perfect (19th), past continuous (19th), verb bracket (28th) and match verb (34th). This kind of skillbased teaching focus does not match our belief that learners in English-dominant societies need more communication skills than declarative linguistic knowledge. This finding matches observations from qualitative research. For example, Masuhara et al. (2008, p. 297) noted that '[t]he emphasis in most (ESOL) courses is on explicit teaching of declarative knowledge followed by controlled or guided practice. ' Tomlinson et al. (2001, p. 84) pointed out that 'in most of the [ESOL] courses, grammar is signalled as the main focus of most units.' This research further extends our understanding of how much grammar 'weighs' in these textbooks; for example, Table 2 shows that many textbook commands call attention to grammar.

Other collocations
A sizeable number of collocations (1,599) were classified as 'others'. Around 64% of the collocations (1,018/1,599) relate to the style of response requested by the textbook commands. These collocations include correct option (sixth place; as in 'Listen again and choose the correct option to complete the sentences' from Open Mind B1, p. 61), correct answer, (sixth), correct order (sixth; as in 'Put the words in a and b into the correct order to make questions' from Outcome B1, p. 45), check answer (eighth; as in 'Read the text and check your answers' from Cutting Edge B1, p. 34), check idea (eighth; as in 'Watch or listen to Part 2 and check your ideas' from Cambridge Empower B1, p. 23), true false (17th), choose option (20th), choose best (31st; as in 'Choose the best verb in the sentences below' from Cutting Edge B2, p. 27), and agree disagree (37th). The cognitive style promoted by these commands seems to be a straightforward, controlled practice style.
There were 286 collocations calling ESL learners to look at colourful photos and discuss them in an ensuring task; they are look photo (seventh) and look picture (seventh).
'Look at the photo. Work in pairs. Discuss the questions' (Outcome B2, p. 22).
'Work in pairs. Look at the pictures. Which of these things from the first decade of the new millennium do you recognise or remember?' (Open Mind B1, p. 9).
Overall, there are more seeing-the-photo collocations in this corpus than in our corresponding EFL corpora (Chan, 2020). We do not have a theory as to why this is so. It may be that photos make it easier to motivate speaking in an ESL classroom, in which learners mostly come from different places and have different prior experience, education and schemas. Photo discussion tasks are much less common in textbooks in Hong Kong. In any case, we hope to show through this report that corpus linguistics can usefully facilitate textbook comparisons.

An extra step: Searching for place names
A corpus of sheer textbook commands consisting of 154,681 words is not small by any standard. One way of making language learning relevant to living in the inner circle is for tasks to 'bring in the everyday world of objects and stories to create meaning' (Pahl and Rowsell, 2012, p. 50). Equally, the researchers would like to know how ESOL textbook tasks, which have been said to transmit Anglo-centric and Western values to learners (McGrath, 2013), do so by incorporating symbolic artefacts (e.g., British place names) in their instructional texts. How do they use these artefacts to mediate learning? In this extra step, the authors performed quick searches of symbolic place names and artefacts in the corpus 11 (using, e.g., 'UK', 'London', 'Birmingham', 'US' and 'Queen' as search words). This search found four instances of 'UK', 10 instances of 'London', one instance for 'US' and none for 'Birmingham' and 'Queen'. Examples are as follows: 'You are interested in doing a summer course in the UK, which combines English with sport. You see this advert for a school in Belfast, Northern Ireland. Make a list of about six things you would like to know about the course and school' (New Language Leader B1, p. 101).
'Goran is a student from Croatia who's going to study English in the UK. He's going to stay with a family. Read the email from Mrs Barnes and complete it with expressions from the list' (English File B1, p. 113).
'Look at the photo of the 2012 London Olympics closing ceremony and discuss the questions' (Cambridge Empower B2, p. 95).
'How much do you know about the London Underground? Try to answer the questions below' (Cutting Edge B1, p. 30).
The purpose of this search is to take us directly to textbook commands so that we can analyse how these tasks are set up to 'pave the way' into the world of ESL. Analysing commands in this way enables us to see how artefacts symbolic of ESL lifestyle (e.g., place names) are used to mediate learners' reading and writing Rowsell, 2010, 2012). Apart from the first instruction, which directly addresses the learner ('You are interested in … ') and nests the task in space and time (a summer course in Belfast), the other instructions tend to treat artefacts as 'things' ('let's analyse the London Underground'; 'let's analyse Mrs Barnes's email'), positioning learners as onlookers in regard to the host (ESL) society, rather than inviting them to be actual participants. These examples evoke Masuhara et al.'s (2008, p. 301) comment: 'The voices of the authors in these [textbooks] are neutral and semi-formal.'

Concluding remarks: Problems and prospects of a corpusbased textbook description
By exploring the interface between lexis and pedagogy, this study infers common ESL teaching practices by examining a corpus of instructional language. Lexical repetitions are vital for expressing alignment with community norms, ethos, and identity (Hyland, 2015); the number of times and the manner in which they occur in textbook commands are important for hermeneutic interpretations. Because current textbooks are used in ESL contexts, the issue of perspective is at the core of our evaluation: Why are we, as outsiders working and living in the expanding circle, interested in these materials? As with many who are in our region, we were curious about what inner-circle pedagogy these materials offer. This special vantage point allows the paper to take into consideration both EAL and ESOL paradigms, which is an unusual perspective, given the relatively secluded areas they each represent. Without question, after reading the foregoing analyses, the reader may query whether or not these textbooks do represent the 'ESL pedagogy' that we purport to find. We ought to restate that the current focus on ESOL textbooks is not a matter of convenience, but rather the result of a realisation of a research and notional impasse: Published EAL materials do not exist. Revisiting the notion of ESL is important at this point, as the term continues to be indispensable and widespread in applied linguistics and ELT literature. Given the current discussion, we have two suggestions for the future use of the term: either to realise that ESL (in terms of EAL) pedagogy does not quite exist in and of itself (i.e., EAL is not an official curriculum), or to revise our expectations of ESL pedagogy by 11 The WordSmith Tools did not find any such words and so we took this extra step. These searches were not searches of collocations. also considering the teaching of English to non-mainstream learners. Only in this way can a reader develop a more accurate understanding of what is meant by 'ESL learners' and 'ESL classrooms' when such terms are seen in the literature.
The current study takes a broader view of ESL pedagogy. Many of the current findings provide further evidence of the qualitative analyses conducted by McGrath (2013) and Tomlinson (2008). These prior works provide grandstand views of the textbook production industry from an emic perspective, focusing more on the cultural content of these materials and less on their pedagogies. Our data give information regarding the character of ESL activities: Many of them are skills-based activities focusing on lexical-grammatical knowledge. There is an uneven emphasis of the four skills. The emphasis on speaking and listening may reflect a basic concern in regard to helping ESL learners to better communicate in inner-circle societies. 12 This may be the first time that such an emphasis has been investigated using corpus tools. Furthermore, investigating artefacts and place names enables us to examine how pedagogic tasks create scenarios to prepare learners for language learning. Using modern literacy-practice standards to examine these books, it does not appear that these textbook activities are effective in drawing learners into the cultural matrix of the English-speaking world. Overall, the current corpus-based approach makes use of bottom-up, quantitative data; this approach facilitates cross-corporal comparisons of textbook features. For example, our data can be compared with another corpus constructed out of another set of textbooks.
There are several limitations to the current study. While the benefit of constructing a corpus is that we were able to use frequency data to explore textbook features and ethos, the finer patterns of the instructional language may not be searched, often because they occur very infrequently in the materials. The current study also did not investigate task content or task difficulty. In the current study, although our quest was to look for 'ESL pedagogy', admittedly, the data we offer is relatively constrained, as we did not look at real classroom data. Our presentation of the 40-item list represents our best effort to characterise ESL teaching practices within the space limitations; the list in fact contains 80 frequent collocations. Overall, this study demonstrates the utility of using corpus-selected collocations as the basis for subsequent interpretations, in order to avoid superficial analyses. While the current corpus only focuses on a small section of textbooks, the frequency data does reveal features that can easily escape the attention of qualitative researchers. Therefore, apart from the usual practice of examining textbooks' lexical quality, corpus data can equally be applied to detect implicit norms, routines and trends embodied in these materials.

Funding
This project was supported by the Faculty Research Grant (FRG2/ 14-15/091) from the Faculty of Social Sciences of Hong Kong Baptist University.