Collocates of ‘Great’ and ‘Good’ in the Corpus of Contemporary American English and Indonesian EFL Textbooks

Previous studies have proven some discrepancies between language presented in EFL textbooks and the real use of English. This fact contradicts the teaching materials’ aim, which is intended to equip the learners to be communicatively competent. This study then investigates the use of adjectives, limited to ‘great’ and ‘good’ in terms of the frequent and strong co-occurrences, i.e., collocations, in a general reference corpus of Standard English to be compared with those in Indonesian EFL textbooks. The data were collected from both the corpus and the textbooks, but the analyses in the textbooks were generated based on the Mutual Information (MI) score of the collocates. Based on the comparison, it is evident that there are some similarities between adjective use in the textbooks and the corpus in terms of verb collocates of the adjectives. The mismatches, however, are quite remarkable, especially in terms of the variability of adverb collocates and the absence of prominent noun collocates in the textbooks. Pedagogically speaking, these results should be taken into consideration in writing the textbooks to enhance the quality of the language content prepared for the learners in the EFL context.


INTRODUCTION
In language teaching, teaching materials play crucial roles. Teaching materials are designed to equip learners to be communicatively competent in real life (Gilmore, 2007). In other words, the language used in teaching materials must correspond to the real use of English (Gilmore, 2007;Römer, 2012;Siegel, 2014). As for inexperienced teachers, teaching materials could help provide ideas on how to plan and teach lessons (Richards, 2001). Teaching materials include textbooks, institutionally prepared materials, and teachers' materials (McGrath, 2013). Textbooks, among others, are the most prominent teaching materials since they provide learning objectives specified in the syllabus and they suggest the intensity of syllabus coverage (Richards & Rodgers, 1986). It means that textbooks supplement the learning activity, serve as the source of language information, and stimulate learners' interest (McGrath, 2013). Richards (1998) stated that the most commonly found elements in second and foreign language teaching are teachers, learners, and textbooks.
For the EFL context, this should be taken more seriously because the writers of teaching materials are most likely non-native speakers of English. As an alternative, checking the textbooks' language can be done by consulting the language content to the native speakers of English. However, this way is problematic since the personal experience of using language is limited, while language use is collective production and experience (Burkette & Kretzschmar Jr., 2018). Therefore, to solve this problem, the use of 'big data' of language, i.e., corpus, is of prominence since it offers a large size of language data from enormous sources which represent language and serves as the representation of real use of the language (Burkette & Kretzschmar Jr., 2018;Römer, 2010). Corpus can be defined as the digital storage of texts, either spoken or written, to assist language analysis (Baker, 2010;Flowerdew, 2015;McEnery & Hardie, 2012). Nowadays, there are many English textbooks, especially for ESL and EFL contexts, written based on corpus consultation (e.g. Browne et al., 2013;Carter & McCarthy, 2006;Conrad & Biber, 2009;McCarthy et al., 2014;O'Dell & McCarthy, 2017).
Concerning ELT (English Language Teaching), there have been many studies comparing the language used in textbooks/course books and language in corpora (Arellano, 2018;Biber & Reppen, 2002;Burton, 2012;Cheng & Warren, 2007;Choi & Chon, 2012;Dongkwang & Chon, 2011;Leung, 2016;Ojanen, 2008;Phoocharoensil, 2017, among others). Interestingly, those studies predominantly show the discrepancies between language in textbooks and language in corpora. Some mismatches between real use and language in textbooks based on corpus investigation cover the use of linking adverbials (Phoocharoensil, 2017), the sharp contrast between the information found in grammar books and real use of grammar (Biber & Reppen, 2002), the expressions of understanding in textbook conversations and real conversations as portrayed in a corpus (Cheng & Warren, 2007), the different collocations between those found in textbooks and in a native-speaker corpus (Choi & Chon, 2012), and the mismatches of the future tense marker usage in textbooks and corpora (Ojanen, 2008). Although there has been some work in this area, this research still has to be conducted regarding different textbooks and regions. Dongkwang and Chon (2011) and Burton (2012) emphasize a lack of studies with respect to the use of corpus-based analysis towards textbooks since many textbook writers do not make use of corpora in designing and developing English textbooks.
Considering the issue in the Indonesian context, a small number of studies were conducted on corpus in ELT. They have studied lemmatization in the textbook corpus (Astika, 2018), the implementation of corpus-aided grammar teaching (Yanto & Nugraha, 2017), the use of corpus in designing English materials for tourism (English for a Specific Purpose) (Fauzi & Suradi, 2018), and the design of learner corpus of elementary school students (Zen et al., 2019). Regarding corpus analysis for textbooks, a corpus study for textbooks in Indonesia is related to the design (considering the lemmatization in textbooks), while the research of Indonesian EFL textbooks focuses more on investigating task analysis (Ayu & Indrawati, 2018), cultural values (Mayangsari et al., 2018;Widodo, 2018), and authenticity of conversation materials in comparison with authentic interactions (Setiaji, 2016). Aside from the previous studies, little is still known about the language of Indonesian textbooks: whether the language in Indonesian EFL textbooks corresponds with the real use of English represented in a general reference corpus. Therefore, this study attempts to examine the language in the textbooks compared to the presentation in a general reference corpus of English.
Regarding language use, it is not only about using words in isolation (Jones & Waller, 2015;Szudarski, 2017), but it is also about combining words with other words, and it is not surprising to find that there is a tendency in word co-occurrence (Gablasova et al., 2017;Szudarski, 2017). A particular word tends to co-occur with a specific word in a certain context, and they are strongly associated with each other, which is known as collocation (Choi & Chon, 2012;Gablasova et al., 2017;O'Dell & McCarthy, 2017). The knowledge of collocation is acquired natively, and the mastery of this knowledge shows the naturalness of language production of the learners. This means that formulaic language, e.g., collocation, is of significance to learn, especially in the EFL context, so that learners can produce natural English as it is usually spoken or written by native speakers. However, EFL learners have various problems in the oral and written production of collocations (Molavi et al., 2014). This emphasizes that collocations should have a specific concern in the EFL context, including in Indonesia. However, due to the wide coverage of language aspects that can be considered for the analysis, this study focuses on adjectives since adjectives also hold a crucial role in building up linguistic construction by attributing qualities to nouns (Dixon & Aikhenvald, 2004).
In regard to corpus analysis, corpus use in investigating collocations related to ELT have been recognized and done in some work (Akıncı & Yıldız, 2017;Orlando, 2009;Wu, 2016). As for the collocations of adjectives, corpus use has been proven very accurate in providing the results of the analysis (Almela, 2011;Ghaniabadi et al., 2015;Takač & Lukač, 2013). However, little is known for the collocations of adjectives in EFL textbooks in Indonesia that are very salient for English mastery in the EFL context. Therefore, this study investigates the collocates of adjectives in a general reference corpus of English to be compared with those in Indonesian EFL textbooks. This study, however, delimits to spoken English since it is believed that spoken and written language has slightly different characteristics due to different mediums that lead to various linguistic features as well as lexical choices (Biber & Quirk, 2012). Textbook conversations seem to have some discrepancies with authentic conversations (Oktavianti et al., 2020;Setiaji, 2016;Siegel, 2014). Since adjectives are lexical categories and have a huge number of members, this study limited the analysis to 'great' and 'good' since they are frequent adjectives in the Corpus of I. N. Oktavianti & J. Sarage,Collocates of 'great' and 'good' in corpus of contemporary American English and Indonesian EFL textbooks | 460 Contemporary American English (COCA) (in the 4 th and the 1 st rank), and they belong to the same type of adjective, 'value' adjective (Dixon, 2005). In particular, this study attempts to answer these questions: 1. What are the collocates of 'great' and 'good' in a general reference corpus as the representative of the real use of English? 2. How are 'great' and 'good' presented in Indonesian EFL textbook conversations for senior high school compared to the use in the corpus?

Corpus Linguistics
In ELT research, the corpus was introduced by the founding of Collins Birmingham University International Language Database (COBUILD) by Sinclair, which later developed the Bank of English, one of English corpora (Sinclair, 2004). The building of the corpus was aimed to assist the writing of dictionaries and grammar books that are prominent in language teaching. Specifically, a corpus can be implemented directly in the classroom to the context of discovery learning in which the learners inductively analyze language data to generate patterns or known as Data-Driven Learning (DDL) (Leńko-Szymańska & Boulton, 2015;Lessard-Clouston & Chang, 2014). Besides, a corpus can also be used indirectly for teaching English, which is in the writing of teaching materials. To this extent, a corpus can be referred as a source for consultation (McEnery & Xiao, 2010;Timmis, 2015). A corpus as the source of natural use of the target language can be consulted regarding the content of textbooks (McCarthy, 2004;O'Keeffe et al., 2007;Timmis, 2015). McCarthy (2004) states that corpus-informed materials are special in terms of the quality of the language content. Corpus-informed materials differ from intuitionbased materials because corpus-informed materials are based on the authentic use of English. The materials are not invented, and the contexts are authentic since the data are collected from empirical usage of language produced or uttered by native speakers of English (e.g., newspapers, magazines, talk shows, fiction, academic texts, and so on). However, corpus investigation results should be adjusted and modified to the teaching and learning needs. It is also argued by Römer (2011Römer ( , 2012, pointing out that corpus can contribute to the pedagogical aspects in terms of its authenticity. Apart from the debate of the term 'authenticity' (Widdowson, 1998), Gilmore (2007) emphasizes that what matters in teaching materials is their ability to equip learners to be communicatively competent. On top of everything, the main idea of being authentic is that the text is not specifically designed for teaching purposes (Timmis, 2015).

Collocations
The term 'collocation' refers to the umbrella term 'formulaic language' that deals with multi-word combinations (Szudarski, 2017;Wood, 2015). Specifically, collocation is the tendency for words to co-occur frequently (Saeed, 2016;Sinclair, 1991). Early work investigating collocations was initiated by Firth in which he classifies collocations into several possible types, such as 'habitual collocations' and 'idiosyncratic collocations' (Wood, 2015). However, according to Stubbs (2002), the combination is not a fixed expression, but it is a habitual co-occurrence. Halliday (1966) was another researcher who studied word relation, and he defines collocations as the syntagmatic relation between words whose co-occurrences are quite regular to form a pattern. Collocation use and other formulaic use of language are vital in daily communication since it reflects the natural production of the language. Therefore, the teaching and learning of collocation are also necessary for the mastery of English.

Collocations of 'Great' and 'Good'
Due to the definition of collocation, the collocates of 'great' and 'good' are the linguistic units (i.e., words) that tend to co-occur with these adjectives. Considering the nature of adjectives, therefore there are several possible patterns of collocations. In linguistic construction, an adjective can occupy a position in phrasal level and or clause/sentence level (Greenbaum & Nelson, 2002). At the phrasal level, an adjective can serve as the pre-modifier of a noun phrase (van Gelderen, 2017). As a head of a phrase, an adjective can also be modified by another word, including an adverb (van Gelderen, 2017). At the clause/sentence level, an adjective can occupy the syntactic function of subject complement following a linking verb (Greenbaum & Nelson, 2002). Based on these functions and distributions, therefore there are some collocation patterns of adjectives, such as adjective + noun, adverb + adjective, and verb + adjective.
The following examples of adjective collocations in Table 1 are taken from COCA.

Previous Studies on Collocations in ELT
Some previous studies have been conducted related to collocations in language teaching, including the teaching of collocations and the analysis of collocations in teaching materials. The previous work on teaching collocations has proven the use of corpus to teach collocations effectively. Akıncı and Yıldız (2017) examined the effectiveness of using corpus consultation in teaching collocations of verbs + nouns. They figured out that the corpus is more effective in terms of teaching collocations for ELT students. Similarly, Wu (2016) investigated the effectiveness of using a corpus in correcting the errors related to the use of collocations. Not limited to corpus use, the teaching of collocation has been studied by comparing the effectiveness of using two textbooks, another study done by Roohani (2011). Roohani (2011) investigated how lexical and grammatical collocations used in high school and pre-university English textbooks are used in public schools in Iran compared to the New Interchange series used in private language institutes. Based on the score of the t-test, learners using the New Interchange series performed better on both lexical and grammatical collocation tests. It is possible due to the better presentation of collocational patterns in the New Interchange series.
In terms of teaching materials, there has been some previous research analyzing collocations in textbooks. Some of the previous studies were done without referencing any corpora or using corpus analysis. For example, Shahrokhi and Moradmand (2014) examined the use of grammatical and lexical collocations used in Iranian high school English textbooks compared to the American English File books. This study proves that there are major differences in the use of collocations in Iranian English textbooks and American English File books, especially in terms of lexical collocations. In the Indonesian context, Hutapea et al. (2019) analyzed the types of collocations used in an English textbook for grade 10 in Indonesia. The results show that the most frequent collocation type is adjective + noun collocation, and collocations can be found frequently in narrative texts (38%), followed by recount texts (35%) and descriptive texts (27%) of the textbook.
Collocations in textbooks have also been investigated using the corpus approach (e.g. Alfiandita & Ardi, 2020;Choi & Chon, 2012;Kouzouloglou, 2015;Molavi et al., 2014;Tsai, 2015). For example, Orlando (2009) analyzed the collocations of modal verbs in textbooks in comparison to Standard English corpora. The finding of this study indicates that the frequency of modal verbs and frequency of modal verb patterns in the textbooks are different from those in the British National Corpus (BNC). Similar to Orlando (2009), Choi and Chon (2012) compared the collocations in high school 1, a 10 th grade EFL textbook with 2k Graded Reader native-speaker corpus produced, and the comparison results in noticeable differences. This study also found out that the collocations related to learner's life and interest are relatively infrequent in the textbook. Similarly, Kouzouloglou (2015) observed a small number of collocations in textbooks. Kouzouloglou (2015) examined the collocation frequency in EFL textbooks used in Greek state schools. The study compiled a corpus of EFL textbooks used in Greek state schools and processed it in the AntConc corpus tool. The results of the study show that there is some deficiency and unsystematic use of collocation in the textbooks that should be considered by EFL practitioners in Greek.
Another textbook and corpus comparison were done by Molavi et al. (2014). Molavi et al. (2014) examined the distribution of lexical collocations in three textbooks and compared them with Open American National Corpus (OANC). This study found out that there are differences in the patterns of the collocations in the textbook series and the corpus. The textbook series pay less attention to patterns, such as n1 + of + n2, verb + adverb, noun + adverb, and adverb + adjective, different from the corpus. It indicates that the choice of collocations in the textbooks differ from that in the corpus as the representative of the real use of English.
Similarly, Tsai (2015) investigated the collocational profiles (verb + noun) of English textbooks published in Taiwan, EFL students' writing, and native students' writing. The study compared the compilation of the three texts to the BNC. The findings show that the collocational density of the textbooks is comparable to those of native speakers' writing. Compared to the native speakers' writing, EFL learners' writing has limited collocational density and is underdeveloped.
In the Indonesian context, Alfiandita and Ardi (2020) investigated the types and variations of collocations used in students' worksheets (or Lembar Kerja Siswa) for grades 10, 11, and 12 of senior high school students. The study identified that the worksheets contain both lexical and grammatical collocations, but the grammatical collocations are the most frequent ones. However, this study emphasizes that the use of the collocations is not various, which should be taken into consideration by the teachers and the material writers.
The presentation of adjectives in textbooks still shows some discrepancies, such as those explained by Biber and Reppen (2002). They mentioned that adjectives presented in grammar books are not in accordance with the real use of adjectives based on the investigation using Longman Corpus of Spoken and Written English (LCSWE). Regarding the use of corpus in investigating the collocation of adjectives, there are studies by Almela (2011), Antle (2013), Takač and Lukač (2013), and Ghaniabadi et al. (2015). Almela (2011) explored the verb-adjective ('cause' + adjective) collocation patterns in Oxford Collocations Dictionary. This study found out that certain collocation of 'cause' (e.g., 'cause faulty', 'cause defective') cannot be found in the dictionary, although semantically the adjectives match the semantic prosody of the verb 'cause'. Similarly, Takač and Lukač (2013) investigated the use of adjective + noun collocations in a learner corpus in Croatia, Croatian Corpus of English Learner Essays (CELE), in comparison to another learner corpus, International Corpus of Learner English (ICLE), and a native speaker corpus, BNC. The results show that the use of collocations remains relatively similar to non-native corpora, which is different from that in the BNC.
Having a slightly different result, Ghaniabadi et al. (2015) investigated the use of adjective + noun collocations produced by Iranian EFL learners. This study examined 28 students' writing at the university level. The repetitions of highfrequency patterns were contrasted to their frequency in COCA. The results of the study show that Iranian EFL learners more frequently used the adjective + abstract noun collocations. In comparison to COCA, surprisingly, there was no significant difference in the use of adjective + abstract noun collocations between Iranian EFL learners and the corpus. Finally, another study in adjective collocations also deals with the compilation of the collocations, such as a study carried out by Antle (2013). Antle (2013) compiled a collocation list of adjective + noun for intermediate English learners by following several criteria: the nouns and adjectives must be very frequent as individual words in the BNC, and the collocations must occur at least 50 times in the corpus.
None of the studies above examined the collocations of adjectives in textbooks in the Indonesian context. Therefore, this study is important to carry out because it can provide the linguistic perspective on how EFL textbooks in Indonesia should be designed and developed. This study can contribute to the presentation of collocational knowledge in textbooks. The textbook authors can consider improving the language content based on the results of this study to enhance the quality of teaching materials in Indonesia.

METHODS
Corpus linguistics is acknowledged as a methodology rather than a branch of linguistics since it deals with the approach to analyze language (McEnery & Hardie, 2012;Stefanowitsch, 2020). This research is a corpus study since it meets two criteria of a corpus-based study: (1) it compiles a corpus (i.e. a conversation corpus of Indonesian EFL textbooks for senior high school) and uses an existing corpus, and (2) it focuses on one of key analyses in a corpus study (i.e. collocation analysis) (Brezina, 2018;Szudarski, 2017). A collocation analysis can only be done in the large size of linguistic data that a corpus can offer (Stefanowitsch, 2020;Szudarski, 2017).
The data sources are a general reference corpus of English, COCA, and three Indonesian EFL textbooks for senior high school (Bahasa Inggris X, Bahasa Inggris XI, Bahasa Inggris XII). COCA was selected due to its always-updated language data (2019 and ongoing), its huge size (1 billion words), and a wide range of period (from the 1990s) and registers or genres of texts (spoken and written) (Davies, 2008). In addition, COCA was proven to improve the collocational knowledge of English learners (Mansour, 2017). The textbooks were selected since they are curriculumbased textbooks and published by the Indonesian Ministry of Education and Culture.
The data were collected by doing keyword searches (the data collected from COCA) and document observation (the data collected from the textbooks). Since this study prioritized conversation, the corpus search was set to the spoken sub-corpus of COCA and the textbooks' conversation sections. By doing so, the analysis between the corpus and the textbooks was comparable. To collect the data from the corpus, the keyword search was done in COCA by selecting the 'collocates' feature and entering the keyword ('great' or 'good'), and setting the Part of Speech (PoS) into adj.ALL. The collocates were adjusted to the nouns (as the head of a noun phrase being modified by the adjective-right collocates), verbs (as the linking verbs for the subject complement-left collocates), and adverbs (as the pre-modifier of the adjective in the adjective phrase-left collocates). The collocation window (span of words) was set up to be as wide as possible within 1-4. Figure 1 and Figure 2 show the process. As for the document observation in the textbooks, the adjectives used in the conversations were identified along with the co-occurrences of nouns (in noun phrases), verbs (linking verbs), and adverbs (pre-modifiers in adjective phrases).
The data found in the corpus were then analyzed by using an association measurement called Mutual Information (MI) score (Brezina, 2018). The MI score is used to calculate the strength of association between words (Brezina, 2018;Szudarski, 2017). In determining collocation, a raw or absolute frequency is unreliable, so this study used the MI score to measure collocations because it focused on the strength of word associations and need not large size of corpus (unlike t-score which requires large data) (Hunston, 2002). Furthermore, using the MI score is comparable across corpora. Since it concerns low-frequency items that are restricted to specific texts/genres/registers (Gablasova et al., 2017), it is appropriate to use in the one-genre analysis (e.g. spoken/conversation). Below is the calculation of MI applied in COCA.

MI = log ((AB * sizeCorpus) / (A * B * span)) / log (2)
Details: A = frequency of node word B = frequency of collocate AB = frequency of collocate near the node word (e.g., 'color' near 'purple') size  (2) is literally the log10 of the number 2 The collocation with MI-score 3 or above 3 shows that the co-occurrences of the two words are strongly associated (Szudarski, 2017). Therefore, this study uses this parameter to determine the collocates of adjectives. However, it should be noted that the highest MI-score does not always mean the strongest collocate (Gablasova et al., 2017).
After the collocates in COCA based on the MI-score were identified, the results were then checked with the use of adjectives in the textbooks: whether the cooccurrences of nouns, verbs, adverbs with adjectives in the textbooks correspond to the results of collocation analysis in the corpus. The comparison of collocations in the textbooks and COCA was interpreted and described. The results of the comparison were then connected to the results of the previous studies.

RESULTS
This section presents the analysis results regarding the left collocates (nouns, verbs, adverbs) of adjectives in English conversation. To begin with, Table 1 shows the verb collocates of 'great' in COCA.  Table 2 shows that the verb 'look' (morphosyntactic variant 'looks') is the most frequent co-occurrence of 'great', and it also has an MI score above 3 (5.04) which means they are strongly associated. Based on Table 2, there are some other perception verbs and their inflected forms. This is in accordance with the nature of the adjective 'great' that semantically shows the quality of things and syntactically cannot be assigned with semantic roles. In other words, the verbs are not action verbs, but more perception verbs that only have one argument (assigned to the subject). However, along with perception verbs, 'deserves' and 'skated' are included in the list which may not be directly associated with 'great' but with the noun phrase with 'great' as the premodifier of the noun phrases. This is proven in Figure 3. Figure 3 shows that the verb 'deserves' co-occurs with the adjective 'great' because the adjective is the pre-modifier of the noun phrase following the verb (e.g., 'great credit', 'great respect'). As for the noun collocates as the right collocates of 'great', Table 3 shows ten collocates taken from COCA. Based on Table 3, most of the noun collocates are abstract nouns (e.g., 'deal', 'depression', 'pleasure', 'detail', 'humor', 'length', 'difficulty', 'pride') and the other collocates are concrete nouns ('lake', 'grandfather'). This is in regard to the semantic nature of 'great' which shows the value for non-physical things. Meanwhile, 'great lake' and 'great grandfather' have different meanings; one is related to dimension, and the other is related to family relationship indicating one degree upward or downward.
As for the other lexical category, there are some adverb collocates found in COCA. Table 4 presents the top ten adverb collocates of the adjective 'great'.  Based on Table 4, it is evident these adverbs have quite low co-occurrences with 'great'. However, in regard to the MI score, they are strongly associated with 'great'. The most frequent with the MI score above 3 is 'insanely', followed by 'recognizably ', 'undoubtlessly',' expectedly', etc. After analyzing the adjective 'great', there is the adjective 'good' as the most frequent adjective in COCA, and it belongs to the same type of adjective as 'great'. As an adjective, 'good' can also occupy the syntactic function as a subject complement in which it is preceded with a (linking) verb. Table 5 displays the verb collocates of 'good' as identified in COCA. Based on Table 5, it is evident that all verb collocates in the table are perception verbs (syntactically, linking verbs). Those verbs are 'feel', 'look', 'sound', 'smell', 'taste', and the inflected forms ('feels', 'feeling', 'smells', 'tastes', 'sounded'). This is in accordance with the semantic nature of 'good' that shows the quality of things as well as the syntactic nature of adjectives following verbs (subject complement) that cannot be assigned with semantic roles. In other words, the verbs are not actions, but more perception verbs manifested in the forms of linking verbs.
In terms of noun collocates, Table 6 presents the list of ten noun collocates (as heads noun phrases) of the adjective 'good'.  Table 6 shows that the first five collocates are time greetings (i.e., 'good morning', 'good night', 'good evening', 'good luck', and 'good afternoon'). It is plausible since the context is spoken/conversation in which greetings are frequently uttered. The remainders of the noun collocate in the list are all abstract ('shape', 'intention', 'fortune', 'grade', 'housekeeping') as well as the first five noun collocates. It indicates that 'good' is mainly used in companion with abstract nouns to attribute the nouns' quality.  Table 7 shows that 'very' is the most frequent co-occurrence of 'good', and it also has an MI score above 3, indicating a strong association between the two words. This is plausible since 'very' is a general intensifier used among different text types/genres/registers. Thus, it is also commonly used in companion with the adjective 'good'.
After identifying the collocates of 'great' and 'good' in COCA, it is important to confirm the use of adjectives in the textbooks: is it in accordance with the representation of adjective use in COCA as the closest representation of 'real' use of English? Table 8 presents the co-occurrences of 'great' and 'good' in the textbooks' conversations.
Based on Table 8, it is evident that the use of adjectives 'great' and 'good' in the textbooks is relatively low, especially 'great' that can only be found in BI X. Due to its limited use, the adjective 'great' does not have adverb collocates in the textbooks' conversations. On the contrary, although it is negligible, there are noun, verb, and adverb collocates of 'good' in the conversations of the textbooks.
In terms of the similarity between COCA and the textbooks, there are some aspects needed to be underlined. For the verb collocates of 'great', the verbs cooccurring with 'great' in the textbooks are identical to those in the corpus in which most of them are perception verbs. As for the noun collocates, although the manifestation is different and limited, the nouns that co-occurred with 'great' in the textbooks are also abstract ('idea', 'job'). Unfortunately, no adverb co-occurred with 'great' in the textbooks, so it is impossible to compare this aspect. It shows that there is an invariant use of adjectives in the textbooks. Although 'great' is very frequent in spoken English, the absence of 'great' in BI XI and BI XII also affects the incompatible comparison. Unlike 'great', the adjective 'good' can be found in all the textbooks under study. Besides, there are also verbs, nouns, and adverbs co-occurring with 'good' in the textbooks. There is only one verb that co-occurred with 'good', that is 'sound' which is also found in the corpus.
As with nouns, most of the nouns that co-occurred with 'good' in the textbooks are abstract nouns, similar to those in COCA. However, in the corpus, there are nouns related to daily greetings (e.g., 'good morning', 'good night', 'good evening', 'good afternoon') and wishes (e.g., 'good luck'), so those nouns are prominent noun collocates of 'good'. Surprisingly, there is only 'luck' (in the construction 'good luck') found in the textbooks, which shows that daily greetings are rarely used in textbook conversations. This point alone is unacceptable for spoken context since, to start the conversation, people predominantly use greetings, especially time greetings.
Another significant problem is the invariant use of adverb collocates of the adjectives. In the textbooks, there is only one adverb found, i.e., 'very'. Although 'very' is the most frequent adverb collocate in the corpus and has an MI score above 3, introducing learners with one adverb variant is insufficient, which contradicts the aim of teaching materials. Therefore, it is evident that the mismatches' major problems are related to the monotonous use and the absence of prominent collocates which might result in limited vocabulary mastery of the learners.

DISCUSSION
Given the findings of the study, it is evident that there are some mismatches between the collocates of adjectives in textbooks and COCA. Despite the small number of similarities, the discrepancies are related to limited variations of adverb collocates and the absence of prominent noun collocates. These results are significantly observed in the textbooks which are ironic because textbooks are salient elements in teaching English, especially in the EFL context. However, those discrepancies found in this study are in line with those found in previous works, despite the differences of the objects. Similar to the results of this study, Ojanen (2008) points out that the use of future time constructions in textbooks differs from those in the corpus. The contracted form ''ll' is as frequent as the full form 'will' in the spoken corpora. However, textbook conversations avoid the use of this contracted form.
Similarly, Phoocharoensil (2017) mentions that the linking adverbials surveyed in the textbooks do not represent all the common patterns that native speakers of English naturally produce in the academic contexts as represented in the academic writing corpus. For instance, the use of 'hence' is very limited to one textbook among four textbooks studied, while it is the third most frequent linking adverbial in the corpus. The identical results mentioning the mismatches of linguistic constructions used in the textbooks and in the corpora to represent natural use lead to the questionable textbook designs. As for the presentation of adjectives in textbooks, Biber and Reppen (2002) find out that adjectival nouns (nouns used to pre-modify nouns in noun phrases) are less acknowledged in the textbooks, while this type of adjectival unit is quite frequent in LCSWE. Given the similar result, this study set out the fundamental basis of language discrepancies in ELT materials.
Specifically related to conversations in textbooks, Siegel (2014) states that there are some differences in the topics of textbook conversations and real-life conversations among students. In real life, students mostly talk about school life that is not quite frequent in textbook conversations. The mismatches in conversations were also studied by Setiaji (2016), and this study found relatively similar results to that in Siegel (2014). In other words, the conversations in textbooks are not in line with authentic interactions. Before these studies, Cheng and Warren (2007) have identified the differences between the mutual understanding expressions in spoken English and those in textbooks. Having identical results, Oktavianti et al. (2020) claim that conversational features in textbooks do not correspond to those in natural conversations such as the use of discourse markers, response forms, hesitation devices, and interjections. Those features are monotonous in textbooks and some of them did not match the natural use. The results of the previous studies support the claim being argued in this study that conversations in textbooks do not correspond to real conversations.
As for the collocations, this study shows some differences in terms of collocational use of adjectives in the textbooks and COCA. This result, however, is not surprising since this is relatively similar to the previous works. Choi and Chon (2012) formerly identify that the collocations in the tenth-grade textbooks are different from those in the 2k graded reader native speaker corpus. Regarding the collocations of adjectives, this study also confirms the results of the study done by Takač and Lukač (2013), stating that there are mismatches of adjective collocations in EFL textbooks and corpora. However, this study declines the results of the study of Ghaniabadi et al. (2015) because they have contrasting results.
Considering the variants of the collocations, this study also finds notable differences. The variants of collocates are not various and limited in the textbooks which correspond to Tsai's (2015) study. Thus, the naturalness of collocation presentations in textbooks is questionable and should be revisited. Closely related to collocations, Orlando (2009) investigates the modal verbs in a textbook compared to three Standard English corpora, which results in the considerable differences of modal verb patterns (co-occurrences). As this study suggests the use of corpus, Wu (2010) and Akıncı and Yıldız (2017) have proven that the use of corpus in teaching collocation is effective.
The results of this study, as well as the previous ones, emphasize the fact that there are some considerable problems with language contents in the textbooks. The language used in the textbooks does not correspond to the real use of English, which might result in learners' difficulties in using the language in a real-life setting (Gilmore, 2004;Siegel, 2014). These discrepancies occur in many EFL textbooks because the textbooks were written without consulting corpus. Dongkwang and Chon (2011) and Burton (2012) state that, despite the increasing recognition of corpus in teaching materials, there are still many textbook writers that do not consider corpus to consult the language aspects. Three textbooks examined in this study were all written by non-native speakers of English and do not consult the language content to the corpus.
Some plausible reasons might influence the fact that many textbooks do not refer to corpus. Burton (2012) explains two different points of view regarding this issue, namely the publishers and textbook authors. In terms of the publisher, they tend to follow the norms and minimize revolutionary aspects for the marketing point (Burton, 2012). As for the textbook authors, although some of them are familiar with corpus and have made use of it, some still consider corpus new and difficult to use. It is due to the lack of sufficient knowledge and training on how to use corpus (Burton, 2012). In other words, aside from the benefits of corpus use, it is still less recognized in language teaching. Some conditions might influence this state such as the use of technology, unfamiliar terms, and access (Kim, 2019;Kızıl & Savran, 2018;Lai, 2015;Lin, 2016).
Thus, this study pedagogically implies the revisiting of language content in regard to the adjective use in textbooks and the introduction of corpus use in language teaching. This study also suggests textbook writers consider corpora as the native sources to be consulted in terms of language quality. By doing so, it is expected that the quality of the language can naturally represent the use of English in real communication (Gilmore, 2007), and thus the aim of the teaching materials is fulfilled. Suffice to say, to achieve high-quality teaching materials, one of the criteria to fulfill is that they should be corpus informed.

CONCLUSION
This study presents the analysis of collocates of adjectives in COCA and three Indonesian EFL textbooks, which results in some mismatches between them. Some collocates in the corpus are not well presented, or even absent, in the textbooks since adjective co-occurrences in textbooks tend to be monotonous. Corpus as a native representation of language use should be familiarized in language teaching. ELT practitioners can use it in any aspect of teaching and learning English, either directly or indirectly. Regarding future studies, it is expected that more textbooks and corpora will be involved to obtain more robust and reliable results for the improvement of Indonesian EFL textbooks.
However, this study only analyzed three textbooks published by the Indonesian Ministry of Education and Culture, while there are also some other non-government English textbooks used in Indonesian schools. Moreover, the corpus used as a comparison here is only one English corpus, apart from its huge size and updated data, but it primarily compiles data from American contexts. Therefore, future studies should include more textbooks as well as more corpora. As for corpus, it is better to reflect the use of language in textbooks to a wide variety of corpora, such as to a corpus of another popular regional dialect of English (British National Corpus), a corpus of global English (Global Web-based English), a corpus of news texts (NOW Corpus) to get a more comprehensive comparison on collocations. It is also necessary to consult learner corpus, such as International Corpus of Learner English (ICLE), to find a clear illustration of how collocations are used among English learners. In terms of the research object, this study merely investigates two frequent adjectives which can be continued by analyzing more adjectives in the future.