不同母語之英語學習者動名搭配詞錯誤分析

全文

(1)國立臺灣師範大學英語學系碩. 士. 論. 文. Master’s Thesis Department of English National Taiwan Normal University. 不同母語之英語學習者動名搭配詞錯誤分析 Verb Noun Collocation Errors Made by 11 Groups of EFL Learners. 指導教授：陳浩然博士 Advisor: Hao-Jan Chen, Ph.D. 研究生：陳玟均 Wen-Chun Chen. 中華民國 105 年 6 月 June, 2016.

(2) 中文摘要在第二語言學習之搭配詞研究中，動詞與名詞的搭配是十分重要的。先前研究主要是探討英語學習者在動名搭配詞的過度使用、過少使用與錯誤使用。在錯誤使用的研究中，研究者大部分以人工檢查，或者是結合語料搜索工具和人工檢查的方式，在學習者的英語作文中找尋錯誤搭配詞，這些方式需要相當大量的人力與時間。此外，這些研究大部分探討一到兩個非英語系國家之搭配詞錯誤，較少研究討論多國學習者的動名搭配詞錯誤，此現象可能和現有語料搜索工具的限制有關。因此，本研究藉由 Sketch Engine 中半自動化搜索搭配詞錯誤的功能，以此相對節省時間與人力之方式，來探討十一種不同母語之英語學習者的搭配詞錯誤，並且列出普遍型錯誤（六個以上不同母語之英語學習者共同犯的錯誤）和特定型錯誤（六種母語以下或特定幾種母語人士才會犯的錯誤）。本研究討論的學習者語料為 TOEFL 11，此語料含有十一種母語之英語學習者的托福作文，這些英語學習者的母語有阿拉伯文、中文、法文、德文、印地語、義大利語、日文、韓文、西班牙文、泰盧固語（印度東部德拉維人語言）和土耳其文等十一種語言。本研究回答以下三個研究問題：(1)對於這十一種不同母語之英語學習者，最頻繁的動名搭配詞錯誤有哪些？ (2) 這些動名搭配詞錯誤可以分為哪些類別？(3) 在這些搭配詞錯誤當中，哪些錯誤是普遍型錯誤，而哪些錯誤是特定型錯誤？針對第一個問題，本研究找出了 109 類之錯誤，共發生了 807 次，在這些錯誤中，較常發生的錯誤包括 agree, travel, specialize, loose 和 do。agree、travel 和 specialize 的錯誤和省略後面介系詞有關，而學習者也常搞混 loose 和 lose，此現象可能與此兩字字形和字義有某些程. ii.

(3) 度的相似有關。另外 do 屬於虛義動詞（delexical verb），學習者傾向於在虛義動詞後接任何一個名詞，因為虛義動詞較無明確意義，因而造成學習者之搭配詞誤用。而關於第二個研究問題，本研究的搭配詞錯誤可以分為四類，分別是錯誤動詞搭配、介系詞錯誤、錯誤名詞搭配和其他錯誤等等。最後，本研究發現 61 種普遍性錯誤分別是錯誤動詞搭配(*do, *loose, *experiment, *try and *reach), 介系詞錯誤(*agree, *travel, *specialize, *listen, *search, *go, *participate), 錯誤名詞搭配 (*sale) 和其他錯誤。這些錯誤顯示出學習者對於虛義動詞、字義和字型相似之動詞、使用特定動詞時省略介系詞等共同問題。另一方面，此研究也發現 48 種特定型錯誤，這些錯誤以他們的語言系列(language family)做分類，有十五類之錯誤屬於一個語系，二十六類錯誤為兩個語系所共享，七類錯誤為三個語系所共有，印歐語系（Indo-European）傾向產生錯誤動詞搭配，而阿爾泰語語系(Altaic)傾向犯較多介系詞錯誤。另外，韓國學習者和日本學習者常常犯相同之錯誤，此現象可能與他們皆屬於同一語系有關。. 關鍵字：動名搭配詞; 半自動化; 跨語言錯誤; Sketch Engine; Sketch-Diff.. iii.

(4) ABSTRACT The dimension of verb-noun collocations has been important in studies of EFL learners’ collocation use. Much research has investigated the overuse, underuse or misuse of verb-noun collocations in different EFL learner groups. In terms of misuse of verb-noun collocations, these studies usually examined collocation errors in learner production with manual inspection or the combination of corpus tool and manual inspection, which needs a lot of time and efforts. In addition, they mostly observed errors made by one or two nonnative groups. Relatively few studies discussed the cross-linguistic use of collocation errors made by learners of several L1s. The possible reason might be due to the limited function of corpus tools. Therefore, the present study, introducing a semi-automated method of a query system, Sketch Engine, aims to explore the cross-linguistic use of collocation errors in eleven EFL learner groups in a more time-saving and less laborious way. The universal errors, defined as errors shared by more than six L1 groups, and specific errors, shared by certain L1 groups, will be revealed. The learner corpus examined in the study is TOEFL 11, which includes essays written by learners of eleven mother tongues, containing Arabic, Chinese, French, German, Hindi, Italian, Japanese, Korean, Spanish, Telugu, and Turkish. Three research questions will be answered in the present study. These questions are (1) to find out the frequent verb noun collocation errors made by eleven L1 groups, (2) to explore what the error types of these miscollocations are and (3) to discover universal collocation errors shared by more than six L1 groups and specific errors made by certain L1 groups. First, 109 types of miscollocations were found in the present research, with with 807 occurrences in total. Among these miscollocations, the frequent miscollocations containing the deviant use of agree, travel, specialize, loose and do, were shown in the finding. The incorrect iv.

(5) use of agree, travel and specialize is related to missing prepositions. The misuse of the verb loose is attributed to learners’ confusion of lose with loose because they have similar forms and a certain degree of shared meanings. Finally, the misuse of the do, which is a delexical verb, might be due to the lack of specific meanings of delexical verbs. As for the second research question, four groups of error types were found, including deviant verb usages, deviant prepositions of verbs, deviant noun usages and other types. Both deviant verb usages and deviant prepositions contain 46 error types, followed by deviant noun usages, comprising 13 types and the others containing 4 types. Finally, there are 61 types of universal errors shared by over six L1 groups in TOEFL 11, including deviant verb usages (*do, *loose, *experiment, *try and *reach), deviant prepositions of verbs (*agree, *travel, *specialize, *listen, *search, *go and *participate), deviant noun usages (*sale) and others. These results showed the learners’ problems with delexical verbs, verbs sharing similar meanings and forms, and lack of prepositions when using certain verbs. On the other hand, 48 specific errors shared by less than six L1 groups were found. These errors were categorized by language families with 15 types made by one language family, 26 by two language families and 7 by more than three language families. Indo-European language family tends to make more errors of deviant verbs while Altaic makes more errors of missing prepositions. Besides, Korean and Japanese learners tend to make the same mistakes quite often in TOEFL 11, attributed to their similar L1 background.. Keywords: Verb-noun collocations; semi-automated; cross-linguistic errors; Sketch Engine, Sketch-Diff.. v.

(6) ACKNOWLEDGMENT I would like to thank Prof. Howard genuinely for inspiring me to explore this research topic and giving me many suggestions to complete the present research. With the professor’s encouragement and advices, I started to explore the features of The Sketch Engine and fortunately discovered the new method of utilizing the platform, leading to the completion of the study. Without the professor’s guidance, I might not have found such an interesting topic to further investigate. My sincere thankfulness also goes to the committee members, Prof. Tseng and Prof. Chang whose valuable advices make the thesis more complete. I thank professors for reminding me of some neglected details. On the way to the completion of my thesis, there are a lot of people that support me. Thank you, my partner, Willie, for listening to my complaints all the time and supporting me when I struggle with both thesis and teacher examination. I also express my deep thankfulness to my friends, Terri, Ignace, Christine and Boa for chatting with me when I feel impatient and frustrated. Every time after chatting with you, I feel more energetic to go on the journey. Finally, I would like to show my deepest gratefulness to my family, who financially and spiritually supports me in my graduate school study and the teacher examination. Thanks to people who love me and whom I love, this thesis was accomplished on time and without a lot of difficulties.. vi.

(7) TABLE OF CONTENTS 中文摘要................................................................................................................................... ii ABSTRACT .............................................................................................................................iv ACKNOWLEDGMENT ........................................................................................................vi TABLE OF CONTENTS ..................................................................................................... vii LIST OF TABLES ..................................................................................................................ix LIST OF FIGURES ................................................................................................................. x CHAPTER 1 INTRODUCTION ............................................................................................ 1 1.1 Background and Motivation .......................................................................................... 1 1.2 Purposes of the Study ..................................................................................................... 6 1.3 Research Questions of the Study ................................................................................... 8 1.4 Significance of the Study ................................................................................................ 9 1.5 Definitions of Terms ..................................................................................................... 10 CHAPTER 2 REVIEW OF LITERATURE........................................................................ 11 2.1 Collocations ................................................................................................................... 11 2.1.1 Definitions and V-N types of collocations ............................................................... 11 2.1.2 Theories of Collocations .......................................................................................... 15 2.2 Difficulties Faced by EFL learners ............................................................................. 16 2.2.1 Methods to Explore Collocation Knowledge ........................................................... 17 2.2.2 Error Types and Causes of Miscollocations ............................................................. 25 2.3 Cross-linguistic Use of Collocations ............................................................................ 31 Summary of Chapter Two and Research Gaps ............................................................... 32 CHAPTER 3 METHODOLOGY ......................................................................................... 35 3.1 Corpora.......................................................................................................................... 35 3.1.1 BNC.......................................................................................................................... 35 3.1.2 COCA ....................................................................................................................... 36 3.1.3 TOEFL 11 ................................................................................................................ 36 3.2 The Query System––The Sketch Engine ..................................................................... 39 3.2.1 Corpus Creating Function ........................................................................................ 41 vii.

(8) 3.2.2 Sketch Diff Function ................................................................................................ 41 3.2.3 Source Tracing Function .......................................................................................... 44 3.3 Data Collection .............................................................................................................. 45 3.4 Data Analysis................................................................................................................. 46 CHAPTER 4 RESULTS AND DISCUSSION ..................................................................... 52 4.1 Frequency ...................................................................................................................... 52 4.2 Error Types ................................................................................................................... 58 4.3 Universal Errors ........................................................................................................... 59 4.3.1 Deviant verb usages ................................................................................................. 60 4.3.2 Deviant prepositions of verbs ................................................................................... 66 4.3.3 Deviant noun usages and others ............................................................................... 71 4.4 Errors Specific to Certain L1 Groups ........................................................................ 73 4.4.1 Errors made by one language family........................................................................ 74 4.4.2 Errors shared by two language families ................................................................... 79 4.4.3 Errors shared by more than two language families .................................................. 83 4.5 Discussion ...................................................................................................................... 85 4.5.1 Universal errors ........................................................................................................ 85 4.5.2 Errors specific to certain L1s ................................................................................... 87 CHAPTER 5 CONCLUSION ............................................................................................... 90 5.1 Summary of major findings ........................................................................................ 90 5.2 Pedagogical Implications ............................................................................................ 91 5.3 Limitations of the Present Study and Future Research ............................................... 93 REFERENCES....................................................................................................................... 96 APPENDIX 1 ........................................................................................................................ 101 APPENDIX 2 ........................................................................................................................ 114. viii.

(9) LIST OF TABLES Table 2.1 Grammatical collocations adapted from Benson et al. (1997) ................................. 13 Table 2.2 Lexical collocations adapted from Benson et al. (1997) .......................................... 14 Table 3.1 Word Tokens in Each L1 and Word Tokens in Total. ............................................... 38 Table 3.2 The language codes of 11 languages ........................................................................ 46 Table 4.1 The collocation errors in TOEFL 11 (ranked by frequency) .................................... 52 Table 4.2 Error types of 109 collocation errors in TOEFL 11. ................................................ 58 Table 4.3 The error types of universal errors. .......................................................................... 59 Table 4.4 The deviant use of do ............................................................................................... 60 Table 4.5 The deviant use of loose........................................................................................... 62 Table 4.6 The deviant use of experiment ................................................................................. 64 Table 4.7 The deviant use of try............................................................................................... 64 Table 4.8 The deviant use of reach .......................................................................................... 66 Table 4.9 The deviant use of agree .......................................................................................... 67 Table 4.10 The deviant use of travel ........................................................................................ 68 Table 4.11: The deviant use of specialize ................................................................................ 68 Table 4.12 The deviant use of listen and search ...................................................................... 69 Table 4.13 The deviant use of go ............................................................................................. 70 Table 4.14 The deviant use of participate................................................................................ 71 Table 4.15 Deviant noun usages .............................................................................................. 72 Table 4.16 Other types of errors .............................................................................................. 73 Table 4.17 Specific errors categorized by language families .................................................. 74 Table 4.18 Errors made by learners of Indo-European language family ................................. 75 Table 4.19 Errors made by learners of Altaic language family................................................ 76 Table 4.20 Errors made by learners of Sino-Tibetan language family .................................... 78 Table 4.21 Errors made by learners of Dravidian language family ......................................... 78 Table 4.22 Errors made by learners of Indo-European/Altaic language family ...................... 79 Table 4.23 Errors made by learners of Altaic/Sino-Tibetan language family.......................... 81 Table 4.24 Errors made by learners of Indo-European/ Dravidian language family ............... 82 Table 4.25 Errors made by learners of Altaic/Dravidian language family .............................. 83 Table 4.26 Errors made by learners of Indo-European/Altaic /Afro-Asiatic language family 84 Table 4.27 The most frequent types of collocation errors by different language families....... 89 Table 4.28 The number of errors made by Korean, Japanese or both learners ........................ 89. ix.

(10) LIST OF FIGURES Figure 3.1: Number of Essays per Language per Prompt from Blanchard et al. (2013) ......... 37 Figure 3.2: Histogram of essay lengths for all essays (Blanchard et al., 2013)....................... 38 Figure 3.3: Language families in the corpus (Blanchard et al., 2013) ..................................... 39 Figure 3.4: An example of the word sketch function on the Sketch Engine ........................... 40 Figure 3.5: The Sketch Diff interface showing big and large in the BNC ............................... 42 Figure 3.6: The objects of keep and maintain as collocates in the BNC. ................................ 43 Figure 3.7: Part of concordance of maintain the integrity in the BNC .................................... 44 Figure 3.8: The function of tracing source of the concordance ............................................... 44 Figure 3.9 The columns of object of and object from the noun “time” and the verb “do.” ..... 48 Figure 3.10 The source tracing function (After a user clicked on file2263628 before the concordance, the country code, ARA, in the yellow note will pop up). ................................... 49 Figure 3.11 The procedure of data extraction. ......................................................................... 49 Figure 3.12 The data analysis in the present study .................................................................. 51 Figure 4.1 The ratio of error types in TOEFL 11 ..................................................................... 59. x.

(11) CHAPTER 1 INTRODUCTION. 1.1 Background and Motivation The importance of collocations has long been discussed and emphasized in the field of second language learning for many decades by previous research (Biber, Johansson, Leech, Conrad, & Finegan, 1999; Granger, 1998; Lewis & Conzett, 2000; Schmitt, 2004; Schmitt, Ng, & Garras, 2011; Wray, 2002). Collocations, as word combinations in make a decision or commit suicide, play an essential role to help learners think more efficiently and quickly in listening, speaking, reading and writing (Hill, 2000). Collocation knowledge is also significant in vocabulary learning because, as Woolard (2000) has stated, “learning more vocabulary is not just learning new words; it is often learning familiar words in new combination.” Once learners get familiar with these word combinations, their language usages can become more accurate and fluent (Wray, 2002). Three kinds of word combinations, including free combinations, collocations and idioms, have been discussed in studies of collocations. Collocations located in the middle of the continuum between free combinations, which are phrases freely combined depending on rules of syntax, and idioms, which are used in restricted senses and cannot be determined its meanings according to its literal meanings (Howarth, 1998; Nattinger & DeCarrico, 1992). Collocations, therefore, fall around the middle ground of the continuum and are regarded as “habitually occurring lexical combinations that are characterized by restricted co-occurrence of elements and relative transparency of meaning” (Laufer & Waldman, 2011).. 1.

(12) Many researchers have found that foreign language learners have difficulties in the usage of collocations. Some learners may doubt that why people cannot say “conduct a diary” or “maintain a diary”; instead, they should say “keep a diary.” Similar instances occur in everyday life and Benson, Benson, and Ilson (1997) also provided some good examples of difficulties that learners might face: Collocations are arbitrary and non-predictable. Non-native speakers cannot cope with them; they must have a guide. They have no way of knowing that one says in English make an estimate, (but not *make an estimation), commit treason (but not *commit treachery). In English, one says commit fraud and perpetrate fraud. However, only the collocation commit suicide is possible; one does not say *perpetrate suicide. One says bake a cake, but make pancakes (p. 258). Previous research showed that EFL learners usually make collocation errors due to the lack of collocation knowledge in their writing or translation tasks (Bahns & Eldaw, 1993; Laufer & Waldman, 2011; Nesselhauf, 2003; Zinkgräf, 2008). Due to the avoidance of using unfamiliar collocations, many students tend to ‘rely on larger, rarer, and clumsier words which make their language sound stilted and awkward’ (Sinclair, 1991, p. 79). Cobb (2003) also mentioned that learners would produce grammatically acceptable sentences that are unidiomatic and strange to native speakers because of their unfamiliarity of collocations. Likewise, Laufer (2010) found that students had trouble finding the right verbs of collocations, and they thought they knew the collocations without a need to consult dictionaries but actually they did not. When learners face difficulties in finding the right collocations, they tend to rely on lexical simplification strategies, such as avoidance, synonyms, paraphrasing, and L1 equivalence (Mohammed Farghal & Obiedat, 1995). However, learners sometimes find it difficult to paraphrase collocations that they cannot produce (Bahns & Eldaw, 1993). 2.

(13) Sometimes, learners’ misuse of collocations are originated from overusing certain collocations with core verbs (make, do, have, etc.) without utilizing native-like expressions due to their deficiency in collocation knowledge (Altenberg & Granger, 2001; De Cock, 1998; Granger, 1998). The overuse may be attributed to their overdependence on those high frequent verbs, with which they feel comfortable since they learn them early and use them widely (Hasselgren, 1994). Furthermore, since these verbs lack specific meanings, learners tend to overgeneralize them and combine them freely with other nouns (Zinkgräf, 2008), which also explain why learners overuse the high frequent verbs. These phenomena reveal that overusing collocations with core verbs sometimes results in learners’ misuse of collocations, and learning collocations are challengeable for language learners. Learning collocations is a long and tiring process for not only low proficiency learners, but also advanced learners. Several studies have revealed that even for advanced learners, they still make a variety of collocation mistakes in their writing tasks (Bahns & Eldaw, 1993; Mohammed Farghal & Obiedat, 1995; Nesselhauf, 2003). The finding of Gitsaki (1996) might somehow explain the difficulties among advanced learners. Gitsaki (1996) observed that Greek junior high school students’ collocation knowledge did not increase as their grammatical competence improved, which showed that learners’ language competence did not match collocation competence. This implies that even though advanced learners’ grammatical competence is well developed, they could still have trouble dealing with collocations. According to Benson et al. (1997), collocations can be divided into two categories: grammatical and lexical collocations. Grammatical collocations are composed of a dominant word—noun, verb, adjective, participle, and a preposition or a grammatical construction. On the other hand, lexical collocations do not contain a dominant word and they have structures, like verb + noun, adjective + noun, noun + verb, noun + noun, adverb + adjective, adverb + 3.

(14) verb, etc. Studies have revealed that EFL learners have relatively more problems with lexical collocations rather than grammatical ones (Bahns & Eldaw, 1993; Gitsaki, 1996, 1999). Especially, much research revealed that learners have difficulties in the usages of verb noun lexical collocations (Bazzaz & Samad, 2011; Mohammed Farghal & Obiedat, 1995; Granger, 1998; Juknevičienė, 2008; Laufer & Waldman, 2011; Liu, 2002; Nesselhauf, 2003; Todirascu & Gledhill, 2008; Zinkgräf, 2008). Therefore, it seems that verb noun collocations can be a potential area to further explore. Previous research tended to explore verb noun collocations from two different perspectives, conducting collocation tests (Bahns & Eldaw, 1993; Bazzaz & Samad, 2011; Boers, Demecheleer, Coxhead, & Webb, 2013; Jaén, 2007) and exploring learner corpora. In terms of the latter one, most previous studies have extracted EFL learners’ use of verb noun collocations from learner corpora through the method of manual extraction (Mohammed Farghal & Obiedat, 1995; Liu, 1999; Nesselhauf, 2003; Zinkgräf, 2008). The manual extraction of miscollocations is mostly time-consuming and needs substantial humor labor. For instance, Bazzaz and Samad (2011) observed Iranian EFL learners’ verb noun collocations in the writing tasks and the number of verb noun collocations were counted manually. Nesselhauf (2003) extracted German advanced learners’ use of verb noun collocation errors in their free writing manually. Likewise, Zinkgräf (2008) manually searched for instances of verb noun miscollocations in the writing production of 102 Spanish students. On the other hand, some studies attempted to resort to the assistance of concordancer for the purpose of accelerating the process of extraction but researchers still needed to check each concordance to find collocations. Altenberg and Granger (2001), for example, compared French learners and Swedish learners’ lexical patterning of make with native speakers’ usages through the concordancer, Wordsmith Tools and manual examination. Moreover, 4.

(15) Juknevičienė (2008) manually selected the verb noun collocations of five core verbs (have, do, make, take, give) through the aid of Wordsmith Tools as well. From studies above, it can be found that many studies chose to manually search for miscollocations, which means they need to go through learners’ production to achieve the purpose, which is really time-consuming. Even though some studies indeed resorted to some tools, such as Wordsmith Tools, to search for miscollocations, which is more efficient than pure manual extraction but it is still necessary for them to look at each concordance to find the collocation errors after they typed in certain target words, Accordingly, it seems that a more efficient way to retrieve collocation errors is needed. Studies that discussed collocations mostly attributed learners’ collocation errors to L1 interference (Altenberg & Granger, 2001; Biskup, 1992; Nesselhauf, 2003, 2005; Zinkgräf, 2008). Since these studies investigated one or two non-native learner groups, the L1 influence would be more obvious or emphasized more often. However, it seems that those collocation errors may not be totally due to the influence of learners’ mother tongues. Other strategies, such as using synonyms or extension of analogy, which is the process of substituting one lexical item for another known element, might play important roles as well. For example, Källkvist (1999) found that Swedish and Norwegian learners tend to overuse delexical verbs, referring to verbs without specific meanings, such as make, because of L1 influence since these two languages are both Germanic languages. However, French learners also overuse those delexical verbs (Granger, 1996), which revealed that this finding is not totally L1-related. There is a need to further compare collocation errors among different L1 backgrounds to have more insight in the causes of these errors. Although previous studies provided fruitful insight of learners’ misuse of verb noun collocations, two aspects of verb noun collocations seemed to be neglected and unsolved. First, it can be observed that these studies only focused on collocation use in one or two 5.

(16) nonnative groups. Few studies investigated learners’ collocation use across different L1 groups; in this way, a wide range of cross-linguistic differences remain unstudied. Second, most research adopted a manual method to extract the verb noun collocations, which is laborious, tiring and cannot confirm the reliable cover of all miscollocations in learner corpora. While some studies indeed utilize the corpus tools, such as Wordsmith tools, those researchers still need to look at each concordance carefully to find the collocations, which could be time-consuming. These phenomena show that there still remains a need to further explore the cross-linguistic different use of verb-noun collocations in order to discover what kinds of verb noun miscollocations are universal for learners of different first languages (L1s) and what kinds of errors tend to be made only by learners of certain L1s. In order to explore the cross-linguistic differences, the innovative tool of extracting collocations is needed to save time and energy while researchers examine verb noun collocations. Therefore, the present study aims to investigate EFL learners’ verb noun miscollocations across eleven L1 groups in the learner corpus of TOEFL 11 through its comparison with British National Corpus (BNC) and Corpus of Contemporary American English (COCA) in a platform—The Sketch Engine to innovatively and efficiently save time and human labor.. 1.2 Purposes of the Study A query system—The Sketch Engine provides a new direction to explore collocations in different learner groups more easily and effectively. The Sketch Engine can allow users to view word sketches, similar words, and compare differences in corpora. Users can search for. 6.

(17) a keyword in the word sketch so that the concordance of its collocates will show up to provide more information (Kilgarriff, Rychly, Smrz, & Tugwell, 2004). Two special functions, Corpus creating and Sketch Diff make The Sketch Engine different from other concordancers (Kilgarriff et al., 2004). The Corpus creating function allows researchers to upload their own corpus data to The Sketch Engine and they can explore a keyword’s collocates through making use of the functions of the query system in their corpora. On the other hand, the Sketch Diff function can enable users to find out the differences in a keyword’s collocates of various part of speech in two different corpora, such as one native speaker corpus and one learner corpus. Users can compare the frequency of the keyword’s collocates in two different corpora at the same time. Therefore, researchers can create their corpora through the function of Corpus creating, and then use the Sketch Diff function to observe the differences of certain keywords’ collocates between their corpora and other corpora. Moreover, The Sketch Engine can track the sources of the concordances after users zip the data that share the same background as the same file, and upload them to the platform, which can help users know the source of the data when observing each concordance. The efficiency of searching for verb noun collocations can be enhanced through the use of Corpus creating and Sketch Diff, which can be a reliable way to replace the traditional method of extracting collocations manually. Through the use of the Sketch Engine, the present study attempts to compare EFL learners’ verb noun collocation errors in different L1 groups through comparing them with native speakers’ usages in a semi-automated method. The corpus of TOEFL 11, which contains learners’ English production from eleven L1 groups, would be analyzed and compared with BNC and COCA so as to observe the misuse of collocations from eleven L1 groups, including Arabic, Chinese, French, German, Hindi, Italian, Japan, Korean, Spanish, Telugu and Turkish learners of English. Collocations that appear in TOEFL 11 rather than the 7.

(18) BNC and COCA will be possible miscollocations. The possible collocation errors will be checked by native speakers again to make sure that they are unacceptable. When the researcher searches for a keyword’s verb noun collocations in TOEFL 11, the concordances of those collocations will appear by clicking on the collocates. Each concordance has a file name before it; thus, the researcher can click on the file name and then the original source (e.g. languages) will pop up so that the researcher can know learners of which L1 produced the collocations. The source of collocation errors will be found out through this way. The collocation errors will also be divided into four types as deviations in verbs, deviations in noun phrases, deviations in prepositional phrases and others based on Nesselhauf’s (2003; 2005) study, which will be explained in Chapter 2. Three dimensions will be investigated in the present study. One deals with the frequent verb noun collocation errors made by eleven L1 groups. Another is about what the error types of these verb noun miscollocations are. The other is concerned about what universal collocation errors are made by at least six L1 groups, and what errors are only made by certain L1 groups.. 1.3 Research Questions of the Study Based on the purposes of the present study, the following three research questions are proposed by the researcher: 1. What are the verb noun miscollocations made frequently by eleven L1 groups of EFL learners? 2. What are the error types of these miscollocations? 3. Among these miscollocations, what are the universal collocation errors shared by at least six L1 groups, and what are collocation errors made only by specific L1 groups?. 8.

(19) 1.4 Significance of the Study Previous studies tend to neglect the cross-linguistic differences and similarities of verb noun miscollocations among EFL learners. Also, most researchers extracted verb noun collocations manually in previous research, which is time-consuming and laborious (Bazzaz & Samad, 2011; Mohammed Farghal & Obiedat, 1995; Granger, 1998; Juknevičienė, 2008; Laufer & Waldman, 2011; Nesselhauf, 2003; Todirascu & Gledhill, 2008; Zinkgräf, 2008). Therefore, the present study aims to use a semi-automated way to more effectively investigate EFL learners’ universal and specific errors of verb noun collocations among eleven L1 groups through the corpus query system, The Sketch Engine. The present study can offer some insights in four perspectives. Firstly, the study will reveal the frequently misused verb noun collocations from these eleven L1 groups. Instructors can observe what kinds of collocations that learners tend to make mistakes and pay attention to those collocations while teaching them. Secondly, among these errors, the researcher will divide these miscollocations into different error types and discuss whether the result of error types is similar to previous research so as to expand the findings of error types in previous studies. Thirdly, the methodology used by previous researchers are usually time-consuming, the present study thus used an innovative way to extract miscollcations, which is The Sketch Engine. This way can save efforts and time, and at the same time, avoid human misjudgment to search for miscollocations. Last but not least, due to the Sketch-Diff, the researcher can find out the sources of those collocation errors more easily, which can let the researcher discover the universal errors shared by different L1 groups. In that way, the instructors and researchers can know what errors are really problematic for most learners no matter what languages they use. Also, specific errors only made by certain L1 groups can be revealed as well.. 9.

(20) 1.5 Definitions of Terms 1. Verb noun collocations: According to Benson et al. (1997), verb noun collocations belong to one type of lexical collocations, which consist of verbs and nouns, e.g. make a decision and keep a diary. In the present study, verb noun collocations will be extracted by The Sketch Engine semi-automatically. 2. Learner corpus: A learner corpus contains learners’ written or spoken production. Researchers usually compared a learner corpus with a native speaker corpus in order to find learners’ underuse, overuse or misuse of language production. In the present study, the learner corpus—TOEFL 11 will be compared with the native speaker corpora—the BNC and COCA in order to find the misuse of verb noun collocations. 3. The Sketch Engine: It is a corpus query system, which allows users to observe word sketches, thesaurally similar words, sketch differences and create a corpus. The word sketch can provide one list of collocates consisting of various grammatical relations (Kilgarriff et al., 2004). The special functions of creating corpora and Sketch Diff in The Sketch Engine are utilized in the present study to extract verb noun collocations in the eleven L1 groups. 4. Semi-Automated: Traditionally, researchers need to extract verb noun collocations manually by looking at the plain text or each concordance carefully and then checking their pragmatic and semantic appropriateness to determine whether they are miscollocations or not. The Sketch Engine, in the present study, eases the process of extracting and determining the appropriateness of collocations by doing the part of human labor, which is considered as the semi-automated way.. 10.

(21) CHAPTER 2 REVIEW OF LITERATURE 2.1 Collocations In this section, the definitions of collocations will be firstly reviewed. One specific type of collocations, verb noun collocations, will be defined later. Then, the theories of collocation studies will be discussed.. 2.1.1 Definitions and V-N types of collocations Collocations are firstly defined as “the company words keep together” by Firth (1957, p.196). Researchers proposed different terms to refer to collocations. For instance, Alexander (1984) considered collocations “fixed expressions” and Granger (1998) termed collocations as “prefabricated patterns.” The notion of collocations are related to different types of lexical expression, containing “catch-phrases, clichés, fixed expressions, formulae, free and bound collocations, idioms, lexical phrases, turns-of-phrase and so on” (C. J. Gledhill, 2000, p. 7). The distinction between these category types spurred a lot of debates and discussion among different researchers. Various researchers have their own interpretation about the definition of collocations in previous studies. According to Frath and Gledhill (2005), there are three different perspectives of collocations: (a) co-occurrence, which refer to collocations as textual clusters commonly co-occurring together from a statistical view; (b) construction, in which collocations are considered “obeying inherent semantic and syntactic patterns of particular lexical items” (Frath & Gledhill, 2005, p. 3); (c) expression, a pragmatic view of collocations on the basis of a 11.

(22) relation between a sign and a textual function. However, neither of them can fully describe the notion of collocations since traditionally speaking, collocations are considered as a continuum from free combination to idioms (Howarth, 1998; Laufer & Waldman, 2011; Nesselhauf, 2003; Todirascu & Gledhill, 2008): ‘Free Combination’ ↔ ‘Bound Collocation’ ↔ ‘Frozen Idiom’ Many researchers who hold the view of collocations as a continuum believe that free combinations and idioms are on the two opposite sides of the continuum. Laufer and Waldman (2011) claimed that the elements of restricted co-occurrence and relative semantic transparency differentiate collocations from free combinations and idioms. Free combinations (e.g. want a car), in which the individual words can be replaceable easily following the rules of grammar, differ from collocations based on the concept of restricted co-occurrence. On the other hand, opaque idioms (e.g. kick the bucket) are different from collocations because people cannot determine their meanings from the individual words that compose them, which is concerned about relative semantic transparency. However, sometimes it is hard to differentiate free combinations and collocations. It seems that free combinations occur without restriction but actually they are still bounded in some cases. Take the free combination “accept murder” for example, the word “accept” cannot be replaced by any synonym like “take.” The model proposed by Benson et al. (1997), therefore, might solve the problem of defining collocations. Benson et al. (1997) considered collocations as the following: In any language, certain words regularly combine with certain other words or grammatical constructions. These recurrent, semi-fixed combinations, or collocations, can be divided into two groups: grammatical collocations and lexical collocations. Grammatical collocations consist of a dominant word—noun, adjective/participle, 12.

(23) verb—and a preposition or a grammatical construction. Lexical collocations, on the other hand, do not have a dominant word; they have structures such as the following: verb+noun, adjective+noun, noun+verb, noun+noun, adverb+adjective, adverb+verb. (p.xv) Therefore, in the present study, we follow Benson et al. (1997) definition considering collocations as “fixed, identifiable, non-idiomatic phrases and constructions” (p. xv), such as commit suicide and make a mistake. Some free combinations might be referred to either grammatical or lexical collocations, based on the features they have (Wei, 1999). In order to understand the collocation types more explicitly, the two broad categories of grammatical and lexical collocations are presented in Table 2.1 and Table 2.2.. Table 2.1 Grammatical collocations adapted from Benson et al. (1997) Rules Examples (1) noun + preposition. The blockade of enemy ports by our navy.. (2) noun + to + infinitive. We made an attempt to do it.. (3) noun + that-clause. She took an oath that he would do his duty.. (4) preposition + noun. We discovered it by accident.. (5) adjective + preposition. She was angry at my friends.. (6) adjective + to + infinitive. She is ready to go.. (7) adjective + that-clause. It is necessary that he be replaced immediately.. (8) 19 verb patterns, including:. She continued to write.. verb + to + infinitive, verb +. They enjoy watching TV.. gerund, verb + object 1 + object. The police fined her fifty dollars.. 2, and others. 13.

(24) Table 2.2 Lexical collocations adapted from Benson et al. (1997) Rules. Examples. (1) verb (creation/activation) + noun. She does the laundry every day.. (2) verb (eradication/nullification) + noun The teacher declined our invitation. (3) adjective + noun / noun + noun. The room has a sour smell.. (4) noun + verb (action). Bombs exploded across Bangladesh.. (5) noun (unit) + of + noun. David gave Sandy a bouquet of flowers.. (6) adverb + adjective. They are closely acquainted.. (7) verb + adverb. They argued heatedly in that debate.. Bahns and Eldaw (1993) and Gitsaki (1999) stated that EFL learners have relatively more problems with lexical collocations than grammatical ones in their language production since they tend to make more errors in lexical word combinations. Among various types of lexical collocations, many previous studies found that EFL learners face difficulties especially in verb noun lexical collocations (Bazzaz & Samad, 2011; Mohammed Farghal & Obiedat, 1995; Granger, 1998; Juknevičienė, 2008; Laufer & Waldman, 2011; Nesselhauf, 2003; Todirascu & Gledhill, 2008; Zinkgräf, 2008). According to Cowie (1992), verbs can include three kinds of features in verb noun collocations: 1. Figurative: in “deliver a speech,” the keyword deliver has the figurative or abstract meaning of conveying certain ideas to the listeners, which goes beyond its original meaning of sending an object to others. 2. Delexical: in "make recommendations," the verb make does not have a specific meaning but is rather grammatical and vague. In this type of collocations, their meanings depend on the nouns since delexical verbs do not contain special meanings.. 14.

(25) 3. Technical: in "try a case," the action try whose meaning is constrained, narrow and specific, exhibits the meaning of the collocation.. Since verb noun collocations were found problematic for L2 learners from previous research, the present study aims to focus on this specific type of collocations in learner corpora. Previous studies in verb noun collocations tend to explore learners’ overuse, underuse or misuse of collocations. The use of verb noun collocations and also other types of collocations is usually explained by theories of the Open Principle and the Idiom Principle, which will be further explained in the following part.. 2.1.2 Theories of Collocations Sinclair (1991) suggested that both spoken and written human discourse are governed by two major principles: the Open Choice Principle (OP) and the Idiom Principle (IP), which are two contrary camps of thoughts. The OP, based on Chomskyan School of linguistics and the theory of Universal Grammar, claims that people have the ability to grammaticalize meaning in natural language in accordance with limitation rules of sub-categorization and selection in a language while there can be parametric dissimilarities between different languages. People construct the meanings of collocations by adding up the meaning of each individual word according to the OP. The IP, on the other hand, is related to a grouping of socio-lexical conventions in any language, which reveals that collocation meaning is constructed through the “whole phrase” of a collocation. The similar idea is also revealed in the statement of Sinclair (1991, p. 110), “the principle of idiom is that a language user has available to him a larger number of semi-preconstructed phrases that constitute single choices, even though they might appear to be analyzable into segments.”. 15.

(26) For instance, in English greeting, “good morning” resorts to the IP rather than the OP since the expression of “wonderful morning” or “excellent morning” as greeting is unavailable. The lexeme morning can be freely combined with adjectives, such as nice, terrible, wonderful and excellent based on the OP; however, in the greeting, the only acceptable and frozen expression is “good morning,” which can be explained by the idea of the IP instead of the OP (M Farghal & Al-Hamly, 2007). Bazzaz and Samad (2011) examined Iranian learners’ use of collocations and supported the importance of the idiom principle in their finding, indicating that the role of collocations should be highlighted since collocation knowledge is more beneficial than knowledge of individual words for EFL learners. According to Granger (2011), the IP considers each word form as having its phraseology, which explains learners’ difficulties since learners are not capable of resorting to this principle in their language production. This phenomenon was also shown in the finding of Zinkgräf (2008), who stated that learners tended to rely on the OP rather than the IP and ignored some restrictions of word combinations when they used collocations. The consequence of resorting to the OP for learners is that collocation errors add some foreign flavor to learners’ speech and writing, and make them differ from native speakers’ production. Mohammed Farghal and Obiedat (1995), who explored Arabic learners’ use of collocations, also reported that the participants were highly dependent on the OP for item selection, leading to wrong and incorrect collocations. These studies supported Sinclair’s (1991) claim that learners tend to rely on the OP while native speakers actually depend on the IP, which leads to EFL learners’ difficulties in collocation usages.. 2.2 Difficulties Faced by EFL learners Regarding learners’ use of verb noun collocations, most studies have indicated EFL learners’ deficiency in collocation knowledge, such as the overuse, underuse or misuse of 16.

(27) certain collocations in their language production (Altenberg & Granger, 2001; Hasselgren, 1994; Juknevičienė, 2008). Some studies indicated that learners’ knowledge of lexical items is better than knowledge of collocations comprised of the same lexical items. For instance, Barfield (2006) found that Japanese learners had better knowledge of single nouns and verbs than the collocations which consist of the same nouns and verbs. Additionally, not only low proficiency learners but also advanced learners have difficulties in the use of collocations (Altenberg & Granger, 2001; Bahns & Eldaw, 1993; Zinkgräf, 2008). These phenomena showed learners’ difficulties in collocation usages; thus, many researchers utilized different methods to conduct collocation studies for the purposes of providing more insight in learners’ use of collocations and offering pedagogical implications. Therefore, in the following two sections, the methods of exploring collocation knowledge will be firstly reviewed and the findings, which especially indicated the error types and causes of miscollocations, will be demonstrated later.. 2.2.1 Methods to Explore Collocation Knowledge Researchers found learners’ problems in verb noun collocation use through three methods: collocation tests, elicitation studies of collocations and error analyses of collocations. The latter two methods are usually related to exploring learner corpora while the first method is to design collocation assessment for learners, which will be reviewed in the following three parts.. I. Collocation Tests Previous studies used a variety of collocation tests, such as translation tasks or cloze tests to examine learners’ collocation competence, and found that learners’ collocation competence was still limited. For example, in a study of Bazzaz and Samad (2011), they found that there is a linear relationship between Iranian EFL learners’ collocation knowledge and their use of 17.

(28) collocations in writing stories by statistically comparing the result of collocation c-test with manually extracted verb noun collocations from the stories. Additionally, the work of Bahns and Eldaw (1993) developed two instruments, a translation task and a cloze test, to test German advanced learners’ productive knowledge of verb noun collocations. The result found that the number of collocation errors was twice more than the number of errors about single lexical items in the translation tasks among these EFL learners. Jaén (2007) not only assessed productive collocational knowledge but also receptive knowledge by an 80-item test consisted of multiple choice questions and cloze questions among Spanish university students. Expectedly, the result revealed that the score of productive knowledge items are significantly lower than test items of receptive knowledge, which shows that EFL learners’ collocation productive knowledge still needs to be improved. Boers et al. (2013), on the other hand, conducted four trials of pretests and posttests, which are fill-in-the-blank tests, in order to compare the effectiveness of collocation matching exercises in textbooks with exercises containing collocations as intact wholes for the purpose of finding out which exercises can better improve learners’ collocation competence. The finding indicated that the negative side effects were found more often in the matching exercises than the exercises where collocations are intact since the matching exercises might lead to error connections of collocations in learners’ memory. In addition to collocation tests, other studies examined learners’ collocation knowledge in their actual usage of collocations through learner corpora, including elicitation of certain collocations and error analyses of all collocations in learner production.. II. Elicitation Studies of Collocations In terms of elicitation of certain collocations, researchers searched for the pre-determined set of collocations in learner production to examine learners’ collocational 18.

(29) competence. These pre-determined set of collocations is usually the collocations with delexical verbs or high frequency verbs, such as have, do, make, take and give. High-frequency verbs contain some characteristics that make them specifically interesting from a cross-linguistic perspective (Viberg, 1996): 1. They express basic meanings and often prevail various semantic fields; 2. They have equivalents that are frequently used in most languages; 3. They contain a high degree of polysemy, resulted from two types of meaning extension: --one universal tendency producing more abstract, general, grammaticalized or delexicalized usages, --a variety of language-specific tendencies leading to specialized meanings, collocations, and idiomatic usages; 4.. They are usually problematic for EFL learners. Zinkgräf (2008, p. 108) stated, “under the assumption that these verbs lack a specific. meaning, learners overgeneralize and combine them with any noun under the illusion that there is no restriction to the way they can be used.” Boers et al. (2013), similarly, mentioned that it is usually the verb in collocations that learners replace by an unconventional choice (*do mistakes), and such replacement might be originated from L1 interference when equivalent L1 nouns can collocate with a different verb (Nesselhauf, 2005; Yamashita & Jiang, 2010). Learners’ dependence on L1 and their tendency to ignore restriction on word combinations could imply that learners construct meanings from individual words instead of from prefabricated patterns, unlike native speakers (Kjellmer, 1991; Wray, 2002), which corresponds to the Open Choice Principle. Thus, learners tend to consider delexical verbs as having unrestricted collocations, which leads to learners’ misuse of verb noun collocations. Two contradictory findings of delexical structures have been discussed by previous studies. Some studies pointed out the overuse of collocations with delexical verbs by EFL 19.

(30) learners. For example, Hasselgren (1994) observed that core words, such as high frequency verbs, are widely overused by Norwegian learners who tend to rely on them, like “lexical teddy bears.” She explained, “core words—learnt early, widely usable, and above all safe (because they do not show up as errors) are usually overused, even among learners sufficiently advanced to have been weaned off them”(Hasselgren, 1994, p. 250). A similar finding is also revealed in the study of Källkvist (1999), which showed the overuse of collocations with delexical verbs among Norwegian and Swedish learners. Granger (1996) found French learners also overuse those high frequency verbs in their collocation usages. On the contrary, other studies revealed that learners tend to underuse collocations with delexical verbs. Sinclair (1991, p. 79) reported, “many learners avoid the common verbs as much as possible, and especially where they make up idiomatic phrases. Instead of using them, they rely on larger, rarer, and clumsier words which make their language sound stilted and awkward.” For instance, Altenberg and Granger (2001) explored French and Swedish advanced learners’ use of grammatical and lexical collocations of MAKE in International Corpus of Learner English (ICLE) through Wordsmith Tools, and found that both Swedish and French learners underuse the delexical combinations, which correspond to Sinclair’s underuse hypothesis. These advanced learners not only underused delexical structures but also made several mistakes in collocates with a high frequency verb, make. Similarly, Juknevičienė (2008) conducted a pilot study to examine Lithuanian learners’ collocations with high frequency verbs, including HAVE, DO, MAKE, TAKE and GIVE in comparison with native speakers’ usages quantitatively and qualitatively by the assistance of Wordsmith Tools. The result showed that Lithuanian learners underused the typical collocations with delexical verbs in academic writing compared to native speakers, which can be explained through their deficiency of collocation knowledge. To compensate for the lack of knowledge, learners tended to resort to their L1 to create their own collocations, which 20.

(31) were rarely used by native speakers. Nesselhauf (2004), on the other hand, analyzed German learners’ miscollocations of delexical verbs, such as have, take, make and give. Although she did not discuss the dimension of underuse and overuse of miscollocations, she raised some typical learners’ difficulties of verb noun combinations and divided them into different error types, such as wrong verbs (*take changes) and wrong nouns (*make a trial) in the finding. From these studies, although the overuse or underuse of delexical verbs in collocations is still paradoxical and inconclusive, these phenomena can sometimes lead to the misuse of verbs in verb noun collocations, which can explain the partial causes of verb noun collocation errors. Besides, the importance of delexical verbs in collocation usages can be revealed in these studies.. III. Error Analyses of Collocations In addition to the studies eliciting certain collocations with delexical structures, other studies adopted a broader approach to examine the overall collocations of complete texts created by students. Some of them resorted to the manual extraction of collocation errors in learner corpora while others observed learners’ collocation usages through the assistance of corpus tools and manual inspection. Traditionally speaking, researchers tend to investigate learners’ use of collocations manually. For example, Liu (1999) collected 94 learners’ writings and 127 exam papers among Taiwanese EFL learners and manually checked learners’ miscollocations. The finding showed that 63 collocation errors were found in their written production and most of them are verb noun collocation errors. According to Chen (2002), she manually checked learners’ collocation errors in 90 compositions from 30 Taiwanese senior high school students through consulting the BNC. If a collocation appeared in the BNC, it would be considered as the correct one; however, if no occurrence of the collocation appeared in the BNC, the 21.

(32) collocation would be identified as the wrong one. In the end, 147 grammatical collocation errors and 125 lexical errors were found in her study. Nesselhauf (2003) manually counted the verb noun combinations in German advanced learners’ 32 essays from The International Corpus of Learner English (ICLE) and categorized the verb noun combination mistakes into three types as free combination errors, collocation errors and idiom errors through looking up the Oxford Advanced Learner’s Dictionary (OALD 2000) and the Collins COBUILD English Dictionary (CCED 1995). She found that among collocation mistakes, the most frequent type of mistakes is the wrong choice of verb. Two years later, Nesselhauf (2005) went on to analyze a larger German corpus from ICLE. She searched for verb noun miscollocations manually by using the BNC, dictionaries and consultation from native speakers. First of all, the collocations in learner production would be considered acceptable if they appeared in the following dictionary: Oxford Advanced Learner’s Dictionary 2000, the Collins COBUILD English Dictionary 1995, The BBI Dictionary of English Word Combinations 1997, and Oxford Dictionary of English Idioms 1993. The BNC would also be consulted to assist the researcher to determine the acceptability of the collocations. If the researcher still could not decide whether the collocation was right or wrong, she would consult native speakers to find out the answers. Finally, her result revealed that among 2078 verb noun collocations, 744 of them were misused by German learners. Zinkgräf (2008) also extracted the number of verb noun miscollocations manually from 13 written assignments by 102 high-intermediate Spanish students through consulting several dictionaries, and he found that the wrong choice of verbs and nouns accounted for 70% of the corpus. These studies checked for miscollocations manually in learner production, which is time-consuming and laborious.. 22.

(33) Recently, more and more researchers attempted to make use of some corpus tools, such as Wordsmith Tools or Part-of-Speech (POS) taggers to ease the process of searching verb noun collocations. Laufer and Waldman (2011), for instance, investigated the collocation use in both the Israeli Learner Corpus of Written English and Louvain Corpus of Native English Essays (LOCNESS). They used the word list function of Wordsmith Tools to retrieve the list of frequent nouns in the LOCNESS and then analyzed the list in the Vocabulary Profile to produce 220 most frequent nouns in English. Later, the researchers created the concordances for each noun in both corpora so as to extract verb noun collocations through checking each of them in The BBI Dictionary of English Word Combinations and The LTP Dictionary of Selected Collocations. Combinations that occurred in the two dictionaries would be considered as collocations. The collocation errors and the proficiency differences in the learner corpus would be analyzed as well. Their result indicated that L2 learners underused verb noun collocations in their essays compared to native speakers. While advanced learners used more collocations than low and intermediate level students, they made more mistakes as well, which implies the more collocations they produced, the more errors they made. Todirascu and Gledhill (2008) attempted to develop a semi-auto method to extract collocations through analyzing the Acquis Communautaire Corpus (ACC), a parallel corpus available in all European languages (Steinberger, Pouliquen, Widiger, & Ignat, 2006), through conducting various POS taggers, such as TreeTagger and TTL – a complex tool to pre-process texts (Ion, 2006). After tagging the corpus, the researchers applied a statistical module, using Log-likelihood (LL) scores, from the studies of Steinberger et al. (2006) and Todirascu and Gledhill (2008) in order to identify verb noun constructions. A list of verb noun co-occurrences was created through the statistical module. The finding, however, showed that the statistical methods alone were not enough to identify verb noun collocations. 23.

(34) and a manual inspection was needed to sort the data of the result into verb noun collocations and irrelevant combinations. Lin (2010) also explored verb noun collocations in tagged corpora, including a Taiwanese learners’ corpus, a Chinese learners’ tagged corpus and the BNC, through the tools of Antconc, Monoconc and Perl. The researcher firstly generated a list of thirty-three verb noun collocation patterns in the corpora through the Antconc program. Later, Monoconc was used to retrieve these patterns of verb noun collocations in the three corpora. Then, she compared and deleted the same patterns between two learner corpora and the BNC by Perl. In this way, only the potential miscollocations were remained since these usages did not appear in a native speaker corpus, the BNC. The potential miscollocations were inspected manually to delete the acceptable and correct ones and the real miscollocations were found in her study. The result listed several collocates that were frequently misused by Taiwanese and Chinese learners. Most miscollocations were attributed to L1 interference and the wrong choice of verb could be found in most of the collocation errors. More innovatively, Liu (2013) utilized The Sketch Engine, a platform capable of performing the usages of Antconc, Monoconc and Perl in Lin’s (2010) study, to explore Taiwanese and Chinese verb noun collocation errors and their causes through the comparison with the BNC. The Sketch Engine enabled the researcher to create his own corpora, provide a keyword’s collocates and spot the difference of collocates between two corpora. The researcher found the potential miscollocations through recording collocations that only appeared in the learner corpus instead of the BNC through the function of Sketch Diff, which can spot differences of collocates in two corpora. In the end, a total of 134 types of miscollocations were found in the finding and its possible causes were described by checking them in several dictionaries and online resources.. 24.

(35) From error analyses of collocations in learner corpora, one can observe that most previous research investigated collocations in a time-consuming and laborious way. Even though some research examined collocations through corpus tools and tagged corpora, some methodological limitations can be revealed in those studies. For example, although Laufer and Waldman (2011) utilized Wordsmith Tools to ease the process of searching for collocations, they still needed to observe a concordance after a concordance in the corpus. Todirascu and Gledhill (2008) can retrieve a list of collocations based on the tagged corpus and LL scores semi-automatically but this method cannot help researchers identify possible collocation errors. Lin (2010) indeed explored collocation errors in learner corpora through the comparison with the BNC; however, her procedure was really complicated and a lot of corpus tools, such as Antconc, Monoconc and Perl were needed. Therefore, The Sketch Engine, which can save time and efforts in the work of Liu (2013), will be a more optimal option for the present study to further discuss learners’ cross-linguistic differences and similarities of collocation usages. From the results of previous studies, some researchers categorized the collocation errors into different types and attempted to describe the possible causes of the miscollocations, which will be further elaborated in the following part.. 2.2.2 Error Types and Causes of Miscollocations Nesselhauf (2003) categorized German learners’ verb noun combination errors into three types as free combination errors, collocation errors and idiom errors. Regarding collocation errors, he divided verb noun miscollocations into nine types: 1. Deviant verbs: *carry out races (hold races) 2. Deviant nouns: *close lacks (close gaps) 3. Inappropriate combinations: *take notice (to notice) 25.

(36) 4. Not existed combinations: *hold children within bounds (show children where boundaries lie) 5. Inappropriate prepositions of verbs: *fail in one’s exam (fail one’s exam) 6. Inappropriate prepositions of nouns: *raise the question about (raise the question of) 7. Determiners (lack of or extra determiners): *get the permission (get permission) 8. Numbers (nouns used in the singular instead of the plural): *pass one’s judgments (pass judgment) 9. Wrong syntactic structures: *make sb. friends (make friends with sb.). Nesselhauf (2005) went on to analyze a larger German corpus from ICLE and categorized the miscollocations into deviations in verbs, deviations in noun phrases or prepositional phrases and more global deviations, which contain stretched verb construction instead of the corresponding verb, whole collocation inappropriate and deviations in the structure of the collocation. Most of the types are similar to her study in 2003. Only the categorization of more global deviations is somewhat different. Summarized from Nesselhauf (2003; 2005), error types which appear in both studies will be the standard of error types in the present study since these types are more common. Therefore, in the present research, the researcher will categorize the collocation errors into deviations in verbs, deviations in noun phrases, deviations in prepositional phrases and others to understand the general distribution of the error types. The reasons why learners misuse verb noun collocations can be attributed to two main aspects, interlingual errors or intralingual errors. Many researchers found learners’ atypical collocations are mostly interlingual, that is, their errors are influenced by their mother tongues (Altenberg & Granger, 2001; Mohammed Farghal & Obiedat, 1995; Granger, 1998; Laufer & Waldman, 2011; Nesselhauf, 2003; Zinkgräf, 2008).. 26.

(37) For instance, Nesselhauf (2003, 2005) suggested that German learners’ verb noun miscollocations were mostly attributed to their mother tongue. Altenberg and Granger (2001) concluded that collocation errors with delexical verbs made by French and Swedish learners were due to the influence of their mother tongues. Likewise, Laufer and Waldman (2011) searched for verb noun miscollocations in Israeli Learner Corpus of Written English and demonstrated that more than half of the collocation errors were influenced by learners’ L1. From the result of Juknevičienė (2008), the researcher explained that Lithuanian learners resorted to their L1 to compensate for their deficient collocation knowledge, which usually resulted in the misuse of collocations. In the study of Zinkgräf (2008), the finding showed that 61% of Spanish learners’ atypical collocations were generated from the interference of L1. Biskup (1992) indicated that the wider the meaning scope of a lexical item, the more the L1 interference in the collocations that lexical items took part in. Likewise, the more synonyms a lexical item had, the more difficulties learners faced when producing a restricted collocation. In addition to L1 interference, when learners fail to find appropriate verbs or nouns of collocations, they will resort to some lexical simplification strategies, such as the use of synonyms, avoidance, paraphrasing, extension by analogy or reliance on the Open Choice Principle (Mohammed Farghal & Obiedat, 1995; Zinkgräf, 2008), which can explain learners’ intralingual errors shared by learners from different L1 backgrounds. Howarth (1998) analyzed several studies and summarized a few strategies that learners frequently use when facing difficulties in collocation knowledge: 1. Avoidance: When learners are unable to produce collocations or they have difficulties in paraphrasing, they tend to avoid using those collocations. For instance, Biskup (1992) found that Polish students were more inclined to avoid answering questions while German students tended to paraphrase the unknown collocations.. 27.

(38) 2. Experimentation: From the study of Biskup (1992), the author concluded that German learners were risk-takers and when they did not know the correct form of collocations, they could find some synonyms to replace them. This kind of learners’ experimentation is also found in the results of Zinkgräf (2008) and Mohammed Farghal and Obiedat (1995). 3. Transfer: Learners’ risk-taking behavior might lead to L1 transfer even though learners may not be aware of problems in using L1 equivalence in L2 collocations. *State a record (set the record) by Polish learners and *lead a bookshop (run a bookshop) by German learners from Biskup’s (1992) study could be good examples. 4. Analogy: The process of substituting one lexical item for another known element in a collocation can be considered as L2 intralingual transfer. As Howarth (1996) indicated, a nonnative speaker can use the following verbs, including adopt, base on, bring in, discuss, implement and modify as the collocates of method, which can lead to a great degree of variation. However, analogy may also give rise to overgeneralization of collocation usages, such as *adopt a way (adopt an approach). 5. Repetition: A nonnative writer would repeatedly use a limited number of collocations when they lack confidence to produce collocations with analogies. Granger (1998), for example, concluded that French learners in her study tended to overuse certain adjective collocations with very, and some constructions like deeply-rooted appeared several times in the finding. These learner strategies could evolve into the causes of leaners’ atypical collocations. Based on the discussion of Liu (1999), investigating miscollocations from 94 copies of general writings and 127 exam papers written by college students, and Chang and Yang (2009), exploring Chinese learners’ causes of collocation errors modified from the research of James (1998), several possible causes of EFL learners’ verb noun collocation errors were summarized as follows: 28.