工作記憶和英語為外語的閱讀兩者間的關係--以嘉義市一所國中為個案 - 政大學術集成

全文

(1)國立政治大學英國語文學系碩士在職專班碩士論文. 指導教授: 余明忠教授 Advisor: Dr. Ming-chung Yu. 工作記憶和英語為外語的閱讀兩者間的關係--以嘉義市一所國中為個案 Relationship between Working Memory and EFL Reading Comprehension: A Case Study of Chiayi Junior High School Students. 研究生: 歐雅婷撰 Name: Irene Ou 中華民國一百零五年七月 July, 2016.

(2)

(3) RELATIONSHIP BETWEEN WORKING MEMORY AND EFL READING COMPREHENSION: A CASE STUDY OF CHIAYI JUNIOR HIGH SCHOOL STUDENTS. A Master Thesis Presented to Department of English, National Chengchi University. In Partial Fulfillment of the Requirements for the Degree of Master of Arts. by Irene Ou July, 2016. iii.

(4) iv.

(5) Acknowledgements The thesis would not have been completed without the support and encouragement of the following people. First, my highest and sincerest appreciation goes to my advisor, Dr. Ming-chung Yu. Without his inspiring guidance and encouragement throughout my research for this work, I could not be able to identify the flaws and refine the thesis. My appreciation extends to my other committee members, Professor Chieh-yu Yeh and Dr. Yi-ping Huang, who spent time reviewing the paper and provided valuable comments. My gratitude is especially given to those who have helped me stay sane through these difficult years. Their consistent support and care helped me overcome setbacks and stay focused on my graduate study. I greatly value their friendship and deeply appreciate their belief in me. Finally, I want to deliver my thanks to my parents and my beloved husband whose passionate encouragement made it possible for me to complete this research. With their love and support, I could devote all my efforts to the thesis writing. They stood by me and encouraged me through the days in completion of the thesis.. v.

(6) Table of Contents Acknowledgements ............................................................................................................ v Table of Contents .............................................................................................................. vi List of Tables.....................................................................................................................vii List of Figures ..................................................................................................................viii Chinese Abstract ............................................................................................................... ix Abstract ..............................................................................................................................iii CHAPTER ONE INTRODUCTION ............................................................................... 1 CHAPTER TWO LITERATURE REVIEW .................................................................. 7 Reading Comprehension .............................................................................................. 7 Working Memory Capacity (WMC) .......................................................................... 13 Theories Concerning WM and Reading Comprehension .......................................... 19 Measurement for WMC—Reading Span Test (RST) ................................................ 26 CHAPTER THREE METHODOLOGY ....................................................................... 31 Participants ................................................................................................................. 31 Instruments ................................................................................................................. 32 Procedure ................................................................................................................... 41 Data Analysis ............................................................................................................. 42 CHAPTER FOUR RESULT........................................................................................... 45 Introduction ................................................................................................................ 45 Correlation Analyses of the Reading Comprehension and WMC ............................. 45 Regression Analyses of the Reading Comprehension and WMC .............................. 47 CHAPTER FIVE DISCUSSION .................................................................................... 53 The Influence of WM on Literal Comprehension ...................................................... 53 The Influence of L2 WMC on L2 Inferential Comprehension .................................. 56 RST Types and WMC ................................................................................................ 59 CHAPTER SIX CONCLUSION .................................................................................... 65 Summary of the Study ............................................................................................... 65 Limitation ................................................................................................................... 68 Pedagogical Implication............................................................................................. 79 Suggestions for Future Studies .................................................................................. 81 REFERENCES ................................................................................................................. 85 APPENDIX A RST—LIST OF RST STIMULUS SENTENCES ................................ 97 APPENDIX B The Reading Comprehension Test with Four Passages Applied in the Study.................................................................................................................................. 99 vi.

(7) List of Tables Table 3.1 Summary of Literature Review on RST............................................................. 33 Table 3.2 Example Sentences of RST with WMC=3 ......................................................... 34 Table 3.3 Pearson and Johnson’s Taxonomy of Reading Comprehension Questions ....... 40 Table 4.1 Correlation Analyses of Recall-WMC and RC .................................................. 46 Table 4.2 Correlation of Recognition-WMC and RC ........................................................ 47 Table 4.3 Regression Analysis of WMC and L2 Literal RC ............................................. 48 Table 4.4 Regression Analysis of WMC and L2 Inferential RC ....................................... 49 Table 4.5 Regression Analysis of WMC and L2 Overall RC ............................................ 50 Table 6.1 Descriptive Analysis on Reading Span Task (RST) of the 190 Participants ...... 71. vii.

(8) List of Figures Figure 2.1 The multiple-component system of WM........................................................ 15 Figure 3.1 Procedure Flow Chart ....................................................................................... 42. viii.

(9) 國立政治大學英國語文學系碩士在職專班碩士論文提要. 論文名稱:工作記憶和英語為外語的閱讀兩者間的關係 --以嘉義市一所國中為個案指導教授:余明忠教授研究生: 歐雅婷論文提要內容: 本研究旨在探討工作記憶對英文閱讀理解的相關性，而其中更細分去看工作記憶對字面理解和推論理解的影響。另一方面，測量工作記憶的閱讀廣度作業(RST) 在研究工作記憶和閱讀理解時常被忽略不同作業測量是否對研究結果有所影響，而廣為研究採用的是辨識閱讀廣度(recognition-RST)以及再認閱讀廣度(recall-RST)。因此，本研究除了探討工作記憶對英文字面和推論理解的相關性，也欲比較此兩種不同閱讀廣度作業對於工作記憶和閱讀理解的關係是否有所不同。實驗對象是嘉義市立研究者任教的一所國中，全校 37 個班級，其中抽出可以配合研究過程的學生，總計 190 人，來自七個班級。這些班級的學生都有各自原本班級要進行的課程，因此為了施測方便，以班級為單位，不同班級完成不同的工作記憶廣度，會使研究有辦法進行，因此，本研究並無設定特定的分組規準，七個班級會分兩頭取得各自的工作記憶廣度—也就是四個班級完成再認閱讀廣度(recall-RST) 以及閱讀測驗; 另外三個班級辨識閱讀廣度(recognition-RST)以及閱讀測驗。其中使用的閱讀測驗旨在評量參與者字面閱讀理解和推論閱讀理解的能力。研究結果顯示再認閱讀廣度的工作記憶對字面閱讀理解的相關性、和推論閱讀理解的相關性、和整個閱讀理解的相關性皆未達顯著相關性。另外，辨識閱讀廣度的工作記憶對字面閱讀理解的相關性、和推論閱讀理解的相關性、和整個閱讀理解的相關性也皆未達 ix.

(10) 顯著相關性。此次研究結果和 Ruppe 等人(2006)以及 Sweller(1994) 的研究一樣，主張工作記憶和字面閱讀理解關係不大。至於推論閱讀理解部分，本研究針對工作記憶和推論閱讀理解的相關性和其他研究則不符，許多研究主張工作記憶對高階認知學習(high-level cognition activities) 有極大的幫助 (Anderson et al., 1996; Altepkin & Ercetin, 2001; Baddeley, 2012; Conway & Engle, 1994; Daneman & Hannon, 2007)。此處和其他研究主張意見分歧，可以從 Gathercole 和 Alloway (2007 ) 的說法來解釋，工作記憶對於學習認知活動的相關性是有所限制的。談及影響工作記憶和推論理解或是與其他較具挑戰性的認知活動時，尚有很多相關因素在研究中需要考量，例如參與者的閱讀技巧、專注力和背景知識。最後，本研究欲探討辨識閱讀廣度 (recognition-RST)以及再認閱讀廣度(recall-RST)對於工作記憶再閱讀的影響力是否有所差別，迴歸分析表示此兩種閱讀廣度測到的工作記憶對閱讀的影響力無顯著差異; 針對這點，有些學者主張閱讀廣度雖不同，但都能測驗到相同的工作記憶(Turner & Engle, 1989)，然而也另外有學者主張辨識和再認閱讀廣度測驗到的工作記憶有所不同(Alptekin & Ercetin, 2009; Unsworth & Engle, 2007)。因此未來需要更完善的研究設計和實施，排除其他影響工作記憶和閱讀相關的因素，例如分組規準、參與者的閱讀技巧、背景知識等，才能更加確定工作記憶對字面和推論理解的影響力，以及不同的工作記憶測量工具對工作記憶影響力的差異。. x.

(11) Abstract This study examined the influence of working memory (WM henceforth) on literal and inferential comprehensions in second language (L2) reading. WM refers to individuals’ cognitive process in which new and old information is temporarily stored and simultaneously processed. The strength of WM enables individuals to accomplish complex tasks, such as reading and reasoning. It sounds invincible, but it has its limitation (Baddeley, 2000; Conrad & Hull, 1964). The capacity of WM (WMC) represents the abstract concept of WM and is often measured with reading span tasks (RSTs). Also this study was aimed to investigate whether the influence of WM on reading comprehensions (RC henceforth) was different when WM is measured with different RSTs. According to Alptekin and Ercetin (2009), the difference of measurement tasks about WMC is often not taken into consideration in research. Thus, the investigator measured the participants’ WMC with two main RSTs, a recall-RST and a recognition-RST, and the results were later analyzed with their performance of L2 literal and inferential comprehensions. The participants in this study were 190 students from 7 classes in a Chiayi City Junior High School with a total of 37 classes. Due to the limitation of course schedules at school, participants had to attend their own classes. The investigator decided to group entire classes into same WM group in order to carry out the study. Thus, no specific grouping criterion was applied—the participants from three classes accomplished a recognition-RST and a RC test; the participants of the rest accomplished a recall-RST and a RC test. The test of RC contained four passages with five literal and five inferential RC questions, and was designed to assess participants’ ability to read literally and inferentially. The findings showed that WMC measured with a recall-RST had no correlations with participants’ literal, inferential and overall RC. In the recognition-RST group of WM, iii.

(12) the correlations of WM and RC was not significant, either. The current study and previous studies all suggested that WMC had no influence on literal RC (Rupp et al., 2006; Sweller, 1994). However, the result of WMC in this study did not successfully demonstrate positive correlation with inferential RC, which was against previous findings (Anderson et al., 1996; Altepkin & Ercetin, 2001; Baddeley, 2012; Conway & Engle, 1994; Daneman & Hannon, 2007). This result suggested that it was not easy to relate WM to the performance of inferential RC. Though the strength of WM is highly related to complex tasks, according to Gathercole and Alloway (2007), WM is unfortunately limited in certain ways regarding its influence on cognitive activities. Relevant factors, such as participants’ reading skills, attention to each current task and background knowledge, if not controlled, might interfere with the positive influence of WM on RC. Further research is warranted to fully examine the relationship between WM and RST in terms of one’s ability to read literally and inferentially.. iv.

(13) CHAPTER ONE INTRODUCTION “The eye sees only when the mind is prepared to comprehend.” –Henri Bergson Many researchers have argued that working memory (henceforth WM) is predictive of reading comprehensions and accounts for language performance (Altepkin & Ercetin, 2010; Anderson, Reder, & Lebiere, 1996; Baddeley, 2012; Conway & Engle, 1994; Daneman & Carpenter, 1980; Daneman & Hannon, 2007; Miyake & Friedman, 1998; Walter, 2004). WM refers to a systematic mechanism or an ability to maintain information active while processing the very or another piece of information simultaneously (Baddeley & Hitch, 1974; Baddeley, 2003; Daneman & Carpenter, 1983). WM flexibly supports daily cognitive activities that require storage and manipulation of information, such as mental mathematics and reading comprehension (henceforth RC). WM correlates with RC mostly because it helps readers to efficiently store information from texts, such as syntactic or semantic messages, and to coherent all the information with coming messages in following texts (Daneman & Carpenter, 1980). Though the relationship between WM and RC has been extensively studied, most works in this field study seldom involve EFL (English as a foreign language) learners or and has assessed RC as a single construct, providing limited data and neglecting the multiple levels of RC (Daneman & Carpenter, 1980; Baddeley, Logie, Nimmo-Smith, & Brereton, 1985; Engle et al., 1992; Ericsson & Delaney, 1999). It was only until recently that more researchers have come to pay attention to “its multilevel representational architecture and the role played by each level in reading comprehension” (Alptekin & Ercetin, 2009, p.628). It could be concluded that WM has significant influence on reading comprehension as one single construct and also on it as a multilevel construct. But further investigation is necessary in terms of the relationship between WM and language comprehension as a multilevel construct. In Taiwan, the Basic Competency Test for junior high school (BCT) was 1.

(14) implemented from 2001 to 2013. In 2014 it was replaced by a new entrance examination to senior high school, Comprehension Assessment Program for High School Students (CAP). One main difference between the English reading comprehension tests of the CAP and those of the BCT is the proportion of comprehension questions ranging from basic ones to difficult ones; more challenging comprehension questions, such as making inference, are included. According to the official report of Item Difficulty Description for CAP English Comprehension Test from Ministration of Education (MOE) in 2014, this reform is aimed to better discriminate the English proficiency of future senior high students and to assess test-takers’ English competence in a more comprehensive way. By increasing the proportion of inferential reading comprehensions, the CAP could assess test-takers’ abilities to infer information which is not literally stated in texts. Making inferences requires high-order thinking ability in which readers must decode details and infer implicit messages beyond literal messages. In sum, the CAP English test measures reading comprehension both as a global construct and as multilevel representations. I am a 7th and 9th grade teacher in a small junior high school in Chiayi City for almost 9 years and the students from the community vary in their academic interest and proficiencies. Along with the reform of CAP, a strong urge arises to improve students’ overall reading comprehension abilities. In the previous English exams of BCT, some students who had taken the BCT exam had claimed that it was effortless to choose the right answers because some reading comprehension questions only required them to recognize some factual clues from the text. Others had claimed they would be able to get high or satisfying scores based on their instinct or common sense. As an English teacher in junior high school where most students find it difficult to survive the newly-reformed English test of the CAP, it is my interest in this study to investigate whether WM could benefit junior high students in terms of making successful and comprehensive understanding about the CAP English comprehension test. Hence, the predictive power of 2.

(15) WM on RC intrigued me to learn about the essential elements which would foster students’ comprehension abilities. Though many studies show the strong correlation between WM and RC, inadequate research has involved Taiwan EFL high school students as participants. The research regarding the interaction of WM and L2 reading have mostly investigated advanced English learners with TOEFL scores 500 or more or elementary students in Holland, Italy, US, Japan, and Mainland China (Alptekin & Ercetin, 2010; Carretti, Borella, Cornoldi, & De Beni, 2009; de Jonge, 1996; Friedman & Miyake, 2004; Sawyer & Harrington, 1992). Some studies have involved pre-readers (Hannon & Frias, 2012), children (Cain, Oakhill & Bryant, 2004), adolescents (Cromley & Azevedo, 2007), and seniors (Hannon & Daneman, 2009). Due to the limited sampling in this field of WM and RC, I think it would be worth investigating the power of WM on reading comprehension by involving my students as participants in studies about WM and RC. Besides the scant studies involving high school students, English is learned as a second language for the target participants in my study. Studies have shown significant differences between first language (L1) and second language (L2) when it comes to the interaction between WM and language performance. It is indicated that the discrepancy or the difference between one’s L1 and L2 working memory capacity (WMC1) decreases when one’s L2 knowledge increases (van den Noort, Bosch, & Hugdahl, 2006; Service Simola, Metsänheimo, & Maury, 2002). Junior high school students begins to gain more L2 knowledge than what they learned in elementary school English courses—English learners start to learn more about using the language in junior high school English. Due to the reasons stated above, it was the researcher’s interest to investigate how WMC interacts with RC performance of students with elementary or intermediate English. 1. Here WM refers to the abstract mental system that is devoted to information maintenance and process while WMC refers to the specific capacity that represents how much mental space individuals possess. 3.

(16) proficiency, particular those junior high school students from my school in Chiayi City. In my English class, I teach students about language usage and language use. I include lots of communicative activities to involve students to become active language users. I have joined many as TESOL workshops or seminars as possible in order to improve my instruction and the efficiency of student’s learning. To involve students in joy of learning English is always one of my expectations in designing my courses. Some students following every step and suggestion in my class sometimes still found it overwhelming to fully understand the challenging comprehension texts in newly-reformed CAP exams. Besides teaching the knowledge about English and including creative teaching activities, I wondered what else I could do to help students feel less frustrated and more confident in RC. As a teacher in high school, I was intrigued by the predictive power of WM on reading comprehension discussed among numerous studies. To date, several studies have shown that WM is a limited-capacity system that serves two functions—to store information temporarily and to process information simultaneously (Baddeley & Hitch, 1974; Baddeley, 2003; Daneman & Carpenter, 1983). To measure the capacity of WM or WMC, a dual-task is often adopted to measure the functions of storage and process and is known as reading span tasks (RSTs). However, studies investigating the influence of WM on RC have adopted different scoring methods and measurements. In other words, the difference of WMC measurement tasks is often not taken into consideration in research. In other words, the factor of different degrees of working memory resources underlying different cognitive tasks is usually overlooked in research (Alptekin & Ercetin, 2009). Some researchers adopt recall-RSTs which require participants to memorize and later to spell the ending words (Harrington & Sawyer, 1992; Miyake & Friedman, 1998; Osaka & Osaka, 1992) while other researchers adopt recognition-RSTs which require participants to memorize and later to recognize ending words of each sentence from a multiple-choice test (Chun & Payne, 2004). These two 4.

(17) types of WM measurement, a recall-RST and a recognition-RST, both require participants to store and to process information simultaneously. Irrespective of WMC measurement types (recall-RST and recognition-RST), according to Turner and Engle (1989), RSTs of all kinds could tap the same variation in reading comprehension. When taking recognition-RST, participants judge the grammaticality of presented sentences and recognize the endings words of the presented sentences from a provided multiple-choice test. In this case, a recognition-RST is to tap the storage function of WMC in which participants identify the externally presented cues. On the other hand, recall-RSTs also tap the storage function of WMC. When taking a recall-RST, participants also judge the grammaticality of presented sentences but they have to memorize and write down the ending words. Hence, in a recall-RST, the ending words in each sentence that participants memorize are internally generated cues for later free recall. In a recognition-RST, participants recognize the ending words from externally presented cues (Unsworth & Engle, 2007). Since recall-RSTs and recognition-RSTs are commonly used in studies about WMC and its influence on RC, this study was aimed to investigate whether these two types of RST would make a difference on the influence of WM on RC. This paper aimed to study the relationship between L2 WMC and L2 reading comprehension in the dimensions of literal and inferential comprehensions when L2 WMC is measured with two measurement task types. The research questions are: (1) Is there a significant relationship between L2 WM recall reading span score and English comprehension accuracy in terms of L2 literal and inferential reading? (2) Is there a significant relationship between L2 WM recognition reading span score and English comprehension accuracy in terms of L2 literal and inferential reading? (3) Will the influence of L2 WM on L2 RC be different if WMC is measured via these two types of RST, recall-RST and recognition-RST? Based on the review of literature, hypotheses regarding the research questions are as 5.

(18) follows: (1) L2 WMC2 measured in a recall-RST and a recognition-RST is not expected to have a significant relationship with L2 literal RC. (2) L2 WMC measured in a recall-RST and a recognition-RST is expected to have a significant relationship with L2 inferential RC. (3) Compared with a recognition-RST, MWC measured in a recall-RST is expected to be significantly correlated to the L2 RC performance.. 2. The capacity of WM is called WMC, which represents the abstract mental concept of WM. And WMC is often measured with reading span tasks (RSTs). 6.

(19) CHAPTER TWO LITERATURE REVIEW As mentioned above, some researchers treat RC as one single construct when it comes to WM and RC. Instead of treating RC as one single construct, the first part of literature review demonstrates the urge to treat RC as a multilevel construct when it comes to the relationship between WM and RC. This chapter first provides theoretical concepts of RC and its multiple levels. Then, the following part presents thorough and relevant literature review concerning WM. Three aspects in the research questions—L2 WM, two commonly used WM tasks along with L2 literal and L2 inferential comprehensions—were well elaborated. Reading Comprehension Overview of Reading Comprehension What makes a good reader a good reader? What exactly happens in readers’ mind during reading processes? Readings are written messages from the text producers and are aimed to be read and comprehended by readers (Just & Carpenter, 1992). Good readers are interactive readers. They interact with texts and writers. They decode the surface information and utilize their own background knowledge in order to form comprehensive, if successfully, and correct understanding. Comprehending a text sometimes seems to be easy and effortless for readers; however, some texts are difficult to apprehend and “demands extensive storage of partial or final product in the service of complex information processing” (Just & Carpenter, 1992, p.122). When readers read, readers basically process information in which the visual information is transformed in successive processing states in forms of visual, phonological and episodic memory until the information is ultimately comprehend in readers’ semantic system (Hannon, 2003; LaBerge & Samuels, 1974 ). Some reading tasks could be complex. Readers use genre structures, linguistic 7.

(20) features and other factors provided by the text producers to build up semantic system or the mental representation. In other words, when comprehending spoken or written texts, listeners or readers must quickly “retrieve some earlier presented words and phrases” (Just & Carpenter, 1992, p.122). During the comprehension process, listeners or readers retain in mind the mental representation or the themes from the previous sentences. Also they might need to generate multi-faceted propositions about what is currently read or comprehended. Readers sometimes comprehension texts literally and sometimes need to infer the implicit messages between lines. Namely, besides readers’ perception of what messages a text literally delivers, readers construct mental representations using their world knowledge and their newly-formed comprehensions about the text at hand (Duke, Pearson, Strachan & Billman, 2011). Thus, successful RC takes efforts and requires robust cognitive strength or abilities. Instead of treating RC as a global construct, this study delves into the multi-level representational nature of RC. As Alptekin and Ercetin (2009) observe, researchers mostly consider RC to be “a global construct, paying little attention to its multilevel representational architecture and also the role played by each level in comprehension” (p.628). As mentioned above, RC of various reading tasks involves different degrees of contribution from surface coding and implicit messages (Kintsch, 1998). Due to the complexity underlying different comprehension levels, there is a need to scrutinize different levels of comprehension when it comes to investigating the relationship between WM and RC instead of treating RC as a whole. Some researchers have highlighted the need for further research to investigate the influence of L2 WM on RC in two dimensions—L2 WM on literal and inferential RCs (Alptekin & Ercetin, 2009; Jeon & Yaashita, 2014). First of all, in research concerning cognitive activities and WM, researchers mostly treat RC as a global proficiency, paying little attention to its multiple levels behind RC. Studies support a positive correlation between WM and RC but little 8.

(21) research in this field considers different levels behind RC (Daneman & Carpenter, 1983; Jeon & Yamashita, 2014; Just & Carpenter, 1992). Thus, RC should not be regarded to be one performance or proficiency as a whole when it comes to research studying the influence of WM on RC. Secondly, the influence of WM on RC would vary with the complexity of the given texts—some texts are easy to comprehend while some texts are demanding to apprehend (Sasaki, 2000). Moreover, some texts require both literal and inferential comprehensions. It is possible that WM is involved in the multiple levels of comprehensions to different degrees; therefore, it would be reckless to acknowledge the influence of WM on RC without investigating the power of WM on different levels of RC. The following part would be devoted to elaborating the multiple facets behind the global construct of RC from Kintsch’s construction-integration model of comprehension. Construction-Integration Model of Reading Viewing reading comprehension as a mental representation, Kintsch (1998) suggests that reading comprehension involves at least three main levels of knowledge representations—surface representation, textbase and situation model. At the surface representation, the readers must decode the words from texts and sometimes parse them into chunks so as to grasp the grammatical relationship among the chunks. At this level of surface understanding, readers might need to understand the referential identity or the implicit causal links based on their linguistic knowledge. A text, once comprehended, becomes a system of interconnected propositions. The next level of textbase representation refers to the combinations of propositions across sentences. Next at the second level of textbase, inside the working space of readers’ WM, readers temporarily hold the propositions which are decoded from the surface representation. This level of information storage is regarded as textbase, and includes both global and local understanding of texts. Local comprehension refers to literal 9.

(22) understanding about texts while global comprehension requires readers to put the literal understanding in a context of the entire texts. Both the global and local comprehensions would serve later as resources for readers to retrieve information from in order to generate a more thorough understanding. According to Kintsch (1998), fluent readers tend to retrieve information from more important propositions which have more connections among meanings of sentences (e.g., lexical decoding, word-to-text interpretation, and syntactic parsing). In other words, more competent readers are better at selecting relevant or important clues from the texts in order to comprehend the texts more correctly. On the contrary, less competent readers usually fail to select important messages and would be easily overwhelmed by too many literal messages. Hence, less competent readers usually are less able to grasp the main ideas of texts. Furthermore, readers extract meanings from sentences, gradually accumulate meanings or propositions from successive sentences and finally generate a coherent discourse understanding. Their abilities to select relevant information and to inhibit minor information are important for this level of comprehension. Last but not least, at the next level of situation model, readers not only understand texts within contexts (textbase) but also activate their background knowledge in order to understand authors’ interests or goals. Situation model refers to readers’ global comprehension which requires readers’ knowledge about linguistic elements and topics of texts. The strong connections of readers’ prior knowledge, experiences, interests or goals are essential for readers to generate situation model successfully. To sum up, surface representation is linguistic proposition; textbase is a symbolic and verbal structure; and the last comprehension level of situation model goes beyond the text (Kintsch, 1998). Like literal comprehension, the surface representation and textbase mean that readers decode the surface or literal meanings of the texts. On the other hand, situation model, like inferential comprehension, requires readers to understand messages 10.

(23) beyond the surface linguistic forms. As Alptekin and Ercetin (2009) indicated, it is highly likely that the influence of WM on RC would vary in power and qualities since different reading tasks require different degrees of cognitive activation and complexities. It could be concluded that it is worth investigating RC as a multilevel representative construct when it comes to WM and RC. Literal and Inferential Reading Comprehensions Literal reading comprehension requires the readers to recognize or recall exactly what is stated from the original text. Inferential reading comprehension, on the other hand, demands the readers to “interpret the author’s meaning through connecting information that is implicit in the text” (Dennis & Barnes, 2001, p.352). Both literal and inferential comprehensions require readers to understand and interact “with the text to different degrees” (Dennis & Barnes, 2001, p.352). However, literal comprehension is independent of higher-level processing because they differ in their underlying constructs (Hannon, 2000). In terms of complexities involved in RC, inferential reading comprehension was found to be more demanding than literal reading comprehension (Rupp et al., 2006; Sweller, 1994). Studies show that different cognitive tasks put different demands on readers’ WM in terms of the intrinsic cognitive load required by each task (Gough et al., 1996; Hannon, 2000). Hannon (2000) further argues that literal and inferential comprehension exhibit weak correlation with each other because these two types of cognitive actions are actually two independent variables. Literal reading comprehension is mostly characterized to be automatic recognition of information stated literally from texts and therefore is less demanding than inferential reading comprehension. As such, literal reading comprehension means that readers fully understand what messages authors intend to deliver. But at this level reads might not be able to generate further understanding beyond the text itself. From this perspective, literal reading comprehension is usually regarded as an inability to deeply understand texts and 11.

(24) thus is different from inferential reading comprehension (King, 2007). The difference between literal and inferential comprehension is clear in the exemplary sentences suggested by Duke et al., (2011): 1. Roberto desperately wanted to buy a bicycle. 2. He took an after-school job sweeping out the bodega around the corner from his family’s apartment. Literal comprehension requires readers to understand the literal message, including connecting pronouns to the antecedents. In the sentences above, successful literal comprehension means readers know the “he” in the second sentence refers to “Roberto” in the first sentence. At the level of inferential comprehension, readers use their world knowledge or to make logical connection among the text messages. In the whole process of reading, they not only apply relevant and useful background knowledge but also have to keep in mind the mental representation of what they have read. It could be inferred that Roberto’s desire for a new bicycle makes him work part-time at a bodega, a storehouse for wines or a small grocery store downtown. Furthermore, readers would infer that Roberto must be diligent and self-determined for he works for what he wants with his bare hands, because the word, bodega, reminds readers of their background knowledge about neighborhood’s grocery store. In general, writers usually do not literally state the motive behind characters’ actions. Instead, they expect readers to use their world knowledge or experiences to infer the personalities hidden behind characters’ actions (Duke et al, 2007). However, under certain circumstances, readers might comprehend literal messages successfully but do not make correct inferential comprehension. It would be worth scrutinizing how WM interacts with underlying facets beneath RC before it is concluded that WM has significant influence on RC. Thus, there is an urgent need to investigate the influence of WM on literal and inferential RC. All in all, instead of regarding reading 12.

(25) comprehension as a global construct under the influence of WM, the current study is aimed at investigating the correlation between L2 WM and L2 RC in two dimensions—literal RC and inferential RC. Working Memory Capacity (WMC) What is WM? As it literally implies, individuals’ memory is working and functioning in an active and available state in order for individuals to fluently cope with immediate cognitive tasks. It is active and not as static as short-term memory (STM3). Moreover, it is more than an ability to store information. Instead, Baddeley (2002) has suggested WM acts more like a cognitive skill or ability. It helps people temporarily store and process information in an efficient way in order to accomplish complex cognitive tasks, like learning, reading and reasoning. In order to understand the underlying constructs when it comes to the relationship between WM and RC, this part of literature review was mainly dedicated to the theoretical constructs of WM. First of all, working memory has been regarded as a limited-capacity system that manipulates and stores information. It is regarded as a central component in human learning activities by cognitive neuroscientists. Scholars haven not come to an agreement about the specific definition of WM; however, there is no doubt that Baddeley’s multi-component model is one of the most influential (Miyake & Shah, 1999). According to Baddeley’s multi-component WM model (Baddeley, 2000, 2003), WM is a multi-component system which is composed of one supervisory attentional component (central executive) and three other subsystems—one processing phonological memory (phonological loop), anther processing visual memory (visualspatial sketch pad) and the other processing information related to long-term memory (episodic buffer). The following part included the historical background of WM, Baddeley’s multi-component. 3. STM stores individuals’ moment-to-moment thoughts and perceptions and the content endures over a short term only if individuals try to rehearsal (Craik & Lockhart, 1972). 13.

(26) model of WM, and other renowned theories about WM. Historical Background of Working Memory and Its Multiple Components Historical Background Atkinson and Shiffrin (1968) proposed the two-component model of STM and long-term memory (LTM4): the former served as an antechamber to a durable system that stores and processes information over a long period of time—LTM. This two-component model was unquestionable until Shallice and Warrington (1970) learned that STM performances of neuropsychological patients still remained functioning even with the damage to their medial temporal lobes which had been believed to contribute to STM. It implied that STM was not one component which simply stored information but a multileveled system with more than one storing functions. It led to the assumption, later verified and empirically tested by Baddeley and Hitch (1974), that STM was composed of three components—“one was a very efficient secretary, another a taxi driver, while a third ran a shop” (Baddeley, 2003, p. 190). They broadened the concept of verbal short-term memory by including visual and spatial temporary memory (visuo-spatial sketchpad ), a range of control processes (central executive) that accounts for selection and implementation of strategies and a serial ordered verbal memory (phonological loop) (Logie, Osaka, & D’Esposito, 2007). The most important component, the “master system” of central executive, is supplemented by the other subordinating “slave systems”—phonological loop, visuo-spatial sketchpad and episodic buffer (Gathercole & Baddeley, 2014, p.4). Baddeley borrowed the term, master-slave system, from control engineering. It refers the WM characteristics of taking charge and maintaining information though he later discovered that the slave system of phonological loop is also capable of “providing a means of action control” (Gathercole & Baddeley, 2014, p.11) 4. Information stored in long-term memory (LTM) is stable and remains over a long period of time. It is easily accessible with meaningful retrieval structures from short-term memory and working memory. (Ericsson & Kintsch,1995). 14.

(27) (See a clear diagram of WMC as a multiple-component system in Fig. 2.1).. Figure 2.1.. The multiple-component system of WM. Since the idea of WM came from STM, it is worth clarifying the difference between STM and WM regarding the definition and functions during the cognitive process? STM stores individuals’ moment-to-moment thoughts and perceptions and the content endures over a short term only if individuals try to rehearsal (Craik & Lockhart, 1972). In comparison with STM, WM is an immediate and limited mental workspace that manipulates cognitive information which is temporarily stored. Besides it is responsible for a wide variety of complex cognitive activities, such as “verbal reasoning, comprehension, long-term leaning” along with language learning and reading (Baddeley, 2003; Baddeley, 2007, p.5). While WM stores and processes information through sensory stimuli from reading texts for instance, it mediates among perceptions, LTM and actions. The process of mediation involves retrieving relevant information from LTM (Baddeley, 2012). The information undergoing manipulation in WM will be either forgotten or stored inside LTM (Rai et. at, 2011). The active process in WM takes place mainly in the prefrontal context and could be influenced by emotional context, such as anxiety (Gray, Braver, & Raichle, 2002). More importantly, WM is limited by individuals’ different 15.

(28) capacity and the process time of completing cognitive tasks. Unlike STM, which passively stores information, WM not only stores but also actively makes connection between the current information and old knowledge from LTM while individuals strive to apprehend the cognitive tasks on a grand scale (Jerman, Reynolds, & Swanson, 2012). Readers connect what they know from current texts to any relevant information they have already known. At the same time readers move toward the goal of comprehending the reading tasks at hand effectively. As Baddeley suggests (2012), when it comes to the function of multi-component working memory model, STM could be regarded as “a simple storage of information, contrast to long-term memory” (Baddely, 2012, p.4). Namely, STM stores immediate information, WM actively delegates necessary information, new or old, and relevant background knowledge from LTM in order to understand reading tasks. Recent evidences suggest that WM goes beyond the function of information storage (Baddeley, 2012). Children in a study by a renowned Russian psychologist Alexander Luria on investigating kids’ learning process were found to gradually learn independently by combining their background knowledge and new information from new texts when they were fully engaged. In this case, children demonstrate progress in learning for they held new information temporarily and made meaningful connection to background knowledge. The importance of executive attention and the dual function of WM were emphasized in that study. The information stored in the slave systems of WM lasts for a short period of time, “unless constantly rehearsed” or delegated in task performance under the control of central executive (Jerman, Reynolds & Swanson, 2012, p.144). It suggests that successful cognitive tasks require not only the temporary storage of current information but also the function of the central executive to manipulate the task-relevant information so as to accomplish complex cognitive tasks. Phonological Loop 16.

(29) These three slave systems all have their own specialized missions in order to help individuals temporarily store and process information. One of the components, phonological loop, is a temporal verbal-acoustic storage system which is in charge of immediate retention of sequences of digits, or “verbally coded information” (Gathercole & Baddeley, 2014, p.4). It “comprises a brief store with a means of gathering information by vocal and subvocal rehearsal” (Baddeley, 2012, p.7). Factors, such as word length, irrelevant sound effect and verbal suppression, would influence how people remember information by using sound clues. To prove the significance of phonological loop in learning, Adams and Cathercole (1995) compared the recall of nonwords—loddenapish and contramponist. The latter nonword had similar word structures and sounds in comparison with the former one. It turned out that learners could remembered the word, contramponist, better since the underlying English structure of words were matched to learners’ prior knowledge about English words. The results showed that WM plays an important part in vocabulary acquisition with resources from the knowledge in LTM. In terms of reading comprehension, awareness of phonological structures in words is a primitive predictor of word recognition performance (Bishop & Adams, 1990; Deavers & Brown, 1997). If readers’ pholonogical loop in WM functions well, readers would have sufficient knowledge about the relationship between graphemes and phonemes, which greatly helps reader to understand texts better. Also the ability to fluently recognize words alleviates readers’ cognitive loads, which spares more cognitive space in WMC for further demanding cognitive activities. Hence, Rapala and Brady (1990) suggests that high-level cognitive tasks would be difficult if readers fail to have ideal control of phonological processing. Taken together, the function of phonological loop enables readers to read fluently and correctly, which makes phonological loop important and essential for reading comprehension and other cognitive tasks. 17.

(30) Visuo-spatial Sketchpad On the other hand, visuo-spatial sketchpad is viewed as a parallel visual subsystem to phonological loop which mainly maintains and manipulates spatial or visual information. Readers recall items or messages easily if they could recall the specific location in the source text—human brains remember things using clues of sounds, visual and spatial clues printed in paper for example. In vocabulary learning, children learn words faster if visual clues of words are provided. Baddeley (2012) indicated that duration of word recall was facilitated with visual aids. Central Executive Last but not least, all the cognitive behaviors and processes are controlled by central executive. It is regarded as “a purely attentional system” and also the most important component in WM (Baddeley, 2012, p. 14). Central executive functions as a shop owner or a company CEO that actively and automatically makes decisions and performs functions under human sub-consciousness (Baddeley, 2001, 2003, 2012). Individual receive new information through sensory system—hearing, seeing, and reading. The understanding of cognitive tasks at hand requires the relevant and helpful knowledge from LTM. Individuals generate thorough understanding by integrating old and new information, and central executive is greatly responsible for this process of information manipulation (Baddeley, 1996). Besides the function of controlling information, central executive is also responsible for inhibition which is essential for successful RC. Inhibition refers to the ability to select relevant information and ignore the irrelevant messages. When individuals deal with overwhelming information from complex cognitive tasks, such as inferential comprehension, their ability to inhibit irrelevant information is quite important (Kane et al., 2001). Simply put, central executive is responsible for coordinating other WM components and is also in charge of controlling information. The efficiency of central executive and the demands or abstractions vary inversely. Because 18.

(31) WM is a limited-capacity system, demands from each cognitive task would compete for mental space in WMC with WM efficiency. Thus the more demanding the current tasks are, the less efficient WM would be. Episodic Buffer Later in a case of an amnesic patient, Baddeley (2000) introduced the last and fourth component, episodic buffer, which plays the role of “binding together information from a number of different sources into chunks or episodes” and served as a buffer to “combine information from different modalities into a single multi-faceted code” (Baddeley, 2003, pp. 191-203) (See fig. 1 for WM as a four-component system). Episodic buffer is able to convert received stimuli into meaningful chunks meaningfully and creatively. This binding function allows readers to imagine something new, such as an “ice-hockey-playing elephant” and recall new information easily (Baddeley, 2012, p.16). Also it consumes less cognitive space in WMC without excessive focus on linguistic messages, which makes the integration of information work more smoothly in WM. Collectively, visual, phonological and semantic information is integrated by central executive in the workspace of episodic buffer. Episodic buffer can be viewed as a buffer, or a booster station, between LTM and central executive (Baddeley, 2000, 2003). Theories Concerning WM and Reading Comprehension Working Memory and Attention—Engle and Cowan’s Model Engle and his colleagues suggest that difference in individuals’ WMC is mainly about their ability to control their own attention to the tasks, rather than how much information ones can store in mind (Engle, Tuholski, Laughlin, & Conway, 1999). Readers’ attentional control refers to the ability to focus attention on task-relevant information so as to shift information between different levels of mental activities (Engle & Kane, 2004). Besides the ability to focus attention, Engle and Cowan emphasize the importance of the ability to inhibit irrelevant information. Inhibition means the ability to 19.

(32) ignore irrelevant interference, including information from the tasks at hand or other possible disturbance, such as fatigue. If inhibition functions well, individuals would be able to successfully withdraw relevant information or background knowledge from LTM and neglect distractions. From this perspective, WM is more like the efficiency or the ability to execute attention, rather than the ability to maintain or remember information for a short period of time (Kane et al., 2001). Studies have indicated that these two elements in WM (abilities to focus attention and to inhibit distraction) are more important than readers’ space or capacity of temporarily holding memory (Engle, 1996, 2002; Engle et al., 1999; Kane et al., 2001). In other words, WM is regarded as a multi-component system that actively holds information when individuals process immediate information along with possible distractions (Kane, 2005). To understand a text correctly, readers must focus on immediate information consciously and attentively and inhibit irrelevant messages. By doing so, they would be able to formulate mental representation of the very text correctly. This theory has experimental supports. Studies have shown that persons with higher WMC (working memory capacity) are better at controlling their attention in comparison with those with lower WMC (Kane et al., 2001). Some possible distractions might not come from the outside environment but from the cognitive activities or the texts themselves. When individuals find what they comprehend from the former text against what they understand in the later text, those with high WMC would eliminate the irrelevant information and move on searching for other useful or coherent clues. Those with higher WMC can more successfully and automatically resist interference of all kinds under normal situations. Having better function of inhibition over information enables readers with high WMC to process information more efficiently and more successfully. However, those with low WMC might either not be able to discriminate irrelevant information or overwhelm themselves with minor information. Eventually, without 20.

(33) inhibition, readers with low WMC would comprehend texts less ideally than those with high WMC. Working Memory and Long-term Memory—Ericsson and Kintsch’s Model Though the literature has emphasized the importance of WM in terms of RC, there is doubt about how some demanding cognitive activities, such as inferential RC, can be processed within limited space storage (Ericsson & Kintsch, 1995). WM functions like an active mental workspace to process and synthesize information when individuals encounter cognitive tasks. Though the influence of WM sounds invincible, it has its own limitation—a rather limited cognitive space. Thus adequate space is necessary for demanding tasks. WMC could support in dealing with cognitive tasks only if the space is not consumed excessively. To compensate the restriction, according to Ericsson and Kintsch (1995), background knowledge from LTM is essential. WM could be expanded with resources from background knowledge from individuals’ LTM (Ericsson & Kintsch, 1995). Baddeley (2002) focuses on interaction between tasks at hand and the four WM components. On the other hand, Ericsson and Kintsch (1995) focus on the interaction of WM and LTM when it comes to understanding texts at hand. The predictive power of WMC on cognitive tasks not only depends on the short-term working interface of storing and processing information but also relies on the resources from individuals’ background knowledge of LTM (Ericsson & Kintsch, 1995). As mentioned above, while WM stores and processes incoming information through sensory stimuli, from reading texts for instance, it mediates among perception, LTM and actions, which includes retrieving relevant information from LTM and depends on retrieval structures in memory system (Baddeley, 2012). The retrieval structure, a stable structure in memory, enables readers to link their knowledge to the immediate information at hand. The information, like propositions from 21.

(34) reading, would become cues. Those cues are available items in WM that could activate or stimulate memory inside LTM. When individuals have robust retrieval structures, individuals would be more capable of facing demands of complex cognitive tasks like experts, making successful reading inference for example. According to studies, retrieval structures, which support participants’ performance of cognitive tasks and its relationship with WMC, come after some prerequisites (Chase & Ericsson, 1982; Ericsson & Kintsch, 1995). The first prerequisite is that a large body of knowledge and problem-solving pattern concerning the undergoing information is required. The second prerequisite is that individuals must be able to anticipate or be prepared for the coming demands for information retrieval via recognizing the tasks. Recognizing the tasks and knowing the types of cognitive thinking pattern might help individuals comprehend and solve the tasks efficiently. Last but not least, the third prerequisite is that individuals must have his or her own encoding strategies, developed through continuous practices, so as to fluently retrieve information from LTM and store selective information from tasks at hand into LTM. Ericsson and Kintsch (1995) suggest that WMC has great influence on RC and so is background knowledge of domain field. Likewise, Brown and Hulme (1992) point out that L2 readers’ comprehension would be improved when they have adequate background knowledge about the language or the text content. When individuals are involved in L2 input, their L2 WMC tends to be reduced in comparison with their L1 WMC. The reduction of their WMC is caused by the lack of adequate LTM contributions to L2 WMC. It implies that LTM fuels the work efficiency of WM. Knowledge in domain field is stored in an organized structure in LTM. This organized structure of domain knowledge makes it highly accessible when readers withdraw relevant knowledge to support their comprehension. Namely, while the relevant knowledge is stored in LTM, LTM would be more accessible and supportive to RC and other cognitive activities (Ericsson & Kintsch, 22.

(35) 1995). In brief, three elements are mutually beneficial—WM, LTM and RC. Capacity Constrained Comprehension Theory and Working Memory—Just and Carpenter’s Model Though WM plays an important role in language comprehension as mentioned above, it has its own constrains. Daneman and Carpenter (1980) have suggested that WM has stronger relationship with high-order cognitive activities, such as language comprehension, reasoning, and problem solving, rather than with simple cognitive tasks. During the process of comprehension or cognitive activities, WM can be viewed as a “pool of operational resources that perform the symbolic computations and thereby generate the intermediate and final products” (Daneman & Carpenter, 1980, p.122). Listeners or readers, according to Just and Carpenter (1992), will “construct and integrate” ideas from the spoken or written discourse (p.122). During this process, listeners or readers interpret the immediate information and temporarily store it as a meaningful mental representation—WM plays a crucial role in storing and processing information. Furthermore, according to Just and Carpenter’s WMC theory, comprehension performance “declines with an intrinsic memory load beyond one’s capacity, such as retaining information across successive sentences of a text, or an extrinsic burden” (Just & Carpenter, 1992, p.135). And the decline is even more severe for low-span readers. Due to the capacity limitation, readers with low WMC need to have more reading time to process overwhelming information and are more likely to make wrong comprehension. Readers with high WMC, on the other hand, might also have poor comprehension even if they have higher WMC. For readers with high WMC, reading skills, such as skills to select useful clues, are important—their WMC would be consumed extremely when readers tend to interpret sentences from multiple angles and to generate more than one syntactic ambiguity. Briefly speaking, based on Just and Carpenter’s WMC capacity theory (1980), how readers comprehend a text depends on their WMC. When the tasks at 23.

(36) hand exceed or overload readers’ WMC, “the storage and computation” functions of WM would decline (Just & Carpenter, 1980, p.124). To briefly summarize these theories mentioned above, WM is responsible for temporarily storing and processing information, which makes WM essential for a wide range of complicated cognitive activities. Some (Daneman & Carpenter, 1980) propose that each individual has limited WMC while some other scholars (Anderson, 1983) define that WM is not necessarily limited in capacity. Responding to both different perspectives, Baddeley (2002) suggests in his latest review on working memory that WM should be regarded as a theoretical “framework for the analysis of the contribution of working memory to languages” (Baddeley & Hitch 1974; Baddeley 2002; Gathercole & Baddeley, 2014, p.2). Despite the nuance of differences among WM theories and different opinions on its influence on cognitive activities, the researchers mentioned above all share the idea that WM helps store and process information. Also, with WM, individuals would be more able to focus and switch attention when dealing with immediate cognitive tasks—either to make comprehension or to discard irrelevant information (Baddeley, 2007; Engle, 2002). It is generally agreed that during the working process in WM, knowledge from LTM (e.g., grammar knowledge, vocabulary knowledge, phonological knowledge) will be retrieved and withdrawn into an active mental workspace. Without WM, RC is impossible (Jeon and Yamashita, 2014). Briefly speaking, WM definitely influences the quality of cognitive tasks regardless of the differences among WM theories. WM and RC—L1 and L2 Though Fodor (1986) commented on the WM model of Baddelely that WM was “unhelpful and neuropsychological implausible” (Baddeley, 2012, p.7), WM is still influential on cognitive tasks, like RC (Alptekin & Ercetin, 2011; Nassaji, 2002; Kane, Bleckley, Conway, & Engle, 2001; Unsworth & Engle, 2007). Responding to this, 24.

(37) Baddely admitted the incompletion of his model and urged others to complete his WM theory with more studies. Instead of focusing on flaws of his WM theory, Baddeley (2012) suggested that researchers should consider WM as “a relatively loose theoretical framework rather than a precise model that allows specific prediction” (p.7). Even though the WM model is not concrete and specific, Baddeley (2012) suggests that central executive should play an important role of information manipulation when it comes to cognitive activities and WM. When more than one tasks or cognitive activities are required to be processed, the resources or space would be compromised under the competition among the needs to deal with tasks at hand. In the long run, it increases the burden on central executive and slows down the efficiency to complete the immediate task. Due to this, it makes WMC related to cognitive tasks and the efficiency of accomplishing them. WM is highly related to RC basically because WM helps maintain immediate information in mind. More importantly, WM enables readers to construct meanings from multifaceted resources under conditions of full engagement (Tierney, 1990). In terms of engagement, one important construct of WM is the automatic operation of inhibition, which controls the content or mental representation of WM by preventing irrelevant stimuli from overloading WM capacity (Borella & De Beni, 2008; Engle, 2001; Kramer, Humphrey, Larish, & Logan, 1994; Rosen & Engle, 1998). From this perspective, it is clearly that both RC and operation of WMC tap the same construct, which explains the high correlation between the two shown in numerous studies. Also, Daneman and Carpenter (1980) found that the result from RSTs, individuals’ WMC (working memory capacity), is able to predict participants’ prose comprehension skills in their college participants. They continued to find more details of the ways which WMC seemed to underpin components of WMC, such as the ability to make inference and to extrapolate beyond the given literal information. 25.

(38) Furthermore, Just and Carpenter (1992) noted that the dual functions of storage and process in WM make humans capable of conducting various linguistic (speech, sound, and visuo-spatial) and conceptual tasks (semantic and episodic information) smoothly and automatically. During the automatic cognitive process, comprehending a text for instance, relevant information from LTM is retrieved and would be combined with the dynamic newly-received information in WM so as to establish a meaningful mental representation as a whole (Alptekin & Ercetin, 2009; Daneman & Carpenter, 1983; Daneman & Merikle, 1996; Walter, 2004). And this automatic cognitive process makes WM relevant to RC performances. When it comes to L1 or L2 in terms of WM and RC, studies show that L2 RC is more relevant to L2 WM than to L1 WM (Daneman & Hannon, 2007). It is suggested that cognitive resources underlying L1, L2 and even L3 are closely related to each other. L1 WMC is found to be strongly or moderately correlated to L2 WM capacity. The correlation was both over .70, which suggests a strong relation force (Osaka & Osaka, 1992; Osaka, Osaka & Groner, 1993). Other studies found a moderate correlation between L1 and L2 WM capacity ranging from .39 to .68. That led to the burgeoning studies concerning the relationship between WM and L1/L2 RC for that WM is said to be a good predictor of RC and scant studies investigate the cross-language interplay of WMC and comprehension (Danemen & Carpenter, 1980; Harrington & Sawyer, 1992). As indicated in several studies, L2 WM capacity is strongly correlated to L2 RC rather than L1 WM capacity to L2 RC (Chun &Payne, 2004; Walter, 2004; Miyake & Friedman, 1998). Hence this paper is aimed at studying English as L2 RC under the influence of L2 WMC. Measurement for WMC—Reading Span Test (RST) Individuals’ WMC is usually measured with a dual-task which requires participants to complete both processing and storage tasks. Span tasks for measuring WMC usually 26.

(39) “include a dual-task paradigm which combines a memory span measure with a concurrent processing task” (Alptekin & Ercetin, 2010, p.206). Dual-tasks are designed to measure the two functions of WMC—to temporarily store and process information at the same time. It comes in a variety of types, such as operating span, counting span, and reading span. The WM theory emphasizes the functional importance of “an immediate-memory system that could briefly stores a limited amount of information in the service of ongoing mental activities” (Conway et al., 2005, p.769). The design of dual-tasks could assess individuals’ ability to store and process information at the same time. In comparison with other span tasks which do not include readings in procedure, a RST, according to research would have better validity and reliability when it comes to the relationship between WMC and RC (Conway, et al., 2005; Friedman & Miyake, 2004). What is a dual-task precisely? In completing a RST, participants need to complete two tasks simultaneously. According to Conway et al. (2005), the WM system would fail to demonstrate its function fully if it is encountered with one single task of simply storing information or rehearsing factual information. Instead of one single task, dual-tasks or complex tasks have been developed to measure WMC and the efficiency of central executive inside WMC. Complex tasks demonstrate to-be-remembered stimuli, such as letters or numbers, and those stimuli are spread between interfering components, such as reading sentences (Daneman& Carpenter, 1980) or problems to solve (Turner & Engle, 1989). Complex tasks are to measure WMC and are more related to high order cognitive tasks, such as making reading inference (McCabe et al., 2010; Unsworthy & Brewer, 2009). Regarding the trade-off between processing and storage as interdependent components operating inside WM, the tasks are assumed to show that “increase in the amount of processing demands leads to a decrease in the number of storage items and vice versa” (Alptekin & Ercetin, 2010, p.206). The literature has emphasized the importance of WM when it comes to its influence 27.

(40) on RC; however, there is little consensus on the measurements and assessments of WM (Friedman & Miyake, 2004; Juffs, 2004; Waters & Caplan, 1996). Difference WMC tasks have different procedures or tasks (Waters & Caplan, 1996). For instance, a recall-RST requires readers to judge the grammaticality of sentences and write down cues afterwards (Harrington & Sawyer, 1992; Walter, 2004). A cognition-RST requires readers also to judge the grammaticality of sentences but to identify ending words of sentences from a list of provided options. A recognition-RST measures the participants’ abilities to recognize the ending words from the provided options or “externally presented retrieval cues” (Unsworth & Engle, 2007, p.112). Unsworth and Engle (2007) suggest that performances in recognition-RST are mainly driven by two independent mechanisms: fast-pacing and automatic process of recognition based on familiarity. Likewise, Alptekin and Ercetin (2009) pointed out that the performance of recognition-RSTs could be “partially contingent on a strategically controlled search process of long-term working memory or on the automatic retrieval of information from long-term working memory” (p.635). During the process of completing a recognition-RST, participants can quickly target ending words from externally presented cues, which is called the automatic process of recognizing words. It operates based on participants’ recognition of familiarity or words which they have read from the RST. In the dual-task of grammaticality judgment in recognition-RST participants need to respond to externally presented retrieval cues. The provided list of sentences or the multiple-choice questions provide participants with the chance to retrieve relevant information from their long-term working memory by recognizing cues automatically. However, when completing a recall-RST, participants remember the ending words from memory, which resembles free-recall tasks. To complete a recall-RST, the participants recall their memory of ending words from their internally generated cues. Short-term working memory is activated in the controlled search process inside long-term 28.

(41) working memory. Instead of automatically recognizing words, participants internally generate retrieval cues by using relevant cues to the presented ending words with knowledge from their long-term working memory. Altepkin and Ercetin (2009) suggest that there is a qualitative difference between recall-RSTs and recognition-RSTs in terms of their different cognitive constructs. A recall-RST requires the participants’ abilities to inhibit irrelevant information and to focus on relevant information in order to retrieve items using internally generated cues. Thus, free-recall tasks or recall-RSTs would be more demanding than recognition-RSTs and the cognitive demands posed by these two different RST types on individuals should be different. It is expected to see the influence of WM on RC vary when WMC is measured with these two RST types. Regardless of RST types, researchers proposed that WM span tasks of all kinds all account for similar or the same variance in comprehension (Turner and Engle, 1989). However, Chun and Payne (2004) found no significant correlation between WMC and RC when using recognition-RSTs in their research. According to Chun and Payne (2004), the cognitive tasks in recognition-RST in which readers recognize correct options of ending words from RST do not correspond to the construct behind RC. To investigate the relationship between WM and RC, studies should consider the factors of different WMC tasks and dimensions of RC (Chun & Payne, 2004). That is, different amounts and types of working memory cognitive resources involved in different cognitive tasks is usually not taken into consideration in studies (Alptekin & Ercetin, 2009). Given the scant consensus concerning which RST is better to measure WM, the investigator was interested to find out whether the influence of WM on RC would vary when WM is measured with a recall-RST and a recognition-RST. To sum up, this paper was aimed to find out the relationship between L2WM of English learners in high school and their abilities to make literal and inferential comprehensions. 29.

(42) 30.

(43) CHAPTER THREE METHODOLOGY Participants The participants in this study were 190 students from 7 classes in a Chiayi City Junior High School with a total of 37 classes. The participants ranged from 8th graders to 9th graders (aging from 12 to 15 years old). They had learned English since 1st or 3rd grade in elementary school (for about seven to nine years). These 190 students in the current study did not have similar language proficiency. There were some obstacles for the feasibility of the study. First of all, under the restriction of course schedules at school, it was unlikely to withdraw students of the same language proficiency from my school, which would have stopped them from attending their scheduled classes. In order to include as many participants as possible in the research, I asked as many classes as possible to participate in the study and only the teachers of these 7 classes agreed to spare some periods for conducting this experiment in class. Secondly, the number of students passing the GEPT elementary level in the researcher’s school was not many—there were about 30. If the participants were only those 30 students passing the GEPT elementary level, the number of participants would be too few and would sabotage the power of analytic generalization. Thus, the study included these 190 students from the 7 classes and the number of participants was appropriate for analytic generalization. Therefore this convenience sample of 190 students from 7 classes was finally invited as participants. Although the language proficiency among the participants varied in levels, the individual difference was statistically canceled among one another under the number of 190. Also Chinese counterparts for some difficult English words were provided in the RC test in order to avoid the reading difficulty from unknown words. As for the grouping criteria (recall-RST and recognition-RST), due to the limitation of course schedules at school, participants had to attend their own classes. The 31.

(44) investigator decided to group entire classes into same WM group in order to carry out the study. Thus, no specific grouping criterion was applied—the participants from three classes accomplished a recognition-RST and a RC test; the participants from the rest accomplished a recall-RST and a RC test. The test of RC contained four passages with five literal and five inferential RC questions, and was designed to assess participants’ ability to read literally and inferentially. Instruments The instruments of this study included (a) recall- and recognition-reading span test (RST) to measure working memory capacity (WMC); (b) a RC test with literal and inferential comprehension questions. Each participant completed the whole procedure in the following order: a WMC RST and a RC test. 104 participants took a recall-RST and a test of RC; the other 86 participants took a recognition-RST and a test of RC. The RST was administered in one class period of their English class in the classroom and the RC test was done in another period. Reading Span Task (RST)—Measurement of Working Memory Capacity (WMC) The reading span task (RST) or a working memory test developed by Daneman and Carpenter (1980) is the most commonly used tool for WMC. Other span tasks for WM include operation span and counting span. Among all types of WM tools, RSTs are found to be highly related to RC. Daneman and Carpenter (1980) define RSTs as dual-tasks that require participants to complete two simultaneous missions. RSTs require participants to “fulfill both processing and storage requirement” simultaneously (Conway et al., 2005, p.581). Namely, two tasks have to be done by the participants simultaneously: to read a series of sentences and to recall or recognize the final words later on. Participants’ WMCs were the maximum number of ending words that he or she recalled or recognized from multiple-choice options. A RST adopted from the study of Harrington and Sawyer (1992) was administered in 32.