Summary of the Study

This study tried to duplicate the influence of WM on L2 RC, especially the literal and inferential RC.

Besides, since the difference of WM measurements (RSTs) has often been neglected in studies, it was one of the interests to testify whether the influence of WM would vary mainly when WMC is measured with a recall-RST and a recognition-RST. The result of the current study indicated that L2 WMC of the participants in the current study did not have significant correlations with their literal and inferential RC. The final results regarding two types of RSTs showed that the influence of WM on RC did not vary to a significant level. Thus, the findings of the study could be interpreted with the caveat that the efficiency or performance of one’s cognitive tasks depends not only on WMC but also other relevant factors, such as reading skills, background knowledge and readers’

devotion to reading. The positive influence of WM is prone to deterioration and it was not easy to relate WM to RC. The lack of stability of WMC suggests that one person’s WMC might change from task to task. The lack of stability in individuals’ WM might also happen in the case of taking RST and a RC test; one could have high WMC when taking a RST but demonstrate lower WM when taking a RC test. Their poor reading skills and inadequate background knowledge consume his or her WM. It remains uncertain whether recall-RSTs could better assess WM than recognition since there are two sides of

theoretical discussion as mentioned above.

In sum, L2 WMC is not correlated to L2 literal RC based on the results of this study and other previous research (Altepkin & Ercetin, 2010; Chiappe, Hasher & Siegel, 2000;

Daneman & Hannon, 2007). In other words, the function of WMC works as a highly-efficient system of information manipulation. Thus it has little influence on

66

automatic recognition of information without the need to further syntheses or inference, such as L2 literal RC.

When it comes to the robust correlation between WMC and inferential RC, it is difficult to testify with simple one-time research design in the current study—a WMC RST and a test of students’ comprehension performance (Altepkin & Ercetin, 2010;

Anderson et al., 1996; Baddeley, 2012; Conway & Engle, 1994; Daneman W. & Hannon, 2007). First of all, individuals’ WMCs change. WM is not a stable strait or characteristic for an individual reader. Some readers might demonstrate high WMC in one task but lower WMC in another task. Their reading skills, background knowledge or attention carry severe influence on the influence of WMC on L2 inferential RC. Seemingly, WMC is unfortunately limited in certain ways regarding the power of WMC on learning or cognitive activities (Gathercole & Alloway, 2007). Distractions or overload of information that happen during the process of undertaking cognitive tasks would definitely weaken the influence of WMC on individual’s learning performance, like RC (Blalock & McCabe, 2011; Conway, el at., 2007; Just & Carpenter, 1992; Kane et al., 2001). The influence of WMC on cognitive tasks would deteriorate when individuals are faced with far too demanding tasks. Too demanding or difficult tasks would reduce the mental space of WMC, even for readers with high WMC. In this case less cognitive space is left for high-level cognitive tasks.

Secondly, other relevant factors are participants’ background knowledge and reading skills as mentioned in the previous discussion section (Daneman & Carpenter, 1980;

Ericsson & Kintsch, 1995). Those relevant factors, if not controlled, might interfere with the duplication of WM being predictive of RC. How the previous studies successfully correlate WM to RC without controlling more variables is unknown and intriguing to find out. However the correlation between WM and RC was difficult to be confirmed in a simple one-time procedure like the research design of this study. More thorough research

67

design in future studies is needed to testify the predictive power of WMC on L2

inferential RC. Other variables, such as background knowledge and reading skills, can be included as independent variables.

When I first worked on the literature and research design of this study, I reviewed lots of studies supporting the predictive power of WM on RC. I therefore had come to a misunderstanding that those who performed well in RST, or readers with higher-WMC, would always end up with better comprehension about any text. In the data analysis phase of this study, I felt frustrated and devastated to find out the statistics did not show any significant correlation between WMC and RC. Going through lots of literature reading and struggling with myself, I suddenly realized that WMC changes to be higher or lower under certain circumstances or under different reading tasks (Daneman & Carpenter, 1980;

Ericsson & Kintsch, 1995). Those who have ever gained higher scores in RST might demonstrate malfunction of their high WMC. As mentioned above, WMC is not a stable trait for individuals, like gender or nationality, and WMC might change over time, place or under other situations. A reader could demonstrate high WMC in one text but lower WMC in another. Reading skills and resources form background knowledge would account for the changes of WMC in the very reading text and also for the performance of RC. Many researchers agree that WMC plays an essential role in inferential RC, but as previous discussion suggests a person’s WMC changes with immediate texts (Daneman &

Carpenter, 1980; Ericsson & Kintsch, 1995). In previous studies that support the influence of WM on RC, participants take a RST and a test of RC. Chances are the condition of one’s WMC when they take RST is not the same as in comparison with his or hers WMC on another later test of RC. To rule out the possibility whether WM might vary from task to task, future studies could try to gain the WM and RC performance in on task, rather than comparing the WMC in one task (a RST) and their performance in another task (a RC test) . On the other hand, doubts are how the previous studies could successfully

68

relate the WMC results from RST to participants’ RC performance. My reasonable speculation is the stable reading ability and reading condition, whether participants’

attention or their background knowledge, are quite the same in both RST and the follow-up RC test.

In sum, WM can positively influence the outcome of demanding cognitive activities in that it is responsible for information manipulation which requires skilled operation to integrate new and old knowledge under attentive situation. Scholars suggest relevant factors that would deteriorate the influence of L2 WMC on L2 inferential comprehension, like readers’ background knowledge, attention control and inhibition of irrelevant

information, participants’ attitude and motivation toward the experiments (Brown &

Hulme, 1992; Chun & Payne, 2004; Daneman & Carpenter, 1980; Ericsson & Kintsch, 1995; Hannon, 2012). L2 WMC is seen as a mechanism that efficiently manipulates incoming information and forms a comprehensive understanding with new and old knowledge. Relevant factors to WMC, factors mentioned above and others related to readers’ mindset in dealing with demanding reading tasks, should be controlled in research design in order to testify the relationship between L2 WMC and L2 inferential RC.

Limitation

The result of the present study should be considered tentative and the power of generalization should be taken with restrictions. Several relevant factors restricting the result of this study were elaborated as follows.

The first limitation is the sample size.

The investigator included these 190 students without similar language proficiency in the current study. 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 and to stop them from attending their

69

scheduled classes. In order to include as many participants as possible in the research, I asked many classes for permission to participate 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 around 30. If I only had included 30 participants in the study, the number of participants would have been too few and would have sabotaged the analytic generalization.

Furthermore these qualified students were in different classes and grades. Averagely one class had two to four students who had passed the GEPT elementary level. Thus, I had to forgo the idea of including participants above certain English proficiency but to adopt a convenient sample of six classes instead. These students with varying language

proficiency and different reading skills formed a heterogeneous group. The number of 190 students from the 7 classes was appropriate for analytic generalization.

Sample size is crucial when it comes to the influence of WM on RC. Although WM is related to L2 language comprehension and use, studies regarding the magnitude and effect of WM have somehow been inconsistent among studies. Some suggest the inconsistency might result from variation of population effect size (e.g., measurement errors or sampling error); some suggest the possible reason is the systematic difference regarding WM theoretical framework. A similar study result of non-significant correlation between WMC and inferential RC was found in the study of Chun (2000). According to Chun (2000), the result suggested that the individual difference among readers, such as readers’ L2 WMC in this study, does not or might not affect RC performance. Chun (2000) suggests, small sample sizes of participants, participants’ attention to tasks, and the

authentic nature of research-conducting situation might lead to masked research

results—no significant correlation between L2 WMC and L2 inferential comprehension.

In Chun’s study, participants were allowed to have access to annotation resources about the research reading texts if necessary when completing comprehending tests with a

70

multimedia access in that study. When discussing the implication behind the research result, Chun recalled that some participants only completed parts of the research procedure. Few of the participants in that study did not “take the test very seriously” as observed by administration instructors (Chun, 2000, p.498). In that study, Chun indicated that factors like sizes of samples and participants’ attitude tremendously deteriorated the relationship of L2 WMC and L2 inferential comprehension.

The second limitation is the participants’ awareness of the importance of the studies.

When it comes to participants themselves, one of the crucial experimental factors indicated by Chun in her study, the current research was conducted at the end of the semester, right after the final examination and before the winter vacation. Just finishing the final exam, most participants lacked intrinsic motivation toward taking the reading comprehension test—it was not a high-stake examination and no immediate reward was provided. During the process, some of the participants gave up trying to understand the reading texts and claimed it was too difficult for them to read. Specifically, about 1/6 participants scribbled down their answers without prudent reading and thinking while other peer struggled to finish the test at the first 20 minutes of the procedure. Furthermore, for those who were willing to try overcoming this reading test might not use proper

reading strategies. Many of the participants were found to be stuck at one reading for quite a while. To be specific, the testing period lasted for 45 minutes but many

participants spent almost 20 to 30 minutes on the first text leaving inadequate time for other 3 texts. All these variables and individual difference greatly masked and

compromised the research result to a certain extent and led to the inconsistent result that L2 WMC failed to reveal significant correlation with L2 inferential RC.

In Chun’s study, she included 19 undergraduate L1 students enrolled in one of the university in the West Coast of the US. Since Chun intended to investigate four variables besides WMC and RC, participants in his study were allowed to consult Internet

71

resources in case of having difficulty understanding the readings. According to Chun (2004), the non-correlation between WMC and RC might result from the not-regulated reading time and participants’ ignorant attitude toward research tasks. As Chun (2004) noted, “a larger sample with more complete language data on the language performance measures would generate a different set of results” (p.497). In other words, as Chun indicated, the failure of demonstrating participants’ language competence and engaging participants in completing research tasks could deteriorate the relationship between WMC and reading comprehension. Likewise, Jeon and Yamashita (2014) pointed out that most studies on L2 WMC and L2 reading comprehension included low or not enough numbers of effect sizes, which means that accumulated evidence gleaned from researches is insufficient to generate trustworthy implication regarding L2 WMC and L2 reading comprehension. Hence, the sample size and the participants’ attitude are two significant factors in influencing the research result.

Table 6.1 Descriptive Analysis on Reading Span Task (RST) of the 190 Participants

number minimum maximum mean SD

recall-RST group

WMC 112 0.20 12.00 8.3168 2.54938

literal RC 104 0 5 1.96 1.343

inferential RC 104 0 5 1.93 1.331

overall RC 104 0 5 3.89

recognition-RST group

WMC 87 3.45 12.00 11.1768 1.39676

literal reading 86 0 5 1.76 1.116

inferential RC 86 0 5 1.71 1.167

overall RC 86 0 5 3.47

To reiterate, participants’ attention and text difficulty of the provided reading

72

comprehension in studies are important factors regarding the power of implication behind the research result. In terms of the participants’ overall RC including literal and

inferential RC, the mean scores of RC were 3.89 for the recall group and 3.47 for the recognition group (Table 6.1 shows the descriptive analysis of participants’ WMC and RC performances). In terms of performances of literal RC, the mean scores of RC were 1.96 for the recall group and 1.76 for the recognition group. In terms of performances of inferential RC, the mean scores of RC were 1.93 for the recall group and 1.71 for the recognition group. Besides, the SD (standard deviation) for the recall groups was 1.343 for their literal RC and 1.331 for their literal RC. The SD for the recognition group was 1.116 for their literal RC and 1.167 for their literal RC. This small gap among these SD implied that the RC performances of all participants in both the recall-RST and the recognition-RST groups were quite close to each other. Standard deviation is used to demonstrate the situation in which data points are spread out in the area of statistic values.

Considering the total score of the experiment reading test being 10 points, the SD values (1.343, 1.331, 1.116 and 1.167) indicated that the discrete degree of RC among

participants was low. Besides the density of the cross-participants RC performances, the general RC performance of all participants seemed to be low and not high. As for other descriptive analyses, the maximum scores of RC were 10 for the recall group and 9 for the recognition group; the mean scores of RC were 3.89 for the recall group and 3.47 for the recognition group. To sum up, the descriptive statistic of mean scores and SD

indicated that this reading test was difficult and challenging for most participants regarding the poor RC performance indicated by the mean scores and SDs mentioned above.

Likewise, several researchers emphasize that individual differences in WMC are less contributable to real storage of information but more to one’s ability to control attention in dealing with cognitive tasks (Engle, 2002; Engle & Kane, 2004). According

73

to Engle and Kane, readers with high WMC have better attentive control to process new and old information when dealing with cognitive task demands. Literature suggests that the interaction or interplay of storage and process within WMC would be mediated by participants’ selective attention—the function of inhibition or oppressing the minor information (Engle et al., 1999; Conway, 1998). In other words, participants’ control of their attention to relevant information regarding the immediate tasks at hand is essential in terms of generating a successful comprehension and performance. Thus, participants’

attention control is highly related to central executive component inside WMC and greatly accounts for the situation how individuals deal with cognitive tasks, such as RST and RC tests. Readers’ attention is emphasized when it comes to WM and cognitive tasks.

The information stored in the slave systems of WM would last and maintain in readers’

mind in an active form 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). Successful cognitive actions require temporary storage and manipulation of current information along with relevant resources from LTM background knowledge. The insufficient level of attention deteriorated the influence of L2 WMC on L2 inferential comprehension performance.

Specifically, according to Engle and Kane (Engle, 2002; Engle & Kane, 2004), tasks to which participants hold full attention, such as RC, would fully reveal or demonstrate the predictive power of WMC, either in the case of L1 and L2. Conversely, WMC would be revealed to be unrelated to tasks in which participants simply complete with automatic processes instead of fully activating their mind power to infer further comprehension.

When it comes to automatic process for comprehension, one example is literal RC. Literal reading comprehension requires readers simply to choose the words they recognize from the text. In doing so, some fail to connect the proposition from sentences into a coherent meaningful representation which might not be stated explicitly from the print text. Most

74

important of all, it is the participants’ abilities that could wield the influence of central executive to manipulate the task-relevant information so as to accomplish complex cognitive tasks.

The third limitation is the study design of reading skills and background knowledge.

First of all, the reading texts were adopted from the standardized exam of CAP (Comprehensive Assessment Program of Junior High School Students) and were found to be too difficult in comprehending for most of the participants in this study. The mean scores of overall RC performance were 3.89 for the recall WMC group (N=104) and 3.47 for the recognition WMC group (N=86) while the total reading score was 10. When it comes to L2 inferential RC, the mean scores were1.93 for the recall WMC group and 1.71 for the recognition WMC group under the premise of total inferential RC score (=5) . In order to ensure the reliability and validity of reading texts, the texts of reading were adopted from the English standardized test of CAP (Comprehensive Assessment Program for Junior High School Students) in 2014. Four reading passages with 5 literal and 5 inferential RC questions were included. According to an official report regarding the item difficulty of English test from CAP (2014), this newly-reformed reading exam requires test-takers not only to understand reading literally but also to go beyond the text and form coherent comprehension within the context. In this study, the participants ranged from 8th to 9th graders and did not complete the whole three-year English course in junior high school. In order to alleviate possible reading difficulty or problem from unknown English vocabulary, necessary Chinese counterparts for some English vocabulary, which had not included in their textbooks, were provided in the tests. The difficulty understanding the texts greatly overloaded the participants, which resulted in little space to process and

First of all, the reading texts were adopted from the standardized exam of CAP (Comprehensive Assessment Program of Junior High School Students) and were found to be too difficult in comprehending for most of the participants in this study. The mean scores of overall RC performance were 3.89 for the recall WMC group (N=104) and 3.47 for the recognition WMC group (N=86) while the total reading score was 10. When it comes to L2 inferential RC, the mean scores were1.93 for the recall WMC group and 1.71 for the recognition WMC group under the premise of total inferential RC score (=5) . In order to ensure the reliability and validity of reading texts, the texts of reading were adopted from the English standardized test of CAP (Comprehensive Assessment Program for Junior High School Students) in 2014. Four reading passages with 5 literal and 5 inferential RC questions were included. According to an official report regarding the item difficulty of English test from CAP (2014), this newly-reformed reading exam requires test-takers not only to understand reading literally but also to go beyond the text and form coherent comprehension within the context. In this study, the participants ranged from 8th to 9th graders and did not complete the whole three-year English course in junior high school. In order to alleviate possible reading difficulty or problem from unknown English vocabulary, necessary Chinese counterparts for some English vocabulary, which had not included in their textbooks, were provided in the tests. The difficulty understanding the texts greatly overloaded the participants, which resulted in little space to process and

在文檔中 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 (頁 77-98)