(PDF) Bui, S. (2016). Implicit and explicit grammatical knowledge in predicting reading comprehension of Vietnamese EFL learners

This is the pre-print (before peer-review) version of the article. Citation and DOI link of the published version: Bui, S. (2016). Implicit and explicit grammatical knowledge in predicting reading comprehension of Vietnamese EFL learners. Korean Journal of Applied Linguistics, 32(1), 175-201. http://doi.org/10.17154/kjal.2016.3.32.1.175 Korean Journal of Applied Linguistics Implicit and Explicit Grammatical Knowledge in Predicting Reading Comprehension of Vietnamese EFL Learners Bui Thien Sao Korea University ____________________________________________________________________________________________ This study explored the predictive power of implicit and explicit grammatical knowledge in explaining reading comprehension of Vietnamese EFL students and examined how differently these two types of grammatical knowledge could predict for students of different reading ability. The participants of the study comprised 51 Vietnamese adult learners of English at different proficiency levels. They completed three English tests including a timed grammatical judgment tests, an untimed grammatical judgment test, and a reading comprehension test. The results indicated that both implicit and explicit grammatical knowledge were two significant predictors of reading comprehension. However, the two types of grammatical knowledge lost their predictive power when it came to the higher reading ability group. For the case of the lower level group, explicit grammatical knowledge did not contribute to reading, but implicit grammatical knowledge remained a significant contributor. Keywords implicit grammatical knowledge, explicit grammatical knowledge, grammaticality judgment test, reading comprehension I. INTRODUCTION It has been well documented that grammatical knowledge plays an indispensable part in reading comprehension. According to Grabe (2009) and Givón (1995), knowledge of grammar serves as the cues for text coherence, thereby assisting readers’ understanding of the text structures, meaning, and inferences. In Nation and Snowling’s (2000) explanation, grammar can aid readers in grasping the idea conveyed in each clause or sentence, integrating and unifying them all to comprehend the whole content. There has been an ample body of research providing empirical evidence for the essential role of grammatical knowledge to L1 reading comprehension. For instance, it has been found that grammar knowledge was a strong predictor of reading in comparison with other predictors, remaining its direct contribution when controlling for other variables, and the deficit in syntax skills was even the main cause of poor readers' problems (Bowey, 1986; Urquhart & Weir, 1998; Grabe, 2005; Nation, 2005; Snow, Griffin & Burns, 2005; Nation & Snowling, 2000; Waltzman & Cairns, 2000; Cain, 2007; Brimo et al., 2015; Klauda & Guthrie, 2008). In L2 context, much less work notwithstanding, the significance of grammar is recorded in Bernhardt (1991), Alderson (2000), Koda (2005), Morvay (2012), Shiotsu (2010), Shiotsu and Weir (2007). Besides, the role of grammar was revealed to depend on EFL readers’ ability levels. Nonetheless, the findings are inconsistent: Kim and Cho (2015) found that grammar knowledge contributes more in the case of advanced L2 readers while the reverse was disclosed in Shiotsu and Weir’s (2007). The gap in L2 research along with the inconclusive results indicates a need for further work, which is also the motivation for the present study. Additionally, it is noteworthy there is a lack of studies investigating how the contribution varies depending on the types of grammatical knowledge. Few works such as Cain (2007) and Brimo et al. (2015) delved into syntactic awareness and syntactic knowledge. Whereas, other researchers often treated grammar knowledge as a single dimensional unit, or did not define the construct clearly (Bowey, 1986; Shiotsu & Weir, 2007; Shin & Kim, 2012; Kim & Cho, 2015; Morvay, 2012). Therefore, this study sought to scrutinize the grammar role in a different and more well-defined way of categorizing, that is, grammar in terms of implicit and explicit knowledge. Although the concept of these two types of knowledge was coined long ago by Krashen (1981), little was known about their contribution to reading comprehension. Thus, this study attempted to provide more understandings about the contributions of implicit and explicit grammatical knowledge to reading comprehension of L2 learners and examined whether the relationships are the same or different for learners of different reading ability. II. LITERATURE REVIEW In the first place, language knowledge can be categorized into two types: implicit and explicit knowledge. As postulated by Ellis (2005) and Bowles (2011), “intuitive awareness”, “procedural knowledge”, and “automatic processing” are some of the salient characteristics of implicit knowledge. This type of knowledge is supposed to be typical for mature native speakers who can use their first languages spontaneously and effortlessly (Zhang, 2012). Conversely, explicit knowledge necessitates consciousness, declarative knowledge of language rules, and effortful or controlled processing (Ellis, 2005; Hulstijn, 2005). Utilizing their explicit knowledge, learners can verbalize or explain the underlying linguistic properties or patterns of the presented information (Ellis, 2005; Hulstijn, 2005). Though classified as two discernible dimensions of knowledge, a positive relation has been proved present between them (Ellis, 2005; Zhang, 2012). Furthermore, some scholars such as N. Ellis (1994, 2005) and R. Ellis (2005) even conceded that one type can facilitate or become the other type, that is, there is an interface between the two. In the light of implicit and explicit knowledge, grammar knowledge can also be viewed to consist of two types, namely implicit grammatical knowledge and explicit grammatical knowledge. These two kinds of grammar knowledge bear similar attributes of implicit and explicit knowledge mentioned above, however, with the confinement to grammatical rules and features of a language. It should be noted that for L1 speakers, the implicit knowledge is supposed to substantially outweigh their explicit knowledge in using their native language; therefore, research in the field was not much concerned with L1 speakers, but rather L2 learners. However, even in L2 context, there has been only a handful of research dealing with the role of both or either of grammatical knowledge types in L2 proficiency in general and reading in particular. Some researchers were interested in only explicit grammar knowledge which was referred to by using the term “metalinguistic knowledge” or “metalinguistic awareness” in several studies; nonetheless, their results were conflicting to some extent. While Renou (2001), Roehr (2007), and Butler (2002) approved that explicit knowledge correlated significantly with L2 proficiency, Alderson, Clapham, and Steel (1997) unveiled only a weak correlation and affirmed that possessing better knowledge of grammar and ability to explain grammar rule did not mean faster L2 learning. Meanwhile, the correlations discovered by Elder and Manwaring (2004) were mediated by their starting age of L2 learning. More specifically, there were more significant and strong relationships between metalinguistic knowledge and L2 Chinese proficiency found for the adults who were beginners of L2 Chinese at that time than for those who had learned the language since secondary school. The same was true when written tests (reading and writing) were considered separately. Moreover, it is worth noting that the correlations were greater with the written test scores than the oral tests. Observably, the relationship with reading was not discussed or reported in these studies, and there existed an effect of students’ ability on the relationship which deserved more concerns and investigation. To date, there have been few studies taking both types of grammatical knowledge in consideration, two of which also aimed at relating learners’ grammatical knowledge to their L2 English proficiency. The first one was conducted by Han and Ellis (1998) with the attention paid to only verb complementation in English. The two proficiency tests employed in the study were the Secondary Level English Proficiency Test (SLEP) and the Test of English as a Foreign Language (TOEFL) which did not measure speaking skills or actual writing skills. As a result, these tests were not precisely about the general proficiency, but rather dominated with receptive skills. Han and Ellis (1998) concluded that their study suggested a favor of both types of grammar knowledge in L2 proficiency. The results for reading part, however, were not uniquely provided. The second one shed light on not only the relations but also the contributions of both knowledge types to all four skills tested the IELTS tests: listening, speaking, reading and writing (Ellis, 2006). Another different point was that this research investigated into each of 17 grammar structures’ predictive power in L2 proficiency. Consistent with Han and Ellis (1998), Ellis’s (2006) findings were that that both implicit and explicit knowledge correlated and contributed significantly with English proficiency which was indicated by the total IELTS scores. Reading skill was reported to have a stronger correlation with explicit grammar knowledge than the implicit one. In addition, though each grammar knowledge type had its own structures that can predict reading in separate regression models, when the relative contribution was examined, only the grammar structure of explicit knowledge became the significant predictor for this receptive skill. Overall, the studies favored the role of explicit type over the implicit one in predicting reading comprehension. However, hierarchical regression was not employed, so the partial contribution of each type was not specified. Moreover, it is not satisfactory to take one or two out of 17 grammar structures for each knowledge type and make a claim about implicit and explicit grammar knowledge’s accountability in general. Differing from above researchers, Erçetin and Alptekin (2013) and Zhang (2012) only took reading comprehension of advanced EFL students into account. Like Ellis (2006), a significant relationship between merely the explicit grammatical knowledge and reading comprehension was revealed by Erçetin and Alptekin (2013). However, no conclusion on the predictive power of grammar knowledge can be drawn from this correlation analysis. By contrast, Zhang’s (2012) is the only research that gave no support to the role of explicit knowledge in reading comprehension to date. With regression analysis on a large sample size (172), Zhang (2012) observed a significant contribution made by only implicit grammar knowledge to reading comprehension even when controlling for explicit grammatical knowledge and vocabulary size. However, the explicit knowledge instrument, which was extracted from real TOEFL test, was criticized by Kim and Cho (2015) to be possibly adulterated with a vocabulary test. Besides, Zhang (2012) also admitted the limitation of having only three reading subskills tested. Hence, the findings should be read with caution. Conspicuously, the scarcity of studies particularly on L2 reading and the absence of studies looking at different levels of ability were undeniable. Moreover, the small array of research keeps producing incongruent findings to some extent. As a result, there is a need for further research in the field. Another issue in the realm of implicit and explicit knowledge is concerned with the validity of the measurements. Possibly, the reason why this area did not flourish in the past was the absence of rigorous tools to assess learners’ implicit knowledge separate from their explicit knowledge. Recently, there has come a test battery developed specially for measuring implicit and explicit grammatical knowledge by Ellis (2005). The whole set of the tests was composed of an imitation test, and oral narrative test, a timed grammatical judgment test which targeted at measuring the implicit type and an untimed grammatical judgment test, a metalinguistic knowledge test for explicit knowledge. This test battery was proved reliable and valid in various studies (Bowles, 2011; Ellis, 2004, 2005, 2006; Gutierrez, 2013; Loewen, 2009), thereby paving the way for other researchers interested in students’ L2 proficiency. It is some of these established tests or the tests designed in similar approach which were the instruments in the studies by Han and Ellis (1998), Ellis (2006), Erçetin and Alptekin (2013) and Zhang (2012). In a similar vein, the present study also employed the timed and untimed grammatical judgment tests as the tools. Regarding only these two judgment tests in previous studies, the former has been shown to primarily measure the implicit grammatical knowledge and the latter was a robust instrument in explicit knowledge assessment. Han and Ellis’s ones in 1998 can be understood as the predecessor to Ellis’s (2005, 2006). The implicit knowledge test by Han and Ellis (1998) allowed the test takers to judge each sentence’s grammaticality in precisely 3.5 seconds irrespective of the item’s difficulty, which was inappropriate to some extent. Therefore, Ellis (2005) decided to establish the item time limit based on native speakers’ actual speed. It was this method of time setting which was used in Zhang’s (2012) and the present study. In addition, there is some room for improvement in the explicit grammar knowledge test as well. The small number of choices (only two: grammatical and ungrammatical) along with unlimited test time in Ellis’s (2005, 2006), Erçetin and Alptekin (2013) can make the students’ answers susceptible to wild guess. What is more, Ellis (2005) and Loewen (2011) showed that the ungrammatical items in the untimed test can measure more exactly the students’ explicit knowledge than the grammatical ones. Hence, it is probably more effective and sensible for the untimed test to include only ungrammatical items and require students to provide error corrections. This is also the way that the author of the present study, with the intention to maximize the test’ validity, adapted Ellis’s (2005) and Loewen’s (2009) tests. To make a long story short, there is a gap in the field of implicit and explicit grammatical knowledge and their relationships with reading comprehension; the literature about the effects of reading ability levels on the relationships is scarce as well; and finally, the measurements utilized in previous works still have some room to be desired. All of these form a void for this study to fill in. III. RESEARCH METHOD 1. Research questions The present study was conducted to explore the relative importance of implicit and explicit grammatical knowledge in L2 reading comprehension of Vietnamese adult learners of English. It further examined whether the roles of these two types of grammatical knowledge differed according to students’ reading abilities. Briefly, the study sought to answer two research questions: 1. Between implicit and explicit grammatical knowledge, which has relatively stronger predictive power in explaining reading comprehension of Vietnamese adult EFL learners? 2. Does the relationship among implicit and explicit grammatical knowledge and reading comprehension depend on students’ English reading abilities? 2. Participants 51 Vietnamese adult EFL learners at a private language institute were recruited as the participants of the study. This sample was composed of 34 students from 10 colleges and 17 employees from 17 companies in Vietnam. Totally, there were four males and 47 females. Their ages ranged from 20 to 24 with the average age of 21.94 (SD= 1.10). Vietnamese was the learners’ mother tongue, and the starting age of learning English was around ten (mean = 9.8, SD = 1.68). On average, the participants had studied English for 12.12 years (SD = 2.28). It was also reported that none of the learners had lived in a country where English was widely spoken. Besides, some participants (22) studied Chinese, Korean, Japanese, or French as other foreign languages. Although these learners came from different universities and companies, their majors and working fields were mainly related to commerce and business (38/51). At the institute, the students were in four different classes, having three one-hour-and-a-half English sessions per week. In general, the participants were a mixed group of English proficiency. Some had just entered the course while others had taken two consecutive courses at the institute. A few even majored in English or English for commerce (7). Moreover, the fact that these adult learners went to different schools and workplaces also contributed to the variation in the participants’ background of English. This heterogeneity could make the sample better representative of the population and enhance the generalizability of the findings. To address the second question, three groups were drawn from the participants based on the scores of their reading tests. The high reading ability group (HG) was made up of the learners with reading scores equal or higher than the mean score, and those getting the scores lower than the mean score formed the low ability group (LG). By this way, 24 learners were assigned in HG, and the number for the LG was 27. A t-test was run, and it showed that there was a significant difference between the two groups (t = 9.816, p < .01). (See table 2) 3. Instruments The test battery included a timed grammaticality judgment test, an untimed grammaticality judgment test, and a reading comprehension test which were designed to measure the students’ implicit grammatical knowledge, explicit grammatical knowledge and reading ability respectively. The timed grammaticality judgment test (TGJT) contained 34 English sentences taken from the grammaticality judgment test by Loewen (2009). There were equal numbers of grammatical and ungrammatical sentences, that is, 17 and 17. All the sentences were decontextualized, simple in wording and quite short (5 to 14 words). These items pertained to 17 grammatical structures which were argued to be problematic to English learners, ranging from early to late acquired grammatical knowledge and representing elementary to advanced competence levels (Ellis, 2005). Each of these grammatical features was covered in two items of which one completely abided by English grammar and one violated. The TGJT was a computer-delivered test using Microsoft PowerPoint. The instructions with an example were explained orally by the teacher in charge of the class and presented in Vietnamese on the screen as well. The sentences were presented one by one on the screen, and each test taker was required to click on the button “true” or “false” to indicate whether each sentence was right or wrong in terms of written English grammar. To activate the test takers’ implicit knowledge, every single test item had a distinct time limit which was specified based on the native speakers’ speeds. For each item, the response time of five native speakers were averaged out and then 20% of the mean time was added to induce the final time limit for the participants, who were L2 learners with slower processing speed (Ellis, 2005; Zhang, 2012). The resulting time for the sentence judgment ranged from 2.70 to 6.92 seconds. Moreover, to relegate the order effect, four versions with different randomized item orders were created and equally distributed. Regarding scoring, one point was given for each correct judgment; the wrong items and no response ones received zero. The maximum possible score a student could get was 34. The test reliability was .804. The untimed grammaticality judgment test (UGJT) consisted of 17 ungrammatical sentences adapted from Loewen’s (2009) GJT and metalinguistic knowledge tests. The specific test items were different from the ones in the TGJT; however, they had the same basic characteristics. That is, they were decontextualized, simple in wording, and short (5-11 words per sentence). Both tests were identical in tested structures of grammar knowledge. The teacher also explained briefly the task and, the direction was written in both English and Vietnamese. In addition, unlike the TGJT, the UGJT was a paper-based test, and as a measure of the explicit grammar knowledge, there was no time limit. However, all the students handed in the test paper less than 20 minutes. The participants were instructed to locate the only ungrammatical element among three underlined choices and correct it. The total possible score for the UGJT was 34. Each item was worth two points: one for the correct detection and one more for the appropriate correction. No response or incorrect answers for a specific item received no point. The Cronbach alpha for the test was .822. On the reading comprehension test, there were five passages and 20 multiple choice items all of which were from the reading part of the actual TOEIC tests. The length of the passages was between 151 and 341 words long. With respect to the readability, the Flesh-Kincaid Grade indices ranged from 8.9 to 14 across five passages. This variation could help to avoid ceiling or floor effects (Nagy, Anderson & Herman, 1987) and better assess the reading ability of the mixed- level group in the present study. The passages also varied in genres: they were articles, advertisements, announcements, memos or emails. There were no esoteric topics found in the passages, but rather general ones which were also related to the participants’ fields of economy and business. Each of the first two passages was followed by three questions; the third passage had four questions, and five items followed each of the last two passages. These questions purported to measure the readers’ understanding of main ideas, detailed information, co- reference, and inference. The students were asked to select the best response among four choices to indicate their answer. The test terminated in 25 minutes. In terms of scoring, each correct option received one point, which made the maximum possible score 24. The test had the Cronbach alpha at .733. 4. Procedures The tests were administered in the order of TGJT, UGJT, and reading comprehension test. The TGJT was delivered prior to the UGJT in order to avoid a negative test effect: Ellis (2005) and Bowles (2011) explained that doing the error correction task before could make the test taker more alert to the ungrammatical elements and activate their explicit knowledge more when they took the TGJT which covered identical structures of grammar with the UGJT. The three tests were separately included in three different regular English periods of four classes. The instructions for all tests were presented both orally and visually in Vietnamese. All the participants did three tests individually. After that, they were required to answer a background questionnaire which was written in Vietnamese, asking about the test taker’s age, gender, disciplines/working fields, starting age of learning English, the number of years they had learned English, sojourn in an English speaking country, and foreign languages other than English they studied. 5. Analysis The data were coded and put into SPSS 16. Descriptive statistics for the three tests were calculated. The internal reliability indices were determined using Cronbach alpha. Pearson product-moment correlations were run to analyze the relations among the tested knowledge and ability. Furthermore, hierarchical regression analyses were carried out to examine how strong the implicit and explicit grammatical knowledge were as a predictor of reading ability. The second research question delved into the relationship with reading ability serving as a moderator variable. To solve this, two groups of students at the high and low level of reading ability were formed on the basis of their reading scores. The cutoff scores were the mean score. A t-test was run to check the significance of the score difference. After that, the correlations and multiple regression analysis were computed for each group. IV. RESULTS 1. Descriptive statistics The data collected from all three tests were checked with the skewness and kurtosis indices, which then totally confirmed a normal distribution for each test. Other descriptive statistics such as mean score and standard deviation were summarized in table 1. A matched t-test indicated that the participants scored significantly higher on the untimed grammatical judgment test which measured their explicit knowledge than on the timed one which measured their implicit knowledge (t = 2.755, p < .01). TABLE 1 Descriptive statistics for test scores of the whole sample (N = 51) Minimum Maximum Mean SD Implicit grammatical knowledge 6.00 33.00 19.8627 5.68865 Explicit grammatical knowledge 6.00 34.00 22.4902 6.83629 Reading comprehension 4.00 18.00 13.3137 3.38520 As mentioned previously, the participants were divided into two groups (HG and LG) based on the mean score of the reading test. Comparing the mean score of each group with a t-test showed that the HG significantly outperformed the LG on reading comprehension (t = 9.816, p < .001). Therefore, it is probably safe to say two groups were at two different reading ability levels. Table 2 provided the descriptive statistic information on all three tests of each group and the t-test results. It can be seen that compared with the lower reading ability group (LG), the group of higher reading ability students (HG) scored significantly better in two grammar tests as well (t = 2.742, p < .01 and t = 4.303, p < .001 for the implicit knowledge test and explicit knowledge test respectively). TABLE 2 Comparing mean scores of three tests between HG and LG HG (N=24) LG (N=27) T Min Max Mean SD Min Max Mean SD RC 4 13 16.21 1.41 14 18 10.74 2.38 9.816*** Implicit GK 16 33 22.04 5.35 9 27 17.93 5.35 2.742** Explicit GK 13 34 26.23 5.48 9 30 19.15 6.22 4.303*** ** p < .01, *** p < .001 RC: reading comprehension GK: grammatical knowledge 2. Predictive power of implicit and explicit grammatical knowledge to reading comprehension for the whole sample Before computing the predictive power of implicit and explicit grammatical knowledge to reading comprehension, Pearson product-moment correlations were conducted (See table 3). TABLE 3 Pearson correlations among scores on implicit and explicit grammatical knowledge, and reading comprehension for the whole sample (N=51) GK Implicit GK Explicit GK RC GK 1 Implicit GK .810** 1 Explicit GK .873** .420** 1 RC .600** .487** .523** 1 ** p < .01 GK: grammatical knowledge RC: reading comprehension The results of the correlation analyses revealed that reading comprehension was significantly and positively interrelated with grammar knowledge in general which was indicated by the combined scores of two grammar tests (r = .600, p < .01) In other words, the students with better grammatical knowledge tended to comprehend what they read better. In addition, there was a significant and positive correlation found between implicit and explicit grammatical knowledge. However, the r value was not high, which signified that the timed and untimed tests measured different types of grammatical knowledge. More importantly, both implicit and explicit types were correlated positively and moderately with reading comprehension at a significant level. Put differently, the students who had a better understanding of the given texts had a tendency to possess greater implicit and explicit knowledge of grammar. The indices illustrated that the relationship between explicit grammatical knowledge and reading was slightly stronger than that between implicit grammatical knowledge and reading. TABLE 4 Hierarchical regressions analyses predicting reading comprehension of the whole sample (N=51) Step Predictor R R2 ΔR2 ΔF 1 Implicit GK .487 .237 .237 15.258** 2 Explicit GK .600 .360 .123 9.226** 1 Explicit GK .523 .274 .274 18.448** 2 Implicit GK .600 .360 .087 6.520* * p < .05, ** p < .01 GK: grammatical knowledge In order to examine which, among the two kinds of grammar knowledge, is a stronger predictor of reading, hierarchical regression analyses were conducted twice with implicit grammatical knowledge and explicit grammatical knowledge entered in reversed orders (See the bottom vs. top panels of table 4). When implicit grammatical knowledge was entered first (see the top panel), it explained approximately 24% of the variance in reading, making significant contribution (p < .01). Controlling for the effects of implicit grammatical knowledge, the explicit one contributed significant unique variance to reading comprehension, explaining an additional 12% of the variance (ΔF = 9.226, p < .01). Subsequently, the order was switched (see the bottom panel of table 4). Entered at the first step, explicit grammatical knowledge significantly accounted for about 27% of the variance in reading comprehension. When controlling for the effects of explicit grammatical knowledge, the implicit one significantly explained a unique proportion of variance of reading comprehension (ΔR2 = .087, ΔF = 6.520, p < .05). Together, these two types of grammatical knowledge made a significant contribution to reading comprehension, that is, 36% of variance of reading comprehension. Generally speaking, both types of grammatical knowledge were significant predictors of reading comprehension. When comparing the relative predictive power of implicit and explicit grammatical knowledge, the latter turned out to be a stronger predictor of reading (beta = .325, p < .05, and beta = .386, p < .01, respectively) 3. How the relationships depend on the students’ English reading abilities? To address the second research question, correlation and hierarchical regression analyses were run for each of two groups, that is, the higher reading ability group and the lower one. TABLE 5 Pearson correlations among scores on TGJT, UGJT and reading comprehension for HG and LG HG (N=24) LG (N=27) 1 2 3 4 1 2 3 4 1. GK 1 1 2. Implicit GK .806** 1 .762** 1 3. Explicit GK .815** .314 1 .830** .271 1 4. RC .174 -.082 .358 1 .430* .597** .125 1 * p < .05, ** p < .01 GK: grammatical knowledge RC: reading comprehension Firstly, the correlation results were presented in Table 5. Chief among striking features for higher reading ability group was that no significant relations were recorded between reading comprehension and grammatical knowledge in general, or each type of implicit and explicit knowledge in particular. The correlation was higher for the explicit knowledge than the implicit one; nonetheless, they were all insignificant. Besides, the result also demonstrated an insignificant relationship between two types of grammatical knowledge. In general, the findings implied that for the students at higher reading ability, their grammatical knowledge was not a meaningful predictor of reading comprehension. By contrast, the correlation between grammar knowledge in general and the reading score was significant in the case of lower ability group. The numeric value was also much higher than the one for higher ability group. However, for the less-skilled reading group, only implicit grammatical knowledge rather than explicit knowledge was interrelated to reading comprehension at a significant level. Furthermore, its magnitude was much greater than the one for explicit type and reading (r = 0.597 vs. r = 0,125). Unlike the whole sample, neither of two groups had a significant relationship between two types of grammatical knowledge, which again gave evidence for the separation of two measures of implicit and explicit knowledge. Subsequently, the hierarchical regression analyses were carried out for each student group in the same manner as for the whole sample. The outcomes indicated that implicit and explicit grammar knowledge were not powerful predictors of higher ability readers’ comprehension (See table 6). Together they could explain only a small and insignificant proportion of variance in reading (17%). Additionally, no matter what entry orders were applied for implicit and explicit grammatical knowledge, the unique variance of reading each type can account for was neither high nor significant. Put it in another way, when controlling or not for one factor, the other remained insignificant to reading. Table 6 Hierarchical regressions analyses predicting reading comprehension for HG and LG HG (N=24) LG (N=27) Step Predictor R R2 ΔR2 ΔF R R2 ΔR2 ΔF 1 Implicit GK .082 .007 .007 .148 .597 .356 .356 13.832** 2 Explicit GK .412 .170 .163 4.127 .598 .358 .001 .055 1 Explicit GK .358 .128 .128 3.231 .125 .016 .016 .396 2 Implicit GK .412 .170 .042 1.056 .598 .358 .342 12.781** ** p < .01 GK: grammatical knowledge Differently, for the group at the lower level of reading ability, only implicit grammar knowledge illustrated a noticeable and significant contribution to reading comprehension. When entered prior to explicit grammatical knowledge, implicit one significantly accounted for roughly 36% of the variance in reading comprehension. After controlling for the explicit knowledge, implicit aspect could still have a unique shared variance of 34% with reading (ΔF = 12.781, p < .01). Meanwhile, explicit type merely explained a tiny proportion of variance in reading, and even when entered in the first step, its accountability was not significant. In short, it was implicit grammatical knowledge that could serve as a meaningful predictor of reading comprehension of lower level students. V. DISCUSSION This study was implemented with the purpose of examining the predictive power of implicit and explicit grammatical knowledge in explaining reading comprehension of adult Vietnamese learners of L2 English. Overall, the findings were that grammatical knowledge in general and each type of explicit and implicit grammar knowledge significantly correlated and contributed to reading comprehension. The relationships were also found to vary according to the students’ reading ability levels. More specifically, the first research question was about the predictive power of implicit and explicit grammatical knowledge to reading comprehension of all the participants who were at different levels of proficiency. Significant relations and contributions were revealed as the answer to this question. Firstly, grammar knowledge in general, which was indexed by the combined scores of both grammar tests on implicit and explicit knowledge, correlated significantly with reading comprehension and accounted for a large proportion (60%) of the variance in reading. This completely confirmed the pivotal role of grammatical knowledge in reading discovered by many of other researchers such as Bernhardt (1991), Alderson (2000), Koda (2005), Morvay (2012), Shiotsu (2010), Shiotsu and Weir (2007) to name but a few. When considered each type of grammatical knowledge separately, both implicit and explicit knowledge types were significant predictors of reading comprehension. One remained significant even when controlling for the other. The findings were consistent with previous studies either on explicit grammatical knowledge alone (Butler, 2002; Renou, 2001; Roehr, 2007) or on both types of grammatical knowledge and their relationships to L2 proficiency (Han & Ellis, 1998; Ellis, 2006) and reading skill (Ellis, 2006). It was evident that the findings supported the contention: “effective reading requires a type of grammatical knowledge that can be rapidly accessed in reading for syntactic parsing and word integration” (Zhang, 2012, p.570). Furthermore, the importance of explicit grammatical knowledge in theory was also corroborated. Grabe (2005), Urquhart and Weir (1998), and Zhang (2012) supposed that controlled processing could make syntactic parsing easier and help readers monitor their comprehension more efficiently. Besides, concerning the relative roles of two knowledge types, the correlation and contribution of the explicit grammar knowledge was slightly higher than the implicit one. This outcome accorded closely with Ellis’s (2006) conclusion which favored the explicit knowledge more than implicit knowledge in explaining the receptive skills. However, the result of the present study differed from Erçetin and Alptekin’s (2013) and Zhang’s (2012) which investigated both types of grammar knowledge but found only either of them a significant predictor for L2 English reading comprehension. Erçetin and Alptekin (2013) reported a significant correlation between reading comprehension and explicit grammar knowledge, but not implicit grammar knowledge. This might be due to the fact that their reading test actually loaded on the same factor as the untimed GJT, which meant it inherently prioritized explicit grammar knowledge over implicit knowledge. Additionally, the timing of the reading test was not specified, so it was unclear whether it was a power test where students had enough time to think and finish all the items or a speeded test that exercised time pressure. If there was little time pressure, then explicit knowledge would be activated and made the most by the test takers. In the present study, time constraint did exist in the reading test; thus implicit knowledge was needed to solve the task. Moreover, while the participants in the present study formed a mixed group, the sample in Erçetin and Alptekin’s (2013) was at an advanced level. More exactly, their participants were being trained to become English teachers who are supposed to be able to explain plainly language rules to their English learners. That their explicit knowledge might be activated more compared with normal learners may skew the result. One more point to mention, Erçetin and Alptekin (2013) did not run regression analysis after finding out the correlations; hence, the accountability of implicit and explicit knowledge in reading was beyond their scope. Contrastively, Zhang’s findings (2012) only support the role of implicit grammatical knowledge. Similar to Erçetin and Alptekin (2013), this researcher dealt with advanced learners rather than a diverse group of participants as in the current study. However, the cause of the discrepancy between Zhang’s (2012) and the present study may rest in the instrument difference. The explicit and implicit grammar knowledge tests designed by Zhang (2012) did not seem equivalent. Regarding item number, there were 196 items for the timed GJT, whereas, the untimed GJT only had 20 items. This huge disparity made the latter, in fact, less reliable than the former. Additionally, two GJTs were possibly different regarding the readability: the implicit knowledge one was described as short and lexically simple while the explicit knowledge one was extracted from TOEFL, thus being likely to demand a high level of vocabulary knowledge to solve the grammar task. This might increase the difficulty of the explicit grammatical knowledge task and even make the construct of the test overlap with a vocabulary test to some degree as criticized by Kim and Cho (2015). Obviously, Zhang’s (2012) measurements were somewhat in favor of implicit knowledge more than the explicit one. Furthermore, as Zhang (2012) mentioned, the limitation of his research may come from the reading test which aimed to assess only three reading subskills. Generally speaking, differences in the outcomes might be attributable to the variations in procedures and measurements across studies. The above mentioned factors might make the results obtained in a less objective manner than the current study. Delving deeper into the relationship between reading comprehension and explicit and implicit grammar knowledge, the second question focused on its dependence on the students’ reading levels. The result proved that the relationship did change as a function of reading ability. In the first place, no significant correlations or contributions were found for the case of the higher reading ability group. Probably, other variables such as vocabulary knowledge, metacognitive skills, etc., played a more influential role in reading than grammar knowledge at this level. For the higher reading level students, the results in a sense can be related with Erçetin and Alptekin’s (2013) and Zhang’s (2012) because they all worked on the advanced students. However, it can be seen that the results are not conclusive at all. Erçetin and Alptekin (2013) gave evidence for the explicit knowledge, Zhang (2012) implicit knowledge, and the present study neither of them. As discussed above, findings in the two studies conducted might stem from the biased instruments. Meanwhile, what was revealed in this research can be explained firstly by the threshold of explicit grammar knowledge. As displayed in the result section for the whole sample, the students comprehending the texts better tended to score higher in the explicit knowledge test. In this study, they, in fact, significantly outperformed the lower ability students on the untimed test as pointed out in table 2. One fourth of them (6 out of 24 students) got high scores (31/34 points or more) while no one in the other group could reach this level. Their better performance might be enhanced by how the explicit knowledge test was delivered to some extent. The test was untimed, so it allowed those who excelled in manipulating their explicit grammar to have even more than enough time to study the sentences carefully and recognize more errors. Whatever the reason was, the group’s performance probably surpassed the threshold for grammatical error detection and correction in L2; thus no longer predicting their reading comprehension. This speculation was also suggested in explaining the disassociation of students’ explicit knowledge and text comprehending by Zhang (2012). Secondly, participants might not have developed and make the most of their implicit knowledge. In an environment that English is not used widely even in the mass media and in daily life as Vietnam, the students do not seem to have plenty of chances to exposure constantly to English and they have little urge for understanding English on the spot in daily life. As a result, the development of implicit knowledge was ignored to a large extent. Moreover, it is possible that lack of chances for practice will lead to no habit formation and the ignorance of how to take advantages of it. When conjecturing the reasons behind the weak relationship between implicit grammar knowledge and reading, Erçetin and Alptekin (2013) also contended: “It is likely that the transferability of implicit linguistic knowledge to the ability to apply its proceduralized routines to L2 reading comprehension is quite narrow...” (p.745). Diverging from the higher reading ability group, the lower level students had grammar knowledge correlate with their reading comprehension, but of two types, only implicit knowledge made a significant contribution. Never before did a study delve into the relationships for the case of students at low level; therefore, no study can be discussed in relation to the current one. However, the result can still be explained in some ways. To begin with, it was possible that the participants who were at low level of both types of grammar knowledge had to judge the sentences primarily according to on their “feel” rather than explicit rules. Put simply, their judgments were based on whether the sentence sounded right or not, which was supposed to be the way learners operationalize their implicit knowledge (Ellis, 2005). Another factor was that the reading test was timed to some degree, and this succeeded in activating implicit knowledge of lower level readers better than the other group. Probably, they needed more time to tackle the reading task than the skillful students, and they could not really resort to their insufficient explicit knowledge, either. Consequently, time pressure can exert more influence on their implicit knowledge. The last speculation for the lower level group was concerned with the difficulty levels of two grammar tests. Though timed, the implicit knowledge test might still be a little easier than the explicit one, which may be due to the fact the they did not have to correct the error, but just clicked on “true” or “false”. Other reasons for divergences of findings can be attributed to different methods of sorting participants into groups. While Zhang (2012), Erçetin and Alptekin (2013), Han and Ellis (1998) use learners’ proficiency as the criterion, this study used reading comprehension scores interpreted as the students’ reading ability. Asides from grouping factor, there existed an influence of grammar structures. In other words, learners’ reading comprehension may be correlated with and explained by some certain grammar structures, but not others (Ellis, 2006; Zhang, 2012). For instance, Han and Ellis (1998) solely looked at verb complementation, Erçetin and Alptekin’s (2013) and the current study worked on 17 structures proposed by Ellis (2006), and in Zhang’s (2012), it was not stated clearly. An additional point to mention was the issue of two grammar measurements. The correlation between the two test scores was significant, but not high for the whole sample, and for each group, the correlation turned out insignificant. This supported Ellis’s (2005) assertion that implicit and explicit grammatical knowledge are two different aspects of knowledge that need measuring by separate tools, which were two grammar judgment tests in the case of this study. VI. CONCLUSION In brief, it was observed in this research that the students’ implicit and explicit grammatical knowledge were two significant predictors of reading comprehension for a mixed level group. However, those types of grammar knowledge lost their predictive power for the case of higher reading ability students whilst the lower ones had their implicit grammar knowledge contribute significantly to reading comprehension. Some pedagogical implications can be drawn from these findings. First of all, grammatical knowledge should remain a crucial part of EFL instruction, especially for students at low level. Secondly, as discussed above, in reading comprehension, both explicit and implicit grammatical knowledge come into play; hence, L2 teaching and learning needs to aim at improving both types of knowledge. Given the lack of implicit knowledge development especially in an English-as-foreign-language milieu, pedagogy and curriculum developers should actively take their role. As N. Ellis (2005), Grabe (2009) and Zhang (2012) suggested, the premise of implicit knowledge is implicit learning, which can be formed by an extensive exposure to print resources written in L2. Additionally, according to Ellis (2005) time pressure and attention on meaning were two of criteria for operationalization of implicit knowledge. Therefore, asides from extensive reading, reading under time constraint, or meaning focused interactions embedded in L2 curriculums might be helpful. Although every effort was made to control all the procedures of the present study, some limitations still existed. Admittedly, the number of participants was small, which was often observed in the research of this field. Moreover, it would be better if there were more items in two grammar tests. Given the limitations of the present and previous studies, there is a big gap for further research to fulfill. It can be seen that studies on the field were rarely available, much less research with students’ proficiency as a moderator variable. In addition, the relationships found between implicit and explicit grammatical knowledge and reading were still inconclusive. These cause a pressing need for the studies in the future. It will be desirable for the future project to have a greater sample size, and to apply or adapt the entire test battery developed by Ellis (2005) in measuring two types of knowledge. Finally, the influence of grammar structures on the relationships also deserves more concern. REFERENCES Alderson, J. C. (2000). Assessing reading. Cambridge: Cambridge University Press. Alderson, J. C., Clapham, C., & Steel, D. (1997). Metalinguistic knowledge, language aptitude and language proficiency. Language Teaching Research, 1(2), 93-121. Bernhardt, E. B. (1991). Reading development in a second language: theoretical, empirical, and classroom perspectives. Norwood, NJ: Ablex. Bowey, J. A. (1986). Syntactic awareness in relation to reading skill and ongoing reading comprehension monitoring. Journal of Experimental Child Psychology, 41, 282-299. Bowles, M. (2011). Measuring implicit and explicit linguistic knowledge: What can heritage language learners contribute? Studies in Second Language Acquisition, 33(2), 247-271. Brimo, D., Apel, K., & Fountain, T. (2015). Examining the contributions of syntactic awareness and syntactic knowledge to reading comprehension. Journal of Research in Reading. doi: 10.1111/1467-9817.12050 Butler, Y. G. (2002). Second language learners’ theories on the use of English articles: An analysis of the metalinguistic knowledge used by Japanese students in acquiring the English article system. Studies in Second Language Acquisition, 24(3), 451-480. Cain, K. (2007). Syntactic awareness and reading ability: is there any evidence for a special relationship? Applied Psycholinguistics, 28, 679-694. Elder, C., & Manwaring, D. (2004) The relationship between metalinguistic knowledge and learning outcomes among undergraduate students of Chinese. Language Awareness,13(3), 145-162. Ellis, N. (2005). At the interface: Dynamic interactions of explicit and implicit language knowledge. Studies in Second Language Acquisition, 27, 305- 352. Ellis, N. (Ed.). (1994). Implicit and explicit learning of languages. London: Academic Press. Ellis, R. (2004). The definition and measurement of L2 explicit knowledge. Language Learning, 54(2), 227-275. Ellis, R. (2005). Measuring implicit and explicit knowledge of a second language: A psychometric study. Studies in Second Language Acquisition, 27, 141- 172. Ellis, R. (2006). Modelling learning difficulty and second language proficiency: The differential contributions of implicit and explicit knowledge. Applied Linguistics, 27(3), 431–463. Erçetin, G., & Alptekin, C. (2013). The explicit/implicit knowledge distinction and working memory: Implications for second language reading comprehension. Applied Psycholinguistics, 34(4), 727-753. Givón, T. (1995). Coherence in text vs. coherence in mind. In M. A. Gernsbacher & T. Givón (Eds.), Coherence in spontaneous text (pp. 59-115). Philadelphia: John Benjamins. Grabe, W. (2005). The role of grammar in reading comprehension. In J. Frodesen & C. Holton (Eds.), The power of context in language teaching and learning (pp.268-282). Boston: Heinle & Heinle. Grabe, W. (2009). Reading in a second language: Moving from theory to practice. Cambridge: Cambridge University Press. Gutierrez, X. (2013). The construct validity of grammaticality judgment tests as measures of implicit and explicit knowledge. Studies in Second Language Acquisition, 35, 432-449. Han, Y., & Ellis, R. (1998). Implicit knowledge, explicit knowledge and general language proficiency. Language Teaching Research, 2(1), 1-23. Hulstijn, J. H. (2005). Theoretical and empirical issues in the study of implicit and explicit second language learning. Studies in Second Language Acquisition, 27, 129-140. Kim, J., & Cho, Y. (2015). Proficiency effects on relative roles of vocabulary and grammar knowledge in second language reading. English Teaching, 70(1), 75-96. Klauda, S. L., & Guthrie, J. T. (2008). Relationship of three components of reading fluency to reading comprehension. Journal of Educational Psychology, 100, 310-321. Koda, K. (2005). Insights into second language reading: A cross-linguistic approach. New York: Cambridge University Press. Krashen, S. (1981). Second language acquisition and second language learning. London: Pergamon. Loewen, S. (2009). Grammaticality judgment tests and the measurement of implicit and explicit L2 knowledge. In R. Ellis, S. Loewen, C. Elder, R. Erlam, J. Philp, & H. Reinders (Eds.), Implicit and explicit knowledge in second language learning, testing and teaching (pp. 94–112). Bristol, United Kingdom: Multilingual Matters. Morvay, G. (2012). The relationship between syntactic knowledge and reading comprehension in EFL learners. Studies in Second Language Learning and Teaching, 2(3), 415-438. Nagy, W. E., Anderson, R. C., & Herman, P. A. (1987). Learning word meanings from context during normal reading. American Educational Research Journal, 24(2), 237-270. Nation, K. (2005). Children's reading comprehension difficulties. In M. Snowling & C. Hulme (Eds.), The science of reading (pp. 248- 265). Malden, MA: Blackwell. Nation, K., & Snowling, M. J. (2000). Factors influencing syntactic awareness skills in normal readers and poor comprehenders. Applied Psycholinguistics, 21(2), 229-241. Renou, J. (2001). An examination of the relationship between metalinguistic awareness and second-language proficiency of adult learners of French. Language Awareness, 10(4), 248-267. Roehr, K. (2007). Metalinguistic knowledge and language ability in university- level L2 learners. Applied Linguistics, 29(2), 173-199. Shin, Y., & Kim, Y. (2012). Assessing the relative role of vocabulary and syntactic knowledge n reading comprehension. Korean Journal of Applied Linguistics, 28(2), 169-198. Shiotsu, T. (2010). Components of L2 reading Linguistic and processing factors in the reading test performances of Japanese EFL learners. Cambridge: Cambridge University Press. Shiotsu, T., & Weir, C. J. (2007). The relative significance of syntactic knowledge and vocabulary breadth in the prediction of reading comprehension test performance. Language Testing, 24(1), 99-128. Snow, C., Griffin, P., & Burns, S. (2005). Knowledge to support the teaching of reading. San Francisco, CA: Jossey-Bass. Urquhart, S., & Weir, C. (1998). Reading in a second language: Process, product, and practice. New York: Longman. Waltzman, D. E. & Cairns, H. S. (2000). Grammatical knowledge of third grade good and poor readers. Applied Psycholinguistics, 21(2), 263-284. Zhang, D. (2012). Vocabulary and grammar knowledge in second language reading comprehension: A structural equation modeling study. The Modern Language Journal, 96(4), 558-575. THE AUTHOR Bui Thien Sao is currently a graduate student at Korea University, majoring in English Education. She received her BA from University of Languages and International Studies, Vietnam National University, Hanoi, Vietnam (Email:

) THE AUTHOR’S ADDRESS Bui Thien Sao Korea University Anam Global House, Room G626-2 Seoul, 136-701 Korea Mobile phone: 010-9531-0228