Effects of extensive reading and translation activities on.pdf

(1)

Effects of extensive reading and translation activities on grammar knowledge and attitudes for EFL adolescents

Juhee Lee

^a^,^*

, Diane L. Schallert

^b^,¹

, Eonsil Kim

^c^,²

aDepartment of Foreign Language Education, The University of Texas at Austin, Austin, USA

bDepartment of Educational Psychology, The University of Texas at Austin, Austin, USA

cDepartment of English Education, Korea University, Seoul, South Korea

a r t i c l e i n f o

Article history:

Received 30 October 2014

Received in revised form 16 April 2015 Accepted 23 April 2015

Available online 22 May 2015 Keywords:

Incidental learning Learning L2 grammar Extensive reading Translation activities EFL adolescents

a b s t r a c t

Despite evidence of an association between L2 reading and grammar knowledge by way of research on the role of grammar in reading comprehension, few scholars have investigated the reverse relation, the contribution of reading to grammatical knowledge. In this study, we investigated the effects of two types of reading instruction,extensive reading and translation, on knowledge of general grammar and specific syntactic features (articles and prepositions) as well as learner attitudes. Participants (N¼124) were adolescent EFL learners in South Korea, who received either extensive reading or translation instruction for two academic semesters. From analysis of their responses to linguistic tests and an attitude survey, results suggested that extensive reading and translation activities had differentiated effects on learners' grammar knowledge and attitudes depending on their L2 proficiency. Although both forms of instruction showed positive gains in grammar knowledge from pretest to later tests, extensive reading seemed to have a negative impact on attitude measures for students of low proficiency but produced positive outcomes for high level students on both attitudes and linguistic measures. Alternatively, translation activities seemed beneficial to all level learners for attitudes, whereas the instruction seemed to have more positive effects for mid-proficiency learners on grammar measures.

1. Introduction

Grammar instruction has been a controversial issue in second language (L2) research and teaching. Although the rise of communicative language teaching led to a downturn of attention to grammar, both by researchers and L2 instructors, recent work has reexamined the important role of grammar in L2 learning and demonstrated that teaching grammar helps learners reach a higher level of language competence (e.g.,Ellis, 2002; Norris&Ortega, 2000). Despite this empirical support, it seems unclear how grammar can best be taught. Additionally,Petraki and Hill (2011)found that although many teachers perceive grammar instruction as necessary and effective, they report insufﬁcient knowledge of grammar or inability to explain grammar adequately. It is not surprising that many teachers rely heavily on the explanations and exercises provided in the text materials (Al-Mekhlaﬁ&Nagaratnam, 2011); thus, some teachers feel tension because of the gap between beliefs that

*Corresponding author. Department of Foreign Language Education, The University of Texas at Austin, 1912 Speedway, Stop D5700, Austin, TX 78712- 0379, USA.

E-mail addresses:[email protected](J. Lee),[email protected](D.L. Schallert),[email protected](E. Kim).

1 Department of Educational Psychology, The University of Texas at Austin, 1 University Station, D5800, Austin, TX 78712, USA. Tel.:þ1 512 471 0784.

2 Department of English Education, Korea University, 145 Anamro, Sungbukgu, Seoul, South Korea.

Contents lists available atScienceDirect

System

j o u r n a l h o me p a g e :w w w . e l s e v i e r . c o m / l o ca t e / s y s t e m

(2)

grammar should be taught in context and their actual practices that focus on grammar rules taught in isolation (Phipps&

Borg, 2009). Perhaps as a consequence, many L2 learners perceive grammar instruction as unsatisfying and boring (Jean&

Simard, 2011).

To date, several approaches to grammar instruction have been examined, such as explicit instruction (e.g.,Macaro&

Masterman, 2006; Tammenga-Helmantel, Arends, & Canrinus, 2014), output-based instruction (e.g., Morgan-Short &

Bowden, 2006), feedback (e.g.,Sheen, 2007), task-based instruction (e.g., Ellis, 2003), and input-based instruction (e.g., VanPatten, 1996, 2002), but most of these have focused on sentence-level grammar. As an exploratory and experimental study, some scholars have also examined other ways of teaching grammar through listening to target structures with visual stimuli (de Jong, 2005) or through watching subtitled foreign movies containing particular grammatical features (Lommel, Laenen,&d'Ydewalle, 2006). In the present study, expanding on previous research on grammar, we examined reading instruction as an instructional approach for teaching grammar.

In contexts where English is a foreign language (EFL) such as South Korea, explicit grammar instruction with a heavy focus on discrete grammatical features has been a dominant instructional approach (Klapper &Rees, 2003). This traditional approach to teaching grammar may enhance learners' knowledge about grammar rules and terminology even though their ability to apply grammar rules in actual language use may not commensurately increase (Al-Mekhlaﬁ&Nagaratnam, 2011). In this respect, learning grammar by reading may suggest a new perspective. Indeed, learning grammar by reading is in line with recent approaches to grammar instruction that reject discrete sentence-level grammar teaching and emphasize using numerous examples of a target structure in context to develop knowledge of form-meaning connections (e.g.,Celce-Murcia&

Olshtain, 2000; Nassaji&Fotos, 2004).Celce-Murcia (2002)argued that“all naturalistic learning ofﬁrst and second languages takes place in context and at the level of discourse rather than the abstract sentence level”(p.119). Reading extended text, which occurs at the level of discourse, provides learners with real or possible contexts in which cultures, minds, and values of a society are embedded, rather than simply decoding words or sentences (Gee, 2001). Furthermore, recent research indicates that grammar knowledge and reading ability are strongly associated in that grammar knowledge is a signiﬁcant predictor of reading comprehension (Jung, 2009; Zhang, 2012). Whereas these studies focused on the role of grammar in reading comprehension, we were interested in the reciprocal effect, the contribution of reading to grammatical knowledge.

Thus, we examined two reading approaches, extensive reading and translation, focusing on their effects on speciﬁc grammatical features and attitudes toward each reading instruction. Extensive reading is characterized as reading a large amount of text at a relatively faster speed with the focus on meaning, not language, and on readingﬂuency (Day&Bamford, 1998), whereas translation involves reading and translating short texts from the target language into the native language, and aiming for accurate reading. These two positions on L2 reading have developed along different historical paths. Extensive reading has been favored as an innovative reading approach by recent researchers for improving L2 learners' linguistic competence (e.g.,Day&Bamford, 2002; Krashen, 2007). The claims are based on the concept that repeated exposure to patterns or structures from reading are cumulatively registered in the implicit learning system and lead to incidental learning of linguistic knowledge (Grabe, 2009). Although other aspects of extensive reading have been investigated, its contribution to grammatical knowledge has received little attention. In contrast, translation has been neglected, even rejected, by L2 researchers and educational leaders due to its close association with the grammar-translation method and the rise of the communicative approach, despite its continued popularity with language teachers around the world (G. Cook, 2010). Yet, most recently, there has been a call for re-examining the role of translation in L2 acquisition (G. Cook, 2010).

Our overarching question, therefore, was the following: Would reading a substantial amount of input (extensive reading) or reading short texts accurately (translation) enhance knowledge of grammar? If so, does such reading help improve general knowledge of grammar as well as of particular syntactic features that are challenging to learn even through formal instruction? To answer these questions was the purpose of the study.

2. Literature review

2.1. Incidental learning

One theoretical rationale for enhancing L2 grammar knowledge through reading comes from studies of incidental learning. The termincidental learningrepresents several meanings. It can refer to a) learning one thing as a by-product while engaging in another activity, b) learning without the intention to learn, or c) engaging in an experimental condition without explicit instruction that a test will follow (Gass, 1999; Hulstijn, 2003). Here, we operationalized incidental learning as learning grammar as a by-product of reading while the learner's attention is focused on language meaning rather than language form.

Many scholars (e.g.,Hulstijn, 2005; Krashen, 1989; Reber, 1996) have claimed that a proportion of L2 acquisition occurs incidentally(without intention to learn the language but immersed in an environment in which the target language is used) and implicitly(without taking language courses or consciously attending to regularities of a language). Although they are not identical, incidental learning is similar to implicit learning in the sense that both are unintentional and uncontrolled (Reber, 1996). The importance of incidental learning comes from its ubiquity in individuals' lives.Marsick and Watkins (2001) argued that informal incidental learning is the main source of adult education because it takes place anywhere and anytime, even when individuals are not conscious of learning. Furthermore,Reber (1989, 1996) andReber and Allen (2000)argued that implicitly learned knowledge tends to exceed explicitly learned knowledge. Similarly,Krashen (1982, 1989)claimed that incidental natural L2 acquisition surpasses consciously learned language in terms of its utility and the amount ultimately acquired.

(3)

Despite some controversy over Reber's arguments (e.g.,P. Robinson, 1996) and Krashen's claims (e.g.,Larsen-Freeman, Long,&Jiang, 1991) as well as researchfindings showing that both incidental (implicit) and intentional (explicit) learning modes are necessary for L2 acquisition (Dekeyser, 2003; Ellis, 2004), incidental learning has been recognized as a significant contributor, especially for vocabulary learning. Because vocabulary acquisition depends crucially on repeated and extensive exposure, and because classroom instruction (intentional and explicit) cannot provide sufficient encounters with new words, the argument has been advanced that much vocabulary learning may occur from reading (e.g.,Eckerth&Tavakoli, 2012;

Schmitt, 2008). By contrast to the rich work on vocabulary acquisition, there have been few empirical studies that have explicitly examined incidental L2 grammar learning from reading, hence our focus here.

2.2. Extensive reading: Effective for all learners?

Despite the numerous studies of extensive reading that have reported positive effects (Haﬁz&Tudor, 1990; Lai, 1993; Lao

&Krashen, 2000; Mason&Krashen, 1997), its contribution to grammar knowledge has received little attention. Among the

few studies that examined the effect of extensive reading on grammatical knowledge, the association was addressed only as a passing reference or the learning gains were assessed through insufficient measurements. For example,Elley and Mangubhai (1983) remarked briefly that fourth graders learning English as a second language (ESL) showed significantly better knowledge of English structures after a seven-month extensive reading program, whereasfifth graders showed no such improvement. One year later, students in both grades outperformed control groups who had been taught by the audio-lingual method. InSheu's (2003)two-semester-long study of the effects of extensive reading on grammar knowledge among junior high school EFL students in Taiwan, those who read either graded readers or books designed for native-speaking children showed significant improvement from pretest to posttest in grammar. In contrast, such improvement was not detected for a control group who had review classes of previous English lessons. However, the grammatical knowledge was assessed with only seven multiple-choice questions to demonstrate linguistic improvement from extensive reading, which may not have been a thorough enough measure to assess overall grammar knowledge.

Additionally, some studies with relatively comprehensive assessment revealed contrastingfindings in terms of the effects of extensive reading on grammar.Yang (2001)assigned adult EFL learners in Hong Kong to an experimental group that read two English novels for 12 weeks, or to a control group that received regular instruction, with a focus on reading, writing, speaking, and listening. Results indicated that the extensive reading group achieved significantly higher scores than the control group on a test of grammar knowledge. By contrast,Yamashita (2008)found that EFL Japanese university students who read English books across 15 weeks did not show significant improvement from pretest to posttest that was designed to measure micro-level syntactic knowledge, suggesting that exposure to a large amount of L2 input may not always be linked to improving one's grammar knowledge. Among native speakers of English who were learning Spanish as a foreign language in a U.S. university,Rodrigo, Krashen, and Gribbons (2004)compared the effects of three approaches: extensive reading, extensive reading plus oral discussion, and traditional instruction with a heavy focus on grammar and vocabulary (serving as a control).

Results showed that both reading groups had higher gain scores than the control group on multiple-choice grammar tests, whereas cloze test measures revealed no signiﬁcant group differences. In addition to mixedﬁndings, previous studies did not specify particular domains of grammar knowledge. It seemed to us necessary to examine not only whether incidental learning of grammar would occur from extensive reading, but also which particular grammar features would improve.

Furthermore, previous studies of extensive reading did not track whether effects depended on learners' proﬁciency levels.

In terms of vocabulary learning from reading,Zahar, Cobb, and Spada (2001)claimed that the degree of incidental learning of new words from reading may depend on learners' proficiency because learners at a higher proficiency level may acquire new words in fewer exposures. Likewise, the effect of extensive reading on grammar may be different depending on learners' L2 proficiency, a claim that had not heretofore been investigated.

2.3. Translation: Really all that bad?

In contrast to the widespread support for extensive reading, the issue of whether the use of translation activities should be minimized has been controversial (V. Cook, 2001). Due to the growing importance of communicative competence, govern- ments and educational leaders around the globe have vigorously implemented the communicative approach through nationwide policies and educational reforms; as a consequence, translation has been widely denigrated in the L2 classroom (Nishino, 2012). However, the top-down educational initiatives have not been uniformly sustained in L2 classrooms because teachersﬁnd it difﬁcult to hold onto the principles of the communicative approach and often return to traditional instructional approaches, such as the grammar-translation method, due to contextual constraints in implementing communicative language teaching (Sato&Kleinsasser, 1999). Thus, translation has continued to be used for teaching and measuring linguistic competence for many years, especially in foreign language teaching contexts (Tsagari&Floros, 2013).

Yet, because of long-standing criticisms against the grammar translation method in the second language acquisition (SLA) ﬁeld, the beneﬁts of translation may have gone unrecognized, particularly the merits that translation activities may offer when integrated with other reading components. For example,Malloy (2001)suggested that translation can be a student-centered and compelling literacy activity for adolescents in foreign language learning contexts. By examining transcripts of translation activities in real classrooms, Malloy argued that translating events seemed to help U.S. middle school learners of German read and understand L2 texts containing vocabulary and structures that were beyond their current level. Additionally, through

(4)

think-aloud interviews of U.S. college learners of French,Kern (1994)found thatmental translation(the quick mental noting of what a word or phrase means in one's native language) was inevitable and served to scaffold understanding L2 texts. He claimed this was particularly true for beginning level learners, because“it facilitated semantic processing and permitted consolidation of meaning that would otherwise remain fragmented if represented in L2 form”(p. 447). Although the activities in our study were closer to the conventional use of translation in involving actual written products, Kern's participants were nevertheless engaging in the same cognitive processing of moving from one language to the other by making meaning.

Recently, some scholars have continued this emphasis on the positive role that translation can play in language teaching and assessment. There have been entire chapters in three recent books devoted to translation-related issues in L2 acquisition (G. Cook, 2010; Tsagari&Floros, 2013; Witte, Harden,&de Oliveira Harden, 2009), with some report on the few available empiricalfindings on whether L2 learners' grammar skills could improve through translation. For instance,Korosec (2013) investigated the contribution of translation exercises to grammatical knowledge among EFL university students in Slovenia. Although preliminary results showed no significant effects, Korosec suggested that more positive effects of translation may be evidenced in a forthcoming, more complete analysis. Similarly,Marlein (2009)examined whether English- speaking university students taking a German class benefited from word-by-word translation when learning specific syntactic structures. Despite carefully designed measures and conditions, participants who completed all measures and treatment were so few that results, though promising, seemed equivocal. Thus far, despite the recent attempts to show the positive effects of translation activities on the development of grammatical knowledge, the extant research has not provided clear and convincing evidence to support this claim. Research with adequate numbers of participants involved in a sufficiently long period of instruction seem warranted to assess the role of translation in learning grammar.

2.4. Attitudes toward extensive reading and translation

In evaluating the adequacy of a particular reading instruction approach, it is crucial to examine whether the approach improves not only linguistic skills but also engenders positive attitudes because reading depends on positive attitudes to become a frequent activity in a language learner's life (Lee&Schallert, 2014).Day and Bamford (1998, 2002)have argued that the reason extensive reading promotes positive attitudes towards reading is because students choose what they want to read at their levels, and reading is its own reward. Learner attitudes related to extensive reading have been empirically studied, suggesting that extensive reading, in general, generates feelings of joy while reading and improves motivation to read (e.g., Al-Homoud&Schmitt, 2009; Judge, 2011; Taboada&McElvany, 2009). In this extensive literature, few studies have made a direct test of different reading instruction approaches on attitudes. One such study is that ofAl-Homoud and Schmitt (2009) who demonstrated that EFL university students in an extensive reading group reported more positive attitudes towards reading than the intensive reading students who were required to read short texts and answer comprehension questions. Yet, a contrastingﬁnding came fromRobb and Susser (1989)who reported that students in an EFL university extensive reading class did not hold signiﬁcantly better attitudes toward the reading approach than the skills-based group who read short texts and did exercises designed to learn reading skills. Thus, there is a need for more studies that examine and compare students' affect in response to extensive reading and to reading many short texts accurately, particularly for adolescent students who have less experience in extensive reading and are at an early stage of L2 learning.

In addition to examining attitudes toward extensive reading, we were also interested in examining learner perspectives and attitudes associated with translation. Despite the widespread criticism of the grammar-translation method discussed in previous work, many of its features have been recognized as pervasive in language teaching classrooms throughout the ages, and thousands of language teachers continue to regard translation as“a safe and reliable option”(G. Cook, 2010, p. 14). For our comparison group that included translation exercises as a component of reading instruction, we were interested in the attitudes that students would report toward translation activities, as no study to our knowledge has addressed such learner attitudes.

2.5. Research questions

Grounded in this literature, we addressed the following two research questions:

1. What are the relative effects of two different reading instruction approaches, extensive reading and translation, on L2 grammar development for EFL adolescents, and does the impact of these two reading approaches differ depending on L2 proﬁciency levels?

2. What are students' attitudes towards extensive reading and translation, and do attitudes differ depending on proﬁciency levels?

3. Method

3.1. Participants and settings

Participants were 124 middle school students (61 boys, 63 girls; age range 13e14) (equivalent to U.S. 8th grade), all native speakers of Korean and learning English in South Korea. The participants had received approximatelyﬁve years of formal English instruction (3rde7th grades). Because of limited class time, learners' beginner level reading ability, and large class size

(5)

(30e35 students per class), most reading instruction in middle school tends to focus on interpreting short passages of English textbooks in a teacher-centered manner. Accordingly, reading for pleasure in English is not likely to occur for adolescent learners, particularly in this urban middle school that drew from the surrounding low socioeconomic neighborhoods. Indeed, according to the background questionnaire, 73% of the students indicated that they almost never (53%) or no more than once per month (20%) read something in English for pleasure, indicating that our participants rarely read English texts outside of school. Additionally, 94% of the students had never (91%) or for less than a month (3%) visited an English-speaking country.

At the beginning of the study, participants numbered 142, one-half of the entire 8th grade. Of the 142 participants, however, 124 completed the year-long treatments and the post-survey (attitudes). Among these students, 120ﬁnished all grammar measures.

3.2. Treatments

The treatment contrast involved replacing a portion of the regular English instruction for two semesters, with each treatment acting as a control for the other. Because the study took place in the context of regular instruction, it was important that all students receive some sort of instruction rather than having a control group engaged in no instruction whatsoever, a choice we saw as unethical in this situation. All instruction, including that associated with the two treatments (extensive reading and translation activities), was in the hands of the same teacher who taught English to all participants. English was taught four times per week for 45 min each time. One day per week, students received either extensive reading or translation instruction, with students randomly assigned to treatment in intact class groupings, three classes to the extensive reading group (n¼75) and two classes to the translation group (n¼49). For the other three days, regardless of treatment assignment, students received the same regular instruction with the same textbook mainly focusing on practicing conversational dialogs and interpreting short reading passages with explanations of vocabulary and grammar points.

3.2.1. Extensive reading

This treatment involved helping students choose an English book based on their interest and proﬁciency level from the school's library where approximately 900 English graded readers were available. Although no strict criterion deﬁnes what it means to read extensively, one book per week is typically recommended (Day&Bamford, 2002), and this goal was discussed with the students. After reading their English book in class, students wrote short summaries or responses in Korean in a notebook, as a check that they were actually reading. The teacher typically returned students' notebooks, one week later, with encouraging comments focusing on the book's content. The number of books read for two semesters ranged from 28 to 32.

3.2.2. Translation

Every week, students in the translation group received one short English reading passage (on average 100e200 words) that they translated into their native language. These passages were selected by the teacher from study materials that she judged as containing vocabulary and structures slightly advanced for the students. It should be noted that the translation activities given to the students in this study were different from what is typically involved in Grammar Translation Method (GTM). Whereas with GTM model, the teacher presents target language structures and then students practice these grammatical rules and apply them by translating sentences between the native language and the target language (Chang, 2011), in the present study the teacher did not teach specific grammar structures, nor did she ask students to apply particular rules in translation tasks. Consistent with the extensive reading treatment, the students were responsible for their own learning, but had a portion of the class session devoted to asking the teacher questions about their translation difficulties. Thus, they worked on their own for 30 min, translating the passage into Korean with help from dictionaries or nearby classmates. During the remaining 15 min, the students asked the teacher to explain difficult vocabulary or sentence structures. Class concluded with students submitting their individual translations that the teacher checked for accuracy, generally quickly and globally, and returned the following week with feedback on any obvious errors.

3.3. Instruments

3.3.1. Background questionnaire and attitude survey

A questionnaire was administered at the beginning of the study to gather demographic information (gender, experience of visiting English-speaking countries, and frequency of pleasure reading in English). After the yearlong treatment, a post-survey was distributed to assess students' attitudes. This survey included 16 statements that students rated on 5-point Likert scales (1¼“I strongly disagree”; 5¼“I strongly agree”). We used the same items for both groups, changing only the name of the instruction mentioned appropriate to their assigned treatment (“My English reading ability improved due to the extensive reading program” became“My English reading ability improved due to the translation program”). Adapted from previous studies on extensive reading (Takase, 2007; Yang, 2001), items addressed students' self-assessment of their achievement, overall feelings and attitudes towards extensive reading (or translation), and their preferences about teacher-dependent and independent reading.

3.3.2. Grammar tests

Rather than relying solely on a global undifferentiated measure of grammar knowledge as common to previous studies, we used measures of two particularly difﬁcult grammatical features as well as a general grammar knowledge measure. We were

(6)

informed by the categories ofBitchener, Young, and Cameron (2005)andFerris and Roberts (2001)of grammar features recognized as difficult for learners of English. Specifically,Bitchener et al. (2005)identified prepositions and the combination of indefinite and definite articles as the most frequent error categories in ESL students' writings. Another reason that we chose prepositions and articles was that these grammatical features remain challenging to master even when they are explicitly taught through formal instruction. When a learner'sfirst language does not have prepositions or articles, as with Korean, errors are even more likely (Braine, 2002). For example, Hendricks (2010) stated, “prepositions are often conceptually different from one language to the next, and direct translation cannot be relied on”(p. 24). That is, teaching English prepositions and their correct usage is demanding due to changeable meanings that depend on context and to metaphorical usage related to directions and relationships. Similarly,Robinson (2010)argued that the English article system is one of the most challenging areas for learners, a difficulty that can persist even at later stages. In spite of the availability of some rules governing their use, correcting or teaching articles is difficult because usage depends on context and on nuances of text meaning.

For our study, tests measured three aspects of English grammar, with 70 items at each testing point, pretest, mid-test, and posttest. Thus, we tested separately participants' knowledge of the proper use of prepositions (20 items) and articles (20 items) and collapsed other grammar categories asgeneral grammar knowledge(30 items), including tense, verbs, modals, passive, inﬁnitive, gerund, and relative pronouns. All tests were distributed in paper form and used multiple-choice formats, a testing method often used in previous extensive reading studies to measure knowledge of grammar (Elley&Mangubhai, 1983; Sheu, 2003; Yang, 2001). All items were selected from ESL/EFL testing websites and English teaching materials and reviewed by two English teachers for level and appropriateness, with items modiﬁed based on their feedback. Also, items were carefully chosen to be valid measures of their particular domain. For example, the preposition tests included prepositions indicating place, time, direction, cause, manner, and amount, but excluded idiomatic, vocabulary-dependent prepositions, and prepositional phrases, such ascapable ofandresult in, that simply need to be memorized on an item-by-item basis:

A box of books was ______ the table./The ball is ______ the table.

A. on - under B. towards - on C. up - under D. at - of

Similarly, the article tests measured appropriate usage of deﬁnite, indeﬁnite, missing, and unnecessary articles in sentences, rather than requiring advanced vocabulary ability:

Amy is looking for ____ job now./Did Amy get ____ job she applied for?

A. the - the B. (nothing) - (nothing) C. a - the D. a - a

Finally, the tests of general grammar excluded questions on articles and prepositions and included a variety of other grammar rules based on a list of most frequently occurring errors among English language learners (Bitchener et al., 2005). To ensure the equivalency of the pretest, mid-test, and posttest, we simultaneously assigned items to the three tests by considering vocabulary, length of each context sentence, and equivalent grammar point. It should be noted that like most English teachers in Korea, the teacher's regular instruction was heavily guided by the textbook, which for grammar sections included approximately 20% of the linguistic forms in the general grammar tests, indicating that 80% of the grammar points that we tested were not explicitly taught in regular classes. Furthermore, little focus was paid to prepositions and no attention to articles in regular instruction.

3.4. Procedure

At the beginning of the year, students had been assigned to each of theﬁve 8th grade English classes in the study based on their English grades for the previous semester. We used a one-way ANOVA to establish that each class grouping did not differ in mean English scores,F(4, 119)¼.25,p>.05. We then randomly assigned each class grouping to either treatment, three to the extensive reading and two to the translation treatments, and tested via an independent t-test whether there were initial differences between the two treatment groups. No statistical differences in English scores were found,t(122)¼.44,p>.05.

The pre-measures were administered during three regular 45-min-classes across two weeks: ﬁrst, the articles and prepositions test; second, general grammar; andﬁnally, the background survey. Mid- and posttests were conducted similarly with a four-month approximate interval between tests. The post-survey was administered after the school year had ended.

Except for the one-month summer vacation and the period of mid-terms andﬁnals, students received either extensive reading or translation instruction for 45 min every week for seven months.

3.5. Data analyses

This study included three factors, one within-subject (time) and two between-subject factors (treatment and proficiency), with a pretest-treatmentemid-test-treatmenteposttest design. We operationalized L2 proficiency as the students' English grades from their previous school semester reported as percentages and reflecting assessment of listening, speaking, reading, writing, and grammar skills. The reliability and validity of the proficiency groupings based on school English achievement had been established in a previous study (Lee&Schallert, 2014), showing a high correlation (r¼.89) between previous semester

(7)

English grades and the mean scores of two English proﬁciency tests (100 items total). Thus, we classiﬁed the students into three levels based on school English grades: Low (range: 66e85.2;M: 80.2), Middle (range: 85.3e91.7;M¼88.4), and High (range: 91.8e100;M¼95). The cut points were identically applied to both treatment groups.

For each grammar measure, we conducted a repeated-measure 3 (Time)2 (Treatment)3 (Proﬁciency) analysis of variance (ANOVA). We followed up all signiﬁcant interaction effects with post hoc analyses using a Bonferroni correction. For the post-survey, we used factor analysis of students' attitudes toward their particular treatment instruction. As we report below, the factor analysis reduced the 16 items to four factors that then became dependent measures submitted to a MANOVA.

4. Results

Table 1displays descriptive statistics of the tests, including overall mean scores and standard deviations. Skewness and kurtosis indices indicated that the data were normally distributed. Test scores of general grammar were converted to percentage equivalents. Because the results of the article and preposition tests had been summed at pretest with subscores permanently unrecoverable, we treated these two measures differently, converting the mid-test and posttest scores on a scale of 50% correct so that their combined scores were on a 100% metric. Thus, the pretest scores for knowledge of prepositions and articles were reported as one summative score, whereas scores on the mid-test and posttests of articles and prepositions were separately calculated to test for improvement in each grammar domain.

4.1. Research question 1

Descriptive statistics of the grammar test scores are displayed by group and proficiency level inTable 2. In terms of general grammar, results of a repeated-measures ANOVA indicated that the effect of time was statistically significant, whereas there was no significant time-by-treatment interaction, suggesting that students in both groups significantly improved in general grammar knowledge over time at similar rates.

With regard to the combined article and preposition measure, however, there was a significant time effect, and a significant time-by-treatment interaction effect, favoring the extensive reading group. A decomposition of the mid-test and posttest scores suggests that the benefits to the extensive reading group came from their improved knowledge of articles, not prepositions (see bottom two rows ofTable 2). The effect size of the time-by-treatment interaction was relatively small (partialh²¼.03), with approximately only 3% of the total variance in the combined articleþpreposition scores accounted for by the different treatments over time.

We also examined whether the impact of instruction was different depending on English proficiency levels. Regarding general grammar, results of a repeated-measures ANOVA indicated a significant two-way interaction between time and proficiency with a medium effect size (partialh²¼.05), whereas the three-way interaction among time, treatment, and proficiency was not significant. Similarly, for the combined articleþpreposition measure, the time-by-proficiency interaction was significant with a medium effect size (partialh²¼.07), whereas the timetreatmentproficiency interaction was not statistically significant. Thus, students at different proficiency levels showed different patterns of improvement on tests of knowledge of general grammar and of prepositions and articles across time, although the patterns were not statistically different between the extensive reading and translation groups.

To examine further the significant time-by-proficiency interactions, post hoc comparisons revealed that for both general grammar and combined articleþprepositions tests, middle and high level learners improved substantially, whereas low level students showed no significant improvement across time. To explore whether the improvement of the middle and high level students differed by level and by treatment group, we examined the decomposed three-way interaction coming from the overall repeated-measures ANOVA. For extensive reading, high and middle level students improved significantly in both

Table 1

Descriptive statistics: grammar tests (N¼120).

Tests Items MineMax^a Mean SD

Pre-general grammar 30 21e90 42.7 11.8

Mid-general grammar 30 15e79 46.5 12.7

Post-general grammar 30 16e79 48.5 13.4

Pre-combined^b(artþprep) 40 20e65 40.7 8.5

Mid-combined^b(artþprep) 40 25e73 45.4 10.8

Post-combined^b(artþprep) 40 25e78 49.1 13.1

Mid-article 20 10e38 23.0 5.3

Post-article 20 13e48 27.1 7.4

Mid-preposition 20 8e38 22.5 7.4

Post-preposition 20 8e40 22.0 8.4

Note. Min: Minimum score; Max: Maximum score;SD: Standard deviation.

aIn percentages.

b The combined score of articles and preposition tests was used in analyses.

(8)

general grammar knowledge and knowledge of combined articlesþprepositions. In particular, the change scores from pretest to posttest for high level learners were almost twice the change scores of middle level learners for both grammar measures.

However, for the translation group, middle level students signiﬁcantly improved on both grammar measures, whereas high level students showed signiﬁcant improvement only on the combined articleþprepositions test.Fig. 1depicts these different patterns by group and level.

4.2. Research question 2

At the end of the year, students' attitudes toward extensive reading and translation instruction were surveyed using 5- point Likert ratings. Factor analysis with Varimax rotation was conducted on the post-survey responses of 124 students who had experienced either treatment for two semesters. With a minimum eigenvalue of 1 and a factor loading cut point of .40 or above (Stevens, 2009), four factors were extracted, accounting for 66.8% of the total variance.Table 3presents the factor loadings of items, and the mean scores and standard deviations of each factor.

Factor 1 received strong loadings from six items related to students' self-assessed linguistic improvement after partici- pating in their respective treatment. Factor 1, thus, was labeledPerceived improvement. Factor 2 was namedPositive affect Table 2

Means (standard deviations) of grammar tests by groups and proﬁciency levels.

Tests Levels Extensive reading (n¼73) Translation (n¼47)

Pre Mid Post Pre Mid Post

General grammar Low 38.4 (9.3) 37.0 (9.2) 40.1 (12.5) 40.2 (12.1) 38.2 (10.9) 42.2 (7.1)

Mid 40.7 (11.1) 47.7 (10.8) 45.9 (11.8) 38.7 (7.7) 45.9 (6.6) 47.6 (9.1)

High 50.1 (13.0) 57.3 (12.1) 59.9 (12.4) 47.6 (11.8) 50.9 (12.0) 54.4 (14.0)

Total 43.3 (12.3) 47.6 (13.6) 49.0 (14.8) 41.8 (11.1) 44.8 (11.0) 47.8 (11.2)

Combined (artþprep) Low 37.8 (9.2) 37.6 (8.4) 41.7 (11.2) 37.8 (5.8) 40.9 (8.3) 40.5 (9.0)

Mid 42.1 (7.3) 44.0 (9.0) 49.4 (11.7) 37.4 (6.0) 46.2 (7.3) 46.8 (9.1)

High 44.0 (10.4) 51.4 (12.4) 58.6 (13.5) 44.8 (6.7) 54.5 (7.7) 56.4 (11.6)

Total 41.4 (9.4) 44.5 (11.6) 50.1 (14.0) 39.7 (6.9) 46.9 (9.3) 47.5 (11.6)

article 22.4 (5.5) 27.8 (7.7) 23.8 (4.9) 25.9 (6.7)

preposition 22.1 (7.8) 22.3 (8.6) 23.0 (6.7) 21.6 (8.0)

Note. In the extensive reading treatment,ns for proficiency groups were: low¼24, mid¼23, and high¼26. In the translation treatment, low¼16, mid¼17, and high¼14; F tests indicated the following: for general grammar, there was an effect of time,F(2, 228)¼12.7,p<.05, time-by-treatment interaction,F(2, 228)¼.34,p>.05, and time-by-proficiency interaction,F(4, 228)¼3.2,p<.05; for the combined article and preposition measure, there was an effect of time,F(2, 228)¼37.3,p<.05, time-by-treatment interaction,F(2, 228)¼3.7,p<.05, and time-by-proficiency interaction,F(4, 228)¼4.6,p<.05.

n o i t a l s n a r T g

n i d a e R e v i s n e t x E

30 35 40 45 50 55 60

Pre Mid Post

General grammar

High Mid Low

30 35 40 45 50 55 60

Pre Mid Post

General grammar

High Mid Low

30 35 40 45 50 55 60

Pre Mid Post

Combined (arti+prep)

High Mid Low

30 35 40 45 50 55 60

Pre Mid Post

Combined (arti+prep)

High Mid Low

Fig. 1.Grammar achievement over time by groups and proficiency levels;Ftests from multivariate simple effects analyses of time within each level indicated the following: for the extensive reading group, the time effect was significant on both general grammar knowledge (high group:F(2, 113)¼7.8,p<.05, partial h²¼.12; middle group:F(2, 113)¼4.1,p<.05, partialh²¼.07) and knowledge of combined articlesþprepositions (high group:F(2, 113)¼23.6,p<.05, partial h²¼.29; middle group:F(2, 113)¼5.5,p<.05, partialh²¼.09). For the translation group, the time effect was significant for the middle group on both general grammar,F(2, 113)¼4.6, partialh²¼.08, and combined (article and preposition) test,F(2, 113)¼9.0, partialh²¼.14, and for the high group only on the combined (article and preposition) test,F(2, 113)¼10.1, partialh²¼.15.

(9)

because three items associated with reported interest, enjoyment, and engagement loaded on this factor, with one item (#11) reverse-coded. Factor 3, labeledTeacher-dependent reading, included three items that assessed whether students preferred teacher-dependent and language-focused activities. Note that we decided to keep Item 14, even though it indicated cross- loading on two factors (Factor 2 and Factor 3), because of its conceptual accordance with Factor 3 and a larger factor loading value (.53) on Factor 3. Finally, Factor 4 received factor loadings from four items that described whether students prefer independent reading without teacher help and without translation. This factor, thus, was namedIndependent reading. Factor reliabilities demonstrated reasonable internal consistency ranging from .59 to .93 (seeTable 3).

Table 4displays the means and standard deviations for each attitude factor by group and proficiency level. To test whether the mean differences were statistically significant, we conducted a MANOVA, including treatment and proficiency as independent variables and the mean scores of each factor as dependent variables. Results indicated that the two groups were significantly different for all factors except Teacher-dependent reading. Specifically, in terms of Perceived improvement, the translation group reported being helped by the translation activity to a greater degree than the extensive reading group reported about their extensive reading, and this difference was significant. The treatment-by-proficiency interaction effect on this factor was also significant, indicating that self-perceived improvement was different depending on treatment groups and L2 proficiency levels. In particular, students in the low proficiency level of the translation group reported highest means onPerceived improvement, whereas their counterparts in the extensive reading group showed lowest average score. Similarly, forPositive affect, there was a significant main effect for treatment, with the translation group reporting that they enjoyed the given instruction more than students in the extensive reading group. Here, the treatment-by-proficiency interaction was not significant.

Another interestingﬁnding related to students' preferences for reading method. Whereas both groups had similar levels of preferences as indicated by their scores on theTeacher-dependent reading, they reported different feelings aboutIndependent reading, with the extensive reading group reporting a higher mean than the translation group. Thus, students in the extensive reading treatment seemed to like reading independently and extensively and to believe that they could improve their reading ability without teacher help, more so than students in the translation group. This group difference was statistically signiﬁcant.

5. Discussion

Before discussing our results, we re-address the kinds of instruction represented in our study. For these Korean adolescent students, the extensive reading instruction was a relatively novel task. Likewise, the way that translation was incorporated into classroom activities was also new, with students working either alone or with peers as they produced aﬁrst translation of a passage, followed by a teacher-led explanation. Thus, both forms of instruction challenged the students' expectations about typical English instruction, and our study addressed whether such reading instruction would have an effect on growth in students' grammar knowledge and their attitudes toward reading instruction.

5.1. General grammar knowledge

First, on the general grammar test, students in both groups improved similarly and signiﬁcantly over time with a large effect size. That is, averaging across proﬁciency levels, the learning gains from extensive reading were similar to those from

Table 3

Five-factor Solution for students' Attitudes towards reading instruction (N¼124).

Statement Loading Mean SD

Factor 1: Perceived improvement(a¼.93) 3.01 .81

3. My English writing ability improved due to the extensive reading (or translation) program. .86 2.82 .94 1. My overall English ability improved due to the extensive reading (or translation) program. .85 2.98 .92 2. My English reading ability improved due to the extensive reading (or translation) program. .84 3.17 .89 5. My English vocabulary knowledge improved due to the extensive reading (or translation) program. .83 3.13 .99 4. My English grammar knowledge improved due to the extensive reading (or translation) program. .80 2.82 .97

10. The program was helpful for my English learning. .74 3.16 .96

Factor 2: Positive affect(a¼.59) 3.34 .76

11. The program lowered my interest in English reading. (reversed) .76 3.57 .95

7. The program was fun and enjoyable. .67 2.87 1.1

6. I was engaged in the program sincerely. .53 3.57 1.1

Factor 3: Teacher-dependent reading(a¼.68) 2.81 .88

8. I like my teacher to explain and interpret English texts, including vocabulary and grammar, rather than reading books on my own.

.83 2.89 1.1

12. I like a teacher who pushes me to take vocabulary tests regularly and practice reading with assignments. .70 2.61 1.21 14. I like to read short English passages intensively through translating and studying vocabulary and grammar in details. .53 2.94 1.1

Factor 4: Independent reading(a¼.71) 2.97 .75

9. I can improve English reading ability even without teachers' help. .80 2.75 .99

13. I like to read English books (materials) by myself, rather than interpreting it with teachers. .79 2.91 1.0 15. I like a teacher who allows me to read books by myself and helps me when I ask for help. .61 3.56 1.0

16. I like to read in English without translating it into Korean. .55 2.65 1.0

Note. Percentage variance explained: total variance, 66.75; Factor 1, 27.57; Factor 2, 15.68; Factor 3, 11.75; Factor 4, 11.75.

(10)

translation. It is interesting to note that the extensive reading group was not explicitly taught general grammar but only asked to read a large number of books. Similarly, the translation students were not receiving explicit grammar instruction, although they did engage with their teacher's explanation, provided in Korean, to clarify any questions the students had following their translation work. Our results may suggest that incidental learning of general grammar can occur from both reading a large amount of text for meaning (extensive reading) and reading short texts for accuracy (text translation). Although there is always an expectation of improvement across a seven-month span for students engaged in any kind of classroom instruction and we must acknowledge that the improvement we saw was at least in part attributable to other aspects of English instruction, we are encouraged by the effectiveness of these two reading approaches, especially as 80% of the general grammar features we tested did not appear in regular classroom instruction. Even the 20% of linguistic forms taught in class did not exactly match the forms tested in the general grammar exams, though they fell under the same grammar categories. Because grammar teaching in school was typically limited to explaining certain rules included in the textbook and providing drill-and- practice of these rules for school exams, it would have been difﬁcult for students to recall all they had learned and be able to apply the knowledge to our mid-test and posttest. Even though some students were attending private English institutes, the instruction they encountered there was likely to focus on the content covered in school and on interpreting short texts.

Indeed, some institutes in the low-income neighborhood where the school was situated served more as after-school care centers where supervision of students' self-study and homework was offered with no direct instruction provided. Thus, the improvement we saw for both types of reading instruction seems at least in part attributable to these treatment activities added to regular instruction.

Thisﬁnding contrasts with reports of the few studies showing that extensive reading groups reported higher gains on structure knowledge than control or comparison groups. However, in these previous studies, the comparison groups did not engage in reading per se but received either regular instruction (Rodrigo et al., 2004; Yang, 2001) or language instruction taught by an audio-lingual method (Elley&Mangubhai, 1983). Moreover, inYamashita's (2008)study, there was no control group, and participants showed almost no improvement from pretest to posttest of grammatical knowledge through extensive reading. Unlike these previous studies, our study compared the effects of extensive reading on growth in general grammatical knowledge to another viable reading approach. There have been a few studies that compared extensive reading to other reading approaches, such as reading skills-building instruction (Robb&Susser, 1989) or intensive reading (Laufer- Dvorkin, 1981), in the development of reading comprehension, and in these studies, extensive reading was as effective, but not more so, than other reading approaches. In line with these studies, the positive effect of extensive reading in our study was similar to that of translation activities in improving knowledge of general grammar.

One reason ourfindings may prove particularly interesting to SLA researchers is not only because there has been little research of the effects of extensive reading on grammar knowledge but also because recent research has rarely discussed the impact of translation activities on L2 learning. Our study offers empirical evidence that text-level translation exercises may be useful in improving students' general grammar knowledge. It is worth considering why translation activities benefitted general grammar knowledge. Studies of translation have suggested that university-level foreign language students may learn grammar from sentence- or word-level translation activities (Korosec, 2013; Marlein, 2009), although these studies were either preliminary or involved very few participants.Malloy (2001)also indicated that American adolescent learners of German benefitted from translation activities in understanding complex structures. Likewise, our participants, Korean EFL adolescents, may have grown in general grammar knowledge through text-level translation because they had to work through different grammar structures to write what they understood into Korean.

Such translation activities seem similar to interpretation tasks (e.g.,Ellis, 1995; VanPatten, 2002) in that both activities require learners to process language forms as they construct meaning. VanPatten (2002) explained that as L2 learners interpret, they can notice certain linguistic features. Despite differences between translation and interpretation tasks, we assume that the learning gains from translation may be explained by VanPatten's hypothesis that comprehension-based activities for which learners need not produce target structures can still promote the noticing of grammatical features.

Table 4

Means and standard deviations of four Attitude factors for treatment groups and proﬁciency levels.

F1

Perceived improvement

F2

Positive affect

F3

Dependent reading

F4

Independent reading

Groups Level n M SD M SD M SD M SD

Extensive reading Low 24 2.25 .99 2.89 .86 2.56 .89 2.74 .83

Mid 24 2.97 .48 3.39 .77 2.72 .96 3.11 .47

High 27 3.25 .70 3.31 .71 3.02 .90 3.37 .75

Total 75 2.84 .85 3.20 .80 2.78 .93 3.09 .74

Translation Low 17 3.40 .41 3.41 .70 2.69 .95 2.79 .79

Mid 18 3.17 .80 3.59 .62 2.81 .77 2.69 .74

High 14 3.27 .71 3.64 .63 3.17 .64 2.91 .67

Total 49 3.28 .66 3.54 .65 2.87 .81 2.79 .73

Note.n: number of students in each level;M: Mean;SD: Standard deviation; F tests indicated the following: forperceived improvement, there was a treatment effect,F(1, 118)¼12.0,p<.05, partialh²¼.09, and a treatment-by-proﬁciency interaction effect,F(2, 118)¼7.0,p<.05, partialh²¼.11; forpositive affect, there was a treatment effect,F(1, 118)¼6.9,p<.05, partialh²¼.06; and forindependent reading, there was a treatment effect,F(1, 118)¼4.4,p<.05, partial h²¼.04.

(11)

5.2. Knowledge of articles and prepositions

In terms of knowledge of articles and prepositions, the effects of extensive reading and translation activities were different from those found for general grammar. Although both groups signiﬁcantly improved over time on knowledge of English article usage, the extensive reading group outperformed the translation group. Yet, neither group showed improvement on use of prepositions. Our results suggest that reading a large amount of input may help learners incidentally acquire knowledge about article usage better than translation activities. Extensive reading may provide numerous examples of articles in context, allowing their usage to be implicitly or incidentally registered in an individual's knowledge system (Grabe, 2009), whereas short text translation activities may not provide adequate exposure to the complex realities of article usage.

However, even repeated exposure to prepositions did not help with knowledge of prepositions.Ferris and Roberts (2001) classiﬁed errors of articles astreatableerrors and errors of prepositions asuntreatableerrors, suggesting that untreatable errors are more likely to be resistant to instruction. Later,Ferris (2011)suggested that different types of interventions, such as explicit grammar instruction and feedback (indirect and direct), should be provided to deal with different error types. In terms of prepositions, explicit instruction as well as extensive repeated exposure to proper usage may be essential for learning.

5.3. The role of proﬁciency in gaining grammar knowledge

In addition, taking proficiency into consideration revealed interesting variations in these results. Because the interaction effect of time-by-proficiency was significant on both general grammar and combined prepositions and articles measures, we proceeded to our post hoc and exploratory analyses to examine the differences among proficiency levels in each treatment group. Results indicated that the high and middle proficiency groups benefitted more in grammar knowledge from extensive reading than did the low group, and the high proficiency group's improvement was greater than that of the middle group.

However, results suggested that, for the translation group, the middle proﬁciency group improved as much as or more than the high level group in their grammar knowledge.

The high proficiency group's significant improvement from extensive reading is in line withfindings ofZahar et al. (2001) that more advanced learners tend to learn new vocabulary within a few encounters from extensive reading, whereas lower level learners cannot so easily learn new words from reading. Additionally, students at higher proficiency levels may read more quickly and thus have more opportunities to encounter new language forms, both vocabulary (in the case of participants in Zahar et al.) and grammatical structures in text (in the case of our participants), leading to greater gains when compared to lower level learners within the same period. Thisfinding also suggests that readers may need to have reached a certain level of L2 proficiency, alinguistic thresholdin the words ofCummins (1979), to experience gains in grammar knowledge from extensive reading. Low levels of lexical and syntactic knowledge may prevent readers from experiencingfluent and plea- surable reading, and impede novice learners from learning linguistic skills, such as vocabulary, from extensive reading (Coady, 1997). In contrast, translation activities may be more effective for L2 learners with mid-level proficiency because translation can be used to clarify text meaning and structure (Kern, 1994). Considering that the middle group showed significant gains in both grammar measures from translation activities whereas the high level group significantly improved only on the combined articles and preposition measure, we speculate that for high proficiency students, translating every sentence into the native language may be a redundant process that does not advance their L2 knowledge to the same degree as extensive reading does.

5.4. Attitudes toward reading instruction

The second research question addressed gains in students' attitudes attributable to the two kinds of instruction, again differentiated by proficiency levels. Students in the low level proficiency group who had received the extensive reading instruction reported significantly lower ratings in perceived improvement, positive affect, and preference for independent reading compared to the high level group. Even though a number of beginner level English books were available, the content of the graded readers appropriate for these proficiency level tended to be too simple for adolescent readers, who preferred to read complicated mystery stories in Korean, and would have wanted to readHarry Potterrather than the children's books to which they were funneled. Consistent with their negative attitudes toward extensive reading and independent reading, students in the low level group did not show significant improvement in grammar knowledge from extensive reading. It should be noted that although translation did not have a significant impact on grammar achievement for students in this group either, their perceptions of improvement and positive affect were significantly higher than those of their counterparts in the extensive reading group. That is, for low-level achievers, reading books written in a foreign language may have seemed a frustrating task because of their limited L2 proficiency, whereas translating a short foreign text and clarifying any difficulties with the teacher's explanation at the end of class may have engendered a sense of achievement.

It is interesting that students at all levels in the translation group self-evaluated that their linguistic skills had improved and expressed positive feelings towards translation, suggesting that these students perceived the translation activities to be beneﬁcial and enjoyable regardless of L2 proﬁciency levels. Because the students knew that the materials encountered in these activities would not appear on regular school tests, the degree of engagement depended solely on their enjoyment and