A comparison of the impact of extensive and intensive reading approaches on the reading attitudes of secondary EFL learners

Extensive reading (ER) which encourages second or foreign (L2) learners to engage in a great deal of reading, has long been recognized as an efficient approach in L2 reading pedagogy. While many attempts have been made to understand the effect of ER on the cognitive domains of L2 learners, there has been insufficient investigation into how ER influences their affective domains. Particularly, reading attitudes, one of the key elements of affective factors involved in L2 reading, have received little attention. This classroom-based intervention study investigated the impact of ER on English as a foreign language (EFL) learners’ attitudes toward English reading compared to the influence of the traditional intensive reading (IR) approach. In addition, this study explored whether the impact of the ER approach on EFL learners’ reading attitudes is different depending on L2 proficiency. The study included two intact classes of EFL secondary learners (N = 72) who received either ER or IR instructional treatments for a 12-week period. For the results, ANCOVA showed that the ER approach fostered positive reading attitudes significantly more than the IR approach. In addition, the analysis indicated that the participants’ proficiency levels did not have a significant effect upon changes in their reading attitudes. That is, regardless of proficiency level, the ER approach demonstrated a significantly positive effect on participants’ reading attitudes in comparison with the IR approach.


Introduction
Affective factors have been identified as key components in predicting the success or failure of L2 reading (Lu & Liu, 2015;Macalister, 2014;Yamashita & Kan, 2011). However, in spite of the critical role played by affective factors in second or foreign language (L2) reading, there has been limited investigation into the affective dimensions in L2 reading (Lee, Schallert, & Kim, 2015). This research trend, reflecting a lack of research into affective dimensions, is also evident in extensive reading (ER) studies. The ER approach encourages L2 learners to read long and easy material in quantity based on each individual learner's interests. ER has been viewed with increasing interest in the field of L2 reading pedagogy as effective reading instruction. While there is a great deal of research exploring the effect of the ER approach on the L2 learners' cognitive dimensions such as listening (Renandya & Farrell, 2011;Ware, Yonezawa, Kurihara, & Durand, 2012), writing (Sun, Yang, & He, 2016) and reading skills (Cha, 2009;Huffman, 2014;Park, 2017;Stephens, 2016;Shih, Chern, & Reynolds, 2018), as well as vocabulary knowledge (Park, Isaacs, & Woodfield, 2018;Rashidi & Piran, 2011;Webb & Chang, 2015), the affective dimensions remain comparatively under-researched. In particular, reading attitude, one of the primary affective dimensions involved in L2 reading (McKenna, Cramer, & Castle, 1994), has received inadequate attention. Since the primary goal of the ER approach is to promote L2 learners' enjoyment of reading through nurturing a positive reading attitude, it is necessary to investigate how L2 learners' reading attitudes change after an ER session.
Additionally, little attention has been paid to whether the ER approach affects L2 learners' proficiency levels. In terms of the cognitive domains of L2 learners, such as vocabulary, reading fluency, and grammar knowledge development, a number of empirical studies suggest that learners at different proficiency levels benefit differently from the ER approach (e.g., Lee, Schallert, & Kim, 2015;Park, Isaacs, & Woodfield, 2018;Yamasitha, 2013). These findings yielded supportive evidence for linguistic threshold theory (Cummins, 1976), which contends that L2 learners should acquire a threshold level of L2 proficiency to obtain functional target language ability. In other words, low levels of lexical and syntactic knowledge may prevent readers from experiencing fluent and pleasurable reading, and can impede low proficiency level learners who are using the ER approach from learning linguistic skills (Lee, Schallert, & Kim, 2015;Morvay, 2012). Therefore, further investigation is needed in order to determine whether linguistic threshold theory can also be applied to the affective domain (e.g., reading attitudes).
To bridge the gap in the aforementioned existing literature, this study examined the effect of the ER approach, in comparison with the traditional intensive reading (IR) approach, on Korean secondary learnes' attitudes toward reading. The IR approach, which has been widely accepted in English as a foreign language (EFL) classroom contexts, requires learners to construct the precise meaning of reading material through close analysis of the language and translation guided by the instructor. Moreover, the current study attempted to expand previous ER research to reading attitudes by exploring learners at different English proficiency levels (advanced, intermediate, low) to investigate which learners stand to benefit the most from each instructional approach.

L2 reading and reading attitudes
A considerable amount of recent literature suggests that affective factors perform a significant role in evaluating the success or failure of L2 reading (Lu & Liu, 2015;Macalister, 2014;Yamashita & Kan, 2011). However, when compared to the number of studies of L2 reading focused on cognitive factors, relatively few studies have been devoted to affective factors (de Burgh-Hirabe & Feryok, 2013;McKenna, Cramer, & Castle, 1994;Yamashita, 2013). This is somewhat surprising considering that many L2 reading researchers, including McKenna et al. (1994, p. 3), argue that "affective aspects of reading are equal in importance to cognitive aspects." Reading attitude has generally been considered to be one of the important affective factors of L2 reading (McKenna, Cramer, & Castle, 1994;Sainsbury & Clarkson, 2008;van Schooten & de Glopper, 2002;Yamashita, 2013). In fact, Smith (1990, p. 215) describes reading attitude as "a state of mind, accompanied by feelings and emotions that make reading more or less probable." In other words, attitude is one of the important affective factors influencing the decision to read (Krashen, 1994). Therefore, L2 readers who have a positive reading attitude are more likely to continue self-directed reading, which enhances their L2 acquisition (Al-Homoud & Schmitt, 2009;Takase, 2009). Considering that the main purpose of ER is to encourage L2 learners to read for pleasure and foster a positive attitude towards reading, this study examined the impact of the ER approach toward reading attitudes of L2 secondary learners (Day & Prentice, 2016;Lee, Schallert, & Kim, 2015).
Despite the importance of reading attitude in developing reading fluency (Shin & Ahn, 2006;Yamashita, 2008Yamashita, , 2013, few studies have attempted to integrate reading attitude into a reading model (e.g., McKenna et al., 1994;McKenna, Stratton, Grindler, & Jenkins, 1995;Yamashita, 2013). One study that did just that is van Schooten and de Glopper (2002), which proposed the reading attitudes model. The model has been widely accepted in research aimed at investigating L2 learners' attitudes toward reading in relation to diverse variables (Mizokawa & Hansen-Krening, 2000;Yamashita, 2008Yamashita, , 2013. Van Schooten and de Glopper's (2002) reading attitudes model was suitable for understanding learners' attitudes toward reading in the current study, because it helped to elucidate the complex hypothetical construct of reading attitudes from diverse perspectives, incorporating cognitive, affective, and conative angles (Yamashita, 2008(Yamashita, , 2013. The next section will explore van Schooten and de Glopper's (2002) reading attitudes model in detail in order to further illuminate what comprises reading attitude and how reading attitude has been defined in the current study. Van Schooten and de Glopper's (2002) model was adopted to define how reading attitude would be conceptualized in the current study. The model was selected because it effectively illustrates L2 learners' reading attitudes utilizing the three-dimensional reading attitude framework. In addition, van Schooten and de Glopper's (2002) model explains effectively how these dimensions form a series of causal links. The first component in their model is cognitive attitude (belief in and evaluation of an outcome), which influences affective attitude (liking and enjoyment), which, in turn, influences conative attitude (reading), with the latter predicting actual reading behavior. The model suggests that the influence of affect (i.e., affective attitude) on behavioral intention (i.e., conative attitude) is direct and very strong ( van Schooten & de Glopper, 2002). While affective attitude influences conative attitude significantly, the influence of cognitive attitude on conative attitude is indirect and weak. Therefore, cognitive attitude has a non-significant influence on conative attitude. Van Schooten and de Glopper (2002) conclude that in terms of promoting reading among L2 learners, the most important aspect of reading attitude is the affective aspect, since it directly influences learners' intention to read (i.e., conative attitude), and ultimately initiates the act of reading itself. In other words, one of the most effective ways to encourage reading is to empower L2 learners to enjoy reading by way of promoting their affective attitude in a positive way. The importance of cultivating such attitudes toward reading is also emphasized within the pedagogical perspective, as shown in Nuttall's (1996, p. 127) virtuous circle. According to Nuttall (1996), the virtuous circle operates as follows: If learners "enjoy reading," it leads them to "read faster" and then to "understand better," which in turn leads them to "read more." Coady (1997, p. 233) suggests that fostering "enjoyment and increasing the quantity of reading," that is, using the ER approach, draws learners into this virtuous circle. This is why, many researchers recommend ER for fostering positive affective attitude, and, in turn, for L2 development (Day & Prentice, 2016;Ehri, 2005;Nation, 2015;Wolf & Katzir-Cohen, 2001). Accordingly, it is meaningful to explore the positive impact of the ER approach on learners' affective attitude compared to that of the IR approach.

ER studies on reading attitude
Previous ER studies have largely focused on the effect of the ER approach on learners' cognitive domain, which includes reading, vocabulary, and writing (Yamashita, 2004(Yamashita, , 2015. By primarily focusing on the cognitive domain, existing ER literature has paid little attention to affective issues. However, research into the affective domain is important because it is one of the critical factors in motivating learners to engage in the actual act of reading (Day & Prentice, 2016;McKenna et al., 1994;Takase, 2009). For example, Elley (2000) argues that learners' attitudes toward reading closely relate to their reading habits, which, in turn, influence their L2 development. Since it has only been in recent years that ER researchers have attempted to explore the affective domain in regard to areas such as reading attitude (e.g., Powell, 2005;Shin & Ahn, 2006;Takase, 2009;Yamashita, 2007Yamashita, , 2013Yamashita & Kan, 2011), more empirical studies are required in order to develop greater understanding of the impact of the ER approach from diverse perspectives. In addition to learners' attitudes toward reading, learners' motivation is another well-known affective domain that plays an important role in successful L2 reading development (Shih & Reynolds, 2015;Shih et al., 2018). Day and Bamford (1998, p. 27) refer to motivation as the L2 learners' emotional drive, or lack thereof, that "makes people do (or not do) something." However, motivation toward L2 reading is not discussed, being beyond the research scope of the current study.
In terms of the effect of ER on L2 reading attitude, few ER studies have specified in detail how participants' reading attitudes were measured or why a particular measurement tool was chosen. As Yamashita (2004) points out, most of the ER studies on reading attitude used instruments with diverse methodological errors (Shin & Ahn, 2006;Ware et al., 2012). For example, Takase (2007) interviewed participants to measure how their reading attitude changed after an ER intervention. However, the questions used in the interviews were not clearly reported in the study and there was no explanation of how the questions were designed and selected for the interview. Moreover, even where ER studies made use of a methodologically sound reading attitude measurement tool, participants' reading attitudes were measured only after the intervention, without carrying out an initial measurement (Takase, 2009). For example, Al-Homoud and Schmitt (2009) used an in-house developed questionnaire to tap the reading attitudes of 70 EFL Saudi Arabian university students. Their study explained in detail how they designed and developed the questionnaire based on the diverse aspects of reading attitude. Nevertheless, the questionnaire was administered only after the intervention. In fact, it would seem that the questionnaires served more as a course evaluation than reading attitude measurement. Without conducting a survey of reading attitude before the intervention, convincing evidence that there was a statistically significant change in reading attitudes after the ER intervention could not be provided. A recent study by Lee et al. (2015) systematically measured the effect of the ER approach on learners' attitudes before and after ER intervention compared to the previous ER studies. However, the main focus was on learners' attitudes toward various ER approaches rather than their reading attitudes.
In summary, a review of the existing ER literature found that relatively few studies have examined the effect of the ER approach on the affective dimensions of L2 learners compared to the cognitive dimensions. To bridge the gap in the existing ER literature, the current study explored the impact of the ER approach on the affective domain of EFL learners; more specifically, it measured participants' attitudes toward English reading and how those attitudes changed after an ER intervention. The present study contributes to the existing ER literature on reading attitudes through addressing the following research questions (RQs): RQ1: To what extent do ER and IR approaches affect Korean secondary EFL learners' attitudes toward reading after a 12-week instructional period? RQ2: Does English proficiency or a particular reading instructional approach affect Korean secondary EFL learners' attitudes toward reading after a 12-week instructional period?

Participants
Two intact classes (N = 73) of a suburban secondary school in Korea were included in this quasi-experimental study. The participants were Korean high school students, whose ages ranged from 15 to 16 years old ( Mage = 15.3 years). A class of 36 secondary students received ER treatments and another class of 37 received conventional IR treatments. Both ER and IR treatments were performed once a week for two hours over one 12-week academic term. Since the current study was carried out in a regular academic context, it was inevitable that all students received some kind of instruction. Consequently, instead of having a control group receiving no instruction, each group performed as a control group for the other. Although all of the participants had studied English as a compulsory subject for more than 6 years, their English proficiency levels varied due to different degrees of exposure to external learning experiences including private English instruction or tutoring. The participants in both ER and IR groups were classified into three groups based on their proficiency levels in English: low, intermediate, and advanced. Proficiency levels were determined on the basis of the participants' overall scores on a practice College Scholastic Ability Test (CSAT) which is designed to measure general English proficiency.

Treatment
The ER group took reading instruction based on a class library of graded readers that contain controlled vocabulary items and sentence structure to afford gradual difficulty and complexity for readers with different proficiency levels (Crossley & McNamara, 2008;Hill, 1997Hill, , 2001. Four sets of fifty graded readers (i.e., a total of 200 graded readers) from the "Oxford Bookworm" series were introduced to the ER group. This graded readers series consisted of fiction literary classics and non-fiction works. Participants received reading lessons once a week for two hours for an academic term (i.e., 12 weeks). They were guided to read one graded reader per week, both in and outside of class. This amount of reading is regarded crucial for an effective ER approach in a number of studies (Day & Prentice, 2016;Nation, 2013;Waring & Takaki, 2003). As Nation and Ming-Tzu (1999, p. 355) assert, "learners need to read about one graded reader per week in order to meet repetitions of the new words soon enough to reinforce the previous meeting." In terms of book selection, the participants were guided to choose books based on their interests and proficiency level. According to the books they had selected in the graded readers collection, each participant made a wish list of books that they felt like reading in the ER course for the 12 weeks. Participants were instructed to read extensively but not to concentrate on new vocabulary or grammatical structures (Day & Bamford, 2002). In every ER lesson, they were encouraged to finish the remainder of their reading as an assignment. Additionally, participants were requested to write a reading log to check whether they had completed the assigned reading. This reading log was an important tool for monitoring the ER participants' reading, since how much written input participants are exposed to is one of the key elements of success in the ER approach (Day & Bamford, 1998).
While the purpose of an ER lesson is to provide learners with reading material that they can comfortably comprehend for rapid and enjoyable reading, the main goal of an IR lesson is to "recycle and reinforce language items through intensive micro-linguistic analysis of the texts" (Bell, 2001, p. 3). Taken in sequence, the IR lessons heavily focused on analyzing and interpreting reading material once a week over a 12-week academic term. In each two-hour IR lesson, the participants were instructed to read four reading materials (700-800 words each) and complete the accompanying exercises. At the end of each lesson, four new reading texts were given to the participants as a reading assignment. Unlike the graded readers used for the ER group, the reading texts chosen for the IR group were texts above the participants' proficiency levels (Day & Bamford, 1998;Prowse, 2002;Renandya, 2007;Schörkhuber, 2009). Accordingly, the texts used for the IR group contained many unfamiliar words and grammar items that the participants were required to learn in order to understand the texts. As a result, the teacher's assistance was required for the IR group to comprehend the specific meaning of the reading texts (Carrell & Carson, 1997;Rashidi & Piran, 2011;Renandya, 2007).
According to each group's reading logs, the average amount of reading outside of class was calculated. Every week, the ER group spent 164 minutes on average for reading outside of class. This period amounts to 240 minutes of class reading (20 minutes*12 weeks) plus an average of 1968 minutes of reading outside of class (164 minutes * 12 weeks). Meanwhile, the IR group spent an average of 192 minutes (16 minutes*12 weeks) of reading outside of class plus an average of 240 minutes of class reading (20 minutes*12 weeks) during the 12 weeks of intervention.

Data collection instruments
To measure the participants' attitudes toward English reading, the present study used a modified version of Yamashita's (2004) questionnaire. Yamashita (2004) developed a questionnaire successfully applying van Schooten and de Glopper's (2002) three-dimensional framework to analyze the reading attitudes of L2 learners in a manner similar to that used in the current study. Furthermore, each of the question items possessed its own distinct variance, not overlapping with items representing the other two components in the questionnaire. Therefore, the current study adopted Yamashita's (2004) instrument, which is firmly grounded in van Schooten and de Glopper's (2002) three-dimensional framework of reading attitude.
The questionnaire contained 5-point Likert-scale items, ranging from strongly disagree (1) to strongly agree (5), enquiring about the students' affective (Section I), cognitive (Section II) and conative (Section III) reactions towards reading in English (see Appendix). Each participant's scores for all the items which corresponded to the three categories were summed up. Yamashita's (2004) adopted questionnaire was examined by three experts in Teaching English to Speakers of Other Languages (TESOL) for the current study. Each expert was asked to assess and provide feedback on the preliminary list of the reading attitude questionnaire items. The TESOL experts suggested that since Yamashita's (2004) reading attitude questionnaire focused on university students, a revision of each question was needed to make it appropriate for the high school reading context. Accordingly, the questionnaire was modified to make it suitable for the target population in the current study. For example, a section asking for participants' background information, such as formal and informal English education study time per week and a mock College Scholastic Ability Test (CASAT) score, was added.
The questionnaire items' internal consistency of in each domain was estimated by calculating Cronbach's alpha and the values were as follows: cognitive domain (.886), affective domain (.788) and conative domain (.889). A Korean version of the reading attitudes questionnaire was administered for the main study, since the purpose of the questionnaire was to explore participants' English reading attitudes and not to measure their L2 reading comprehension ability. In other words, to prevent comprehension problems from affecting the reliability of the collected data, the participants' mother tongue was used.
The present study also made use of a placement test to identify participants' proficiency level. The test used was the mock CSAT for English, which is one of the most widely used placement tests among secondary schools in Korea due to its practicality and accessibility (Shin & Ahn, 2006). CSAT for English is a high-stakes university entrance test approved by the Korean Administration of Education, which is designed to test listening, speaking, reading, and writing skills of English. All the participants in the current study had already been placed according to test results by the participating school before the intervention.

Data analysis
To answer RQ1, an independent samples t test was carried out to examine whether there was a significant difference between the pre-intervention questionnaire responses for the participants in the ER and IR groups. The independent variable was the type of the treatment (the ER or IR approach), and the dependent variable were the responses to the questionnaire at Time 2 (post-intervention). Furthermore, a one-way ANCOVA adjustment was separately applied for each of the three aspects of reading attitude (cognitive, affective and conative) to examine whether there were any differences in this respect. ANCOVA was applied to control for initial differences between the two groups, to attain more statistically accurate comparisons and to reduce the sampling error. The current study adopted convenience sampling using two intact EFL classes rather than random sampling. Although the English proficiency level of Experimental Group 2 (the IR group) was equivalent to that of Experimental Group 1 (the ER group), the two groups were not necessarily equivalent in terms of reading attitudes. ANCOVA is an effective tool for examining the significance of a treatment impact (i.e., the ER and IR approaches) on the dependent variable (i.e., the post-intervention reading attitudes results), while accounting for another variable (i.e., the initial difference in the two groups' pre-intervention reading attitudes results) that may also impact the dependent variable, but not part of the treatment (Pallant, 2010).
For RQ2, a two-way ANCOVA was carried out to scale the effect of the ER and the IR approaches on reading attitude based on the proficiency levels of participants. The participants were placed in one of three English proficiency groups based on the placement test results: Group 1 = advanced; Group 2 = intermediate; Group 3 = low. The independent variables were the type of treatment (the ER or the IR approach) and proficiency level (advanced, intermediate, or low). The dependent variable were the reading attitude questionnaire mean scores of each group taken at Time 2 (post-intervention). The mean scores from the reading attitude questionnaire carried out at Time 1 (pre-intervention) were employed as the covariate to adjust initial group differences. The probability value was set at p < .05 for all statistical tests.

Results
For RQ1, as displayed in Table 1, the ER group scored significantly higher (M = 71.14, SD = 7.4) than the IR group (M = 62.11, SD = 11.87) in the pre-intervention questionnaire (t(58.64) = 3.87, p < .001, two-tailed). This significant difference might have arisen as a result of this study's use of two intact classes rather than random sampling (Pallant, 2010). The independent t-test results showed a significant initial difference between the mean scores of the ER and IR group at Time 1 (pre-intervention). To control for this initial difference and reduce sampling error, a one-way ANCOVA was used (Field, 2009). In this analysis, the independent variable was the style of approach (the ER or IR approach), and the dependent variable was the reading attitude questionnaire responses at Time 2, after 12 weeks of treatment. The responses to the reading questionnaire at Time 1 (pre-intervention) were employed as a covariate. Covariates value of reading attitude mean score at Time 1 (pre-intervention) in each group was estimated at 66.63. The maximum possible score was 120 and the minimum possible score was 24. .14 * Notes. ER = Extensive Reading, IR = Intensive Reading * a significant difference between the groups after a one way ANCOVA adjustment a covariates value of the questionnaire response of mean score at Time 1 (pre-intervention) in both groups is evaluated at 66.63 * p < .05 After applying ANCOVA analysis to the questionnaire responses at Time 1 (pre-intervention), the adjusted mean score (Estimated Time 2) for the questionnaire responses was significantly higher in the ER group by 10.28 than in the IR group at Time 2 (post-intervention) as presented in Figure 1.

Figure 1 Estimated means of reading attitude at Time 2 (post-intervention)
In addition, further analysis was conducted to explore where these significant main effects held among the three main aspects of reading attitude, that is cognitive, affective, and conative ( van Schooten & de Glopper, 2002). For this purpose, a one-way ANCOVA was carried out separately for each aspect. Covariates value of reading attitude mean scores at Time 1 (pre-intervention) in both groups was evaluated at 28.14 for cognitive attitude, 22.79 for conative attitude and 15.69 for affective attitude (see Figure 2). An asterisk designates a significant difference between the two groups after a one-way ANCOVA adjustment. The maximum possible score was 40 and the minimum possible score was 8.

Figure 2
Estimated means of reading attitude in terms of cognitive, affective and conative attitude at Time 2 (post-intervention) As presented in Figure 2, the result of the ER group was significantly different from that of the IR group at Time 2 (post-intervention) in terms of affective and conative attitude, while non-significantly different in terms of cognitive attitude. The results indicated a non-significant improvement in cognitive reading attitude and a significant improvement in affective and conative reading attitude for both the ER and the IR group.
In order to address RQ2, the mean scores from the reading attitude questionnaire taken at Time 1 (pre-intervention) were employed as the covariate to adjust the initial group differences. After adjusting the difference in reading attitude scores at Time 1 (pre-intervention) using a two-way ANCOVA, there was no significant interaction effect, f(2, 65) = 2.42, p = .09, with a minimal effect size (partial eta squared = .07). In addition, there was no significant main effect for proficiency level, f(2, 65) = 2.42, p = .1. However, there was a negligible effect size (partial eta squared = .007). The results suggest that the participants' proficiency level did not have a significant influence on the change in reading attitude. In other words, the ER approach promoted the participants' positive reading attitude compared to the IR approach, regardless of their proficiency level.

Discussion
RQ1 aimed to investigate the effects of the ER and IR approaches on Korean secondary EFL learners' reading attitude, specifically focusing on three aspects: (a) cognitive (personal, evaluative beliefs), (b) affective (feelings and emotions), and (c) conative (action readiness and behavioral intentions). These are the three main components of reading attitude (McKenna et al., 1994) measured by the current study. The results showed that the ER approach significantly promoted positive attitudes toward reading compared to the IR approach. More specifically, among the three components of reading attitude, the affective and conative aspects increased significantly, while the cognitive aspect showed a relatively small increase.
Firstly, the impact on reading attitude is discussed from the perspective of cognitive reading attitude. The results showed that the ER group's cognitive attitude had improved after the intervention in comparison with that of the IR group. Although this improvement was non-significant, the ER approach had more of a positive impact than the IR approach. This enhancement might have resulted from "the intellectual satisfaction that students experienced from gaining new knowledge and information through reading" (Yamashita, 2013, p. 257). Meanwhile, the IR group showed almost no increase in positive cognitive attitude, despite also having had the chance to read diverse texts with new information. This might have been due to the fact that the participants in the IR group could not select their own reading materials. Since the reading materials were less likely to appeal directly to each participant's personal interests or curiosity, it is reasonable to assume that the students in the IR group may have found the reading experience to be less intellectually satisfying than those in the ER group.
Secondly, the findings revealed the ER approach had a significantly positive effect on promoting affective attitude compared to the IR approach. This finding was different from the results for cognitive attitude, which was not significantly promoted by the ER approach. More specifically, among the three components of reading attitude (cognitive, affective, and conative), affective reading attitude increased the most. This is not a surprising finding, since the main goal of ER is to foster the conditions for enjoyable reading (Day & Bamford, 1998, 2002Day & Prentice, 2016;Jacobs & Farrell, 2012;Schörkhuber, 2009), which allows nurturing positive affective attitudes. This finding was meaningful, because according to van Schooten and de Glopper's (2002) reading attitude model, affective attitude is the component of reading attitude that is most likely to initiate the act of reading itself. In other words, based on van Schooten and de Glopper's (2002) reading attitude model, it can be said that an effective way to encourage EFL learners to read is to foster their positive affective attitudes towards this skill. Thus, cultivating the conditions for enjoyable reading helps to create positive affective attitude, as was successfully done by the ER approach in the current study. Furthermore, this improvement is noteworthy as it suggests that the participants from the ER group will tend to continue reading by themselves, which could continue to improve their English knowledge and skills (van Schooten & de Glopper, 2002;Yamashita, 2004).
Thirdly, comments are in order on the impact on reading attitude from the conative perspective. Similar to the findings related to affective attitude, those related to conative attitude also showed the ER approach had a more positive impact. In other words, the participants' conative reading attitude improved significantly more in response to the ER approach than to the IR approach. However, it is interesting to note that while the ER approach did enhance participants' cognitive reading attitude, the improvement was not as significant as it was the case with their affective and conative attitudes. This difference can be explained by van Schooten and de Glopper's (2002) model, which suggests that the affective attitude is directly linked to conative attitude; that is, affective attitude strongly influences conative attitude. On the other hand, cognitive attitude and conative attitude are indirectly linked; that is, cognitive attitude weakly influences conative attitude. For example, if affective attitude improves, conative attitude also improves automatically due to the influence of the former (Yamashita, 2013). At the same time, if cognitive attitude improves, it makes little difference for conative attitude as presented. Therefore, Van Schooten and de Glopper's (2002) model offers an explanation for why both affective and conative reading attitude were significantly promoted, while cognitive reading attitude was non-significantly impacted. Consequently, the results of the study provide support for the reading attitude model, which, it should be recalled, holds that in order to promote reading among EFL learners, the most important is the affective aspect of reading attitude, since it directly influences the intention to read (i.e., conative attitude) and finally initiates the act of reading itself (Day & Prentice, 2016).
When it comes to RQ2, the results revealed that the ER approach cultivated positive reading attitude of participants significantly more than the IR approach, irrespective of their proficiency levels. In other words, learners' proficiency did not turn out to be a significant factor in promoting positive reading attitude irrespective of the instructional approach used. Although there are few ER studies related to L2 reading attitude development according to learners' proficiency, these results are in line with those of Yamashita (2004Yamashita ( , 2007Yamashita ( , 2013 or Rashidi and Piran (2011). However, they stand in contrast to Lee et al.'s (2015) research which demonstrated a correlation between L2 learners' proficiency levels and their attitudes toward reading. More specifically, in Lee et al.'s (2015) study, the ER approach led to a negative impact on reading attitude for learners with low proficiency level. Lee et al. (2015) suggested that the story line of graded readers designed for learners with low proficiency tended be too easy for secondary learners and, consequently, participants in the low proficiency group developed negative attitudes toward reading. One of the possible reasons why the current study yielded different results is that it employed "language learner literature" (Day & Bamford, 1998, p. 74), which is deliberately written for language learners. Such literature is a good example of how to satisfy learners' interests while remaining within their reading level. Unlike graded readers, which are simplified versions of original texts, language learner literature can do so without trivializing the content of an original story in order to control readability. In effect, in this study even the low proficiency level learners in the ER group seemed to enjoy the readings.
In regard to reading attitude improvement according to proficiency level, there are two interesting points to note in the findings. Firstly, contrary to the significant impact of learners' proficiency levels on their L2 development in terms of reading fluency, vocabulary, grammar development (included in the cognitive domain) in previous ER studies, attainment did not turn out to be a significantly influential factor in the development of reading attitude (included in the affective domain) in this investigation. In other words, the ER approach promoted the participants' positive reading attitude when compared to the IR approach, irrespective of proficiency levels. Secondly, because learners' proficiency was not a significant factor in developing reading attitude, linguistic threshold theory does not support the effect of the ER and IR approaches on the promotion of participants' reading attitude. However, this suggests that even learners at a low proficiency level are likely to develop a more positive reading attitude to reading through the ER approach than through the IR approach. A possible reason for this might be found in Powell's (2005) suggestion that in the ER approach learners can choose their own reading materials from a range of texts within their own proficiency level. For this reason, learners with low proficiency do not need to be "embarrassed by not getting abreast of learners with more advanced proficiency" and are thereby "less likely to get frustrated and demotivated" (Powell, 2005, p. 29). This kind of positive experience provided by the ER approach might consequentially promote a more positive reading attitudes in different learner groups. This result lends support to Schörkhuber's (2009) and Renandya's (2007) claim that even low proficiency learners are able to draw pleasure from the very experience of reading books in the target language.

Conclusion
The findings of the study showed that the ER approach significantly promoted positive attitudes toward reading compared to the IR approach. Among the three components of reading attitude (i.e., cognitive, affective, and conative), the affective and conative reading attitudes showed a significant improvement, with the affective attitude improving the most. This result is important because affective attitude is directly connected to conative attitude, which in turn leads to the act of reading, as shown in van Schooten and de Glopper's (2002) reading attitude model. Thus, the investigation showed that the ER approach is an effective way of enhancing learners' attitude toward reading and thereby promoting the act of reading itself. In addition, unlike ER studies focusing on different aspects of the cognitive domain (i.e., reading fluency, vocabulary knowledge, grammar), the analysis revealed that proficiency level did not significantly influence learners' reading attitudes toward English reading after the intervention.
The present study provided evidence that the ER approach successfully promotes EFL learners' enjoyment of reading in English, with affective reading attitude (i.e., feelings and emotions) showing the largest improvement. This suggests that, in order to help EFL learners achieve the greatest benefit from the ER approach, practitioners need to foster environments in which students can genuinely enjoy reading. Specifically, when designing reading lessons, teachers in Korea should begin by diagnosing L2 learners' interests and needs. For example, before an ER course begins, the practitioner should survey the learners' preferred genres and proficiency levels. Only on the basis of such information should reading materials be selected. In this way, learners can read books that interest them while remaining in their "comfort zone" (i.e., within their proficiency level) and therefore they can enjoy the act of reading. Furthermore, this result can serve as a reminder to reading material developers in Korea of the importance of designing books that can appeal to learners within an appropriate reading level such as language learner literature.
This study suffers from several limitations that require careful attention in follow-up studies. The first is associated with the absence of qualitative data. Through triangulation with the qualitative data, the current study's quantitative data could have supported the findings more strongly. The second limitation relates to the experimental design. This quasi-experiment compared Experiment Group 1 (the ER group) with Experimental Group 2 (the IR group) before and after the intervention. However, there was no true control group involved in the study that did not receive any treatment. Thus, further research would surely benefit from the inclusion of qualitative data obtained through interviews or diaries, as this would offer greater insight into the effectiveness of ER and IR approaches in shaping EFL learners' attitudes towards reading in English. In addition, future investigations should include control groups to strengthen the research design.