EFL learners’ reading traits for lexically easy short poetry

Abstract Researchers have recommended that English-as-foreign-language (EFL) classrooms should use lexically easy short poems. However, no studies have examined EFL learners’ actual reading processes for these poetry texts, the insights of which are vital for corroborating future poetry-teaching methods. Hence, this study investigates how upper-intermediate Japanese EFL learners read lexically easy short poems and describes their process for constructing a literal comprehension of the text and an interpretation of the poet’s intent based on reading time, confidence level in poetry-reading performance, subjective assessment of textual difficulty and postreading verbal report. The main findings are that (1) these lexically accessible poems do not guarantee seamless reading, (2) poetry texts of the same lexical level are read at varying speeds with different confidence and difficulty levels, (3) grammatical deviations disrupted the participants’ literal-comprehension and (4) learners encountered various challenges in understanding the poet’s intent and could not decipher it. Based on these findings, this study offers the following pointers for teaching procedures: teachers’ assistance is indispensable for interpretation, learners benefit from pair work using individuated reading difficulties, scaffolding learners’ comprehension of literary languages such as deviations are necessary and teachers must ensure that learners enjoy poetry reading.

vocabulary, grammatical deviations, old-fashioned expressions or complicated metaphorical language (Paran & Robinson, 2015). In addition, they often mention cultural topics or representations that are foreign to learners (e.g., religion, mythology or history). As a result, teachers often cast these texts in a negative light: some are either reluctant to use them (Bouman, 1983) or not interested in this genre (Akyel, 1995), and others are critical about their use (Pullinger, 2012;Tomlinson, 1986). Some teachers also distance themselves from these materials, as they are not confident about using them in classrooms (Hughes & Dymoke, 2011): they did not learn relevant teaching methodologies or techniques from training or methodology handbooks (Gönen, 2018;Paran, 2008).
Scholars have proposed different teaching procedures based on the abovementioned merits of poetry texts (Ariyanto, 2021;Gönen, 2018;Hanauer, 1997;Lin, 2006;Paesani, 2005;Paran & Robinson, 2015;Rosenkjar, 2006;Scott & Huntington, 2007;Tomlinson, 1986Tomlinson, , 1998Viana & Zyngier, 2017). Gönen (2018) devised a four-step teaching procedure (schema activation, basic comprehension, detailed analysis [literary appreciation/interpretation and linguistic analysis] and personalisation of the poetry text) with 28 language activities for not only integrating the linguistic, cultural and personal growth aspects of poetry texts (Carter & Long, 1991) but also comprehensively fostering four language skills (listening, speaking, reading and writing). Meanwhile, Cushing (2018) proposed the following four-step model for integrating teaching grammar and poetry, which is applied to first-language (L1) teaching but can also be implemented in EFL: (1) engaging individual learners in freely constructing meaning without providing the title, (2) giving learners the title of the poetry text and having them discuss whether meaning changes (if yes, in what way and why), (3) encouraging them to write and justify their meaning on a sheet of paper and (4) letting them consider what grammatical traits facilitated meaning construction.
These teaching models, which presuppose learners' textual engagement, must be corroborated by insights into learners' actual poetry-reading processes (Hall, 2015;Nishihara, 2016). Despite the increase in investigations on native speakers' individual poetry-reading processes (e.g., Jacobs, 2018;Müller et al., 2017;Xue et al., 2019), only a few have focused on foreign-language learners, and little is known about how they construct meaning, what reading problems they face and what reading behaviours and strategies they exhibit during poetry reading. 1 Among these few studies were Hoffstaedter (1987), Chesnokova and van Peer (2016), and Nishihara (2016). Hoffstaedter (1987), which required native German speakers and American learners of German to underline poetic expressions they felt they were during German poetry reading, showed that learners were less sensitive about poetic expressions and less consistent than native speakers regarding poeticity judgment. Meanwhile, in Chesnokova and van Peer (2016) study, novice and advanced EFL learners, including teachers, were asked to read either the original text of an E. E. Cummings poem or its manipulated version which excluded all deviant expressions. After reading the first, middle and final lines of the texts, the participants were told to answer questionnaires and describe how their reactions changed. The primary results were that both groups noticed the original texts' aesthetic structures, advanced learners performed better than novice learners in evaluating these structures, both novice learners and teachers considered the altered text more emotive and the most significant differences in learners' reactions to the two texts emerged after they read the first line (and tended to decrease as learners read the texts further). In another study by Nishihara (2016), Japanese university EFL learners with limited English poetry knowledge reported on the textual features they noticed during their meaning-making processes of a Robert Frost poem, from which three poetryreading strategies were identified: (1) application of expository reading strategies (see also, McCarthy, 2015), (2) examination of repetitive expressions and (3) construction of meaning based on semantic intuition and explicit knowledge of English.
These empirical studies are vital in that they revealed foreign-language learners' poetry-reading traits, which could be considered when designing teaching procedures and language activities that use poetry texts. However, these studies also explored how inexperienced foreign-language learners process rather difficult poetry texts such as those of E. E. Cummings and Robert Frost. Although studies have suggested that learners start with short, simple poetry texts (e.g., Paran & Robinson, 2015;Peskin, 2000;Rosenkjar, 2006;Tomlinson, 1986), a knowledge gap exists regarding the manner in which they read these lexically accessible poems, and there is no guarantee as to whether they would succeed in constructing the meaning of such poems. To corroborate or improve poetry-teaching procedures, studies must use short, simple poems to describe their meaning-making processes.
Therefore, this study explores learners' reading responses to lexically easy short poems focusing on two research questions: • Can upper-intermediate EFL learners decipher the meaning of lexically easy short poems seamlessly?
• How do these learners construct the meaning of these poetry texts?
This research examines learners' responses to the texts during both literal-comprehension and interpretation based on conventional ideas in EFL and ESL reading instruction, which distinguishes these two reading processes (e.g., Koda, 2005;Nuttall, 2005;Richards & Schmidt, 2002). It seeks to answer the abovementioned research questions using four types of data: reading time, confidence level in poetry-reading performance, subjective assessment of textual difficulty and postreading verbal reports on meaning construction. Reading time refers to learners' online response traits, while poetry-reading confidence level and subjective assessment of textual difficulty cover their offline reading traits (Castiglione, 2017), using postreading verbal reports on meaning construction as complementary evidence. Furthermore, this study regards reading time as a three-pronged index: real-time processing effort or difficulty (e.g., Avery & Marsden, 2019;Castiglione, 2017), the wrap-up effect of semantic processing (Hirotani et al., 2006;Müller et al., 2017) and the savouring effect (i.e. processing aesthetically preferred input; Menninghaus & Wallot, 2021). Simply put, reading time is used not only for understanding the texts or their partial textual traits that hindered participants' meaning construction processes but also for examining whether they spend time building their ideas or enjoying the texts' aesthetic appeal.

Participants
The details of this experiment, which was conducted outside the curriculum, were explained to all students in the department division of the present author, and a total of 24 Japanese TESOL majors (17 undergraduate students and 7 postgraduate students; 12 male and 12 female; age range 21-26 years) provided informed consent and participated in this study. They were upperintermediate EFL learners who had studied English under the Japanese educational system and whose Test of English for International Communication (TOEIC) scores ranged from 545 to 925 (mean: 787.61; standard deviation: 106.71; perfect score 990), which placed them under B1-B2 in the Common European Framework of Reference for Language (CEFR). The participants spent fewer than 60 minutes a day reading Japanese and fewer than 30 minutes a day reading English (for both leisure and academic purposes). They were averse to poetry-reading whether in Japanese and English, and most of them do not consume even one piece of poetry a year. They also possess a limited knowledge of English and American poets and were unfamiliar with the authors of the poems in this study (with a mean and standard deviation of 0.63 and 1.13, respectively, according to the Author Recognition Test [ART]). This study was conducted with the approval of the institutional review board of the university, which the author of this study and recruited participants belong to No. 2019539. All participants were given a financial reward after the experiment.

Poetry texts used in this study
This study followed a three-step text selection procedure. First, short poetry texts were collected from poetry anthologies (Bradford, 1993;Brooks & Warren, 1978;DiYanni, 2000;Hoover, 1994;Kacian et al., 2013;Kennedy & Gioia, 2002;Kennedy & Kennedy, 1999;Maley & Moulding, 1985;McRae & Vethamani, 1999;Ramazani et al., 2003;Riddell, 1972). Second, texts with low-frequency vocabulary and complicated syntactic structures were excluded. Finally, poetry texts with unfamiliar topics to participants (e.g., religious or region-specific themes) were removed, as the effects of background knowledge on poetry reading were out of this study's scope. In addition, this study also excluded texts with strong sexual connotations to avoid upsetting the participants during the experiment. A total of 19 poetry texts were selected as potential research material.
These 19 texts underwent stylistic analysis to confirm whether different types of texts were included in the shortlisted research materials. Consequently, the 19 texts were categorised into five text types. Type A texts were shape poetry (visual or concrete poetry), which used figures of letters, words or lines to create visual images. Type B texts were poems in which each line corresponded to clauses or phrases so that participants could conclude textual understanding at each line. Type C texts consisted of ordinary sentences cut into several lines and cast into stanza structures. These texts have enjambments, which prevent the confirmation of textual understanding at each line. Type D texts, which lack sentence structure, consisted of noun or adverbial phrases. Type E texts included a deviant syntactical structure. This study selected two texts from Types A, B, C and D and one text from Type E (the only candidate found in the text selection procedure) as research materials. Table 1 lists the chosen texts' sources; authors' names; line, word and syllable counts and average word frequency levels (calculated based on SVL 12000 2 ).
Fisher's Exact Test showed no significant difference in lexical frequency levels among these nine texts (p = .588, Cramer's V = .18). Considering the participants' English proficiency levels, all texts were lexically easy to read, that is, accessible, for the participants.

Questionnaire and ART
Before the experiment, the participants completed a questionnaire to evaluate their English proficiency, English-language learning background and reading habits. To check whether they were familiar with the poets' names of the research materials, they also took the ART (Stanovich  Castiglione's (2017) version using the poets' names for this study. The participant cohort based on these materials was already described above.

Procedures
Each participant individually took part in the laboratory experiment. This study used a computer with PPT2TTL software (WAWON DIGITECH), which controlled the material presentation. The participants sat in front of the computer screen, which presented instructions and randomly showed poetry texts. After receiving instructions and conducting practice sessions, the participants were provided with a vocabulary list of the words used in the nine poetry texts to eliminate the effects of unfamiliar words on poetry-reading processes. The participants then opened an experiment file and performed two reading tasks for each poetry text: literal comprehension and interpretation. 3 They were initially told to report their responses to the poetry texts in an interview at the end of the experiment. The texts were presented without titles and poets' names to exclude the effects of extratextual information on poetry-reading processes, which is out of scope in this study.
In the literal-comprehension task, each text was presented line by line until the entire text was displayed, and the participants must understand their literal meaning step-by-step. Since this was a self-paced reading task, they pressed the space bar to display the next line on the screen when they thought they had understood one line. Once they had been presented a line, they could not hide it. In this task, A1 and A2 (shape poems) were presented in a simple line form. When the participants finished understanding the final line of the text, they proceeded to the interpretation task. When they started the interpretation task for the same text, they could not revisit the literalcomprehension task anymore.
In the interpretation task, the participants reread the whole text to determine the poets' intended meaning. As discussed by Guy et al. (2018), Stockwell (2016), and Zyngier (1999), readers' consideration of authorial intention is natural and important in their meaning construction processes. To complete the task, the participants must interpret the lines and consider the poets' implicatures or emotions. Upon deciding that they have grasped the poet's intended message, or upon giving up on the task, they pressed the space bar again to start the literal-comprehension task for the next text. In this task, the participants were presented layouts of A1 and A2 for the first time.
After finishing all the tasks, the participants were interviewed with reference to a copy of the research materials. For each text in their reading order, they were asked to report on (1) the difficult line(s) in the literal-comprehension task, (2) the order of difficulty if they encountered more than two such lines in one text, (3) their confidence in their accuracy in understanding the literal meaning of the text, (4) their thoughts about the poet's intended message to the reader, (5) their confidence in their interpretations of the poet's intent and (6) other concerns that emerged when engaging with the text. The first three questions were relevant to the literal-comprehension task, while the fourth and fifth items were for the interpretation task. At the end of the interview, the participants were asked (7) whether they had any difficult texts for the literal-comprehension task (if yes, which one, and, if they had more than two such texts, their order of difficulty), (8) the same question for the interpretation task and (9) whether they had read or known any of the poetry texts used in this experiment.

Types of data
This study investigated learners' poetry-reading processes using four types of data.

Reading time
PPT2TTL automatically recorded the time the participants spent on the two abovementioned tasks when they pressed the space bar (in milliseconds). For the analysis, this study counted reading time per syllable, as the selected texts had different total word counts. For each poetry text, this study calculated reading times for the literal-comprehension and interpretation tasks.

Confidence level in poetry-reading performance
The relevant data for both literal-comprehension and interpretation tasks were collected from the participants' responses to interview questions (3) and (5) (provided in percentages).

Subjective assessment of textual difficulty
Relevant data were collected from the participants' responses to interview questions (7) and (8).

Postreading verbal report
This dataset consisted of the participants' verbal responses to interview questions (1), (2), (4) and (6), which were used to strengthen the description of their poetry-reading traits based on the three abovementioned data.

Results
In the interview, five participants reported reading at least one of the texts used in this experiment (question 9). For the remaining 19 participants who were not familiar with all the research materials before the experiment, their responses in the literal-comprehension and interpretation tasks for each poetry text are summarised in Table 2: reading time (per syllable), confidence level in poetry-reading performance (in percentage) and subjective assessment of textual difficulty (number of participants who found a given text difficult and most difficult).
The first three subsections below discuss the participants' responses in the literalcomprehension and interpretation tasks. The next two subsections highlight texts with long reading times, low confidence and high subjective difficulty and examine what caused these effects. Table 2 shows inconsistent results for the texts under the same category. For example, the participants' confidence level in the interpretation task significantly differed between A1 and A2. Consequently, the research outcomes presented below are based on the data analysis by text (not text type).

Literal-comprehension task
Results of the Friedman test showed a significant difference in reading time with a middle effect size among the nine poetry texts in the literal-comprehension (χ 2 (8) = 58.88, p < .001, η 2 = .39) 4 . As shown in Table 2, E1 and B1 were the slowest and the second slowest texts to understand, respectively, and C1 was the fastest to comprehend. Table 3 shows Conover's post hoc test results, indicating that the participants spent more time reading E1 than other texts except B1 and that they spent longer time reading B1 than C1.
Friedman test results also showed a significant difference in the participants' confidence level in their literal understanding of the texts (χ 2 (8) = 80.75, p < .001, η 2 = .53). According to Table 2, the participants had the lowest confidence in E1 and B1, that is, the slowest and the second slowest texts, respectively. Meanwhile, they felt quite confident in their understanding of C1. These results were consistent with those for reading time. Table 3, which presents Conover's post hoc test, shows that the participants were less confident in E1 than in any other poetry texts except B1 and less confident in B1 than A1, A2, C1 and D1. Also worth noting is that shape poems (i.e. A1 and A2) showed high literal-comprehension confidence as in C1.
Regarding the participants' subjective assessment of textual difficulty, one of them stated that they experienced no challenges during the literal-comprehension task. From the responses of the remaining 18 participants (Table 2), E1 was clearly the most difficult text in the literalcomprehension task, which parallels the results for reading time and confidence level. Only three participants experienced difficulty in B1 (the second slowest and second lowest-confidence text), which may be attributed to the significant numerical gaps in reading time and confidence between these two texts, as shown in Table 2 (according to Table 3, however, their differences were not statistically substantiated).

Interpretation task
Results of the Friedman test showed a significant difference in reading time with a middle effect size among the nine poetry texts in the interpretation (χ 2 (8) = 92.68, p < .001, η 2 = .61). Shape poems (A1 and A2) were the most slowly read texts (ninth and eighth) 5 , with E1, which was the slowest text to read in the literal-comprehension task, placing seventh (Table 2). D2 was also a rather slow-to-read text. Interestingly, B1 was read rather fast this time, and C1 was again read the fastest. Table 4 shows the results of Conover's post hoc test.
Friedman test results also showed a significant difference in the participants' confidence level in their performance in the interpretation task (χ 2 (8) = 39.86, p < .001, η 2 = .26). However, as shown in Table 2, the confidence level did not associate well with reading time unlike in the literalcomprehension task results. A1 and A2, the slowest-and the second slowest-to-read texts, showed the third lowest confidence and the highest confidence, respectively. E1 and D2, the third and fourth slowest texts, were the texts in which participants showed the lowest and second lowest confidence. As presented in Table 4, significant statistical differences were observed between E1 and the two highest-confidence texts (A2 and B2), and D2 and the highestconfidence text (A2). Although the participants still read C1 rather quickly, their performance confidence was low (sixth place), which is quite a different outcome from the literalcomprehension task. Table 2 shows that the participants considered A1, D2 and E1 as difficult texts and regarded D2 and E1 as seemingly more difficult than A1; although seven participants found A1 challenging, only one considered it the most difficult. Table 5 shows a comparison of reading time and confidence level for each poetry text during the two tasks. Unlike Tables 2, 3 and 4, the participants who had read or were familiar with a certain text were excluded only from that dataset; hence, the number of participants differed from one text to another. Table 5 shows that participants spent longer time in the interpretation task than in the literalcomprehension task. Wilcoxon signed-rank test results supported this tendency except for B1 and C1 (a middle effect size was observed for C1, however).

Literal-comprehension task vs. interpretation task
The table also indicates that in all the research materials, the participants were far less confident in their understanding of the poet's intent than their literal-comprehension of the text as the results of the Wilcoxon signed-rank tests show.
Regarding the participants' subjective assessment of textual difficulty, the texts they found difficult and most difficult varied more widely in the interpretation task than in the literalcomprehension task (Table 2). When the participants interpreted the poets' intent, textual difficulty seemed individuated. For instance, A2, which recorded the highest confidence level in the literal-comprehension task, was considered a difficult text in the interpretation task by two participants, with one of them even finding it the hardest to understand. These participants encountered no problems in their literal understanding of this poem, but P24 6 said, 'I think the repeated word symbolises something, but I couldn't understand what it means in this text, so I can't understand what the poet tries to say in this text' 7 . Four participants also reported that C1, the most rapidly read text, was challenging during the interpretation task (although no one considered it the most difficult). For instance, they thought in the last line must have a special meaning, which would help discern the poet's intent, but they could not identify what it was ("I   Note. LC, I, P, D and MD = literal-comprehension task, interpretation task, place, number of participants who reported the text was difficult and number of participants who reported the text was most difficult, respectively. The first rank (1) was assigned to the text with the shortest reading time and the one with the highest confidence level, and the bottom place (9) was given to the one with longest reading time and the one with lowest confidence level. .67 Note. LC and I = literal-comprehension task and interpretation task, respectively. can understand the literal meaning of the text, but I couldn't understand more than this. Perhaps the final line has some special meaning but I couldn't understand what it is". (P15)).

Textual difficulty in the literal-comprehension task
As Table 2 shows, reading speed was slowest and confidence level dropped most sharply with E1 during the literal-comprehension task. In addition, 18 of 19 participants reported E1 was a difficult poetry text, while 17 of them confirmed it was the most difficult text with which to achieve literal comprehension.
To investigate the causes of such difficulty, reading time for each line was compared (see ,  Table 6). Since no participants had read E1 before the experiment, all participant data were analysed. Friedman test results showed significant differences with a large effect size in reading time between the three lines during the literal-comprehension task (χ 2 (2) = 41.33, p < .001, η 2 = .86), while Conover's post hoc test results showed that the participants spent significantly longer time processing Line 3 than Lines 1 and 2, and Line 2 than Line 1.
In the interviews, 21 of 24 participants identified Line 2 as the most difficult. In response to interview question (1), 20 participants reported being surprised at the ungrammatical repetition of the same verb in Line 2, which blocked their literal-meaning comprehension as shown in the following verbal data.
• I was puzzled at the repetition of the same verb in Line 2. (P1) • I failed to understand Line 2, where the same verb was repeated. (P3) • I was not able to understand the grammatical structure of Lines 2. (P13) These outcomes seem to contradict the reading time results, but this can be explained as follows. The participants encountered a grammatical deviation (i.e. the repetition of the same verb at Line 2, which required longer reading time than Line 1, but because they could not solve the problem, they proceeded to the final line. Since they could not still arrive at a solution, they reviewed the entire text to finalise their literal comprehension (the wrap-up effect of semantic processing); as a result, Line 3 took the longest reading time, and participants considered Line 2 the most difficult. While Menninghaus and Wallot (2021) stated that esthetically preferred input could extend reading time (savouring effect), no participants reported enjoying or positively evaluating the grammatical deviation in Line 2 during the literal-comprehension task. Table 2 shows that the participants spent a fairly long time reading A1, A2, D2 and E1. Interestingly, as indicated by the average confidence level value (74.21%), they seemed to construct satisfactory interpretations of A2. In contrast, 7, 12 and 12 participants had difficulty with A1, D2 and E1, respectively; that is, despite spending longer time reading these texts, they gave up or could not interpret them satisfactorily. D2 and E1 seemed generally more difficult than A1; six participants found either text the most difficult, while only one participant considered A1 as such. According to their responses to the interview questions (4) and (6), when the participants performed the interpretation task with A1, D2 and E1, they encountered different reading problems derived from widely diverse textual traits, including letters, punctuation marks, words, grammar and layout (see left column of Table 7). Since they could not address these issues, they showed unsatisfactory performance. Some participants held certain assumptions when they interpreted the poets' intent in these three texts (see right column of Table 7) but reported not being confident as evidenced by their frequent use of Japanese expressions for "possibility", which, in English, correspond to "may", "seem" or "perhaps". Also, while they felt generally confident about their literal interpretations of the texts (E1 being exceptional), they were uncertain of what these texts signified or symbolised. Some participants demonstrated ambitious readings (e.g., P4 for D2 and P16 and P23 for E1), but the elusiveness of meaning (Lamarque, 2009) seemed to prevent them from being fully decisive about their interpretations (e.g., P4 had a 40% confidence for D2, and P16 and P23 were 40% and 50% confident for E1, respectively).

Textual difficulty in the interpretation task
Again, no responses indicated the savouring effect. Although the participants took a long time processing these texts during the interpretation task, they used the time to manage the difficulty they encountered or organise their ideas about the poets' intent (which they seemed to fail at, however, as the verbal data in Table 7 shows) and not to appreciate and evaluate the expressions in the texts.

Discussion
The above findings indicate that lexically easy short poems do not necessarily guarantee smooth understanding, especially regarding the construction of poets' intent (interpretation). Despite already achieving sufficient levels of English proficiency to understand the literal meaning of the research materials, the study's participants encountered different reading challenges with these texts in the interpretation task, lowering their confidence in their performance with a sense of textual difficulty.
This study identified the following lexically easy short poetry-reading traits by upperintermediate Japanese university EFL learners: • Learners read poetry texts of the same lexical level at different speeds with different confidence levels and difficulty.
• Learners spend much time constructing interpretations of shape poems.
• Learners need longer time to construct the poet's intent than to literally comprehend the text.
• Learners were less confident in interpreting the poet's intent than in understanding the text's literal meaning.
• Difficult texts differed individually when considering the poets' intent.
• Texts read at slow speed, texts in which readers were extremely unconfident and texts they found difficult tend to be associated (B1 and E1 in the literal-comprehension task and A1, D2 and E1 in the interpretation task).
• Learners' literal comprehension was disrupted by grammatical deviations.
• The participants encountered various reading problems when constructing the poets' intent but could not resolve them.
• Even when the participants constructed the poet's intent, they remained uncertain about their performance.
• Learners could not find the time to enjoy the text.
Based on these findings, the following provides pointers for using lexically easy short poetry texts in the classroom.
First, teachers must help participants construct interpretations even when the participants find a poetry text linguistically accessible. Assigning participants lexically easy short poetry texts does • I couldn't interpret the meaning of the triangular textual shape. (P8) • The sentence is grammatically incomplete. What word or phrase is omitted after "he last word of the first line"? (P11) • I wonder what is the poet's intent to leave "only one letter" in the final line. (P16) • Are there some meanings for the small space between "letters in the word" of "some"? (P16) • The sentence became shorter and shorter. What does this mean? (P20) • Why is the space between words deleted in this text? (P23) • The letter 'One letter was's were left in the final line, so when the poet thought about his/her best friend, he/she noticed he had one special friend. The "letter" may stand for the name of the poet's best friend. (P2) • not automatically mean they would enjoy constructing nonliteral meaning, such as that of the poets. Difficulty in constructing interpretation is not in proportion to the text's lexical level. Despite the participants' ability to quickly understand literal meaning and their confidence in their performance, they tended to require a long time to construct their interpretations; regardless, they might not possess a sense of confidence in their reading. For instance, P20 reported in the interview, "Although I can understand the literal meaning of the text [D2], I couldn't understand how these three lines are related". Understanding literal meaning and constructing interpretations are different processes, and the latter can be extremely challenging for learners even with lexically easy short poetry texts. 1 Second, poetry texts provide individual participants different reading experiences, which could be adopted by pair-or plenary-reading activities in class (Hanauer, 2001;Scott & Huntington, 2007). The participants read the same materials with different confidence levels and assessments of textual difficulty, especially in the interpretation task. As the verbal data indicates, lexically easy short poetry texts allow for different questions and readings, enabling participants to interact, exchange and negotiate (and collaborate in the construction of) their interpretation of the poet's intent. 2 Third, teachers must observe how participants cope with specific text features. The participants reported several textual features that hampered their reading, and their inability to address their reading challenges resulted in their unsatisfactory performance, making them feel a sense of textual difficulty. For example, their engagement with E1 showed that a syntactically deviant expression severely disrupted learners' literal comprehension despite knowing all the words in the text. 8 Since deviant grammatical structures are known to cause reading difficulties even with L1 readers (Castiglione, 2019), it would be natural for EFL learners to also experience comprehension issues with such text. They also had problems with various textual features during the interpretation task; for instance, they considered what is symbolised by specific expressions and the textual form in D2 and E1 but could not exactly identify them. These features could enable readers to construct a wide variety of interpretations (see, Fialho, 2007;Hakemulder & van Peer, 2016;Hanauer, 2018;McCarthy, 2015), but they also became sources of distress as this study demonstrated. 3 Fourth, learners may require a fairly long time to process shape poems, characterised only by the intensive repetition of some words. This means when planning reading activities with this type of texts, teachers must give learners plenty of time to construct their interpretations. In the literalcomprehension task, the participants quickly read the shape poems (A1 and A2) in a conventional poetry layout, but in the interpretation task, they needed more time to read these texts (even longer than E1, the text they considered most challenging). Before constructing their interpretations, the participants seemed to need time to "re-understand" the literal meaning based on the text's shape in the interpretation task. Shape poetry reading may follow three stages: understanding the literal meaning, mapping the literal meaning onto the text's shape and interpreting the poet's intent. However, as shown in Table 2, the text in which the participants spent the longest reading time does not necessarily mean it is the most difficult, as more participants considered D2 and E1 the most challenging texts. After spending considerable time, they seemed to construct more satisfactory interpretations with the shape poems than with D2 and E1. 4 Fifth, teachers must ensure that learners spend time evaluating or appreciating poetry texts, which is usually included in artwork appreciation (e.g., Frijda & Sundararajan, 2007;Menninghaus & Wallot, 2021). For interview question (6), no participants used words such as "I enjoyed", "I liked" or "this text is interesting". Time seemed to be spent for exclusively managing textual difficulty or finalising the meaning of the text. The present task design, which required participants to press the space bar as soon as they have understood the line or constructed their interpretations, might have prevented them from spending time enjoying the texts. 5 This study provides several opportunities for future research. First, scholars must closely observe learners' poetry-reading processes in natural settings. For instance, the question remains whether EFL poetry-reading processes start with literal-comprehension and advance to interpretation, which is the sequence assumed by the above experiment; indeed, the simultaneous operation of these two processes cannot be ruled out for now (e.g., Frisson & Pickering, 1999;Holyoak & Stamenkovi, 2018). Second, researchers must adopt various research perspectives when examining learners' poetry-reading processes. Learners' poetry-reading mechanisms could be further described using physical data such as brain activation, eye movement or perspiration as well as responses to a battery of questionnaires on emotions during poetry-reading. Third, future studies must identify the driving factors for the participants' reading processes and experiences. This study's participants processed poetry texts with the same lexical levels differently in terms of reading time, confidence level in reading performance and subjective assessment of textual difficulty irrespective of their stylistic categorisation. As suggested by Atler and Ratheiser (2019), researchers must bridge textual traits and their effects on readers. Scholars must also conduct similar studies that involve different texts including manipulated ones (see, Dixon & Bortolussi, 2008Kuzmičová et al., 2017;Miall, 2006 for the merits and demerits of using manipulated texts) to explain how textual traits influence reading processes. 6

Conclusion
This study showed that lexically easy short poetry texts do not necessarily guarantee smooth reading. The findings also clarified the reading traits held by EFL learners during their engagement with these linguistically accessible poems. Those with high English proficiency face a lot of difficulties that they cannot alleviate by themselves and do not feel confident in their reading performance, especially after interpretive processes. Teachers must establish instructional models that support learners' engagement in poetry and provide them a sense of pleasure, which could also facilitate their enjoyment of the proposed learning merits of poetry texts. Constructing such teaching models would require better understanding of their lexically easy short poetry-reading processes, which this study began as an initial step. There needs to be an accumulation of empirical studies that adopt different research methodologies in various contexts to achieve a detailed account of EFL poetry-reading and learners' experience. 7