Language-related Episodes Patterns Engendered in L2 Classrooms: Pedagogical Translation Tasks Versus Monolingual Writing Tasks

This classroom-based longitudinal study examines the language-related episodes (LREs) engendered by pedagogical translation tasks and monolingual tasks. Specifically, the study investigated the discourse features and discussion patterns of LREs yielded in discussions of the two different types of tasks in L2 classrooms. The present study was conducted in an authentic class environment and consisted of two experiments. One group participated in class discussions after being assigned L1-L2 translation tasks, while the other took part in class discussions after working on L2 writing tasks. The LREs elicited in the two group’s class discussions over ten sessions of class discussions were scrutinized. Inductive thematic analysis shows that LREs produced in the concerned experimental tasks included ‘concern-based LREs’ and six different types of ‘response-based LREs’. Striking difference was observed in LRE patterns produced by the translation group and the writing group. The findings revealed that pedagogical translation tasks engendered LREs with deeper level of engagement and more enduring discussions than L2 monolingual tasks. The concern-based and response-based LREs arisen in this study contributes to a new framework for LRE categorization.

evidence showing that linguistic knowledge co-constructed during LREs was retained and reflected in subsequent L2 performance (e.g. Swain, 1998;Swain and Lapkin, 2001;Williams, 2001). Most existing studies in the LRE literature examine the presence of LREs in monolingual tasks, such as direct writing, dictogloss, jigsaw, passage reconstruction, and text editing tasks (e.g. García Mayo, 2002aMayo, , 2002b; García Mayo and Azkaria, 2016; Kim and McDonough, 2011;Swain andLapkin, 1998, 2001;Storch 1998Storch , 2007Storch , 2008. To date, LREs have received different forms of categorization (see Figure 1). One prevailing categorization of LREs is by their nature, such as lexical or grammatical LREs (e.g. Källkvist, 2013a;Swain and Lapkin, 2001;Storch 1998Storch , 2013Williams, 1999Williams, , 2001. Some distinguished LREs by their outcomes, such as whether they are resolved correctly, resolved incorrectly, or left unresolved (e.g. Basterrechea & Lesser, 2019; McDonough, 2008, 2011;Lesser, 2004;Swain and Lapkin, 1998), while some examine LREs by its quality of engagement, which can range from 'elaborate engagement' to 'limited engagement' (e.g. Kuiken and Vedder, 2002;Storch, 2008). The former, as Storch (2008: 100-101) defines, refers to instances where learners 'deliberated over the language items, sought and provided confirmation and explanations, and alternatives', whereas the latter is instances where learners 'simply stated the linguistic item without further deliberation', such as in the case where one participant makes a suggestion but the other does not respond or only responds with a phatic utterance or repeats what was suggested. In investigating the LREs of 22 Asian ESL learners who completed a text reconstruction task, Storch (2008) found that learners' engagement with linguistic choices at both limited and elaborate level contributed to learning or consolidation of his targeted structures, but elaborate engagement was relatively more beneficial for L2 learning. These different forms of LRE categorization arise from the objectives of the studies and presence of these categories in the data generated. They allow for better understanding of the potential strengths or restraints of different pedagogical tasks.
This study is interested in investigating whether pedagogical translation tasks can also elicit high proportions of LREs like the way it did in Källkvist's study (2013a, 2013b) in a completely different language context, where L1 is Chinese and L2 is English. This study also seeks to fills the gap in LRE literature and analyze LREs by their 'discourse features' 2 and 'discussion patterns'.
Specific interest lies in LREs engendered by pedagogical translation tasks, as compared to monolingual writing tasks 3 . Precisely, the following research questions guided the analysis: (1) To what extent can translation tasks be exploited to foster language-related episodes (LREs) in L2 classrooms? (2) How do discussion patterns of LREs found in L1-L2 translation tasks compare with those in monolingual L2 tasks?

Methodology
The present study was conducted in an authentic class environment and consisted of ten sessions of class discussions. The first five sessions were known as Experiment I, while the subsequent five sessions were known as Experiment II. The participants were 26 third-year college students from a higher education institute in Macao and share similar language backgrounds, with Chinese as their first language and English as their second/foreign language ('L2' in this study). Informed consent from the participants were obtained for the Study. Names of the participants in each group were replaced by number S1-S26 to preserve anonymity.
The participants were randomly divided into two groups. In Experiment I (week 1-5), Students 1-13 were assigned to the Tgroup, while Students 14-26 were assigned to the NT-group. The translation assignments allocated to the T-group manifest judicious use of translation materials, with the source texts adapted from authentic L1 news articles consisting of challenging expressions. The writing assignments allocated to the NT-group were prepared using the same topics as the translation assignments as far as possible. The word count for both assignments was 200 word. Both experimental groups were given 30 minutes to complete their corresponding assignment before proceeding to a 20-minute class discussion, where they were invited to discuss language-related challenges, problems or concerns experienced in the corresponding tasks.
In Experiment II (week 6-10), the two groups swapped roles and worked on the other type of tasks and were again invited to corresponding class discussions. The presence of both experiments allows both groups to experience the two different types of task, so that the findings can better reflect the task effects rather than individual. Hence, if the results were attributed to individual differences, they may be equally revealed in the other experiment, where the roles were reversed. If certain students produced more response-based LREs because they were extrovert, this should be reflected when they worked on the other type of tasks, unless such results were not driven by individual difference but task effects, especially when the same trend was evident over repeated sessions.
The class discussions were audio-recorded and the recordings were transcribed verbatim for data analysis. Slightest forms of corrections and information were provided only when necessary. Each LRE was categorized, labeled and coded to identify the discourse features and discussion patterns of the LREs across the data.
The frequency of 'student-initiated LRE turns' was measured quantitatively. Each question, opinion or comment about language use generated by the learner was measured as one LRE turn, regardless of its duration. Clarifications by participants in response to teacher intervention was only treated as one LRE turn, while multiple questions or comments raised concurrently by the same student were regarded as separate turns if they were of completely different functions, nature or content.
The coding was grounded in the data. The themes were reviewed across Data A, refined, and given an informative name. This approach falls in the framework of a qualitative analytic method known as 'thematic analysis', which, according to Braun and Clarke (2006: 79), 'is a method for identifying, analyzing and reporting patterns (themes) within data' and (2006: 87) 'involves six phases: 1) familiarizing with the data; 2) generating initial codes; 3) searching for themes; 4) reviewing themes; 5) defining and naming themes; 6) producing the report'. Comparison of results between experiments and groups was then carried out.

Finding 1a: Concern-based and response-based LREs
By means of inductive thematic analysis, the discourse features of the LREs can be categorized as (1) concern-based LREs and (2) response-based LREs. The first category, 'concerns', covers turns where the students-initiated concerns, questions or doubts about language use, including instances where the students talked about difficulties experienced in the tasks assigned, language-related questions raised or uncertainties about their language usage expressed. The second category refers to turns where the students responded to their peers' concerns and classified into six sub-categories, as illustrated in Table 1.1. The first finding reveals that discussions in the T-group were comprised of many more responses to concerns compared to the NT-group. Within the T-group response-based LREs consistently outnumbered concern-based LREs by a large proportion across the five discussions, accounting for 83.9% (302/360) of the total LREs (see Appendix A). In contrast, the NT-group had very similar proportions of concern-based LREs (76/151) and response-based LREs (75/151), but there was no consistent trend over the five discussions. The T-group produced more response-based LREs than the NT-group in both Experiment I (227) and Experiment II (153) and the differences were statistically significant (Experiment I = p<0.001; Experiment II = p<0.05). On another dimension, when comparing the results for the same group of students in different tasks, the same results emerged. For Students 1-13, response-based LREs accounted for 83.8% and 55.7% of the total LREs in discussions of T-tasks and NT-tasks respectively. As for Students 14-26, 87.9% of their LREs were response-based in discussions of T-tasks, but only 49.6% LREs were responses in discussions of NT-tasks. These differences were all statistically significant (S1-13 = p<0.001; S14-26 = p<0.001).
Such results indicate that both groups of students focused more on responding to each other than on raising concerns and this tendency was much stronger when they were working on T-tasks (translation) than when they were working on NT-tasks (writing).

Finding 1b: Sub-categorization and proportion of response-based LREs
In this study, response-based LREs could be classified into six sub-categories based on their characteristics: (1) showing agreement or disagreement with peers' questions or solutions, (2) seeking for or providing clarification, (3) seeking for or providing explanations, (4) offering suggestions in response to questions or concerns raised by peers, (5) commenting on peers' solutions or opinions, and (6) commenting on peers' solutions or opinions and offering own insights, ideas or suggestions at the same time. [Some people think that celebrities who violate the law set a bad example to society]. I am using the 'the law will set up bad example'. I don't know is it right (if it is right to express it as) 'will set up'?

4
S5 Future tense. 'The stars break the law 4 will build a bad impression for society'. 5 S12 *Interruption: I agree with you.
The second sub-category 'seeking for or providing clarification' (RF) includes turns where students sought or provided clarification in cases of ambiguity, unclear pronunciation or other kinds of lack of clarity. In Example 1.3, student 13 tried to correct the grammatical mistake of student 4, but her comment appeared all of a sudden without clear explanation and therefore confused student 13, who then sought clarification. A clarification was provided by student 10, who noted that what she meant was that 'their donation beyond their need' was a mistake with the absence of the verb 'are'. The third sub-category, 'seeking or providing explanation' (RE), refers to turns where students sought or provided explanation or more detailed information. In Example 1.4, student 14 asked for an explanation from student 1 regarding the decision to use the first conditional, followed by an explanation provided by student 23 in Turn 18.
[If motorists are/were charged for entering city centre during peak hours, some will/would probably consider adjusting their travel patterns.] -This seems like a conditional sentence to me; I am not sure if (I) should use first conditional or second conditional. 16 S23 In my paragraph, I used conditional 1: 'people will consider to change'.

S23
Because I think this sentence is talk about (talks about) the phenomenon of now, so I will choose to use conditional one.
The fourth sub-category 'offering suggestions' (RS) is instances where students shared their opinions with each other, offering suggestions to a concern raised: Example The fifth sub-category, 'making comments' (RC), includes turns that are evaluative in nature or involve critical analysis of others' language usage or opinions fall into this category. As in Example 1.7, the students were evaluating the appropriateness of using terms like 'celebrity', 'famous people' and 'public people' to express the Chinese concept, 公眾人物, which means 'public figures', with one commenting on the usage of another. 244 S12 What about 'hidden name'? 'We can leave a comment hidden name at the internet'… 246 S10 I've written as 'they can hide their identities'. It's the same as 'hidden name', because I think 'identity' is higher level than 'hidden name'; just a name you can name any name, but identity is what is important… Scrutiny into the proportion of response-based LREs shows that in Experiment I the most common type of responses found in the T-group was suggestions (RS), followed by comments (RC) and agreements (RA), accounting for 83.4%, 7% and 4.6% of the response-based LREs respectively (see Appendix B). As for the NT-group, the most common type of response-based LREs was suggestions (RS), comments (RC) and explanations (RE), accounting for 76%, 16% and 5.3% of the response-based LREs. The groups were similar in that the majority of responses were RS and RC, but the actual number of RS and RC produced by the Tgroup was much greater than the number of RS and RC produced by the NT-group.
In Experiment II, the most prevalent type of response-based LREs in the T-group was suggestions (RS), followed by comments (RC) and a combination of suggestions and comments (RCS), accounting for 46.8%, 34.2% and 9.5% of the total number of response-based responses. As for the NT-group, the most common type of responses was also suggestions (RS), followed by clarifications (RF) and comments (RC), accounting for 80.2%, 8.4% and 6.1% of the total number of response-based responses respectively. In general, Experiment II produced similar results as Experiment I in that suggestions (RS) still accounted for the highest proportion of response-based LREs in both the T-group and the NT-group. However, there were some noticeable differences. First, the proportion of RS and RC produced by the T-group was less uneven. Second, a combination of comments and suggestions (RCS) was rare in Experiment I for both groups but became one of the most prevailing types of responses in the T-group in Experiment II.
The same phenomenon held true even in comparing the results for the same group working on different tasks. Students 1-13, who were assigned to the T-group in Experiment I and to the NT-group in Experiment II, produced more RS and RC in discussions of translation tasks than in discussions of writing tasks, but in discussions of writing tasks they produced more RF. On the other hand, Students 14-26, who were assigned to the NT-group in Experiment I and to the T-group in Experiment II, produced a greater number of RS, RC and RCS in discussions of translation tasks than in discussions of writing tasks.

Finding 2: Discussion patterns
To address Research Question 2, the discussion patterns of the two groups were compared and striking difference was observed (see Appendix C and D for details).
The findings show that concerns raised in discussions of T-tasks in Experiment I often prompted responses and the topics discussed were more likely to be addressed by several students. The majority of concerns raised by students were followed by a number of responses, ranging from 1 to 17 (see Table 1.2). Only on three occasions were there concerns left with no responses. The most prevalent patterns 5 were '1C+5R', '1C+3R', '1C+4R' and '1C+8R', accounting for approximately 17%, 14%, 14% and 10% of the overall patterns in the five discussions.

Table 1.2 Patterns of concern-based and response-based LREs in Experiment I
As for the NT-group, almost the opposite trend was observed. Concerns raised in discussions of NT-tasks often failed to prompt responses from peers and even when they were responded to. Only in a few instances was a concern followed by 5 to 6 responses, and there were no instances in which a concern was followed by more than 6 responses, which was common in the Tgroup. The most prevailing pattern was '1C alone', '1C+1R' and '1C+2R', accounting for approximately 55%, 17% and 13% of the overall patterns. In other words, more than half of the concerns raised were not followed by any response and even when there were responses, they were very limited. The results of Experiment II echo that of Experiment I, where all concerns in discussions within the T-Group were followed by response(s), ranging from 1 to 21. Not once were the concerns left unattended (see Table 1.3). The most common pattern was the '1C+4R' pattern, accounting for almost 13% of the patterns, but patterns like '1C+5R', '1C+7R', '1C+8R' and '1C+11R' were very common, each accounting for about 10% of the patterns. There were many instances where the T-group had heated debates. On three occasions, there were patterns with 1 concern plus 13 responses, 16 responses and 21 responses. As for the NT-Group, the most typical pattern was '1C alone', followed by '1C+1R', '1C+2R' and '1C+3R', accounting for approximately 36%, 32%, 12.6% and 12.6% respectively. None of the concerns were followed by more than 7 responses and there were again many concerns left unattended.

Table 1.3 Patterns of concern-based and response-based LREs in Experiment II
The fact that in both experiments concerns in discussions in the T-group were frequently addressed by a number of students while those in discussions in the NT-group were either followed by no response or just a handful of responses suggests that discussions of T-tasks were more enduring and tended to encourage better interaction between students than discussions of NT-tasks.
Such striking contrast can also be seen by comparing the results for the same group of participants across different tasks. Both Students 1-13 and Students 14-26 frequently had one concern plus five or more responses in discussions of T-tasks but rarely did their concerns in NT-tasks prompt more than five responses. The '1C-alone' pattern was rarely found in their discussions of T-tasks but very common in their discussions of NT-tasks.

Concern-based and Response-based LREs
Raising concern is necessary for a discussion because someone has to raise a concern before anyone can respond. However, a discussion cannot be considered as a discussion if it only contains concerns and fails to elicit responses, which allow for circulation of ideas among students. Compare the scenario where a group of students raise concerns and respond to each other, as opposed to the scenario where a group of students simply keep raising their own concerns. The first scenario is preferable because 'more responses' indicates a stronger interaction between students in class discussions. In the light of this reasoning, the finding that there were more responses in discussions among the T-group than among the NT-group suggests that translation tasks can yield better outcomes in engendering interactive class discussions than writing tasks. Interestingly, the proportion of responses and concerns may explain why the difference in the number of LREs generated in the T-group and the NT-group was statistically significant (p<0.05) in Experiment I but not (p>0.05) in Experiment II, despite their parallel trend, while results for each group across tasks were statistically significant (p<0.05). Scrutiny of the t-test results of responses (R) and concerns (C) reveals why three out of four results were statistically significant (see Table 2.1 and Table 2.

2)
In Experiment I, the number of concern-based LREs (mean=5.85) in the NT-group exceeded the number in the T-group (mean=4.46) but the results were not statistically significant (p>0.05), which suggests that the difference was not salient; on the other hand, response-based LREs in the T-group (mean=23.23) outnumbered those in the NT-group (mean=5.77) by a statistically significant proportion (p<0.05). Hence, the fact that the trend of more responses (T>NT) overrides that of more concerns (NT>T) makes it explicable why in total the T-group (mean=27.69) produced statistically significantly more LREs than the NT-group (mean=11.62, p<0.05). In Experiment II, the results of response-based LREs tallied with those in Experiment I, with the T-group producing many more responses (mean=21.85) than the NT-group (mean=10.08) and the difference was statistically significant (p<0.05). However, unlike in Experiment I, concern-based LREs in the NT-group (mean=7.92) outnumbered those in the T-group (mean=3.00) by almost half, and the difference was statistically significant (p<0.05). Here an opposite trend was observed, in which the NT-group had significantly more concern-based LREs than the T-group, whereas the T-group had significantly more response-based LREs than the NT-group. This is one plausible reason why the difference between the LREs in the two groups was not statistically insignificant (p>0.05), though both experiments revealed the same tendency where the T-group produced more LREs than the NT-group. This is also why it is necessary to look at what kinds of LREs were produced in addition to examining the quantity of LREs. As discussed earlier, it is desirable to see responses in discussions and not simply concerns, because responses allow ideas to circulate among students and indicate a stronger interaction between students in class discussions. Therefore, although there was no significant difference between LREs in the T-group and the NT-group in Experiment II, it is worth noting that in fact the difference between the number of response-based LREs occurring in the two groups was statistically significant (p<0.05).
Overall, the results from both experiments across the T-group and the NT-group and results of the same group of participants across different tasks all manifested the same trend, suggesting that translation tasks have greater potential to encourage students to respond to their peers in class discussions than writing tasks.

Response-based LREs
The qualitative data in this empirical study presents a new framework for LRE categorization. That is, in addition to its nature, outcome and engagement level (Figure 1), LREs can be categorized by its discourse features, including 'concern-based' and six types of 'response-based' LREs. An illustration is summarized in Figure 2.
Scrutiny into the proportion of response-based LREs in both groups shows that the dominant responses in the T-group were suggestions (RS) and comments (RC), with 252 RS and 21 RC in Experiment I and 133 RS and 97 RC in Experiment II. The prevailing ones in the NT-group were RS (57) and RC (12) in Experiment I and RS (105) and RF (8) in Experiment II, but they constituted a much smaller proportion of the LREs in the NT-group than in the T-group. One noteworthy result is that a combination of suggestions and comments (RCS) was rare for both the T-group (5) and the NT-group (1) in Experiment I as well as the NT-group (1) in Experiment II, but grew to be the third most prevailing type of responses in the T-group (27) in Experiment II. While all responses were considered good signs of engagements in discussions among students, they may indicate different types of attention. RA, for instance, can be considered to be the one requiring the least critical thinking; however, it is inevitable when students coincidently have the same view as their peers. These situations are only likely to emerge when students share similar problems. RF and RE can be signs of a lack of clarity, where the speakers fail to explain their point clearly. They could also be signs of engagement, where students seek clarification or explanation because they have been listening to their peers and they would like to gain a better understanding of the point made by their peers. Either way, it signals students' interest in the discussions. The rationale is that the students would not bother to ask for clarification if they were not listening to their peers or had no interest in what they were saying. Offering clarifications and explanations are positive training for the students because they provide opportunities to re-organize expressions.
RS can be considered as the standard type of responses that most students may produce because it is an opportunity to share their own solutions to problems that they have pondered over, while RC can be considered as responses that involve more critical thinking because students need to digest the concerns raised and analyze whether they are right or wrong before they can reason with their peers. It also involves more language mediation and elicits more responses from peers. Both RS and RC may emerge when students have encountered similar problems, when students have knowledge of the problem, or when students try to think instantly to respond to their peers. Either way, these two types of responses are more likely to emerge when students have similar problems or the same context. This assumption can be supported by the phenomenon that in this study RS and RC produced by the T-group consistently outnumbered those produced by the NT-group in both experiments over repeated sessions.
Among all kinds of responses, RCS can be considered most welcome for L2 classrooms because it requires critical thinking and involves the students not only commenting on others' language usage, but also offering suggestions of their own. It is worth noting that RCS was rarely found in either group in Experiment I, rarely found in the NT-group in Experiment II, but was prevalent in the T-group in Experiment II. One possible factor for the rise of RCS in the T-group in Experiment II might be the effect of practice, because the students who worked on translation tasks had already become used to the procedures of this type of class discussions (when they worked on writing tasks in Experiment I) and thus were able to be more comprehensive and more critical. Another reason may be that the students showed more interest in the assignment topics devised for Experiment II, which is also why a longitudinal study is necessary to compare task effects. The fact that RS, RC and RCS in the T-group consistently outnumbered those in the NT-group by a large proportion suggests that the nature of translation tasks may create more opportunities for the students to offer suggestions and comments on each other's suggestions than writing tasks do.
These different levels of engagement, according to Storch (2008), can have different impacts on L2 learning. RA seems to fall into what Storch describes as 'limited engagement', where students simply repeat the matter being discussed or respond with phatic utterances. The remaining five types of responses seem to be what Storch categorizes as 'elaborate engagement' as they reflect students' deliberation over the problems being discussed. Having found that students who engaged at the elaborate level demonstrated better use of the targeted structures in a subsequent task, Storch (2008: 111) suggests that although both types of engagement are beneficial, 'elaborate engagement' has a much stronger impact on L2 learning. The findings in this study are consistent with Storch's claim (2008) that LREs which can be described as 'elaborate engagement' may have stronger impact than those that show 'limited engagement'. Although it is beyond the scope of the current work to determine whether and in what way the LREs on the limited and elaborate levels relate to subsequent L2 development, an interesting finding pertaining to the impact of LREs in the RC and RS (and therefore applicable to RCS) categories emerges from observation of the discussion patterns, suggesting that LREs in these categories contribute to more enduring, sustainable and interactive discussions.

Discussion Patterns
The majority of discussions among the T-group in both experiments featured patterns with one concern followed by a number of responses, reaching a maximum of 17 and 21 responses in the first and second experiment respectively. In sharp contrast, the dominating discussion patterns in the NT-group were either a concern alone with zero responses or a concern that was only followed by a few responses, with a maximum of 8 responses in rare cases. The one-concern-alone pattern was rare in the Tgroup but very common in the NT-group. These diverging discussion patterns further explain why there were more LREs in the T-group than in the NT-group and suggest that the nature of translation tasks create more opportunities for enduring and interactive discussions than writing tasks.
In order to understand what leads to more enduring discussions, the three longest discussions in both groups in Experiment I and Experiment II were scrutinized and it was found that RS and RC played a dominant role in the three longest discussions in each group. This suggests that if the tasks are able to invite suggestions or comments, the discussion is more sustainable. This finding seems to provide further support for Storch's (2008) claim that the quality of attention (i.e. whether the LREs are at limited or elaborate level) affects impact. Instead of showing how attention affects students' subsequent L2 performance, the present data suggests how it may affect the discussion itself.

Concluding Remarks
This classroom-based study provides rich empirical analysis on how L2 learners interact when given translations tasks. The results suggest that L1-L2 translation tasks play a more conducive role in engendering class discussions than monolingual writing tasks in so far as they prompt more responses from students and encourage more enduring discussions. The finding that translation tasks can foster LREs with a deeper level of engagement is particularly encouraging considering such kind of 'elaborate engagement' can be useful in fostering L2 learning (Storch 2008). Overall, the findings point to the potential of using translation tasks to foster communication and have positive pedagogical implications for the use of translation in L2 classrooms.
The LRE categorization that arises from this study, where translation tasks are used pedagogically in L2 classrooms, also contributes to the LRE literature. The concern-based LREs and six types of response-based LREs recurred in this study may present a new framework for mapping and categorization of LREs in language studies, Meanwhile, the analysis of learners' discussion patterns also presents an unprecedented method to explore differences between LREs engendered by task types. The findings of the current study also suggest that LREs with deeper level of engagement, such as the 'RS' (suggestions), 'RC' (comments) and 'RSC' (suggestions plus comments) frequently found in translation tasks, can have stronger impact than limited engagement with phatic utterances like 'RA (agreements)' in that they can contribute to more interactive and enduring discussions.
This study compares the overall trends of the two groups involved. Future studies can examine the effect of different tasks on each individual or explore the impact of different genders or proficiency levels on the results of similar experimentation. Empirical research comparing the discourse features or discussion patterns of LREs across a wider range of task types or in different language contexts and educational settings can also offer other pedagogical insights. The major limitation of this study is the sample size as it only focuses on the overall trends of the two groups involved. Future studies can examine the progress of each individual and observe their improvements. Future studies may also use a larger sample size and different language contexts or educational settings.

APPENDIX A
Quantitative Results: Proportion of concern-based and response-based LREs