1 Double-sidedness of collaboration in language learning
Language learning has benefited from facilitative work of peer collaboration, the process of which requires learners’ constant efforts to arrive at some “mutually agreed-upon, or intersubjective, understanding” (Tudge, Reference Tudge1992: 1365). Intersubjectivity refers to shared understanding among the participants of an activity (e.g., Bretherton, Reference Bretherton1991; Kaye, Reference Kaye1982; Rogoff, Reference Rogoff1990; Rommetveit, Reference Rommetveit1979, Reference Rommetveit1985). Collaboration including students’ attempts to establish intersubjectivity develops their communication skills as well as social relationships and group cohesion (e.g., Johnson & Johnson, Reference Johnson and Johnson1999, Reference Johnson and Johnson2002; Koschmann, Stahl & Zemel, Reference Koschmann, Stahl and Zemel2004; Koschmann, Zemel, Conlee-Stevens, Young, Robbs & Barnhart, Reference Koschmann, Zemel, Conlee-Stevens, Young, Robbs and Barnhart2003; Roschelle, Reference Roschelle1996; Stahl, Reference Stahl2006). In addition, collaboration allows students to engage in deeper level learning, develop critical thinking skills, and retain instructed information in the long term (e.g., Garrison, Anderson & Archer, Reference Garrison, Anderson and Archer2001; Johnson & Johnson, Reference Johnson and Johnson1999; Kreijns, Kirschner & Jochems, Reference Kreijns, Kirschner and Jochems2003; Webb & Palincsar, Reference Webb and Palincsar1996). As Jonassen (Reference Jonassen1991, Reference Jonassen1994) and Keen (Reference Keen1992) have suggested, collaborative learning also facilitates co-construction of knowledge and development of learners’ overall competencies. In particular, studies investigating students’ web search processes found that peer collaboration guided the search process and regulated the search results. The collaborative group also showed better results in monitoring and evaluating the search process by employing various types of search strategies (e.g., Caskey, Reference Caskey2003; Lazonder, Reference Lazonder2005; Prekop, Reference Prekop2002). Lazonder (Reference Lazonder2005) found that a group searching the web collaboratively identified targeted information more quickly and more frequently than the group engaging in the same web search task individually.
However, some studies have found that successful collaboration cannot be guaranteed at all times. Students’ collaborative work may lead to incorrect conclusions. As Tudge (Reference Tudge1992) notes, the results of the establishment of shared meaning, or intersubjectivity, could be adverse or advantageous, depending on whose idea is considered as correct initially. In addition, the process of collaboration may include competition, which naturally creates conflicts and power inequality among the participants (e.g., Chan & Chen, Reference Chan and Chen2010; Cheng, Reference Cheng2013; Guzdial & Turns, Reference Guzdial and Turns2000; Leki, Reference Leki2001; Lipponen, Reference Lipponen2002; Lipponen, Rahikainen, Lallimo & Hakkarainen, Reference Lipponen, Rahikainen, Lallimo and Hakkarainen2001). In fact, “collaborative situations are also full of contradictions, competition, and conflicts” in reality (Lipponen, Reference Lipponen2002: 76). Chan and Chen (Reference Chan and Chen2010) identified major sources of conflicts in collaborative learning such as poor management of communication, power inequality, participants’ egocentricity, conflicts of values, and lack of responsibility.
On the psychological front, studies on affective and motivational aspects of collaboration have suggested drawbacks of collaboration in the learning process (e.g., Berejkovskaya, Reference Berejkovskaya2006; Järvenoja & Järvelä, Reference Järvenoja and Järvelä2009; Levykh, Reference Levykh2008; Thompson & Fine, Reference Thompson and Fine1999). In the process of collaboration, conflicts stemming from students’ different characteristics, goals, and demands can cause negative emotions and motivational challenges (e.g., Järvenoja & Järvelä, Reference Järvenoja and Järvelä2005, Reference Järvenoja and Järvelä2009; Järvelä, Lehtinen & Salonen, Reference Järvelä, Lehtinen and Salonen2000; Kreijns et al., Reference Kreijns, Kirschner and Jochems2003). In particular, Järvenoja and Järvelä (Reference Järvenoja and Järvelä2009) suggested that the process of co-construction of knowledge through meaning-sharing could generate a high level of socio-emotional challenges, reducing the learners’ motivation to participate in collaboration.
2 Potential task dependency of facilitatory effects in collaborative corpus consultation
Despite the negative aspects, collaborative learning has gained popularity and been increasingly incorporated into learners’ corpus consultation. A number of studies have found that it improved their lexico-grammatical and linguistic awareness as well as learning autonomy (e.g., Boulton, Reference Boulton2009; Chambers & O’Sullivan, Reference Chambers and O’Sullivan2004; Conrad, Reference Conrad1999; Huang, Reference Huang2014; Lee & Swales, Reference Lee and Swales2006; Yoon, Reference Yoon2008; Yoon & Hirvela, Reference Yoon and Hirvela2004), but that learners experience a significant degree of difficulty and can be overwhelmed in the face of the enormous amount of corpus data (e.g., Ädel, Reference Ädel2010; Conrad, Reference Conrad2005; Cook, Reference Cook1998; Prodromou, Reference Prodromou1996). In order to alleviate the difficulty, some studies have suggested that peer collaboration can work as pedagogic mediation to assist learners’ corpus consultation (e.g., Flowerdew, Reference Flowerdew2008; Gavioli & Aston, Reference Gavioli and Aston2001; O’Sullivan, Reference O’Sullivan2007). These studies noted that the learners’ collaboration enabled co-construction of the learning process of corpus consultation, which allowed for more generalizable interpretation of corpus data. In particular, Gavioli and Aston (Reference Gavioli and Aston2001) pointed out that students often noticed different things and drew different conclusions from the same corpus data. When they worked collaboratively, the students were able to reach more comprehensive conclusions. For instance, when examining concordance lines for food and its Italian equivalent cibo, students in the study made joint efforts to test hypotheses on semantic differences between the two words and come to correct data interpretation, which might have been impossible if they had worked individually.
However, it should be noted that not all collaboration comes with fruitful results in this area. In an attempt to investigate the change in learners’ attitudes toward grammar learning by introducing corpus consultation, Estling Vannestål and Lindquist (Reference Estling Vannestål and Lindquist2007) used a group activity where students worked in pairs to formulate grammar rules from examples they found in corpus data, and pairs of students took turns explaining the grammatical rules to each other. However, the findings of the study noted that the students experienced difficulties, which required a large amount of introduction and support in the analysis. In particular, some students found the corpus work difficult in terms of formulating their queries and interpreting the results. Some even questioned the usefulness of corpus consultation for learning grammar rules, suggesting that peer collaboration might not have been successful for assisting the process of learners’ corpus consultation.
At this point, the mixed results on the role of collaboration in learners’ corpus consultation seem to require meticulous examination of the double-sided effect of collaboration and its relation with task types in corpus consultation (e.g., Berejkovskaya, Reference Berejkovskaya2006; Järvenoja & Järvelä, Reference Järvenoja and Järvelä2009; Kreijns et al., Reference Kreijns, Kirschner and Jochems2003; Levykh, Reference Levykh2008; Thompson & Fine, Reference Thompson and Fine1999). Collaborative corpus tasks can be categorized into two types: those which require students to pool their conceptual competence collectively to reach a generalizable conclusion (e.g., Gavioli & Aston, Reference Gavioli and Aston2001) and those which include a sequence of related tasks exploiting students’ procedural competence (e.g., Estling Vannestål & Lindquist, Reference Estling Vannestål and Lindquist2007). Gelman and Meck (Reference Gelman and Meck1986) distinguished between ‘procedural competence’ and ‘conceptual competence’ assuming that the former relates to procedure performance, whereas the latter refers to knowledge principles. Similarly, Mullins, Rummel and Spada (Reference Mullins, Rummel and Spada2011) note that conceptual instructional tasks trigger elaborate meaning-making processes while procedural instruction draws learners’ attention to strategic problem-solving procedures. In general, a procedural task requires the application of sequential patterns or rules, and involves strategic processes such as monitoring, planning, developing, and testing hypotheses (e.g., Moscovitch, Reference Moscovitch1994; Vakil & Hoffman, Reference Vakil and Hoffman2004).
Given the two fundamentally distinct types of competences, utilizing conceptual and procedural tasks in corpus consultation may create different conditions for the functions of collaboration. In this sense, an investigation of the task-dependent effects of collaboration in learners’ corpus consultation would bring practical implications for incorporating collaborative learning into corpus-based instruction. To that end, this study presents a qualitative examination of the process of conceptual and procedural tasks in corpus consultation by two groups (i.e., individual vs collaborative corpus consultation groups). This study seeks the answers to the two following research questions:
1. How does the collaboration work in a conceptual and a procedural task during learners’ corpus consultation?
2. Does the effect of collaboration show significant differences depending on task types during corpus consultation?
3 Methodology
3.1 Participants
Through an official community website of a four-year university in Seoul, Korea, this study recruited six female students. They varied in terms of age, major, TOEIC (Test of English for International Communication) score, and experience studying abroad. S1, S2, and S3 were in the Collaborative Group (CG), which worked together to complete two corpus consultation tasks. In order to set a baseline to compare with the results of the CG, this study constructed an Individual Group (IG), including S4, S5, and S6, who did the same tasks as the CG students but on their own. A description of the participants is provided in Table 1. S1 was intentionally assigned to the CG because her relatively low English proficiency could be compensated through collaboration, and S2 and S3 were willing to work as a team with S1.
Table 1 Description of participants
Note: According to ETS grading system, TOEIC scores of 860 or higher are graded as level A, 730 or higher as level B, 470 or higher as level C, 220 or higher as level D, and scores less than 220 as level E.
In the CG, S1 was the oldest, as she had changed her major after graduation from a different college. Her English proficiency level was the lowest among the group members based on her TOEIC score. Despite her major (International Relations), which required a high English proficiency level, she said that her English was not good and that she sometimes did not understand lectures in English. S2 majored in Business Administration and noted that her major required high English proficiency. When she was introduced to corpora during the first session of this study, she showed a strong interest in corpus consultation as a novel and innovative source for English learning. Finally, S3 was the most proficient student, who was the youngest but had the highest TOEIC score. She had spent six months in Canada as an exchange student in college, while the other students did not have any experience abroad. Even though her major was not related to English, S3 had a strong interest in learning English. During the corpus consultation, it was found that her reading comprehension of English texts was the fastest and most accurate among the group members.
S4, S5, and S6 were instructed to do the corpus consultation tasks individually. S4 majored in English and had interest in English translation and interpretation. During the pre-interview, she displayed strong enthusiasm for corpus analysis as a tool for improving her English skills. In a written survey of personal information, she noted that her computer skills were at a low level, but during the corpus consultation tasks, it was found that she had good enough skills to search corpus data by herself. S5 was a student in International Relations but, unlike S1 who was in the same major, she seemed to have a good command of English and sufficient computer skills as demonstrated during the corpus consultation tasks. In the pre-interview, she said that participation in this study would be a great opportunity to discover a new way of learning English. S6 had the highest TOEIC score among the participants in this study. In preparation for job seeking, she said that she had taken a series of TOEIC tests to increase her score. During the corpus consultation tasks, it was observed that she read through and analyzed concordance lines very efficiently as she completed the two tasks the quickest among the six participants with accurate answers.
3.2 Tasks
In order to examine task-dependent effects of collaboration, this study employed two types of corpus consultation tasks based on the difference between conceptual and procedural competence (see e.g., Anderson, Reference Anderson1993; Gelman & Meck, Reference Gelman and Meck1986; Mullins et al., Reference Mullins, Rummel and Spada2011). As conceptual competence is related to constructing meaning-making processes by exploiting hierarchical knowledge principles, the conceptual corpus consultation task asks students to draw a generalizable interpretation of corpus data by using their conceptual knowledge of English. In the conceptual task in this study, students consulted a corpus to discern semantic and functional differences between synonymous expressions (Figure 1), as did the students in the study by Gavioli and Aston (Reference Gavioli and Aston2001). For the selection of the search terms for corpus consultation, the students submitted a list of synonymous expressions whose differences were indiscernible to them on the first day of the experiment, and the researcher used the expressions in designing the task.
Fig. 1 The conceptual task of corpus consultation
Procedural competence is the ability to tackle challenges in a step-wise fashion by making strategic decisions. In this study, the procedural task consists of two sub-tasks: targeted corpus consultation and translation by applying results from the prior corpus consultation. This targeted corpus consultation needs to be distinguished from the conceptual task in that the former is highly purposeful and specifically targeted at a subsequent task to be undertaken. In terms of the subsequent task, translation was used in this study because it is a problem-solving task by nature (i.e., finding functional and semantic equivalence and restructuring the message in L2), which can be accomplished by targeted corpus consultation. From a motivational perspective as well, translation seemed to be an appropriate sub-task of this study because the students noted during the pre-interview that they would like to use corpus data as a reference source for L2 writing in the future. As a standardized task of writing for group comparison, translation was chosen to clearly demonstrate differences between the two groups in dealing with the procedural corpus consultation task.
In designing the translation activity, the researcher produced a short Korean passage consisting of three sentences, whose translation into English required corpus consultation to discern semantic differences of synonymous expressions. In addition to the Korean passage, three sets of synonymous expressions (even as/even though/during, worth/worthy, hope/wish) were presented on the worksheet as options for students to use in their translation. Just as in the conceptual task, the expressions in the procedural task were those that the students acknowledged as difficult in discerning differences. In this procedural task, students firstly needed to figure out semantic and functional differences of synonymous expressions through corpus consultation in order to complete the subsequent translation (Figure 2). The conceptual and procedural tasks were administered on two consecutive days.
Fig. 2 The procedural task of corpus consultation
3.3 Data collection procedure
On the first day of the experiment, the students were asked to answer a written survey to gather personal information including age, major, and TOEIC scores to evaluate their linguistic proficiency level. At the end of the survey, the students wrote sets of synonymous expressions whose usage and semantic differences were indistinguishable to them such as affect/effect/influence and hope/wish. The expressions were used in the two corpus consultation tasks in this study. After the survey, an introduction to corpus use was given to improve the students’ basic understanding and hands-on experience of corpus consultation. During the one-hour introductory class, students were introduced to definitions and types of corpora, collocations, and concordancers. In terms of the concordancers, the students were acquainted with the Lextutor software (The Compleat Lexical Tutor: http://www.lextutor.ca), and used parts of the Brown corpus, BNC Written, and BNC Spoken all together (one million words respectively). During hands-on corpus activities, the students learned and practiced some techniques with Lextutor such as sorting and keywords for more effective corpus analysis. After the Lextutor activities, the students were interviewed to ascertain their impressions and expectations of using corpus linguistics for language learning.
The next day, students worked (either CG or IG) to complete a conceptual corpus consultation task. The IG used private computers in corpus consultation whereas the CG was given one computer to share, which was to encourage verbal and nonverbal communication among members (Figure 3).
Fig. 3 Photographs of students sharing one computer and using private computers respectively for corpus consultation
Before the task, the instructor gave each student a worksheet, which included two sets of fairly close synonyms (affect/effect/influence and travel/journey/trip) whose semantic and functional differences were unclear to the students (Figure 1). CG students were required to create hypotheses explaining differences between the items through group discussion following Lextutor searches. The IG students were responsible for their own corpus analysis to complete the task, but they were strongly encouraged to ask the instructor any questions. After the task, the two groups were interviewed to gain their impressions on the activity. Both CG and IG students did the corpus consultation as well as the interview separately so that the IG students would not feel disadvantaged compared to CG by working individually.
On the third day, students were given a procedural task, consisting of two related sub-tasks (i.e., corpus consultation to discern semantic differences between synonyms, and translation of a text into the L2 by using the results of corpus consultation). During the activity, the students needed to find the best answers by discerning the semantic differences between the synonymous items through corpus consultation, and to use them in translation (Figure 2). As in the conceptual task, the CG students engaged in the corpus consultation and completed the translation collaboratively with a single computer while the IG students did the task individually.
Finally, at the end of the third session, the researcher conducted group and individual interviews to examine the students’ impressions of the procedural task as well as any change in their attitudes and impressions of corpus consultation in general. During the semi-structured interview, the students shared both positive and negative responses to their corpus consultation experiences. In particular, they were asked about the degree of various types of difficulties they experienced during the two tasks and how they coped with those difficulties.
3.4 Data analysis
There were three major sources of analysis: verbal and nonverbal communication between the students during the corpus consultation; results of pre- and post-interviews; and the instructor’s observation notes during the tasks. The students’ activity during the tasks and the interviews were video recorded for transcription. The students’ original dialogues in the experiment were in their native language (Korean), which was intended to encourage active peer communication during corpus analysis. The dialogues were transcribed and romanized according to the Yale system at http://asaokitan.net/tools/hangul2yale/. In the excerpts given here, the romanized texts are provided within parentheses along with the English translation (see e.g., Hall, Reference Hall2004; He, Reference He2001; Ochs, Reference Ochs1996). The instructor kept notes throughout the corpus sessions to document observations and reflections on the students’ behaviors. The notes recorded peculiar gestures, positive and negative comments on the activities, the instructor’s impressions, and noticeable group differences during the corpus consultation. The notes were pre-coded mainly by the group and the task differences, which served as a significant reference for the analysis of transcripts of peer interactions and interviews.
In the first stage of data analysis, the transcripts of peer interactions and interviews as well as the instructor’s observation notes were manually coded to reveal distinctive effects of collaboration in two types of corpus consultation tasks. The coded items were categorized into either positive or negative effects, and a dominant effect of collaboration in each type of task was identified. Next, if significant group differences in the two types of tasks were found, the coded data was thoroughly reviewed to find major reasons for the task-dependent effects of collaboration. In an attempt to improve reliability of the study, the transcripts were independently reviewed by two researchers in teaching English as a second language (TESOL) to seek consensus when the data interpretation seemed ambiguous.
4 Findings and discussion
4.1 Conceptual corpus consultation task
The conceptual task asked the students to identify semantic differences among synonymous items through either collaborative or individual corpus consultation. The study found that collaboration had significantly positive effects in the conceptual corpus consultation task as the CG seemed to complete the task more efficiently than the IG. Collaboration allowed CG students to co-construct the data analysis process to reach comprehensive interpretations of data. In many cases, as illustrated in Excerpt 1, the more capable students in the group (S2 and S3) guided the corpus consultation process in the right direction, preventing the less capable one (S1) from making wrong generalizations.
Excerpt 1 Search of the term effect Note: The transcription conventions used in the excerpts here include square brackets [] to indicate simultaneous talk, and double parentheses (( )) to provide contextual information such as explanation of the situation and the participants' gestures. The English translation appears first with romanized Korean in italics below in each turn.
In Excerpt 1, S1 prematurely assumed that effect had a positive connotation (line 2) and attempted to gain consensus (line 3). In response, S2 and S3 jointly guarded against the assumption and suggested further investigation (lines 4, 5, 6). When S1 suggested the same idea again after analysis of the corpus data (line 8), S2 and S3 responded with silence (line 9) as an indication of indirect rejection of S1’s idea. In response to the subsequent request by S1 for agreement (line 10), S2 and S3 expressed their rejection more directly (lines 11, 12), which led to the more generalizable assumption that effect did not have a strong positive connotation and rather it was neutral (line 14). The role of peer interaction to guard against inappropriate data analysis seemed to be crucial for S1 because she might otherwise have believed that her wrong conclusion was correct, and she would not have realized her error without guidance from S2 and S3.
When compared to the IG, the benefits of collaboration become more obvious. During the post-interview as shown below, S3 acknowledged that the peer collaboration was indispensable for reaching a more comprehensive interpretation of corpus data. In highly complex and analytic tasks such as corpus analysis, not only the weaker member such as S1 in Excerpt 1 but also the most capable member such as S3 benefited from the negotiative dialogue through the collaborative learning process. On the other hand, IG students seemed to experience substantial difficulty in the conceptual corpus consultation task. S4 conceded that it would have been helpful if they had worked as a group, and S6 noted the importance of the assisting role of the instructor, hinting that the difficulty of the task was sometimes unmanageable alone.
∙ When analyzing corpus data alone, I came to have a biased view on certain expressions. However, when we analyzed the data together, the group members pointed out many things that I hadn’t noticed from the data. Based on our group discussion, I was able to expand my view to better examine the concordance lines. (S3)
∙ It was challenging for me to figure out the meanings and usage of expressions from the corpus data… If we had worked as a group, it might have been easier. (S4)
∙ I think the role of the instructor was crucial for our corpus consultation. I was able to find important examples when the instructor helped by telling me which concordance lines to pay attention to. (S6)
All in all, the results of the conceptual corpus consultation task by the CG and IG seem to suggest that collaboration had significant facilitative effects. Collaboration provided the students with opportunities to engage in negotiative dialogue, which guided the corpus consultation process so they could reach a more comprehensive conclusion. Psychologically, the collaborative nature of the task seemed to give the students a sense of security as they felt that they complemented each other’s capabilities and knowledge. The benefits of collaboration in the conceptual corpus consultation task were in line with the findings of prior studies (e.g., Flowerdew, Reference Flowerdew2008; Gavioli & Aston, Reference Gavioli and Aston2001; O’Sullivan, Reference O’Sullivan2007). On the other hand, the IG students experienced a degree of difficulty as they felt the need for collaboration and assistance from the instructor during the corpus consultation. It seems that the IG students’ frustration here echoes the findings of previous studies reporting various challenges in learners’ corpus consultation such as difficulty in interpreting and evaluating patterns from an enormous amount of data (e.g., Ädel, Reference Ädel2010; Conrad, Reference Conrad2005).
4.2 Procedural corpus consultation task
The procedural task required the students to tackle a sequence of problems in two major stages: corpus consultation to discern semantic differences among synonyms and L2 writing (i.e., translation as a standardized writing task) using the items chosen from the results of the corpus consultation.
Unlike the conceptual task, results of the procedural task revealed the double-sided effects, or a dark side, of collaboration, which answered the second research question of this study. In completing the procedural task, the CG students seemed to experience a more severe degree of difficulty than the IG students. Most of all, the CG students took twice as long to complete the procedural task as the conceptual task. Also, the CG’s procedural task took longer than the IG, who completed it much faster than the conceptual task. During the post interview, the CG conceded their frustration after the procedural task.
∙ If I had some time to seriously think (study) alone before the collaborative group work, the corpus analysis would have been easier. (S1)
∙ During the pre-interview, I said that the corpus would be the best way to learn English especially in L2 writing, but since I actually used corpus consultation for translation (in the procedural task), I now think the dictionary would be a better choice for L2 learning. (S2)
S1 noted the need for individual corpus consultation before the collaborative work, indicating that she felt great difficulty in the process of collaboration. More interestingly, S2, who showed enthusiasm for the use of corpora in L2 writing during the pre-interview, seemed to be discouraged by the difficulty of the procedural task. In contrast, the IG found the procedural task to be much easier than the conceptual task as shown in the following interview excerpts.
∙ I don’t know the reason why, but I felt today’s activity (the procedural task) was much easier than that of yesterday (the conceptual task). Yesterday, figuring out the semantic differences by myself was very difficult, but today I think it was easier because I had the sentences for translation that I could refer to in the corpus consultation. (S4)
∙ It was a little burdensome to have an additional task (i.e., translation) compared to the conceptual task. But today (the procedural task) was not as difficult as the conceptual task. (S5)
∙ Today, I felt that the sentences given for the translation were like hints for the corpus analysis. (S6)
The IG students (S4 and S6 in particular) specifically noted that the sentence for translation assisted the corpus consultation. This suggests that the IG students kept referring to the sentences for translation throughout the corpus consultation, which set a direction for the corpus analysis. In other words, the students’ constant reference to the subsequent stage in a procedural task could lessen the degree of difficulty of the task. As students engaged in the procedural task individually, they had procedural flexibility of task management, which offered a considerable advantage to the IG.
With an evident group difference in regard to the task type, this study identified two major reasons for the difficulty experienced by the CG in the procedural corpus consultation task: the role of intersubjectivity and significant power inequality between the members of CG. Most notably, unlike their IG counterparts, the CG students did not mention the benefits of the text for translation as a hint for corpus consultation. This indicates that they found it difficult to capitalize on the information given for the subsequent task (i.e., a text for translation) to efficiently complete the current task (i.e., targeted corpus consultation). This seemed to be partially attributable to the establishment of intersubjectivity in the task. Intersubjectivity refers to the shared subjective states and understanding among the participants in an activity (e.g., Kaye, Reference Kaye1982; Rogoff, Reference Rogoff1990; Rommetveit, Reference Rommetveit1979, Reference Rommetveit1985), which generates significant benefits in collaborative learning (e.g., Johnson & Johnson, Reference Johnson and Johnson1999, Reference Johnson and Johnson2002; Koschmann et al., Reference Koschmann, Stahl and Zemel2004; Koschmann et al., Reference Koschmann, Zemel, Conlee-Stevens, Young, Robbs and Barnhart2003; Roschelle, Reference Roschelle1996; Stahl, Reference Stahl2006). In the conceptual task, the established intersubjectivity must have had a favorable effect for the task completion, which assisted the students in negotiating the process of discussion to reach a more generalizable interpretation of the corpus data. For instance, in Excerpt 1 the intersubjectivity helped in discussion of the semantic connotation of effect, which maintained and guided the direction of the discussion.
However, in the procedural task, the strong intersubjectivity of the CG seemed to restrict the flexibility of the task management process. By going back and forth between the two sub-tasks (i.e., corpus consultation and translation), the IG seemed to utilize the information about the translation task to perform the targeted corpus consultation while the CG was forced to complete the tasks one at a time. This is because for the CG, when the intersubjectivity was established while discerning semantic differences between synonyms, it was difficult for the students to divert their attention to the next stage (translation) during the corpus consultation. In this case, figuring out the semantic difference between two synonymous words became the precondition for moving on to the translation task. This limited procedural flexibility in task management made it difficult for the CG students to benefit from the sentences given for translation as their IG counterparts did.
Another reason for the CG’s difficulty in the procedural task relates to the strong control of S3 (the most capable student) in the process of corpus consultation. With only one computer to share, the CG seemed to create power inequality between members. In fact, power inequality in collaborative learning is not a new phenomenon where the most capable member gains power and control while the less capable member is gradually marginalized from group work (e.g., Cheng, Reference Cheng2013; Leki, Reference Leki2001). The strong control that S3 exerted over the search process is well illustrated in Excerpt 2.
Excerpt 2 Search of the term hope
When the students investigated semantic differences between hope and wish (lines 1 and 2) during the procedural corpus consultation task, S2 asked S3 to change the sorting of the concordance to the right-hand side (line 4), and S3 responded to the request (line 5). Acknowledging the high degree of politeness in the request, S1 noted that she had asked S3 in the same way when she wanted to look up terms (line 6). This was an indication not only of the high degree of politeness in the exchanges, but more importantly of the mutual acknowledgement of S3’s strong authority in the corpus consultation process, which led S1 and S2 to ask S3 when they had inquiries to search for in the corpus data. The apparent dominance of S3 in the excerpt is supported by S2’s response during the post-interview that she had no opinion on the difficulty of performing search techniques in corpus consultation because she had almost no chance to access the keyboard or the mouse.
Such dominant control by S3 seemed to have stronger effects in the procedural task than in the conceptual one. The conceptual task in this study required the generalization of semantic connotations of given items from large numbers of concordance lines. Students from both groups appeared not to bother clicking each line but quickly skimmed through them to find a common and distinctive pattern for the search terms. On the other hand, the procedural task required more meticulous analysis of individual concordance lines to make appropriate word selections for translation. In this process, it is imperative for students to examine the co-text of the concordances by clicking them, but for S1 and S2 it seemed to be difficult as the corpus consultation process was governed by strong authority (S3). This is in contrast with IG students, who were able to click on every concordance line they wanted to examine and make lexical decisions for translation.
The investigation of group differences in the procedural task suggested that collaboration was at times a source of difficulty for students when they tried to accomplish problem-solving activities through corpus consultation. The disadvantage of collaboration seemed to be partly attributable to the established intersubjectivity and strong authority of S3 in the process of corpus consultation. In this study, the intersubjective state of corpus consultation allowed only limited procedural flexibility for the students to refer to a subsequent task. In addition, the strong control of S3 in the CG discouraged meticulous examination on individual concordances, which was more important in the procedural task than in the conceptual one. The frustration of the CG students was in stark contrast to the IG students, who seemed to benefit from the private mode of the activity, allowing them to navigate the corpus consultation as they wished.
5 Conclusion
This study presented a qualitative examination of task-dependent effects of collaboration in learners’ corpus consultation. Two groups of students (CG and IG) engaged in a conceptual and a procedural corpus consultation task, the process of which was closely analyzed. The findings of this study revealed significantly different effects of collaboration in two types of tasks. In the conceptual task, the CG seemed to benefit from the collaboration, which guided students through the process of corpus consultation, efficiently leading to a more generalizable conclusion. In contrast, the IG outperformed the CG in the procedural task, which seems to be attributable to the negative effects of collaboration. Analysis of the students’ dialogues and interviews revealed that established intersubjectivity in collaboration allowed less flexibility in task management, consuming a significant amount of time and effort. In addition, the collaborative process gradually created power inequality between members, discouraging personal queries for more meticulous investigation of concordances in the procedural task.
The most significant contribution of this study is its exploration of the double-sided effects of collaboration in learners’ corpus consultation. Given the enormous amount of evidence regarding the benefits of collaborative learning (e.g., Garrison et al., Reference Garrison, Anderson and Archer2001; Johnson & Johnson, Reference Johnson and Johnson1999; Kreijns et al., Reference Kreijns, Kirschner and Jochems2003; Lazonder, Reference Lazonder2005; Webb & Palincsar, Reference Webb and Palincsar1996), it is easy for instructors to assume that collaboration is a silver bullet for solving problems in learners’ corpus consultation. However, along with prior studies on negative aspects of collaboration (e.g., Berejkovskaya, Reference Berejkovskaya2006; Järvenoja & Järvelä, Reference Järvenoja and Järvelä2009; Kreijns et al., Reference Kreijns, Kirschner and Jochems2003; Levykh, Reference Levykh2008; Thompson & Fine, Reference Thompson and Fine1999), the findings of this study caution against researchers’ and instructors’ blind belief in the positive effects of collaboration. Instructors using collaboration as pedagogic mediation in learners’ corpus consultation should keep in mind the double-sided effects of collaboration and make strenuous efforts to exploit the best benefits of collaboration in corpus consultation.
One way to take advantage of collaboration is to pay more attention to different task types during corpus consultation. The findings of this study suggested that different task types produce different effects. In this sense, it is imperative for instructors to enhance their awareness on the task-dependent effects of collaboration and to be more careful in designing collaborative corpus consultation tasks. When a task is designed to improve learners’ procedural competence through corpus consultation, collaboration might not be appropriate pedagogic mediation. However, when a task requires the students to employ and develop their conceptual knowledge through corpus consultation, peer collaboration can assist their navigation through corpus data to complete the task more efficiently as shown in the conceptual task in this study and in that by Gavioli and Aston (Reference Gavioli and Aston2001).
In fact, the task-dependent effect of collaboration has been reported in several studies (e.g., Belenky, Ringenberg, Olsen, Aleven & Rummel, Reference Belenky, Ringenberg, Olsen, Aleven and Rummel2014; Mullins et al., Reference Mullins, Rummel and Spada2011). In particular, Mullins et al. (Reference Mullins, Rummel and Spada2011) noted that collaboration was helpful in a conceptual task while no such effect was observed in a procedural task. Even though they were concerned with mathematical knowledge and skill, the finding is supported in this study, in which IG students outperformed CG students in the procedural task, indicating significant task-dependent effects of collaboration. All in all, it can be concluded that the benefits of collaboration should not be taken for granted in corpus consultation while the benefits of individual work should likewise not go unnoticed, as it allows more flexibility and increased opportunities for students to pursue personal queries in the process of task management.
The findings of this study are, however, not without limitations. Most significantly, due to the experimental design of the case study over a short period, the findings here are not necessarily generalizable. Within a similar qualitative framework, a long-term experiment with multiple types of participants and tasks would provide more robust evidence for the task-dependent effects of collaboration in learners’ corpus consultation. In addition, the collaborative group in this study did not have enough time to build a sense of community before the experiment. Given the important role of trust and bonds in collaborative learning (e.g., Cockburn & Greenberg, Reference Cockburn and Greenberg1993; Gunawardena, Reference Gunawardena1995; Northrup, Reference Northrup2001; Rourke, Reference Rourke2000), future studies to examine the effects of collaboration in corpus consultation could provide students with sufficient opportunities to build a strong sense of community before the experiment. However, it should be noted that in reality, not all classes have enough time to build trust and relationships before they engage in collaborative work. It is possible that instructors rush to start collaborative group work when they have a tight schedule or are unaware of the importance of building trust between students. Thus, it is hoped that this study will trigger future research to investigate the role of trust and relationships between students in collaborative corpus analysis to deepen our knowledge on the effects of collaboration in learners’ corpus consultation.
Despite the limitations, the findings of this study should not be dismissed lightly. Although the experimental case study design has its limitation, it is an exploratory study to open a new path to examine the effects of collaboration in corpus consultation. While the findings may be less generalizable than in large-scale quantitative studies, this study holds value as it pioneers the investigation of task-dependent effects of collaboration in learners’ corpus consultation. In particular, it categorized the learners’ corpus consultation tasks into two types, which invited attention to the distinctive processes and results of different collaborative corpus consultation tasks. Based on the findings of this study, future studies should employ more extensive investigation into task-dependency in collaboration to suggest practical and research implications for collaborative corpus consultation. Also, it is hoped that future studies will develop various types of corpus consultation tasks and instructional strategies to make learners’ corpus analysis more effective and enjoyable in L2 learning.
