Online Learner Self-Regulation: Learning Presence Viewed through Quantitative Content- and Social Network Analysis

This paper presents an extension of an ongoing study of online learning framed within the community of inquiry (CoI) model (Garrison, Anderson, & Archer, 2001) in which we further examine a new construct labeled as learning presence. We use learning presence to refer to the iterative processes of forethought and planning, monitoring and adapting strategies for learning, and reflecting on results that successful students use to regulate their learning in online, interactive environments. To gain insight into these processes, we present results of a study using quantitative content analysis (QCA) and social network analysis (SNA) in a complementary fashion. First, we used QCA to identify the forms of learning presence reflected in students’ public (class discussions) and more private (learning journals) products of knowledge construction in online, interactive components of a graduate-level blended course. Next, we used SNA to assess how the forms of learning presence we identified through QCA correlated with the network positions students held within those interactional spaces (i.e., discussions and journals). We found that the students who demonstrated better selfand co-regulation (i.e., learning presence) took up more advantageous positions in their knowledgeOnline Learner Self-Regulation : Learning Presence Viewed through Quanitative Contentand Social Network Analysis Shea, Hayes, Uzuner Smith, Vickers, Bidjerano, Gozza-Cohen, Jian, Pickett, Wilde, and Tseng Vol 14 | No 3 July/13 428 generating groups. Our results extend and confirm both the CoI framework and previous investigations of online learning using SNA.


Introduction
As online learning continues to grow in higher education, it is critical that we gain a better understanding of the mechanisms by which we can promote its quality. The longstanding community of inquiry (CoI) model (Garrison, Anderson, & Archer, 2000) represents one such mechanism. This model describes the deliberate development of an online learning community, stressing the processes of instructional dialogue likely to lead to successful online learning. It explains formal online knowledge construction through the cultivation of various forms of presence: teaching, social, and cognitive presence ).
The CoI model theorizes online learning in higher education as a byproduct of collaborative work among active participants in learning communities characterized by instructional orchestration appropriate to the online environments (teaching presence) and a supportive, collegial online setting (social presence). The teaching presence construct outlines participant instructional responsibilities such as organization, design, discourse facilitation, and direct instruction (Anderson, Rourke, Garrison, & Archer, 2001) and articulates the specific behaviors likely to result in a productive community of inquiry (e.g., Swan & Shea, 2005). Social presence emphasizes online discourse that promotes positive affect, interaction, and cohesion (Rourke, Anderson, Garrison, & Archer, 1999) that supports a functional, collaborative learning environment. The model also refers to cognitive presence, a cyclical process of interaction intended to lead to significant learning within a community of learners.
More than 10 years of research and a recent two part edited special issue of The Internet and Higher Education (Swan & Ice, 2010), dedicated to CoI and the advances in our understanding of online learning gained through this theory, are testament to its usefulness. However, with more than 6.7 million college students enrolled in at least one credit bearing online course during 2012 and an accompanying growth rate of more than 9% (Allen & Seaman, 2013), it is clear that we will continue to need a comprehensive model that helps describe, explain, and predict how people learn online.
Recently, in an effort to make the CoI model more comprehensive, we (Shea & Bidjerano, 2010;Shea et al., 2012) suggested another dimension of presence in this model. In analyzing student contributions to online courses using the CoI model, we were unable to reliably identify instances of student generated discourse found in collaborative learning activities (such as online discussions and other areas used for Online Learner Self-Regulation : Learning Presence Viewed through Quanitative Content-and Social Network Analysis Shea, Hayes, Uzuner Smith, Vickers, Bidjerano, Gozza-Cohen, Jian, Pickett, Wilde, and Tseng Vol 14 | No 3 July/13 429 group work) using indicators of teaching, social, and cognitive presence (see Shea, Hayes, & Vickers, 2010). Upon further investigation, we considered these student contributions to be examples of online learner self-and co-regulation and applied the term learning presence to describe this interaction. In our most recent CoI research, we presented learning presence (discussed in more detail below) as a new construct that is meant to complement and expand upon teaching, social, and cognitive presences contained in the CoI model.

Learning Presence Defined
Our conceptualization of learning presence is informed by Zimmerman's (2008) wellresearched theoretical construct of self-regulated learning, which refers to "students' proactive use of specific processes [such as setting goals, selecting and deploying strategies, and self-monitoring one's effectiveness] to improve their academic achievement" (p. 167). Self-regulation research conducted in the last two decades has concluded that self-direction (including e.g., setting personal goals, using diverse modes of learning, time management) is predictive of better learning outcomes in classroombased education (e.g., Zimmerman, 2000;Zimmerman & Schunk, 2001). In a similar vein, reviewing studies that investigated online learning (e.g., Bixler, 2008;Chang, 2007;Chung, Chung, & Severance 1999;Cook, Dupras, Thompson, & Pankratz, 2005;Crippen & Earl, 2007;Nelson, 2007;Saito & Miwa, 2007;Shen, Lee, & Tsai, 2007; Wang, Wang, Wang, & Huang, 2006), Means and her colleagues (2009) also concluded that support for enhancing students' self-regulation (such as initiative, perseverance, and adaptive skill) has a positive impact on their online learning.
Our conceptual framing of learning presence reflects learner self-and co-regulatory processes in online educational environments. The coding scheme we developed to delineate this construct aligns with Zimmerman's concept of self-regulated learning and includes phases for forethought and planning, performance, and reflection, with emphasis on the goals and activities of online learners specifically. Under the forethought phase, we include planning, coordinating, and delegating or assigning online tasks to self and others in the early stages of the course, course module, or specific activity. In the performance phase, we include monitoring and strategy use.

Research Questions
Building on this expanded version of the CoI model, we hypothesized that for students who are asked to design and facilitate a portion of an online course (in this case, course discussions), this added responsibility might heighten their self-and co-regulatory behaviors, resulting in higher levels of learning presence. Further, when students collectively focus on knowledge construction in online discussions, they create a network, and the messages they post provide clues to the structure of that network and the relative positions that each student occupies within it. As a result, certain advantageous positions can emerge as indicators of relative prominence among participants (Aviv, Erlich, Ravid, & Geva, 2003;deLaat, Lally, Lipponen, & Simons, 2007a). With this understanding, our second hypothesis was that assigning facilitation roles to students might provide them with increased interaction with their peers, resulting in more prominent roles and network positions influencing the flow of information in the discussions. To test these two hypotheses, we sought to explore online learner self-and co-regulation (learning presence) reflected in quantitative content analysis of student discourse and advantageous positions reflected in social network analysis (descriptions of these methods of analysis are in the sections that follow). With these analyses, we sought to examine the effects of a scaffolded transfer of some instructional roles from the instructor to the learners in online discussions on the expression of learning presence and student location within the resulting network of interaction in those discussions. We theorized that elements of the learning presence construct may possibly be more or less evident in different components of the learning activities designed for the course. For example, we conjectured that we might find more instances of student reflection in activities designed to promote such reflection, such as learning journals. As such, the specific questions we asked were as follows: 1) When part of the instructional role is shared with students (elements of design and facilitation of discourse) to what extent is there an impact on the expression of self-and co-regulation (learning presence) as measured through quantitative content analysis of student discussion postings and learning journals?

Data
The data for this study consisted of students' learning journals and transcripts of their online discussions collected from a doctoral level research methods course that used blended instruction. The course, which was offered during the 2010 fall term at a large state university in the northeastern United States, met face-to-face for three weeks at the start of term then switched to fully online instruction for the remainder of the semester. There were 18 students enrolled in this blended course. The online components of the course consisted of eight modules, with each module lasting for about two weeks. We report on the results from two sets of three concurrent discussions from one of the modules (Module 6) and the learning journals for that module.
Overall, the discussions we analyzed had an aggregated count of 223 student postings, each of which served as our unit of analysis. In each set of discussions, one discussion was required and there were two others from which students could select to participate.
Student postings by discussion were as follows for Weeks 1 and 2 of Module 6: Week 1: In Module 6, there were also a total of 16 journal entries posted to a blog forum. These learning journals were a course requirement and they were available for members of the whole class to read. In their journal entries, students were simply asked to include their Online Learner Self-Regulation : Learning Presence Viewed through Quanitative Content-and Social Network Analysis Shea, Hayes, Uzuner Smith, Vickers, Bidjerano, Gozza-Cohen, Jian, Pickett, Wilde, and Tseng Vol 14 | No 3 July/13 432 comments, questions, insights, concerns, and other reactions to the content of the assigned readings. Although the journal entries were posted to the blog forum, they did not require continuous student interaction. Each student was expected to respond to only one or two other students' journal posts. There were a total of 19 comments made by students to the journals we analyzed from Module 6.
Scaffolding support for shared instructional roles.
Our hypothesis was that having students explicitly share the teaching presence role might foster additional expression of the kinds of self and co-regulatory actions reflected in the learning presence construct. To test this hypothesis, we turned to the online discussion component of the course where students took more responsibility for aspects of teaching presence, specifically the facilitation of the discussions on course topics that they selected.
The online discussions students engaged in (described above) were a requirement in the course and they were scheduled in each of the eight modules. At the beginning of the semester, students divided themselves into teams of two to three students. Each team agreed to be the discussion facilitators for one module of instruction covering one of the course topics. Working with the instructor, each team selected key readings and devised leading questions and activities to facilitate the discussions around these readings.
Following instructor guidelines, modeling, and suggestions, facilitators were expected to guide the class discussions, ask questions, raise issues, and state their agreements and disagreements with appropriate support and evidence from the literature.

Data Analysis
We employed two methods of inquiry to analyze the data: quantitative content analysis and social network analysis (hereafter referred to as QCA and SNA).
QCA includes the process of searching text for recurring trends to identify frequencies (Adler & Clark, 2011). We conducted QCA using a revised version of the original learning presence coding scheme that was developed for a prior study (Shea et al., 2012). At the start of this study, two researchers who developed the original coding scheme refined it to align it more closely with Zimmerman's (1998Zimmerman's ( , 2000 three phases of self-regulation: forethought, performance, and self-reflection. This was accomplished by adding several new indicators and a new reflection category and re-categorizing the existing monitoring and strategy-use sections to sub-categories under a more inclusive organizing principle for self-regulation (i.e., performance, see Appendix A). After the refinement of the coding scheme, additional coders were trained to identify and count every occurrence of a learner presence code in the discussion transcripts and learning journals. No instructor posts were coded because the learning presence construct is specific to students.
In studies that employ QCA, rigorous coding protocols are crucial to reliability. To establish reliability, we began our coding with a test sample of learning journals and Online Learner Self-Regulation : Learning Presence Viewed through Quanitative Content-and Social Network Analysis Shea, Hayes, Uzuner Smith, Vickers, Bidjerano, Gozza-Cohen, Jian, Pickett, Wilde, and Tseng Vol 14 | No 3 July/13 433 discussions from the course with the goal of identifying and negotiating our coding differences. Repeating the coding and negotiation processes with sample texts allowed us to establish an adequate level of inter-rater reliability (IRR), which we calculated using Holsti's coefficient of reliability (CR). This method looks at percent agreement using the following formula: 2M/(N1+N2) where M represents the total agreed-upon observations, N1 represents the number of total observations for coder 1, and N2 represents the total number of observations for coder 2 (Holsti, 1969;Krippendorf, 2004;Neurendorf, 2002). For exploratory research of this nature, an IRR of 0.70 is considered acceptable (Lombard, Snyder-Duch, & Bracken, 2002;Neurendorf, 2002).
Although Lombard et al. (2002)  students' overall network centrality (Freeman degree) by combining measures of indegree centrality, which are counts of inbound ties with other students, and out-degree centrality, which are counts of outbound ties. These same measures, when considered separately, are indicators of network prestige (in-degree centrality) and influence (outdegree centrality). In online discussions, prestige measures the number of incoming responses directed to a student's discussion post and represents the degree to which other students seek out that student for interaction (deLaat, et al., 2007a). Students with high prestige are notable because their thoughts and opinions may be considered more important than others in the class. In contrast, students with high influence are in contact with many other students, as evidenced by the large number of discussion posts that they initiate to others. Students with low influence post fewer messages and are not as actively engaged with building or sustaining relationships with other students.
We used all three measures (Freeman degree centrality, in-degree centrality [prestige], and out-degree centrality [influence]) to quantify students' interactions in three aggregated online discussions and the learning journal entries. We also developed network graphs to illustrate these relationships and to explore the relative measures of students' learning presence found in the discussions and learning journals. To this end, we used a new software tool called SNAPP (Social Networks Adapting Pedagogical Practice) (Dawson, 2008(Dawson, , 2010Dawson et al., 2010;Dawson, Bakharia, & Heathcote, 2010). SNAPP was used to capture student discussion posts from all of the discussions in Module 6. We aggregated these data into adjacency matrices that represented all student interactions across all module discussions, and then we created a separate attribute file containing learning presence frequency counts for each student found in each module's learning journals and discussion posts, as well as individual measures of prestige and influence calculated using UCINet software. Finally, we imported these files into the NetDraw software package to generate a series of network graphs which are analyzed in the Results section.

Results
Research question 1: When part of the online instructional role is shared with students (elements of design and facilitation of discourse) to what extent is there an impact on the expression of self-and co-regulation (learning presence) as measured through quantitative content analysis of discussion postings and learning journals?
When comparing mean learning presence in the combined averaged discussions and learning journals of the Module 6 student facilitators (02, 09, 13, and 19) and the rest of the class, we found that the facilitator group exceeded their peers with an average of 11.3 versus 8.8 learning presence occurrences across the two learning activities. Thus, the facilitators exhibited 31% more learning presence indicators than their non-facilitating peers (see Table 1). Note. Numbers 07 and 14 are not included in this and other tables because one was the instructor and the other was a guest Mann-Whitney U was performed to determine whether student facilitators and nonfacilitators differed with respect to levels of learning presence beyond statistical chance.
Median combined occurrences of learning presence were 12.50 and 8.5, respectively.
Although the student facilitators as a group had a higher average rank (Mrank = 7.0) than the student non-facilitators (Mrank = 10.21), the differences in the distribution of learning presence within the two groups were not statistically significant (Mann-Whitney U = 18.00, n1 = 4, n2 = 14, p =.286 two-tailed).

Research question 2:
What impact does the shared instructional role (learner facilitation of online discussions) have on metrics reflected in social network analysis?
Do facilitators occupy more advantageous locations in the social network?
When we examined student interactions using a network graph (see Figure 1) to visualize the ties that emerged between students as a result of their postings in all of the Online Learner Self-Regulation : Learning Presence Viewed through Quanitative Content-and Social Network Analysis Shea, Hayes, Uzuner Smith, Vickers, Bidjerano, Gozza-Cohen, Jian, Pickett, Wilde, and Tseng Vol 14 | No 3 July/13 436 discussions we analyzed, we found the following students were most centrally positioned in the network: 17, 13, and 09. Two members of this group were student facilitators (students 13 and 09). These three students were most active in initiating posts and responding to other students, as evidenced by the number of ties that connected them to their peers. In contrast, student facilitator 19 was somewhat more central, and student 02 was located on the edge of the network, because he had fewer peer relationships. Overall, the student facilitators demonstrated more prominent network positions for prestige (in-degree centrality) and influence (out-degree centrality) than the rest of the class when these two measures were aggregated and averaged across the group (see Table 2). In terms of prestige, the facilitators had a median of 12.0 incoming ties versus 8.0 for the rest of the class. The median of outbound ties (influence) for the facilitator group was 12.0 versus 9.0 for their peers. In both cases, the facilitators had higher measures than non-facilitators.
Results from Mann-Whitney U, testing differences in prestige and influence between student facilitators and non-facilitators, indicated that although the student facilitators had higher medians of in-bound and out-bound messages than their counterparts, statistically significant differences in the metrics for influence (Mann-Whitney U = 17.00, n1 = 4, n2 = 14, p =.24 two-tailed) and prestige (Mann-Whitney U = 19.00, n1 = 4, n2 = 14, p =.337 two-tailed) were not found.  In comparing the distribution of the three learning presence categories, forethought and planning, performance, and reflection, in the two sets of learning activities in Module 6 (discussions and journals), the monitoring construct was most frequently reported in both discussions (58.4%) and learning journals (51.6%) (see Figure 2). From here Online Learner Self-Regulation : Learning Presence Viewed through Quanitative Content-and Social Network Analysis Shea, Hayes, Uzuner Smith, Vickers, Bidjerano, Gozza-Cohen, Jian, Pickett, Wilde, and Tseng Vol 14 | No 3 July/13 438 patterns diverged. The six discussions accounted for 32.1% of strategy use, with no evidence of forethought and planning, and low levels of reflection (9.5%). In contrast, student learning journals demonstrated more evidence of reflection (22.6%) which occurred more frequently than strategy use (19.4%) and forethought and planning (6.5%). This provides evidence that the categories reflect the intended constructs; one would expect to see more reflection in activities such as learning journals in which students are asked to think about their learning.   To further analyze the effect of learning presence on online activity, a median split was used to identify students with high and low levels of combined learning presence from both discussions and journals (see Table 3). The newly created variable served as grouping to examine differences in centrality, prestige, and influence. As mentioned earlier, we calculated Freeman degree centrality by combining measures of in-degree centrality, which are counts of inbound ties with other students, and out-degree centrality, which are counts of outbound ties. These same measures, when considered individually, are indicators of network prestige (in-degree centrality) and influence (outdegree centrality) (see Table 4). With students' ranks as a dependent measure, learning presence levels (high vs. low) had an effect on the overall centrality of student positions on the network (Mann-Whitney U = 6.50, n1 = 8, n2 = 10, p =.003 two-tailed).    With students' ranks in terms of influence as a dependent measure, the results indicated that students with high learning presence ranked higher on influence (Mann-Whitney U = 10.50, n1 = 8, n2 = 10, p =.008 two-tailed) (see Figure 4). A somewhat similar pattern of network positions found in Figure 3 appears in Figure 4, with a core group comprised of students 05, 09, 13, 17, and 18, all ranking among the highest in both graphs for centrality and influence. The results from independent samples test with prestige ranks as a criterion showed no differences in students' ranks of prestige depending upon high and low levels of LP (Mann-Whitney U = 19.50, n1 = 10, n2 = 8, When we examined combined learning presence found in discussions and learning journals, results from correlation analysis indicated that, as a whole, this measure has a positive and moderate correlation with prestige (Spearman rho (18)  presence did not differ between students with high and low influence in the network, Mann-Whitney U = 20.00, n1 = 4, n2 = 14, p = .385, two-tailed. Again, this suggests that certain students, perhaps those who are less active in public forums do, nonetheless, exhibit elements of learning presence in more private forums, and that asking them to facilitate a module may result in higher expressions of learning presence.

Discussion
With regard to results for our first research question, we found patterns that were suggestive, yet not statistically significant. While student facilitators expressed more evidence of learning presence than their peers, these patterns within a single module were not significant. It seems possible that with a larger sample size, more definitive conclusions could be reached and further research is warranted. In response to our second research question, regarding the occurrence of learning presence among facilitators, we found similarly suggestive patterns of centrality. However, although facilitators occupied more central locations within the network, associated metrics were not significantly different. When we consider our third research question, it is not Online Learner Self-Regulation : Learning Presence Viewed through Quanitative Content-and Social Network Analysis Shea, Hayes, Uzuner Smith, Vickers, Bidjerano, Gozza-Cohen, Jian, Pickett, Wilde, and Tseng Vol 14 | No 3 July/13 444 surprising that students engaged in more reflection in the learning journals than in the discussions. The journals asked students to reflect on their learning processes and they did so. It is somewhat illuminating that students engaged in more learning presence overall in the discussions and that the most frequent form of self-regulation in both journals and discussions was monitoring. Lastly, results for our last research question indicated that metrics of self-regulation evidenced in QCA appear to identify students who are both influential and prestigious as measured by SNA. It seems probable that the capacity to self-regulate in online environments results in more relevant or more sophisticated discourse, making students with better learning presence more attractive interlocutors for their classmates.

Scholarly Significance of the Study
As noted by previous researchers (e.g., deLaat, Lally, Lipponen, & Simons, 2007b) the combination of QCA and SNA may allow for a compatible research approach illuminating some of the qualities of both form and content of interactions in online learning environments. Through the combination of these kinds of analysis, we are able to uncover important patterns bearing on the effects of approaches to new online pedagogy generated from the CoI framework. We have also extended the use of SNA in analyzing a new construct (learning presence) within the CoI framework.
Facilitating learner self-regulation has proven to have advantageous outcomes in much research in classrooms (e.g., Zimmerman, 2000) and in emergent research in online environments (Means et al., 2009). In past research, it has been suggested that providing students with more complex collaborative tasks results in higher levels of self and co-regulatory performance (Shea et al., 2012). This study sought to extend previous findings by implementing learner centered forms of instruction in which we analyzed levels of learning presence of student facilitators and non-facilitators in online discussions and journals through QCA and SNA.
Specifically, in this paper, we analyzed a new element in the CoI model reflecting online learner co-and self-regulatory processes -learning presence. We examined the impact of providing a scaffolded shift in instructional roles in which learners were supported to take on more of the responsibility for design and facilitation of discourse (elements of teaching presence) and observed the resulting variation in associated indicators of selfand co-regulatory performance (learning presence) reflected through QCA of different learning activities. Through research questions 1, 2, and 4 we discovered that lead student facilitators exhibit higher levels of learning presence and occupy more advantageous locations reflected in SNA.
Through the results reflected in our third research question, we disclosed significant and illuminating patterns in categories of learning presence in different learning activities. Perhaps not surprisingly, forethought and planning are not very evident in Online Learner Self-Regulation : Learning Presence Viewed through Quanitative Content-and Social Network Analysis Shea, Hayes, Uzuner Smith, Vickers, Bidjerano, Gozza-Cohen, Jian, Pickett, Wilde, and Tseng Vol 14 | No 3 July/13 445 either online discussions or learning journals where strategy use and reflection are more common. That learners are exhibiting forms of strategy use more during performance (online discussion) and greater monitoring and reflection in journal activities validates the intended categories within the learning presence construct. We would expect to see these patterns, that is, more reflection and monitoring in journals and greater strategy use during performance, and we found them.
Research question 5 is significant in that results suggest that students with high discussion learning presence also have high in-degree centrality, indicating that other students sense that they are valuable partners for interaction and the knowledge building meant to result from it. These results suggest that higher levels of learning presence in online discussions are reflected in important metrics associated with SNA.
Also of note is the finding that learning presence dimensions that are evident in certain activities (learning journals) are not automatically associated with metrics important in SNA.
Overall, these findings are significant in that they support and extend previous research seeking to enhance one of the dominant theories (the CoI framework) that describes, explains, and predicts learning in online environments. Results here represent important support for the validity of learning presence as a complementary construct to this framework. Findings indicating that learning presence can be fostered through shared instructional roles and that this form of self-and co-regulatory performance is associated with advantageous locations in social networks suggest that the construct is useful. We conclude that the long standing belief that online learners require greater self-direction, time management, and the like is supported and better explained through the more inclusive theoretical construct of self-regulated learning and the related construct of online learning presence. We further conclude that the online environment creates demands for new forms of self-regulation that are under articulated in the current CoI model. We believe that the model can be enhanced through additional research into the specific roles of learners qua learners in collaborative online education.
This paper contributes to the literature on constructivist online learning and on SNA.
Specifically, the paper contributes to SNA by adding analysis of a new theoretical construct, learning presence, to it. A weakness of SNA in online educational research has been its lack of a relevant theoretical framing for metrics of centrality. We don't know, for example, based on the numbers of ties between participants in online learning contexts, whether such connections reflect the quality of the discourse or other processes important to learning. We assume that through interaction, learners increase their opportunity to activate processes known to support knowledge construction. For example, in line with constructivist theories of online learning, Chi (2009)  Specifically, these results indicate that students with higher levels of learner presence occupy more advantageous positions, indicating that they are more active and more sought after in networks of interaction. This represents a promising conclusion and additional research into the relationship between learning presence and interaction is warranted.
Finally, we believe that this research continues to provide evidence for the validity of the learning presence construct. Learning presence patterns revealed in this study indicate that student self-regulation as defined here is both logical (the learning presence patterns make sense) and important (learning presence correlates with metrics assumed to be advantageous for interaction). We, therefore, suggest that the inclusion of learning presence in the CoI model may be warranted.