Student Engagement in the Facebook Learning Environment: A Person-Centred Study

Abstract

Facebook has been used not only as a popular social network service among college students but also as a platform for promoting learning and teaching effectiveness in different subject areas. While previous studies have demonstrated the utility of Facebook for enhancing student engagement, little is known about whether there are different profiles of students with different engagement patterns in the Facebook learning environment. Adopting a person-centred approach, we aim to fill this gap in the current study by identifying unobserved sub-populations of students with respect to their engagement patterns in the Facebook as a supplemental learning platform. Latent profile analysis revealed three engagement profiles, a minimally engaged, a moderately engaged, and a highly engaged profile. Results also suggested that academic disciplines and teacher involvement in the Facebook learning activities were significant predictors of student membership in the three engagement profiles. Our findings offer implications for the design and delivery of Facebook learning activities that cater to different groups of learners with different learning needs.

Keywords

Facebook learning platform student engagement latent profile analysis profile membership

Student disengagement and dropout have been critical concerns among higher education institutions, particularly those that adopt a mass lecture mode of knowledge delivery (Dyson et al., 2015; Masserini & Bini, 2020). This mode of delivery is usually characterized by a teacher managing a class of hundreds of students, thus minimizing student-teacher and student-student interactions, reducing students’ concentration, and causing their engagement in the classroom to suffer (Bowman & Akcaoglu, 2014). This concern has become more common in countries where admission to university has shifted from elitism to mass opportunities (Masserini & Bini, 2020). In those contexts, lack of selective admission criteria has resulted in increased enrolments of students with diverse educational needs and varying levels of preparation and motivation, which requires educators to use effective strategies to engage them in the learning process.

In recent years, there has been an increasing interest by both educators and researchers in the use of social media to enhance student engagement (Sheeran & Cummings, 2018). Of the established social network services, Facebook – with almost 2.9 billion monthly active users – is the most popular among university students who use it to maintain social interaction beyond the classroom (Awidi et al., 2019). This pervasive use of Facebook coupled with its capability to afford discussion, communication, and collaboration among students and teachers (Thai et al., 2019) has made it a promising supplemental learning platform to promote student engagement. Studies have found that Facebook can augment different dimensions of student engagement through buttressing their relationship with peers and teachers and their sense of belonging (Sheeran & Cummings, 2018), enhancing their awareness of online presence of fellow students (L. Deng & Tavares, 2013), giving them a greater sense of social connectedness (Thai et al., 2019), satisfying their need for autonomy and reducing boredom (Datu et al., 2018), and increasing mutual trust and sense of teaching presence (Hong & Gardner, 2019).

Although these studies provide rich information on the mechanisms through which Facebook affords student engagement, they mainly adopt a variable-centred approach. That means student engagement is assessed with the assumption that the study participants are homogeneous, being drawn from a single population for which a single set of parameters can be estimated for all variables (Gillet et al., 2019; Wang & Peck, 2013). This might not be the case in the Facebook learning environment because students might have different attitudes towards the use of Facebook for educational purposes and hence engage in Facebook learning to differing degrees of enthusiasm, energy, and affect. As such, variable-centred studies preclude the possibility of detecting multiple sub-populations characterized by distinct patterns of engagement, thus impeding our effort to design individualized intervention programs to support student engagement (Wang & Peck, 2013). Person-centred studies, on the other hand, have potential to identify subgroups of students in the population that show distinct patterns of multidimensional engagement in the Facebook learning environment. These profiles of student engagement then can inform the development of appropriate instructional programs aimed at supporting students with different engagement needs. In addition, there seems to be inconsistencies in the way student engagement in Facebook learning was measured in the previous studies. The majority of those studies measured engagement as an overarching single construct (Dyson et al., 2015; Imlawi et al., 2015; Northey et al., 2015; Thai et al., 2019) or considered only one specific component of emotional engagement (Molinillo et al., 2018). This operationalization seems to be at odds with general consensus in the engagement literature that student engagement is a multidimensional construct consisting of distinct but interrelated sub-components (Fredricks & McColskey, 2012; Kahu, 2013). Failure to account for the multidimensional nature of this construct limits our understanding of the dynamic interaction among different engagement components within individuals and makes it difficult to discern sub-groups of students with unique engagement configurations (Miller et al., 2021).

We aim to address these gaps in the current study by drawing on the self-system motivational perspective (Skinner et al., 2009) to define student engagement and taking a person-centred approach to identifying latent subgroups of students with different engagement profiles in the Facebook learning environment. The self-system motivational perspective was chosen as the overarching theoretical framework in this study because it conceptualizes engagement as a multidimensional construct and considers engagement in an ecological relationship between the context, the self, and relevant learning outcomes. A positive learning context (e.g., the Facebook learning environment) fosters positive self-perceptions which in turn positively affect student engagement and vice versa (Skinner et al., 2009). Therefore, in the following sections, we discuss the definition and measurement of the multidimensional student engagement construct, review previous studies that take a person-centred approach to the examination of student engagement, and describe the analytical approaches employed in the current study.

Student Engagement: Definition and Measurement

Definition of Student Engagement

The research field of student engagement is characterized by considerable variation in the ways the construct is defined and measured (Appleton et al., 2008; Fredricks et al., 2004). For example, Kahu (2013) summarized four different perspectives about student engagement including the behavioral, psychological, socio-cultural, and holistic perspectives while Reschly and Christenson (2012) believed that the diverse perspectives about student engagement depended on the background of those who defined it. In the current study, we use the self-system model of motivational development to inform our conceptualization of student engagement (Connell & Welborn, 1991; Skinner et al., 2009). Within this model, engagement is defined as a student’s energized, directed and sustained action situated in the dynamic interaction between the student, the learning activities, and the environment (Skinner & Wellborn, 1994; Wang & Peck, 2013). Accordingly, student engagement is maintained and promoted to the extent that the learning environment makes students feel competent, autonomous, and related to others (Miller et al., 2021; Skinner et al., 2009; Wang & Peck, 2013). Therefore, a key proposition of the self-system motivational perspective is that student engagement does not exist in and of itself, but rather it is a malleable concept which is responsive to contextual changes (Fredricks et al., 2004; Kahu, 2013; Miller et al., 2021). This person-in-context perspective renders domain specificity an integral part of student engagement (Miller et al., 2021), warranting further exploration of this construct in different learning environments of which the Facebook as a supplemental learning platform is an example.

Despite variations in its definition, student engagement is generally conceptualized as a multidimensional construct encompassing behavioral, cognitive, affective, and social dimensions (Fredricks et al., 2004; Fredricks & McColskey, 2012; Wang et al., 2016).

Behavioral engagement refers to students’ observable behaviors such as effort, persistence, concentration, attention, and contribution to class discussions as they participate in academic activities (J. D. Finn et al., 1995; Fredricks et al., 2004). Students who show positive behavioral engagement are likely to achieve academic success and higher rate of retention (Appleton et al., 2006; Wang, 2009).

Cognitive engagement refers to students’ psychological investment in learning, a desire to go beyond what is required, and self-regulated strategies to understand and master knowledge (Fredricks et al., 2004; Skinner et al., 2009; Wang & Peck, 2013). A high level of cognitive engagement is likely to lead to academic success and more effective use of self-regulated strategies to understand complex ideas (Miller & Byrnes, 2001; Wang & Peck, 2013).

Affective engagement is concerned with students’ positive emotional reactions to learning in the classroom, including their interest, enjoyment, and valuing of learning activities (Fredricks et al., 2004; Skinner & Belmont, 1993). Positive educational outcomes associated with affective engagement include academic success, lower emotional distress, and depressive symptoms (Li & Lerner, 2013; Wang & Dishion, 2012).

Social engagement is defined as students’ willingness to establish and maintain social interaction with peers and teachers during learning as well as the quality of that interaction (R. Deng et al., 2020; Wang et al., 2016). Although social engagement is a relatively new component recently proposed in addition to the established tripartite model of behavioral, cognitive, and affective engagement, we consider it an important component in the context of Facebook learning given that Facebook is a social platform originally developed for social interaction purposes and that it incorporates features that facilitate student-student and student-teacher interactions.

Measurement of Student Engagement

Just as the definition of student engagement is fraught with complexity and variation (Fredricks et al., 2019), the measurement of the construct is also characterized by a variety of different methods, each with its own weaknesses and strengths. Fredricks and McColskey (2012) and Fredricks et al. (2019) reviewed multiple methods for measuring student engagement and concluded that self-report survey was the most popular method due to its practicality, ease of administration, and its capability to assess the unobservable components of cognitive and affective engagement. That said, the use of self-report surveys to measure student engagement also has limitations that hamper research and understanding of the construct (see Fredricks et al., 2019, for a review of these limitations). One issue that has been consistently reported in previous research and construed as a corollary of the definitional confusion of student engagement is the conceptual ambiguity among the multidimensional engagement components (Fredricks et al., 2004; Kahu, 2013). For example, “effort” is included in the conceptualization of both behavioral and cognitive engagement (Fredricks et al., 2004; Kahu, 2013), making it difficult for respondents to make judgment about this aspect of engagement. In some studies, social engagement is subsumed under behavioral engagement (R. Deng et al., 2020), or overlaps with affective engagement in terms of student’s reactions to peer and teacher relationships (Fredricks & McColskey, 2012). The borderline between behavioral and cognitive engagement sometimes becomes blurred as a result of the mismatch between the theoretical conceptions of the two components and the associated questionnaire items (R. Deng et al., 2020; Fredricks et al., 2019; Kong, 2011).

Issues associated with the operationalization of student engagement components, as discussed above, may take a toll on the subsequent analysis of questionnaire data. For example, items crafted to measure one engagement component may simultaneously cross-load on the other components due to their conceptual overlap. Failure to control for item cross-loadings potentially inflates correlations among the components (Asparouhov et al., 2015), leading to biased estimation of the relationships among variables. The conventional confirmatory factor analysis approach that has been employed in previous studies on student engagement (see, for example, Bergdahl et al., 2020; R. Deng et al., 2020) does not account for item cross-loadings because it is based on the restrictive independent cluster assumption (Morin et al., 2016). According to this assumption, each item is forced to load only on the factor it is purported to measure while all item cross-loadings are constrained to be exactly zero. Therefore, in this study, we resort to the exploratory structural equation modelling approach (ESEM, Asparouhov & Muthén, 2009) for the treatment of the psychometric properties of the student engagement questionnaire. This approach builds on the conventional confirmatory factor analysis to take into account item cross-loadings, thereby offering a way to overcome the conceptual ambiguity issues associated with student engagement.

Person-Centred Studies on Student Engagement

Since there is a lack of person-centred studies on student engagement in the Facebook learning environment, we draw on relevant studies in school-based contexts to inform our study. Contrary to the predominant number of variable-centred studies on student engagement, studies that adopt a person-centred approach are few and far between. While variable-centred studies offer useful insights into the extent to which students engage in academic activities and how students’ levels of engagement, on average, are associated with different antecedent and outcome variables, they do not focus on how students vary in their multidimensional engagement profiles and what each unique profile informs us about individual developmental processes (Bergman & Andersson, 2010; Wang & Peck, 2013). By identifying distinct sub-groups of students within a given population that display varying levels and patterns of engagement, person-centred studies enable a richer understanding of the various pathways through which targeted intervention programs and guidelines for practitioners can be devised to manage engagement deficiencies at individual levels (Miller et al., 2021; Pöysä et al., 2020).

Converging evidence from person-centred studies thus far suggests that there indeed exist different sub-populations of students with different engagement profiles, though the number of profiles and their unique configurations of engagement dimensions vary from one study to another. For example, van Rooij et al. (2017); Wang and Peck (2013); Schmidt et al. (2018), and Watt et al. (2017) all discerned as many as three to five profiles in which a profile with uniformly high and a profile with uniformly low engagement across all dimensions consistently emerged. Furthermore, these studies found additional profiles with dominantly high or low level of one engagement component but comparatively moderate level of engagement across other components. For example, Wang and Peck (2013) identified an emotionally disengaged and a cognitively disengaged profile indicated respectively by low level of emotional and cognitive engagement and moderate level across the remaining dimensions. In a similar vein, Watt et al. (2017) discovered a compliant profile of math engagement in which students shown high level of behavioral engagement but low level of emotional and cognitive engagement. This led them to speculate that students in this profile followed requirements in learning mathematics but with minimal curiosity and interest. On the other hand, Bae and DeBusk-Lane (2019) and Bae et al. (2020) found a behaviorally disengaged profile among secondary and elementary students respectively. Students in this profile shown a strong tendency to behaviorally withdraw from academic activities while still maintaining a moderate level of emotional, cognitive, and global engagement.

In addition to identifying sub-populations with different engagement profiles, previous person-centred studies also examined contextual and demographic factors that predict profile membership. For example, gender has been found to be associated with math engagement profile membership with females having higher likelihood of being in the highly engaged profile (Miller et al., 2021) while males being predictive of disengaged profiles (Watt et al., 2017). On the other hand, Bae and DeBusk-Lane (2019) and Bae et al. (2020) found no significant effect of gender as a predictor of engagement profile membership among secondary and elementary students respectively. It would be desirable to examine if gender predicts engagement profile membership in the Facebook learning environment given findings in a previous variable-centred study that showed no differences among males and females in terms of their engagement in Facebook learning (Datu et al., 2018).

Grade is also a potential predictor of profile membership. Bae and DeBusk-Lane (2019) and Bae et al. (2020) reported that students in higher grades were more likely to be classified into less engaged profiles, supporting earlier observations that declines in student engagement emerged as students advanced through their late elementary school and into secondary education (Blumenfeld et al., 2005; Luo et al., 2009). A similar trend was also observed at tertiary education level by Korhonen et al. (2017) who found that the intensity of engagement decreased as students transitioned from first year into their second year at university. Since fostering engagement during first year at university is key to establishing foundations for successful later year study (Krause & Coates, 2008), we consider it important to examine if students in their first year and later years at university have different probability of being classified into different engagement profiles in this study.

Aside from the two individual factors of age and grade, we also examine the contextual factors of academic disciplines and teacher involvement in the Facebook group activities. The former concerns whether students in different academic disciplines differ in their likelihood of being in different engagement profiles. Korhonen et al. (2017) found that students in different academic disciplines showed different levels of engagement because different disciplines attracted different types of students who preferred different learning styles. Teacher involvement in Facebook-based discussion groups might also predict profile membership. Studies found that student engagement improved as they participated in Facebook-based discussion groups established, monitored, and facilitated by teachers (Cunha et al., 2016; Imlawi et al., 2015; Sheeran & Cummings, 2018). Improved interaction and better relationship with teachers as well as higher instructor credibility as perceived by students were identified in these studies as potential factors that promoted student engagement. None of these studies, however, adopted a person-centred approach – a research gap that we seek to fill in the current study.

The studies reviewed above suggest considerable variations in the number of profiles extracted, the configuration of different engagement dimensions within each profile, and the various factors that predict profile membership, depending on the learning contexts and subject areas. This is in line with the self-system motivational perspective that situates student engagement within an ecological relationship between the learners, the learning tasks, and the learning environments. In fact, a common theme that constantly emerges from the literature is that student engagement is a malleable concept varying in intensity and subject to variation in environments (Connell, 1990; Fredricks et al., 2004; Kahu, 2013), and that different dimensions of engagement interact dynamically within individuals (Skinner et al., 2009). Therefore, a person-centred approach to examining the qualitative and quantitative variation in engagement profiles enables a more nuanced understanding of the sophisticated relationships between individuals and contextual factors (Lawson & Lawson, 2013; Watt et al., 2017).

Materials and Methods

Instrument

Student Engagement

Student engagement in the Facebook as a supplemental learning platform was measured by a questionnaire adapted from previous studies (Ben-Eliyahu et al., 2018; R. Deng et al., 2020; Wang et al., 2016). The questionnaire is composed of a demographic information section and 19 items measuring four dimensions of behavioral, cognitive, affective and social engagement along a five-category Likert-type scale. We adjusted the wording of the questionnaire items to make them more suitable for the context of Facebook learning. The final set of items was piloted with an intact class of 49 students and yielded a satisfactory internal consistency measure (Cronbach’s Alpha = .941).

Behavioral engagement was measured by four items about students’ active participation and positive behaviors in Facebook learning (a sample item is “I take notes when I participate in discussions on Facebook). Cognitive engagement was captured by five items about students’ cognitive investment and self-regulated strategies (a sample item is “I try to connect what I am learning on Facebook with what I learn in class). Affective engagement was measured by four items that elicit students’ emotional responses to Facebook learning activities (a sample item is I enjoy learning activities on Facebook). Finally, six items were used to gauge students’ social engagement by asking them about their effort and willingness to establish and maintain interactions with peers and teachers during the learning process on Facebook (a sample item is “I build on others’ ideas during discussions on Facebook).

Teachers’ Involvement in the Facebook Group Activities

Teachers’ involvement in the Facebook group activities was measured by a 4-item questionnaire adapted from A. N. Finn and Schrodt (2016). The questionnaire items mainly assessed students’ perception of the extent to which their teachers facilitated and provided guidance in the Facebook group discussions and activities (i.e., My teacher encourages students to challenge other students’ points of view during the Facebook group discussions). Confirmatory factor analysis conducted to examine the measurement properties of the teachers’ involvement scale yielded acceptable fit indices (χ² = 1.619, p = .445; normed χ² = .809; TLI = 1; CFI = 1; RMSEA = .000; SRMR = .006). All factor loadings were significant and strong ranging from .80 to .87. Mean scores of the perceived teachers’ involvement scale were then used as input data for the subsequent multinomial logistic regression analysis.

Participants and Procedure

The study was conducted at a large university in southern Vietnam that offers a wide range of undergraduate programs across the social sciences and humanities disciplines. Data were collected from the English as a foreign language communication courses offered to students across the university, using the convenience sampling method. Language learners were chosen as participants in this study because engagement is an emerging concept in language education and there is a call for further research on engagement among this particular student population (Hiver et al., 2020). Although the courses covered all four core skills of speaking, listening, writing, and reading, the learning materials, instructional activities, and the number of classroom contact hours varied with respect to students’ academic disciplines. More specifically, while English major students spent an average of 14.5 hours per week practicing the four skills in the classroom, non-English major students only had approximately 3.9 hours per week of direct instruction in the classroom. In addition to the regular learning activities offered in class, the Facebook groups were set up by teachers primarily as a supplemental online learning space where students discussed topics related to the course either prior to or after classroom hours, shared extra learning resources that were not covered in class, got feedback on their works from peers and teachers, and asked for advice on their learning problems. A scenario for the Facebook group activities is as follows: prior to each English class, the instructor posted a topic related to what would be discussed in class and encouraged students to share their ideas and knowledge or to search for and share information about the topic through prescribed learning materials and internet resources. After the class, students were encouraged to share extra learning resources and engage in advanced discussions beyond what had been covered in class. As such, the Facebook groups represented a supplemental virtual learning space augmenting students’ learning in the classroom rather than a stand-alone learning platform employed during the regular classroom hours. The instructor, therefore, served as a facilitator of the Facebook group activities overseeing students’ discussions and providing advice and feedback where necessary. The Facebook groups were set up prior to the start of the academic semester and invitations to join the groups were restricted only to those who were enrolled in the course. Students’ participation in the Facebook groups was voluntary and not counted towards their final grade. At the end of the course, the student engagement questionnaire and consent information documents were uploaded to the Facebook groups. Students were informed about the voluntary nature of their participation, the anonymity of their responses, and their right to withdraw from the study at any time before they submitted the questionnaire. They then indicated their consent for participation by completing and submitting the questionnaire.

Questionnaire responses were obtained from 407 students. The sample consisted of 46.4% first-year students (N = 189), 34.6% second-year students (N = 141), 9.6% third-year students (N = 39), and 9.3% last-year students (N = 38). They were aged between 18 and 27 (M = 19.67, SD = 1.36); 39.6% (N = 161) were male and 60.4% (N = 246) were female. In terms of academic disciplines, 48.2% (N = 196) were English major students who pursued a bachelor’s degree in English language teaching and English translation and 51.8% (N = 211) were non-English major students who learn English as a compulsory subject across different academic disciplines such as finance and banking, economics, social work and education, and international relations. As reported by the respondents, they spent approximately 1.3 hours per day participating in the learning activities in the Facebook groups.

Data Analysis

Data were analyzed in three phases. First, we examined the multidimensional structure of student engagement via the exploratory structural equation modelling approach (ESEM, Asparouhov & Muthén, 2009; Morin et al., 2016 ). This approach was proposed as an alternative to the conventional confirmatory factor analysis (hereafter, ICM-CFA) that precludes the potential effect of item cross-loadings due to its restrictive independent cluster assumption. Figure 1 offers graphical representations of the ICM-CFA and ESEM models.

Figure 1.

Graphical Representations of the ICM-CFA and ESEM Models. A: The ICM-CFA model. B: The ESEM model.

The ICM-CFA depicts a correlated four-factor model in which items are specified to load only on the engagement component they are designed to measure while all non-target loadings are constrained to be zero by default. The ESEM also represents a four-factor model but differs from the former in that each item is specified to load simultaneously on all engagement components, with target loadings being freely estimated and non-target loadings constrained to be as close to zero as possible.

We used the Mplus software version 7 (Muthen & Muthen, 2012) with the weighted least square mean and variance adjusted estimator for the analyses of the measurement models. The oblique target rotation method was specified for the ESEM solution while all item cross-loadings in the ICM-CFA solution were constrained to be zero by default (Morin et al., 2016). Comparison of the two alternative models was based on established global model fit indices, parameter estimates, and theoretical conformity. Commonly used goodness-of-fit indices suggested in the literature include the Comparative Fit Index (CFI), the Tucker-Lewis Index (TLI), the Root Mean Square Error of Approximation (RMSEA) with its confidence intervals, and the Weighted Root Mean Square Residual (WRMR). Excellent and acceptable model fit were respectively indicated by values higher than .95 and .90 for CFI and TLI, values lower than .60 and .80 for RMSEA, and value lower than 1.00 for WRMR (Hu & Bentler, 1999; Marsh et al., 2004, 2005). In addition to improved model fit indices, evidence for the superiority of the ESEM over the ICM-CFA might also include meaningful target factors with strong factor loadings in the ESEM solution and reduced factor correlations from the ICM-CFA to ESEM. The former suggests that the items do indeed tap into the components they are primarily crafted to measure while the latter indicates the presence of item cross-loadings. A failure to control for the effect of item cross-loadings results in higher correlations among the components in the ICM-CFA relative to the ESEM solution (Morin et al., 2016, 2017).

In the second phase, we used factor scores from the retained measurement model in phase one to conduct latent profile analysis. Latent profile analysis, a subset of finite mixture models, can be used to uncover latent sub-groups of individuals that describe heterogeneity in a population given a phenomenon of interest (Masyn, 2013; Nylund-Gibson & Choi, 2018). In the current study, latent profile analysis was used to discern latent sub-groups of learners with different engagement profiles based on their responses to the engagement indicators. To this end, we estimated models ranging from two to six latent profiles, using 5000 random sets of start values, 300 initial stage iterations, and 250 final stage optimizations to avoid identification of local maxima (Hagenaars & McCutcheon, 2002; Masyn, 2013). For the entire model estimation and profile enumeration process, we used the maximum likelihood estimation method with robust standard errors in Mplus. We also adhered to the local independence assumption by constraining item variances across profiles to be equal while allowing for profiles indicator means to be freely estimated (Masyn, 2013).

To determine the optimum number of profiles, we relied on key statistical model fit indices, including the Bayesian information criterion (BIC), the sample-size adjusted BIC (ABIC), the Akaike information criterion (AIC), the Lo-Mendell-Rubin adjusted likelihood ratio test (LMR-ALRT), and the bootstrapped likelihood ratio test (BLRT). The first three indices are information criteria used to compare alternative models with lower values indicating superior fit (Collins & Lanza, 2009; Masyn, 2013; Nylund-Gibson & Choi, 2018). The other two indices are likelihood ratio tests used to compare the K profile model against the K-1 profile model with non-significant p-values favoring the latter (Masyn, 2013). Additionally, we examined the entropy values and the relative sizes of the resultant profiles. Entropy is an omnibus index with values higher than .80 suggesting reliable and precise classification of individuals into profiles while the smallest profile emerging from the analysis should constitute at least 5% of the sample (Nylund-Gibson & Choi, 2018).

In the final phase of data analysis, we fitted a multinomial logistic regression model to examine the predictive effect of covariates on profile membership. We used profiles extracted from the previous phase as the unordered categorical outcome variable and students’ gender, grade, academic disciplines, and perceived teacher involvement in Facebook group discussions as the covariates. For the grade and academic discipline variables, we performed dummy coding to create two binary variables. Specifically, the grade variable was coded as first-year and later-year students while the academic discipline variable was coded as English major and non-English major students. Following the procedure suggested by Asparouhov and Muthén (2014), we used the auxiliary R3STEP function in Mplus to handle the estimation of these predictive relations. Odds ratios were then used to interpret students’ likelihood of profile membership.

Results

Means, standard deviations, and correlation matrix among the variables are presented in Appendix 1. We first contrasted the ICM-CFA against the ESEM to identify the optimum structural representation of students’ responses to the Facebook learning engagement questionnaire. The ICM-CFA achieved an acceptable level of fit to data evidenced by the excellent values of CFI (.977) and TLI (.973), acceptable RMSEA value and its confidence intervals (.069; CI [.061–.077]), though with relatively poor WRMR value (1.123). On the other hand, the ESEM solution achieved a better degree of fit to data as indicated by excellent values across all fit indices (CFI = .990; TLI = .982; RMSEA = .056, CI [.046–.065]; WRMR = .612). In terms of theoretical conformity, although both solutions revealed well-defined factors with moderate to strong target factor loadings (CFA: λ = .665 to .918, M = .796; ESEM: λ = .264 to .924, M = .654), the ESEM revealed some weak target loadings and non-negligible cross-loadings (see Table 1). For example, item B2 (I stay focused during learning activities on Facebook) loaded more strongly onto the cognitive engagement component (λ = .507) than onto its target behavioral component (λ = .425). Similarly, item C1 (I go through learning materials before I participate in discussions on Facebook) and item C3 (I try to find extra learning resources to understand a concept that I find it hard to understand through Facebook discussion) had stronger loadings on the behavioral engagement component (λ = .407 and .380 respectively) than on their target component of cognitive engagement (λ = .333 and .264 respectively). In addition, lower correlations among the engagement components were observed in the ESEM solution (r = .230 to .530, M = .382) relative to the ICM-CFA solution (r = .328 to .918, M = .566) (see Table 2). These results suggested that the engagement components, particularly behavioral and cognitive engagement, were conceptually related; and failure to account for item cross-loadings resulted in inflated factor correlations as revealed in the ICM-CFA solution. Therefore, the ESEM solution, due to its superior model fit and theoretical conformity, was retained as the best fitting model representing the multidimensional structure of student engagement in the Facebook learning environment.

Table 1.

Factor Loadings of the ICM-CFA and ESEM Solutions.

ICM-CFA			ESEM
Items	λ (SE)	δ	Behaviour	Cognitive	Affective	Social	δ
Behavioural
B1	.755 (.025)	.430	.575 (.053)**	.306 (.049)**	.090 (.030)**	−.015 (.036)	.388
B2	.684 (.028)	.532	.425 (.065)**	.507 (.051)**	.002 (.029)	−.085 (.040)*	.443
B3	.742 (.027)	.449	.391 (.058)**	.295 (.048)**	.133 (.035)**	.110 (.048)*	.497
B4	.665 (.030)	.557	.373 (.053)**	.030 (.048)	.096 (.031)**	.287 (.044)**	.600
Cognitive
C1	.718 (.025)	.485	.407 (.047)**	.333 (.051)**	.053 (.034)	.117 (.039)**	.475
C2	.698 (.028)	.512	.270 (.048)**	.434 (.052)**	.070 (.028)*	.111 (.042)**	.507
C3	.817 (.022)	.332	.380 (.039)**	.264 (.047)**	−.024 (.034)	.332 (.037)**	.378
C4	.777 (.021)	.396	.273 (.042)**	.559 (.045)**	−.007 (.027)	.134 (.034)**	.351
C5	.691 (.027)	.522	.052 (.045)	.533 (.050)**	.097 (.028)**	.187 (.037)**	.494
Affective
A1	.888 (.014)	.211	.105 (.028)**	−.104 (.029)**	.912 (.015)**	−.049 (.023)*	.189
A2	.917 (.013)	.159	.044 (.029)	−.085 (.028)**	.924 (.012)**	.000 (.023)	.163
A3	.803 (.019)	.355	−.104 (.034)*	.086 (.035)*	.853 (.019)**	−.071 (.030)*	.299
A4	.802 (.023)	.357	−.061 (.037)	.031 (.035)	.755 (.021)**	.081 (.029)**	.392
Social
S1	.814 (.018)	.337	−.046 (.031)	.164 (.033)**	−.021 (.026)	.749 (.029)**	.343
S2	.840 (.015)	.294	−.093 (.035)*	.147 (.032)**	.021 (.026)	.798 (.028)**	.296
S3	.918 (.010)	.158	−.218 (.030)**	.231 (.030)**	.046 (.021)*	.891 (.025)**	.123
S4	.908 (.010)	.175	−.164 (.033)**	.137 (.032)**	.033 (.021)	.913 (.026)**	.156
S5	.878 (.013)	.229	.259 (.038)**	−.277 (.043)**	.010 (.020)	.900 (.031)**	.102
S6	.807 (.018)	.349	.159 (.045)**	−.197 (.043)**	−.049 (.023)*	.845 (.031)**	.295

*p-value significant at .05 level.

**p-value significant at .01 level.

Table 2.

Factor Correlations of the ICM-CFA and ESEM Solutions.

	ESEM
	Behaviour	Cognitive	Affective	Social
Behaviour		.444(.033)	.230(.037)	.530(.032)
Cognitive	.918(.020)		.259(.036)	.491(.037)
Affective	.371(.045)	.343(.042)		.336(.039)
Social	.682(.029)	.753(.024)	.328(.039)
ICM-CFA

Factor scores derived from the ESEM solution were then used to extract the optimum number of profiles in latent profile analysis. Table 3 presents fit indices for models varying from two to six profiles.

Table 3.

Fit Indices for the Latent Profile Models.

MODELS	Log-likelihood	npar	BIC	Adjusted BIC	AIC	VLMR-LRT	LMR-ALRT	Entropy
2 profiles	−2032.554	13	4143.222	4101.971	4091.107	0.0142	0.0159	0.701
3 profiles	−1943.622	18	3995.402	3938.285	3923.243	0.0159	0.0177	0.813
4 profiles	−1909.189	23	3956.582	3883.599	3864.379	0.0203	0.0223	0.788
5 profiles	−1888.636	28	3945.519	3856.671	3833.272	0.2486	0.2586	0.811
6 profiles	−1872.408	33	3943.107	3838.393	3810.816	0.1149	0.1199	0.838

Note. npar = number of free parameters; BIC = Bayesian Information Criterion; Adjusted BIC = Adjusted Bayesian Information Criterion; AIC = Akaike Information Criterion; VLMR-LRT = Vuong-Lo-Mendell-Rubin likelihood ratio test; LMR-ALRT = Lo-Mendell-Rubin adjusted likelihood ratio test; BLRT = bootstrapped likelihood ratio test.

While the information criteria including BIC, ABIC, and CAIC continued to decrease with the addition of subsequent profiles, thus providing little information for the optimum number of profiles to be retained, the p values of the VLMR-LRT and LMR-ALRT ceased to be significant from the five-profile model onwards. This suggested that solutions with five or more profiles did not provide further improvement in fit compared to the four-profile solution. Therefore, we closely scrutinized the adjacent three-profile and four-profile solutions in terms of their classification uncertainty and relative sizes. The three-profile solution had comparatively higher entropy value than the four-profile solution (.813 versus .788), suggesting that the classification of students into the former was slightly more reliable and accurate than the latter. In addition, the smallest profile in the three-profile solution constituted 5.7% of the total sample while that figure for the four-profile solution was 3.9%. Further inspection of the four-profile solution revealed that the addition of a fourth profile conduced to an arbitrary decomposition of an existing profile in the three-profile solution into two profiles with moderate level of engagement across all indicators. The evidence, therefore, coalesced to suggest that heterogeneity among study participants with respect to their engagement in Facebook learning could be described by three latent profiles.

Figure 2 presents the z-standardized mean scores across four components of affective, behavioral, cognitive, and social engagement in the three-profile solution.

Figure 2.

Z-Standardized Mean Scores of Student Engagement Indicators.

Profile 1, constituting 33.9% of the sample (N = 138), was characterized by scores approximately one standard deviation below the mean across all engagement components. This profile was, therefore, labelled the minimally engaged profile. Profile 2, labelled moderately engaged profile, included 246 students (60.4% of the sample) whose scores on the four engagement components clustered around the mean. The smallest profile constituted 5.7% of the sample (N = 23) and was labelled the highly engaged profile. Students in this profile scored approximately one standard deviation above the mean on all engagement components.

The three profiles were then used as a categorical outcome variable in the subsequent multinomial logistic regression model where gender, grades, academic disciplines, and perceived teacher involvement were added as predictor variables. Table 4 reports the odds ratios of profile membership for the predictor variables.

Table 4.

Multinomial Logistic Regression Results.

	Profile 1 vs. Profile 3			Profile 2 vs. Profile 3			Profile 1 vs. Profile 2
Demographic variables	Coef.	SE	OR	Coef.	SE	OR	Coef.	SE	OR
Gender	−0.326	0.678	0.722	−0.438	0.674	0.645	0.112	0.340	1.119
Grade	−0.442	0.853	0.643	−1.172	1.086	0.310	0.730	0.552	2.075
Academic discipline	0.665	2.792	1.944	1.798	2.791	6.038	1.133 **	0.333	3.105
Perceived teacher involvement	1.488	1.010	4.428	0.540	1.035	1.716	−0.949**	0.299	0.387

Note. Profile 1 = Minimally engaged profile; Profile 2 = Moderately engaged profile; Profile 3 = Highly engaged profile. Coef. = regression coefficient; SE = Standard error; OR = Odds ratios.

**p-value significant at .01 level.

Results showed that only academic discipline and perceived teacher involvement significantly predicted profile membership. More specifically, English major students had higher likelihood of being in the minimally engaged profile compared with the moderately engaged profile while higher perceived teacher involvement in the Facebook learning environment increased the likelihood of being in the moderately engaged profile as opposed to the minimally engaged profiles. No other significant relationships were detected.

Discussion

Drawing on the self-system model of motivational development and adopting a person-centred approach, the current study contributes to the engagement literature by uncovering latent sub-groups of learners with different engagement patterns in the Facebook learning environment. Results of latent profile analysis indicated that there existed heterogeneity in the sample which was manifested via three distinct profiles with different levels of engagement.

The largest profile constituted 60.4% of the total sample and was characterized by moderate level of engagement across all four components. This finding is consistent with previous studies in school-based contexts by Bae and DeBusk-Lane (2019), Bae et al. (2020), Wang and Peck (2013), and van Rooij et al. (2017) who also found the profile with uniformly moderate level of engagement across all dimensions to be the largest in size, ranging from 36.1% to 66.3% of the sample. This salient profile across studies suggests that the various pathways through which a large number of students engage with the learning activities are closely connected and similar across contexts (Bae et al., 2020). This consistency in profile extraction becomes even more pronounced in the current study given that the Facebook learning environment incorporates unique features distinguishing itself from the contexts in which student engagement was assessed in previous studies. Therefore, even though student engagement is posited to be responsive to contextual changes as per the self-system motivational perspective, person-centred studies across settings constantly found a large sub-group of students with moderate level of multidimensional engagement.

In line with previous studies, a minimally engaged profile with uniformly low level of engagement across dimensions was also identified in this study. However, unlike previous studies that found a small proportion of minimally engaged students; for instance, 7.3% in van Rooij et al. (2017); 14% in Wang and Peck (2013); and 7.3% in Watt et al. (2017), the minimally engaged profile in this study accounted for 33.9% of the sample, suggesting that a decent number of students had a tendency to disengage themselves from learning activities on Facebook. The nature of the Facebook learning environment in this study might help explain this finding. More specifically, contrary to studies by Watt et al. (2017), van Rooij et al. (2017), and Wang and Peck (2013) that assessed students’ engagement in mandatory school-based context, student engagement in this study was assessed in the context of Facebook groups as a supplementary learning platform where student participation was voluntary. Students, therefore, had options as to the degree to which their multidimensional energetic investment was prioritized so as to maximize their course learning outcomes. For this group of students, participation in Facebook learning might not give them a cutting edge over the traditional classroom context for the improvement of their academic performance, which in turn impacts on their level of engagement. As unpacked momentarily, membership in this profile is linked to students’ academic disciplines.

Association between students’ academic disciplines operationalized as English major and non-English major students and profile membership was observed. English major students had a higher probability of being in the minimally engaged profile relative to the moderately engaged profile. A plausible explanation for this finding is that these students spent their entire regular class hours practicing English skills, participating in English learning tasks, and having their major learning issues and queries answered to their satisfaction during the learning process in class. This ample in-class contact hours and the English-focused nature of their course design might have rendered the classroom setting and other face-to-face communication channels more engaging for them than the Facebook learning environment. On the other hand, the limited contact hours in class might have predisposed non-English major students to search for other learning environments in compensation. Since the Facebook groups were set up, maintained, and promoted by teachers these students might consider Facebook a more credible and safe place where they can interacted with peers, had their questions answered by teachers, found more learning resources, and shared their learning difficulties, thus engaged in those learning activities at a higher level. An alternative explanation is that students with different academic disciplines used English for different purposes and with different levels of intensity in their respective professional domains. Therefore, they might have differing needs with respect to the amount of time and effort they put in learning English beyond the classroom. As suggested by the current study, non-English major students tended to attach more value to the virtual learning space afforded by Facebook and hence had higher probability of being classified in the group that showed greater engagement in this learning platform. This finding corroborates an earlier observation by Korhonen et al. (2017) that students across different academic disciplines engaged in learning activities to varying degrees, depending on their professional identity and study program requirements. The association between academic disciplines and learner engagement profiles, as found in this study, implied that the design and delivery of Facebook learning tasks should be carefully conducted so as to maximize the learning ecologies and cater to different learning needs among learners across various disciplines. For English major students, for example, the Facebook platform can be used to share extra learning resources that build on, explain, or clarify language learning issues that students find challenging and overwhelming to address in class. On the other hand, non-English major students might benefit from more opportunities for interactions on Facebook via topic discussions, collaborative language projects, and peer tutoring.

Perceived teacher involvement in Facebook-based group discussion was also found to be a significant predictor of profile membership. Higher teacher involvement, as perceived by students, was associated with higher likelihood of membership in the moderately engaged profile as opposed to the minimally engaged profile. This finding is not unexpected given findings reported in previous variable-centred studies that student engagement was higher in Facebook groups established and maintained by teachers (Cunha et al., 2016; Imlawi et al., 2015). Students’ valuing of instructor credibility (Imlawi et al., 2015) as well as higher interaction and better relationship with teachers (Cunha et al., 2016; Sheeran & Cummings, 2018) might have encouraged them to participate in Facebook group discussions more actively. An implication from this finding is that teachers play an important role in promoting the use of Facebook for learning. In addition to their initiative and effort to establish and maintain the Facebook groups, it is their proactive involvement in and creative delivery of learning activities that likely enhance learner engagement.

Finally, the finding that students’ responses to the engagement questionnaire were best represented by an ESEM model might offer some useful implications for the measurement of student engagement. The capability of the ESEM to account for item cross-loadings might help address the measurement concerns regarding the conceptually related and partially overlapping engagement components which have hampered the conceptual understanding of the construct for long (Kahu, 2013). Factor scores obtained from the ESEM then can be considered more reliable indicators for the extraction of latent profiles than scale scores (e.g., sum or average of item scores within a scale). This is because these factor scores better preserve the characteristics of the underlying measurement model of student engagement (i.e. the ESEM representation) and partially control for measurement errors (Gillet et al., 2019). However, due to the convenience sampling method used for participant recruitment in this study, application of the ESEM approach would need to be replicated in future studies to ascertain its utility.

Limitations and Conclusion

This study has several limitations that could be addressed in future studies. First, although the identification of the three engagement profiles proffers useful implications for the promotion of Facebook as a supplemental learning platform, the profiles only show level effect rather than shape effect. Profiles with shape effects in which a single engagement component stands out from the remaining components due to its dominantly high or low score might be more interesting because individualized intervention programs can be designed accordingly to alleviate engagement deficiencies. A potential reason for this lack of shape effect is that the current study did not take the global engagement factor into account. Failure to account for the underlying global engagement factor in addition to the specific components of behavioral, cognitive, affective, and social engagement all assessed from the same set of indicators might have obscured the qualitative differences among the profiles (Morin et al., 2017). Although the co-existence of a global engagement construct and specific components of engagement has been alluded to in some previous studies (Bae et al., 2020; Bae & DeBusk-Lane, 2019; Miller et al., 2021), we tentatively withheld from examining this phenomenon in the voluntary Facebook learning environment until a more solid background is established for this evolving concept based on more robust theoretical, analytical, and empirical evidence. Another limitation of the study is the relatively small sample size which precludes the possibility for split-sample cross-validation of the profiles (Masyn, 2013). Profile validity, therefore, would benefit from more replication studies that cross-validate the profile structure across different populations with different demographic characteristics. Finally, this study did not include students’ academic achievements as a distal outcome in the exploration of profiles’ differences because we were not able to obtain students’ GPA scores at the time of data collection. This leaves a gap that would need to be addressed in future studies.

Despite these limitations, the study shows that latent profile analysis is a promising person-centred analytical approach to examine learner engagement in the context of Facebook learning. The three profiles identified replicate findings from earlier research on student engagement in schools, mathematics, and science learning while introducing some unique features typical of the Facebook learning environment. The design and delivery of Facebook-based learning tasks, therefore, can be adjusted in accordance with the identified profiles to keep learners engaged in terms of their behavioral and emotional responses as well as their cognitive thinking and social interactions.

Footnotes

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The authors received no financial support for the research, authorship, and/or publication of this article.

ORCID iD

Vo Ngoc Hoi

Author Biographies

Vo Ngoc Hoi is a lecturer in the Department of Foreign Languages, Quy Nhon University, Vietnam. He obtained a doctoral degree in education from Queensland University of Technology, Australia. His research interests include educational technology and educational psychology. He has published papers in top-tier educational research journals including Computers & Education, British Journal of Educational Technology, Journal of Computer Assisted Learning, Education and Information Technologies, and Interactive Learning Environments.

Ho Le Hang is a lecturer in the Department of Foreign Languages, Ho Chi Minh City Open University, Vietnam. Her research interests include technology mediated learning and language education. Her research has been published in Asian EFL Journal, Journal of Computer Assisted Learning, and Education and Information Technologies.

Appendix 1

Mean, Standard Deviation and Item Correlation Matrix

Item	B1	B2	B3	B4	C1	C2	C3	C4	C5	A1	A2	A3	A4	S1	S2	S3	S4	S5	S6
B1	1
B2	.499**	1
B3	.472**	.463**	1
B4	.372**	.320**	.448**	1
C1	.474**	.424**	.458**	.371**	1
C2	.405**	.461**	.371**	.319**	.442**	1
C3	.551**	.416**	.436**	.428**	.509**	.433**	1
C4	.485**	.466**	.432**	.347**	.493**	.509**	.547**	1
C5	.353**	.401**	.404**	.275**	.375**	.418**	.392**	.552**	1
A1	.186**	.093	.224**	.150**	.192**	.179**	.152**	.114*	.169**	1
A2	.175**	.049	.219**	.199**	.185**	.175**	.130**	.154**	.168**	.799**	1
A3	.121**	.119*	.143**	.120*	.119*	.091	.073	.087	.127**	.674**	.694**	1
A4	.176**	.054	.214**	.162**	.205**	.132**	.185**	.125*	.183**	.677**	.660**	.662**	1
S1	.340**	.343**	.406**	.400**	.401**	.327**	.488**	.446**	.401**	.126*	.148**	.113*	.185**	1
S2	.371**	.265**	.391**	.357**	.397**	.331**	.535**	.431**	.391**	.175**	.194**	.144**	.218**	.664**	1
S3	.380**	.347**	.421**	.408**	.408**	.430**	.562**	.434**	.441**	.176**	.225**	.144**	.230**	.651**	.726**	1
S4	.347**	.332**	.397**	.432**	.401**	.430**	.478**	.432**	.451**	.154**	.197**	.171**	.196**	.673**	.651**	.780**	1
S5	.379**	.284**	.384**	.462**	.414**	.374**	.528**	.416**	.356**	.207**	.199**	.092	.217**	.617**	.653**	.675**	.716**	1
S6	.325**	.242**	.298**	.383**	.349**	.335**	.464**	.389**	.332**	.099**	.124*	.065	.131**	.556**	.571**	.616**	.649**	.746**	1
Mean	3.457	3.779	3.641	3.459	3.506	3.742	3.106	3.609	3.656	4.700	4.668	4.826	4.683	3.123	2.939	3.086	3.160	2.919	2.823
SD	.820	.753	.768	.852	.868	.772	.988	.884	.824	1.061	1.076	1.001	1.058	.955	1.057	.942	.971	1.046	1.109

*Correlation is significant at the .05 level (2-tailed).

**Correlation is significant at the .01 level (2-tailed).

References

Appleton

J. J.

Christenson

S. L.

Furlong

M. J.

(2008). Student engagement with school: Critical conceptual and methodological issues of the construct. Psychology in the Schools, 45(5), 369–386.

Appleton

J. J.

Christenson

S. L.

Kim

Reschly

A. L.

(2006). Measuring cognitive and psychological engagement: Validation of the student engagement instrument. Journal of School Psychology, 44(5), 427–445.

Asparouhov

Muthén

(2009). Exploratory structural equation modeling. Structural Equation Modeling: A Multidisciplinary Journal, 16(3), 397–438.

Asparouhov

Muthén

(2014). Auxiliary variables in mixture modeling: Three-step approaches using M plus. Structural Equation Modeling: A Multidisciplinary Journal, 21(3), 329–341.

Asparouhov

Muthén

Morin

A. J.

(2015). Bayesian structural equation modeling with cross-loadings and residual covariances: Comments on Stromeyer et al. Journal of Management, 41(6), 1561–1577.

Awidi

I. T.

Paynter

Vujosevic

(2019). Facebook group in the learning design of a higher education course: An analysis of factors influencing positive learning experience for students. Computers & Education, 129, 106–121.

Bae

C. L.

DeBusk-Lane

(2019). Middle school engagement profiles: Implications for motivation and achievement in science. Learning and Individual Differences, 74, 101753.

Bae

C. L.

DeBusk-Lane

M. L.

Lester

A. M.

(2020). Engagement profiles of elementary students in urban schools. Contemporary Educational Psychology, 62, 101880–101813.

Ben-Eliyahu

Moore

Dorph

Schunn

C. D.

(2018). Investigating the multidimensionality of engagement: Affective, behavioral, and cognitive engagement across science activities and contexts. Contemporary Educational Psychology, 53, 87–105.

10.

Bergdahl

Nouri

Fors

Knutsson

(2020). Engagement, disengagement and performance when learning with technologies in upper secondary school. Computers & Education, 149, 103783–103717.

11.

Bergman

L. R.

Andersson

(2010). The person and the variable in developmental psychology. Zeitschrift Für Psychologie / Journal of Psychology, 218(3), 155–165.

12.

Blumenfeld

Modell

Bartko

W. T.

Secada

W. G.

Fredricks

J. A.

Friedel

Paris

(2005). School engagement of inner-city students during middle childhood. In Cooper

C. R.

Coll

C. T. G.

Bartko

W. T.

Davis

(Eds.), Developmental pathways through Middle childhood: Rethinking contexts and diversity as resources (pp. 145–170). Lawrence Erlbaum.

13.

Bowman

N. D.

Akcaoglu

(2014). “I see smart people!”: Using Facebook to supplement cognitive and affective learning in the university mass lecture. The Internet and Higher Education, 23, 1–8.

14.

Collins

L. M.

Lanza

S. T.

(2009). Latent class and latent transition analysis: With applications in the social, behavioral, and health sciences. John Wiley & Sons.

15.

Connell

J. P.

(1990). Context, self, and action: A motivational analysis of self-system processes across the life span. The Self in Transition: Infancy to Childhood, 8, 61–97.

16.

Connell

P., & Welborn

(1991). Competence, autonomy and relatedness: A motivational analysis of self-system processes. In Gunna

M. R.

Sroufe

L. A.

(Eds.), Self processes in development: Minnesota symposium on child psychology (Vol. 23, pp. 43–77). Erlbaum.

17.

Cunha

F. R. D.

Jr. van Kruistum

van Oers

(2016). Teachers and Facebook: Using online groups to improve students’ communication and engagement in education. Communication Teacher, 30(4), 228–241.

18.

Datu

J. A. D.

Yang

Valdez

J. P. M.

Chu

S. K. W.

(2018). Is Facebook involvement associated with academic engagement among Filipino university students? A cross-sectional study. Computers & Education, 125, 246–253.

19.

Deng

Tavares

N. J.

(2013). From Moodle to Facebook: Exploring students’ motivation and experiences in online communities. Computers & Education, 68, 167–176.

20.

Deng

Benckendorff

Gannaway

(2020). Learner engagement in MOOCs: Scale development and validation. British Journal of Educational Technology, 51(1), 245–262.

21.

Dyson

Vickers

Turtle

Cowan

Tassone

(2015). Evaluating the use of Facebook to increase student engagement and understanding in lecture-based classes. Higher Education, 69(2), 303–313.

22.

Finn

A. N.

Schrodt

(2016). Teacher discussion facilitation: A new measure and its associations with students’ perceived understanding, interest, and engagement. Communication Education, 65(4), 445–462.

23.

Finn

J. D.

Pannozzo

G. M.

Voelkl

K. E.

(1995). Disruptive and inattentive-withdrawn behavior and achievement among fourth graders. The Elementary School Journal, 95(5), 421–434.

24.

Fredricks

J. A.

Blumenfeld

P. C.

Paris

A. H.

(2004). School engagement: Potential of the concept, state of the evidence. Review of Educational Research, 74(1), 59–109.

25.

Fredricks

J. A.

Hofkens

T. L.

Wang

M.-T.

(2019). Addressing the challenge of measuring student engagement. In Renninger

K. A.

Hidi

S. E.

(Eds.), The Cambridge handbook of motivation and learning. Cambridge University Press.

26.

Fredricks

J. A.

McColskey

(2012). The measurement of student engagement: A comparative analysis of various methods and student self-report instruments. In Christenson

S. L.

Reschly

A. L.

Wylie

(Eds.), Handbook of research on student engagement (pp. 763–782). Springer.

27.

Gillet

Caesens

Morin

A. J.

Stinglhamber

(2019). Complementary variable-and person-centred approaches to the dimensionality of work engagement: A longitudinal investigation. European Journal of Work and Organizational Psychology, 28(2), 239–258.

28.

Hagenaars

J. A.

McCutcheon

A. L.

(2002). Applied latent class analysis. Cambridge University Press.

29.

Hiver

Al-Hoorie

A. H.

Mercer

(2020). Student engagement in the language classroom. Multilingual Matters.

30.

Hong

Gardner

(2019). Undergraduates’ perception and engagement in facebook learning groups. British Journal of Educational Technology, 50(4), 1831–1845.

31.

L. t.

Bentler

P. M.

(1999). Cutoff criteria for fit indexes in covariance structure analysis: Conventional criteria versus new alternatives. Structural Equation Modeling: A Multidisciplinary Journal, 6(1), 1–55.

32.

Imlawi

Gregg

Karimi

(2015). Student engagement in course-based social networks: The impact of instructor credibility and use of communication. Computers & Education, 88, 84–96.

33.

Kahu

E. R.

(2013). Framing student engagement in higher education. Studies in Higher Education, 38(5), 758–773.

34.

Kong

S. C.

(2011). An evaluation study of the use of a cognitive tool in a one-to-one classroom for promoting classroom-based dialogic interaction. Computers & Education, 57(3), 1851–1864.

35.

Korhonen

Inkinen

Mattsson

Toom

(2017). Student engagement and the transition from the first to second year in higher education. In Kyndt

Donche

Trigwell

Lindblom-Ylänne

(Eds.), Higher education transitions: Theory and research (pp. 113–134). Routledge.

36.

Krause

K. L.

Coates

(2008). Students’ engagement in first‐year university. Assessment & Evaluation in Higher Education, 33(5), 493–505.

37.

Lawson

M. A.

Lawson

H. A.

(2013). New conceptual frameworks for student engagement research, policy, and practice. Review of Educational Research, 83(3), 432–479.

38.

Lerner

R. M.

(2013). Interrelations of behavioral, emotional, and cognitive school engagement in high school students. Journal of Youth and Adolescence, 42(1), 20–32.

39.

Luo

Hughes

J. N.

Liew

Kwok

(2009). Classifying academically at-risk first graders into engagement types: Association with long-term achievement trajectories. The Elementary School Journal, 109(4), 380–405.

40.

Marsh

H. W.

Hau

K.-T.

Grayson

(2005). Goodness of fit evaluation in structural equation models. In Maydeu-Olivares

McArdle

(Eds.), Contemporary psychometrics: A festschrift for Roderick P. McDonald (pp. 275–340). Erlbaum.

41.

Marsh

H. W.

Hau

K.-T.

Wen

(2004). In search of golden rules: Comment on hypothesis-testing approaches to setting cutoff values for fit indexes and dangers in overgeneralizing Hu and Bentler’s (1999) findings. Structural Equation Modeling: A Multidisciplinary Journal, 11(3), 320–341.

42.

Masserini

Bini

(2020). Does joining social media groups help to reduce students’ dropout within the first university year? Socio-Economic Planning Sciences, 73, 1–9.

43.

Masyn

K. E.

(2013). Latent class analysis and finite mixture modeling. In Little

T. D.

(Ed.), The Oxford handbook of quantitative methods (pp. 551–611). Oxford University Press.

44.

Miller

C. J.

Perera

H. N.

Maghsoudlou

(2021). Students’ multidimensional profiles of math engagement: Predictors and outcomes from a self‐system motivational perspective. British Journal of Educational Psychology, 91(1), 261–285.

45.

Miller

D. C.

Byrnes

J. P.

(2001). Adolescents’ decision making in social situations: A self-regulation perspective. Journal of Applied Developmental Psychology, 22(3), 237–256.

46.

Molinillo

Aguilar-Illescas

Anaya-Sánchez

Vallespín-Arán

(2018). Exploring the impacts of interactions, social presence and emotional engagement on active collaborative learning in a social web-based environment. Computers & Education, 123, 41–52.

47.

Morin

A. J.

Arens

A. K.

Marsh

H. W.

(2016). A bifactor exploratory structural equation modeling framework for the identification of distinct sources of construct-relevant psychometric multidimensionality. Structural Equation Modeling: A Multidisciplinary Journal, 23(1), 116–139.

48.

Morin

A. J.

Boudrias

J.-S.

Marsh

H. W.

McInerney

D. M.

Dagenais-Desmarais

Madore

Litalien

(2017). Complementary variable-and person-centred approaches to the dimensionality of psychometric constructs: Application to psychological wellbeing at work. Journal of Business and Psychology, 32(4), 395–419.

49.

Muthen

L. K.

Muthen

B. O.

(2012). Mplus user’s guide (7th ed.). Muthén & Muthén.

50.

Northey

Bucic

Chylinski

Govind

(2015). Increasing student engagement using asynchronous learning. Journal of Marketing Education, 37(3), 171–180.

51.

Nylund-Gibson

Choi

A. Y.

(2018). Ten frequently asked questions about latent class analysis. Translational Issues in Psychological Science, 4(4), 440–461.

52.

Pöysä

Poikkeus

A.-M.

Muotka

Vasalampi

Lerkkanen

M.-K.

(2020). Adolescents’ engagement profiles and their association with academic performance and situational engagement. Learning and Individual Differences, 82, 101922.

53.

Reschly

A. L.

Christenson

S. L.

(2012). Jingle, jangle, and conceptual haziness: Evolution and future directions of the engagement construct. In Reschly

A. L.

Christenson

S. L.

Wylie

(Eds.), Handbook of research on student engagement (pp. 3–19). Springer.

54.

Schmidt

J. A.

Rosenberg

J. M.

Beymer

P. N.

(2018). A person‐in‐context approach to student engagement in science: Examining learning activities and choice. Journal of Research in Science Teaching, 55(1), 19–43.

55.

Sheeran

Cummings

D. J.

(2018). An examination of the relationship between Facebook groups attached to university courses and student engagement. Higher Education, 76(6), 937–955.

56.

Skinner

E. A.

Belmont

M. J.

(1993). Motivation in the classroom: Reciprocal effects of teacher behavior and student engagement across the school year. Journal of Educational Psychology, 85(4), 571–581.

57.

Skinner

E. A.

Kindermann

T. A.

Connell

J. P.

Wellborn

J. G.

(2009). Engagement and disaffection as organizational constructs in the dynamics of motivational development. In Wentzel

Wigfield

(Eds.), Handbook of motivation in school (pp. 223–245). Erlbaum.

58.

Skinner

E. A.

Wellborn

J. G.

(1994). Coping during childhood and adolescence: A motivational perspective. In Featherman

Lerner

Perlmutter

(Eds.), Life-span development and behavior (Vol. 12, pp. 91–133). Erlbaum.

59.

Thai

Sheeran

Cummings

D. J.

(2019). We’re all in this together: The impact of Facebook groups on social connectedness and other outcomes in higher education. The Internet and Higher Education, 40, 44–49.

60.

van Rooij

E. C.

Jansen

E. P.

van de Grift

W. J.

(2017). Secondary school students’ engagement profiles and their relationship with academic adjustment and achievement in university. Learning and Individual Differences, 54, 9–19.

61.

Wang

M.-T.

(2009). School climate support for behavioral and psychological adjustment: Testing the mediating effect of social competence. School Psychology Quarterly, 24(4), 240–251.

62.

Wang

M.-T.

Dishion

T. J.

(2012). The trajectories of adolescents’ perceptions of school climate, deviant peer affiliation, and behavioral problems during the middle school years. Journal of Research on Adolescence: The Official Journal of the Society for Research on Adolescence, 22(1), 40–53.

63.

Wang

M.-T.

Fredricks

J. A.

Hofkens

T. L.

Linn

J. S.

(2016). The math and science engagement scales: Scale development, validation, and psychometric properties. Learning and Instruction, 43, 16–26.

64.

Wang

M.-T.

Peck

S. C.

(2013). Adolescent educational success and mental health vary across school engagement profiles. Developmental Psychology, 49(7), 1266–1276.

65.

Watt

H. M.

Carmichael

Callingham

(2017). Students’ engagement profiles in mathematics according to learning environment dimensions: Developing an evidence base for best practice in mathematics education. School Psychology International, 38(2), 166–183.