International Review of Research in Open and Distributed Learning

verify a Chinese version of Community of Inquiry (CoI) instrument with learning presence and explore the causal relationships of the factors in the instrument. This study first examined the reliability and validity of the instrument. All four presences had acceptable levels of reliability (all Cronbach's α> .765 or higher). The confirmatory factor modeling approach was used to assess its validity. Then, the study used path analysis and regression analysis to explore the causal relationships of the presences. The key findings showed that teaching and social presences directly influenced the perceptions of learning presence. Learning presence was a partial mediating variable of interactional relationship within CoI constructs.


Introduction
China has the world's largest higher education population, with more than 36 million (Ministry of Education, 2016).In recent years, Chinese universities have devoted great effort to the development of online courses such as Massive Open Online Courses (MOOCs) to serve university students and adult learners (Zhang, Perris, Zheng, & Chen, 2015).A large number of online courses have been developed, but the learning experience of online course needed to be examined and improved (Songhe & Xuan, 2014).It is well known that creating and sustaining a learning community is valuable to enhance the online learning experience (Akyol, Garrison, & Ozden, 2009).Therefore, generating a reliable instrument to measure learners' perception of online learning community in China becomes essential.

The Community of Inquiry Framework
The internationally recognized Community of Inquiry (CoI) framework seems to be a valid theoretical framework to understand learners' perceptions of online learning experience (Shea & Bidjerano, 2010; Traver, Volchok, Bidjerano, & Shea, 2014).It has been adopted and adapted by a lot of researchers worldwide (Garrison, Anderson, & Archer, 2010;Swan & Ice, 2010).It provides collaborative-constructivist perspective and methodology for studying online learning experiences through the development of three core presences (Arbaugh et al., 2008;Garrison & Arbaugh, 2007): teaching presence, social presence, and cognitive presence (Garrison, Anderson, & Archer, 2010).
The CoI framework contains three conceptual elements: teaching presence, cognitive presence, and social presence (Garrison, Anderson, & Archer, 2000).Teaching presence refers to the online instructor's "design, facilitation, and direction of cognitive and social processes for the purpose of realizing personally meaningful and educationally worthwhile learning outcomes" (Anderson, Rourke, Garrison, & Archer, 2001, P. 5).Teaching presence depends on instructional design and organization of the curriculum and activities, as well as the maintenance of online discussion and organization of instruction (Ice, Gibson, Boston, & Becher, 2011).Cognitive presence is defined as the degree to which learners are able to construct knowledge through sustained self-reflection and dialogue (Swan, Garrison, & Richardson, 2009).It consists of four interconnected and cyclical phases: triggering event, exploration, integration, and resolution (Garrison, Anderson, & Archer, 2001).Social presence is defined as the ability to which participants demonstrate "real people" socially and emotionally through computer-mediated communication (Garrison & Arbaugh, 2007).It depends on the expressions of personal emotion, the sense of group identity and the group cohesion of learners (Ice et al., 2011).The CoI framework with the three core presences offers a theoretical perspective and methodology for studying the potential and effectiveness of online and blended learning (Garrison et al., 2010).
Recently, some researchers suggested the CoI framework can be further extended (Traver et al., 2014), as it lacks attention to behaviors and attitudes as learners adjust to metacognitive, motivational, and behavioral activities in online learning (Shea et al., 2013).For example, learners' discourse about individual and collaborative efforts to regulate their learning process could not be accounted for by the existing three presences in the CoI framework (Hayes, Uzuner-Smith, & Shea, 2015).Shea and Bidjerano (2010) proposed that learning presence could be a new form of presence in the CoI framework.Learning presence is described as "the proactive stance adopted by students who marshal thoughts, emotions, motivation, behaviors, and strategies in the service of successful online learning" (Shea & Bidjerano, 2012, p. 90).Learning presence has been used to explain self-regulated learning, which includes the planning of learning tasks, monitoring of learning process, and the application of strategy in online learning (Shea & Bidjerano, 2012).The addition of learning presence into the CoI framework helps to enhance the understanding of successful online learning experiences through individual differences of self-regulation (Shea & Bidjerano, 2010).
The viewpoint of whether learning presence should be added into the CoI framework is contested among researchers.Several researchers suggested adding learning presence as a missing dimension into the CoI framework (Hayes et al., 2015;Traver et al., 2014).However, Garrison (2016) believed that the re-conceptualization of the CoI framework must be strictly validated.There seem to be a common acceptable view that regulation is an important factor in the CoI framework (Garrison, 2016;Garrison & Akyol, 2015;Shea et al., 2013).Therefore, whether to incorporate the regulation into the CoI framework through learning presence needs more research.

The Reliability and Validity of the CoI Instrument
Much existing research has attempted to develop instruments to examine the CoI framework and reported its reliability and validity.For instance, the 34-item survey instrument developed by Arbaugh et al. (2008) to measure students' perception of the three presences, has been proved to be a valid and reliable tool for the CoI framework (Garrison, Cleveland-Innes, & Fung, 2004; Shea & Bidjerano, 2009; Swan & Ice, 2010).This instrument is a valid and reliable survey measure of social presence, teaching presence, and cognitive presence (Arbaugh et al., 2008).Swan et al. (2008) reported that the internal consistency reliability of the survey instrument was: a Cronbach's Alpha of 0.91 for social presence, 0.95 for cognitive presence, and 0.9 for teaching presence.Kozan and Richardson (2014a) examined the validity of the CoI instrument conducting exploratory factor analysis (EFA) and found that the final three-factor structure explained 64.83% of the variances in the pattern of relationships among the items (e.g., teaching presence 48.21%, cognitive presence 10.64%, and social presence 5.98%).Confirmatory factor analysis (CFA) found that the hypothesized model of the three factors was verified and exhibited a reasonably good fit for the data (Kozan & Richardson, 2014a).After adding learning presence to the CoI model, the reliability of learning presence was examined.The internal consistencies of self-efficacy and effort regulation scales (called the learning presence) were .95and .75(Shea & Bidjerano, 2010).The CFA examined factorial validity and found that the new model has reasonable factor structure (Shea & Bidjerano, 2010).

Participants
The participants of this study were 350 Chinese undergraduate students, chosen from three majors in the School of Humanities, including 122 Chinese Linguistic Literature students, 111 Primary School Education students, and 117 Educational Technology students.Among them, 184 were sophomores (52.6%) and 166 were juniors (47.4%).The participant had at least a year of blended learning experience and studied two to three blended courses.They were asked to respond to a survey during Regression analysis.The mediating effect of learning presence was tested by regression analysis.According to Baron and Kenny (1986), in the equation Y=cX, M=aX, Y=c'X+bM, if c, a, b, c' were all significant, then M could be seen a significant partial mediating factor between X and Y.If c, a, b were all significant but c' was not, then M could be seen a complete mediating factor between X and Y.In this research, we regarded cognitive presences (Cg) as Y, learning presences (Lp) as M, teaching presences (Tp) and social presences (Sp) as X1 and X2 respectively.Regression equation, standard error, and t value were calculated.
Reliability analysis.SPSS 17.0 was used to validate the reliability of each dimension as well as the total instrument.We used the parameter of Cronbach's α.Since the number of items in each dimension is small, Cronbach's α value higher than 0.7 was accepted.For the whole scale, the Cronbach's α higher than 0.9 could be considered excellent (George & Mallery, 2003).

Results
This part will report the validity of the Chinese version instrument (CFA results) and the relationships of the factors (PA results) respectively.The results of descriptive statistical analysis, the correlation between the four presences, reliability, and mediating effect will also be presented.

Descriptive Statistical Analysis
Table 1 shows descriptive statistics including means, standard deviations, skewness, kurtosis, minimum value, and maximal value of the four presences.For the sample used in this research, no abnormal value was found.In addition, the results presented in the data in this study were normally distributed based on the degrees of Skewness and Kurtosis, because both were less than the absolute value 1.It seems that learners' perceptions of presences were generally lower than that of international studies.
For example, Kozan and Richardson (2014a) reported the average individual item ratings of three presences ranged from 4.17 to 4.57.In this study, the average individual item ratings of the three presences were below 3.61.We noticed that similar results existed in the context of South Korea as well.The South Korean students' perceptions of the three presences were below 3.87.
Three possible reasons may explain this discrepancy.The first is the context differences between North America and China.Previous studies compared the perceptions of students across countries and confirmed the influence of group and cultural differences on perceptions of online learning (Ashong & Commander, 2012).We hypothesized that the inclination of inquiry learning influences students' perception of CoI.Specifically, Chinese learners were found to be less critical and questioning in online discussions than American learners (Thompson & Ku, 2005).They are more likely to get a definite answer from teachers, rather than getting answers from the interaction (Liang & McQueen, 1999).So, the significant differences of students' perceptions of the CoI presences in Chinese and western contexts may depend on the inclination of inquiry learning.The second may be the issue of translation.Although we adjusted the translation content for Chinese students, students may still have difficulties in understanding specific options.The third reason seems to be that the participants' perception of the online learning community could be a little bit lower than that of some previous studies (Swan, Matthews, Bogle, Boles, & Day, 2012).
The strength of the linear relationships among learning presence, teaching presence, social presence, and cognitive presences were examined by Spearman's rho.As Table 2 shows, the four constructs are moderately to strongly correlated.The largest correlations are between cognitive presence and social presence (ρ=.698, p<0.01), and between cognitive presence and teaching presence (ρ=.665, p<0.01).

Validity of the Chinese version Instrument
In    To further test the mediating effect of learning presence, regression analysis was carried out in six steps.Regression equations and results are shown in Table 5.As Table 5 shows, in the first three steps, the regression coefficients c(0.697), a(0.518), c'(0.482), and b(0.417) were all significant, so learning presence (M) was verified as a significant partial mediating factor between teaching presence (X1) and cognitive presence (Y).In the last three steps, the regression coefficients c(0.716), a(0.565), c'(0.499), and b(0.384) were all significant, so learning presence (M) was verified as a significant partial mediating factor between social presence (X2) and cognitive presence (Y).This study provided a reliable instrument for researchers and practitioners in China to measure their students' perception of online and blended learning.It will contribute to the development and application of the CoI framework in other contexts.We hope our validated survey will not only enable

Reliability of the Chinese Version Instrument
Chinese researchers and practitioners to assess the quality of the existing online courses, but also guide them to design and develop more high-quality online courses.
Some research studies attempted to translate and validate the CoI instrument into other languages and use it in non-English speaking countries.For example, Portuguese researchers translated and adapted the CoI instrument into Portuguese, and validated the CoI instrument in a blended learning setting (Moreira, Ferreira, & Almeida, 2013).Yu and Richardson (2015) examined the reliability and validity of a Korean version of the CoI instrument.Through reliability analysis, Cronbach's α of teaching, social, and cognitive presences were .954,.913,and .956respectively.The CFA results showed that the hypothesized model of the CoI instrument was verified as an excellent fit for the data in Korean context (Yu & Richardson, 2015).It is worth noting that the reliability and validity were different between Korean and English versions of the CoI instrument.So, it is necessary to examine the reliability and validity of different language versions of the CoI instrument before using it in another different context.The Causal Relationships Among the CoI Presences Much existing research focused on examining the direct and indirect causal relationships of the CoI presences (Garrison, Cleveland-Innes, & Fung, 2010; Shea & Bidjerano, 2008, 2009).Shea and Bidjerano (2009) found that teaching presence and social presence have direct effects on cognitive presence.Garrison, Cleveland-Innes, and Fung (2010) revealed significant direct effects of teaching presence and social presence on cognitive presence, and a direct effect of social presence on cognitive presence.The above two studies confirmed the direct relationships of the three presences.The indirect relationships focus on the mediating effect of social presence and cognitive presence.Shea and Bidjerano (2009) confirmed that social presence had a partial mediating effect between teaching presence and cognitive presence.The follow-up studies further confirmed that cognitive presence is a full mediator and social presence is a partial mediator in the CoI presences (Kozan, 2016; Kozan & Richardson, 2014b).After the addition of learning presence, there is a lack of research revealing the relationship between learning presence and the other factors of the CoI framework.Shea and Bidjerano (2010) confirmed a correlation existed between self-efficacy (one component of the learning presence) and other presences in the CoI framework.They further suggested that learning presence should be connected with the CoI constructs (see Figure 1) (Shea & Bidjerano, 2010).Based on correlation between factors, researchers may further look for mediators and explore the interdependence behind the relationships (Frazier, Tix, & Barron, 2004).So, the mediators and causal relationships among the presences need to be further validated.Meanwhile, the relationships between the presences may change depending on the learning context (Shea et al., 2014).It is therefore necessary to examine the relationships among the presences in a different learning context.

Figure 2 .
Figure 2. Hypothesized model of the four presences.
time.A total of 325 students completed the survey questionnaire, and the response rate was 92.9%.InstrumentThe CoI survey instrument was originally created byArbaugh et al.(2008), with 34 items used for data collection.It was provided on a 5-point Likert scale, ranging from 0-Strongly Disagree to 4-Strongly Agree.The scale employed in this study ranged from 1 (strongly disagree) to 5 (strongly agree).The numbers of items are 13 (for teaching presence), 12 (for cognitive presence), and 9 (for social presence).The learning presence (self-efficacy and effort regulation) was measured by 14 items(Shea & Bidjerano, 2010; Traver et al., 2014).Two translators who had prior knowledge of the CoI framework and a teacher of English linguistics participated in the translation of the instrument.All three translators were native speakers of Chinese and fluent in both Chinese and English.Four stage translations in Chinese were undertaken from the original English language version of the CoI Survey instrument.Firstly, the translation from English into Chinese was undertaken by one translator.Secondly, another translator translated the reconciled Chinese language version back into English independently.Thirdly, the teacher of English linguistics reviewed the back-translation in meaning and identied problematic items.In the last step, the problematic items were discussed by the three translators until the discrepancies were resolved.Throughout the translation process, the three translators ensured the validity of the Chinese version of the COI instrument.The 34-item of the English and Chinese CoI instruments were shown in the Appendix.Procedure Confirmatory factor analysis (CFA) and path analysis (PA).LISREL v. 8.7 was used to run a full structural equation model.The full structural equation model included four measurement models and a structural model.Confirmatory factor analysis (CFA) was used to examine the four measurement models (relationships among the latent and manifest variables).The main purpose of running CFA was to verify the proposed structure of the CoI instrument along with learning presence.The goodness of fit indices, such as comparative fit index (CFI), Non-Normed Fit Index (NNFI), Root Mean Square Error of Approximation (RMSEA), and the ratio of Chi-Squared to the degree of freedom (χ2/df) were used to examine the predictive validity of the proposed structure of the CoI instrument.According to Schreiber, Nora, Stage, Barlow, and King (2006), for a good-enough fit, χ2/df value should be between 2 to 5. At the same time, above 0.9 of NNFI and CFI, below 0.08 of RMSEA, and all values of t in the key matrixes should be greater than 2. Path analysis (PA) was used to assess the structural model (direct relationships among teaching, social, cognitive presences, and learning presence).Unstandardized path coefficients (B), Standard path coefficients (β), standard error and significance (p) were used to test the significance of the relationships.
this research, the validity of the Chinese version instrument was tested during two rounds.Separate model analysis rounds were conducted on the first 48-item model and the second model.The first round of model analysis.The first round of model analysis was conducted on the 48-item model.In this research, fit indices for the hypothesized structural equation model were as follows: χ2/df =2.44, NNFI=0.927，CFI=0.931，RMSEA=0.0723.All items loaded significantly on the four factors because all the values of t in the matrixes of LX, LY, GA, and BE were greater than 2 (the smallest values of t in LX, LY, GA were 5.752, 5.523, and 2.660 respectively, and the value of t in BE was 3.383).An obviously high modification index (393.21)was found in the matrix of TD, which indicated a high relationship between the item of Sc7 and Sc9.So we decided to delete the item Sc9.The description of Sc7 was more specific.The second round of model analysis.To test whether we could come up with a better fit, the Variables Cognitive presence Teaching presence Social presence Learning presence model analysis was conducted on the modified data.The overall model fit statistics were as follows: χ2/df =2.29, NNFI=0.933，CFI=0.936，RMSEA=0.0674.The CFA results indicated the hypothesized model of the 47-item structure of the four-factor CoI instrument was verified as anexcellent fit for the data.We accepted the 47-item model as our final version (See Fig.3).The results of the CFA confirmed that the model fit is excellent between the proposed model and the observed data.

9. 143
    P<0.001.This study expanded the CoI instrument to a different language and cultural context.Firstly, we verified a Chinese version of the CoI instrument.The result confirmed that the CoI instrument to evaluate social, teaching, cognitive, and learning presence was a reliable tool for measuring the perception of learning experience in China.The study confirmed that the causal relationships between learning presence and CoI constructs.The results extended the CoI framework's components with an additional focus on self-regulatory behaviors and strategies of learners in online and blended learning.The present study engaged a mid-range sample size of 350 students in a university of China.Although this sample from China helped to diversify the research context of the CoI framework, comparatively a mid-range sample size may limit the generalizability of findings.In addition, some researchers suggested that it was necessary to examine the CoI framework across disciplines and institutions(Garrison & Arbaugh, 2007; Garrison et al., 2010).Future research should increase the number of participants and explore the dynamic relationships among the presences across disciplines and institutions.Lastly, exploratory factor analysis should be conducted to further identify the dimensionality of the Chinese version of the CoI instrument.

Table 1
Descriptive Statistics of Four Dimensions

Table 3
Reliability of the Instrument