1-s2.0-s0360131510000035-main

7/31/2019 1-s2.0-S0360131510000035-main

1/10

A study of student satisfaction in a blended e-learning system environment

Jen-Her Wu a , Robert D. Tennyson b, * , Tzyh-Lih Hsia ca Department of Information Management, National Sun Yat-Sen University, 70 Lien-Hai Road, Kaohsiung, 80424, Taiwanb University of Minnesota, 56 East River Road, Minneapolis, Minnesota 55455, United Statesc Department of Information Management, Chinese Naval Academy, P.O. Box No. 90175 Tsoying, Kaohsiung 813, Taiwan

a r t i c l e i n f o

Article history:Received 18 March 2009Received in revised form 23 December 2009Accepted 31 December 2009

Keywords:e-LearningSatisfactionLearner controlInternetTeacher-directedLearner-directedSynchronousAsynchronousFace-to-face

a b s t r a c t

This study proposes a research model that examines the determinants of student learning satisfaction ina blended e-learning system (BELS) environment, based on social cognitive theory. The research model istested using a questionnaire survey of 212 participants. Conrmatory factor analysis (CFA) was per-formed to test the reliability and validity of the measurements. The partial least squares (PLS) methodwas used to validate the measurement and hypotheses. The empirical ndings indicate that computerself-efcacy, performance expectations, system functionality, content feature, interaction, and learningclimate are the primary determinants of student learning satisfaction with BELS. The results also showthat learning climate and performance expectations signicantly affect learning satisfaction. Computerself-efcacy, systemfunctionality, content feature and interaction signicantly affect performance expec-tations. Interaction has a signicant effect on learningclimate. The ndings provide insight into those fac-tors that are likely signicant antecedents for planning andimplementing a blended e-learning system toenhance student learning satisfaction.

2010 Elsevier Ltd. All rights reserved.

1. Introduction

Classroom learning typically occurs in a teacher-directed instructional context with face-to-face interaction in a live synchronous envi-ronment. In contrast to this form of instruction, is an approach that promotes learner-directed learning. With emerging Internet commer-cialization and the proliferation of information technologies, online or electronic learning (e-learning) environments offer the possibilitiesfor communication, interaction and multimedia material delivery that enhance learner-directed learning ( Wu, Tennyson, Hsia, & Liao,2008 ). Although e-learning mayincrease access exibility, eliminate geographical barriers, improve convenience and effectiveness for indi-vidualized and collaborative learning, it suffers from some drawbacks such as lack of peer contact and social interaction, high initial costsfor preparing multimediacontent materials, substantial costs for systemmaintenance and updating, as well as the need for exible tutorialsupport ( Kinshuk & Yang, 2003; Wu et al., 2008; Yang & Liu, 2007 ). Furthermore, students in virtual e-learning environments may expe-rience feelings of isolation, frustration and confusion ( Hara & Kling, 2000 ) or reduced interest in the subject matter ( Maki, Maki, Patterson,& Whittaker, 2000 ). In addition, student satisfaction and effectiveness fore-learninghas also been questioned ( Piccoli, Ahmad, & Ives, 2001;

Santhanam, Sasidharan, & Webster, 2008 ).With the concerns and dissatisfaction with e-learning, educators are searching for alternative instructional delivery solutions to relievethe above problems. The blended e-learning system (BELS) has been presented as a promising alternative learning approach ( Graham,2006 ). BELS refers to an instructional system that combines multiple learning delivery methods, including most often face-to-face class-room with asynchronous and/or synchronous online learning. It is characterized as maximizing the best advantages of face-to-face andonline education.

While BELS has been recognized as having a number of advantages (e.g., instructional richness, access to knowledge content, socialinteraction, personal agency, cost effectiveness, and ease of revision ( Osguthorpe & Graham, 2003 )), insufcient learning satisfaction is stillan obstacle to the successful BELS adoption ( So & Brush, 2008 ). In fact, research ndings from Bonk and colleagues have shown that learn-ers had difculty adjusting to BELS environments due to the potential problems in computer and Internet access, learners abilities and

0360-1315/$ - see front matter 2010 Elsevier Ltd. All rights reserved.doi: 10.1016/j.compedu.2009.12.012

* Corresponding author.E-mail address: [email protected] (R.D. Tennyson).

Computers & Education 55 (2010) 155164

Contents lists available at ScienceDirect

Computers & Education

j ou rna l ho mep age : w ww.e l s e v i e r. com/ lo ca t e / c ompedu
http://dx.doi.org/10.1016/j.compedu.2009.12.012mailto:[email protected]://www.sciencedirect.com/science/journal/03601315http://www.elsevier.com/locate/compeduhttp://www.elsevier.com/locate/compeduhttp://www.sciencedirect.com/science/journal/03601315mailto:[email protected]://dx.doi.org/10.1016/j.compedu.2009.12.012

7/31/2019 1-s2.0-S0360131510000035-main

2/10

beliefs in the use of technology, blended course design, participant interaction, and blended environments integration ( Bonk, Olson,Wisher, & Orvis, 2002 ). These ndings imply that an effective BLES environment should consider the human and technology factors thataffect learner satisfactions withBELS, such as individual attitudes, participant interaction, educational technologies, and course design ( Wuet al., 2008 ). Thus, more careful analysis of learners, educational technologies, and social contexts in BELS environments are needed ( EL-Deghaidy & Nouby, 2008 ).

The adoption of BELS in supporting learning has made it signicant to probe the crucial determinants that would entice learners to useBELS and enhance their learning satisfaction. The degree of student learning satisfaction with BELS courses plays an important role in eval-uating the effectiveness of BELS adoption. Hence, comprehending the essentials of what determines student learning satisfaction can pro-vide management insight into developing effective strategies that will allow educational institution administrators and instructors tocreate new educational benets and value for their students. Because BELS environments differ fromtypical classroom and virtual e-learn-ing, a review of previous research in learning technology shows that there is a lack of studies that have examined the crucial factors thatdetermine learning satisfaction with BELS, such as individual cognition, technological environments, and the social contexts, as statedabove. There is a need for more in-depth research to understand what determines student learning satisfaction in a BELS environmentand to investigate how the determinants inuence student perceptions of BELS contexts and their correlations. This study, therefore, pro-poses a research model, based on the social cognitive theory ( Bandura, 1986 ), to investigate the primary determinants affecting studentlearning satisfaction in a BELS environment. We also empirically validate the proposed model and examine the relationships among thoselatent variables.

2. Basic concepts and theoretical foundation

2.1. Blended e-learning system

Blended learning is described as a learning approach that combines different delivery methods and styles of learning. The blend could bebetween any form of instructional technology (e.g., videotape, CD-ROM, CAI, web-based learning) with classroom teaching. Recently therehas beenan increasing movement toward blending e-learning and face-to-face activities withstudents participating in collaborative learn-ing and interaction with their instructors and classmates. This is called blended e-learning or blended e-learning system ( Graham,2006; Singh, 2003 ).

Graham (2006) dened BELS as a mixing of instruction from two historically separate learning environments: classroom teaching andfull e-learning. The termemphasizes the central role of computer-based technologies (e-learning systems) in blended learning, focusing onaccess and exibility, enhancing classroom teaching and learning activities, and transforming the way individuals learn. From a course de-sign perspective, a BELS course can lie anywhere between the continuum anchoredat opposite ends by full face-to-face and virtual e-learn-ing approaches ( Rovai & Jordan, 2004 ). Kerres and De Witt (2003) identied three critical components of BELS that considers the content of the learning materials, the communication between learners and tutors and between learners and their peers, and the construction of thelearners sense of place and direction within the activities that denote the learning environment. This is an important distinction because itis certainly possible to enhance regular face-to-facecourses withonline resources without displacing classroom contact hours. Accordingly,

we dened BELS as the combination of online and face-to-face instruction and the convergence between traditional face-to-face learningand e-learning environments.Several BELSs, such as WebCT ( www.webct.com ) and Cyber University of NSYSU ( cu.nsysu.edu.tw ) have developed systems that inte-

grate a variety of functions to facilitate learning activities. For example, these systems can be used to integrate instructional material (viaaudio, video, and text), e-mail, live chat sessions, online discussions, forums, quizzes and assignments. With these kinds of systems,instructional delivery and communication between instructors and students can be performed at the same time (synchronously) or at dif-ferent times (asynchronously). Such systems can provide instructors and learners with multiple, exible instructional methods, educa-tional technologies, interaction mechanisms or learning resources and applying them in an interactive learning environment toovercome the limitations of classroom and e-learning. As a result, these online learning systems may better accommodate the needs of learners or instructors who are geographically dispersed and have conicting schedules ( Pituch & Lee, 2006 ). As BELS emerge as perhapsthe most prominent instructional delivery solution, it is vital to explore what determines learning satisfaction in a blended e-learningenvironment.

2.2. Social cognitive theory

Social cognitive theory ( Bandura, 1986 ) serves as an initial foundation in this study for exploring what determines student learning sat-isfaction in a blended e-learning environment. Social cognitive theory is a widely acceptedand empirically validated model for understand-ing and predicting human behavior and identifying methods in which behavior can be changed. Several studies have applied it as atheoretical framework to predict and explain an individuals behavior in IS settings. The theory argues that the meta progress of a humanbeing occurs through consecutive interactions with the outside environment and the environment must be subjected to ones cognitionprocess before they affect ones behavior. It proposes that a triadic reciprocal causation among cognitive factors, environmental factors,and human behavior exists. Behavior is affected by both cognitive factors and environmental factors ( Wood & Bandura, 1989 ). Cognitivefactors refer to the personal cognition, affect and biological events. Environmental factors refer to the social and physical environments thatcan affect a persons behavior.

Environments inuence an individuals behavior through his or her cognitive mechanisms. Hence, social cognitive theory posits twocritical cognitive factors: performance expectations and self-efcacy that inuence individual behavior. It gives prominence to the conceptof self-efcacy dened as ones judgments and beliefs of his/her condence and capability to perform a specic behavior recognizingthat our performance expectations of a behavior will be meaningless if we doubt our capability to successfully execute the behavior in the

rst place. It can enhance human accomplishment and well-being, help determine howmuch effort people will expend on a behavior, howlong they will persevere when confronting obstacles and how resilient they will be in the face of adverse situations. The theory further

156 J.-H. Wu et al. / Computers & Education 55 (2010) 155164
http://www.webct.com/http://cu.nsysu.edu.tw/http://cu.nsysu.edu.tw/http://www.webct.com/

7/31/2019 1-s2.0-S0360131510000035-main

3/10

argues that self-efcacy inuences performance expectations andperformance expectations also inuence behavior. Thus, self-efcacy andperformance expectations are held to be the principal cognitive determinants of individual behavior.

Regarding environmental factors, there is ample educational literature and research that shows the learning environment affects alearners behavior and performance. Traditionally, a learning environment was dened in terms of the physical and social environmentsin a classroom setting. Piccoli et al. (2001) expanded the traditional denition of learning environment and identied ve environmentalfactors that clarify how an e-learning environment differs from classroom-based education, including technology, content, interaction,learning model, and learner control. These factors can be classied into two categories that particularly are relevant to BELS-specic envi-ronments. The rst category relates to the technological environment that includes system functionality and content feature. The secondcategory relates to social environments that include interactions (between learners and instructors or between learners and other learners)and learning climate.

3. Research model and hypotheses

Based on the foregoing theoretical underpinnings, we consider that thesocial cognitive theory is applicable to theBELS learning context.Accordingly, three factors: learners cognitive beliefs (self-efcacy and performance expectations), technological environment (systemfunctionality and content feature), and social environment (interaction and learning climate) are identied and elucidated as the primarydimensions of student learning satisfactions with BELS, as shown in Fig. 1.

3.1. Cognitive factors

Cognitive factors refer to the learners cognitive beliefs that inuence their behaviors in using BELS. Two main cognitive variables: com-puter self-efcacy and performance expectations are believed to be the most relevant factors affecting human behavior in using an infor-mation system (IS) ( Compeau & Higgins, 1995; Compeau, Higgins, & Huff, 1999; Venkatesh, Morris, Davis, & Davis, 2003 ). The socialcognitive theory dened performance expectations as the perceived consequences of a behavior and further noted they are a strong forceguiding individuals actions. The performance expectations are derived fromindividual judgments regarding valuableoutcomes that can beobtained through a requisite behavior. Individuals are more likely to perform behaviors that they believe will result in positive benetsthan those which they do not perceive as having favorable consequences.

Performance expectations are dened as the degree to which a learner believes that using BELS will help him or her to attain gains inlearning performance. The denition is similar to the concepts of perceived usefulness, based on Daviss (1989) technology acceptancemodel ( Venkatesh et al., 2003 ). The inuence of performance expectations on individual behavior of using computer systems has beendemonstrated by Compeau and Higgins (1995), Compeau et al. (1999) and Venkatesh et al. (2003) . Prior research in education or com-puter-mediated learning has found that performance expectations are positively related to students learning performance ( Bolt & Koh,2001 ) and satisfaction ( Martins & Kellermanns, 2004; Shih, 2006 ).

Individual attitudes are a function of beliefs, including the behavioral beliefs directly linked to a persons intention to perform a denedbehavior ( Ajzen & Fishbein, 1980 ). User acceptance is an important indicator that measures a users positive attitudes toward the IS andpredicts their behaviors while using the system, based on theory of reasoned action ( Taylor & Todd, 1995 ). Satisfaction is a good surrogatefor user acceptance and is often used to measure learners attitude in computer-mediated learning studies ( Chou & Liu, 2005; Piccoli et al.,2001 ). Thus, we conceptualize the students attitude toward BELS as the learning satisfaction with the BELS dened as the sum of stu-dents behavioral beliefs and attitudes that result from aggregating all the benets that a student receives from using BELS. Therefore, thefollowing hypothesis is proposed.

H1: A higher level of performance expectations for BELS use will positively associate with a higher level of learning satisfaction with BELS.

The second cognitive factorto beappliedin this researchis self-efcacy. Ingeneral,it refers to an individuals beliefs about his orher capa-bilities to successfully perform a particular behavior. According to social cognitive theory, individuals formtheir perceptions of self-efcacytowarda taskbasedoncuetheyreceivefromthefour informationsources:(1)pastexperience andfamiliaritywithsimilaractivities,(2) vicar-ious learning, (3) social support andencouragement, and(4)attitudes towardthe task. Bandura (1986) noted self-efcacyis task-specicandits measures should be tailored to the targeted domain context. Accordingly, several studies have investigated self-efcacy beliefs towardstasks such as computers and IS-related behaviors ( Compeau& Higgins, 1995; Compeau et al., 1999 ). Derived from the general denition of self-efcacy, computer self-efciency was dened as the individual ability to use information technology to accomplish computer-relatedtasks or jobs ( Marakas, Yi, & Johnson, 1998 ). Computer self-efcacy was also validated as a determinant of IS acceptance and use.

We dene computer self-efcacy as the condence in ones ability to perform certain learning tasks using BELS. Prior research hasshown that increases in computer self-efcacy improve initiative and persistence, which lead to improved performance or outcome expec-tations ( Francescato et al., 2006; Johnston, Killion, & Oomen, 2005; Piccoli et al., 2001 ), including attitude and behavioral intention ( Venk-atesh & Davis, 2000 ). In the context of computer-mediated learning, empirical evidence indicates that increases in computer self-efcacyimprove students condence in their computer-related capabilities, which in turn leads to a perception of positive performance expecta-tions to the learning courses ( Bolt & Koh, 2001; Jawahar & Elango, 2001; Santhanam et al., 2008; Shih, 2006 ). That is, computer self-efcacycould reduce learning barriers in using BELS. If students have higher computer self-efcacy and can control BELS, they will perceive thesystems usefulness and value, which in turn motivates their intention to use BELS. Accordingly, the following hypothesis is proposed:

H2: A higher level of individuals computer self-efcacy will positively associate with a higher level of performance expectations for BELS use.

3.2. Technological environment

The quality and reliability of an e-learning system, as well as easy access to appropriate educational technologies, material content, andcourse-related information are important determinants of e-learning effectiveness ( Piccoli et al., 2001 ). Thus, system functionality and

J.-H. Wu et al. / Computers & Education 55 (2010) 155164 157

7/31/2019 1-s2.0-S0360131510000035-main

4/10

content features are identied as critical technological environment factors for BELS. They are expected to inuence the learner to use andaccept BELS. Prior research has shown that system functionality signicantly affected user beliefs in various computer-related contexts(Igbaria, Gamers, & Davis, 1995; Venkatesh & Davis, 2000 ). For instance, research ndings showed that specic system functionality is acritical factor that inuences e-learning system usage ( Hong, Thong, Wong, & Tam, 2002; Pituch & Lee, 2006 ). Pituch and Lee (2006) denedsystem functionality as the perceived ability of an e-learning system to provide exible access to instructional and assessment media.Accordingly, we dene systemfunctionality as the perceived ability of BELS to provide exible access to instructional and assessment med-ia. Such media, for example, allows students to access course materials and content, turn in homework assignments, complete tests andquizzes online.

In general, content is used to identify various divergent formats and types of information. In this study, content refers to technology-based materials and course-related information that may provide value for learners in the context of BELS. BELS achieves its goals of shar-ing and delivering course content through various forms of media such as tutorials, online discussions, or web-based courses. Due to thediversity of deliverymethods, it is a considerable issue that how to design and represent the hybrid content in appropriate formats or typesbest suited to delivery or access by BELS ( So & Brush, 2008 ). Appropriate BELS content features, as well as effective design, representinghybrid course content and transparent content knowledge transfer, are core components of BELS design ( Piccoli et al., 2001 ). Drawingon the previous research ( Zhang, Keeling, & Pavur, 2000 ), we dene content feature as the characteristics and presentation of course con-tent and information in BELS. Text, hypertext, graphics, audio andvideo, computer animations and simulations, embedded tests, and multi-media information are some examples of content features in BELS environment.

System functionality and content feature have the potential to directly affect perceived usefulness of IS ( Hong et al., 2002; Pituch & Lee,2006 ) that are thought to be similar concepts in performance expectation. Several empirical evidences have argued that both content fea-tures ( Zhang et al., 2000 ) and system functionality ( Pituch & Lee, 2006 ) affects the effectiveness of computer-mediated learning. That is tosay, learners perceiving a higher level of systemfunctionality and content features in BELS will lead to a higher level of performance expec-tations for BELS use. In addition, in the BELS environment, the diverse content features can be delivered and accessed depending upon thesupport of appropriate system functionality BELS facilitated ( Pituch & Lee, 2006; So & Brush, 2008 ). Thus, we consider that the content fea-ture highly depends on the power and quality of system functionality of BELS. Therefore, the following hypotheses are proposed:

H3: A higher level of system functionality of BELS will positively associate with a higher level of performance expectations for BELS use.

H4: A higher level of content features in BELS will positively associate with a higher level of performance expectations for BELS use.

H5: A higher level of system functionality in BELS will positively associate with a higher level of content features in BELS.

3.3. Social environment

In computer-mediated instructional design, there is an increasing focus on facilitating human interaction in the form of online collab-oration, virtual communities, and instant messaging in the BELS context ( Graham, 2006 ). From the group interactions perspective, socialenvironment factors, such as collaborative learning ( Francescato et al., 2006 ), learning climate ( Chou & Liu, 2005 ) and social interaction( Johnston et al., 2005 ) are important antecedents of beliefs about using an e-learning system. Prior research ( Pituch & Lee, 2006 ) showsthat social interaction has a direct effect on the usage of an e-learning system. The interactions among students, between faculty and stu-dents and learning collaboration are the keys to learning process effectiveness. In addition, the emotional learning climate is an importantindicator of learning effectiveness.

Interaction is dened in our study as the social interactions among students themselves, the interactions between instructors and stu-dents, and collaboration in a BELS environment. Learning climate is dened as the learning atmosphere in the BELS context. Johnston et al.(2005) argued that contact and interaction with instructors and learners is a valid predictor of performance. A positive learning climateencourages and stimulates the exchange of ideas, opinion, information, and knowledge in the organization that will lead to better learningsatisfaction ( Prieto & Revilla, 2006 ). That is, when learners believe that BELS provides effective student-to-student and student-to-instruc-tor interactions and improves learning climate, they will be more satised with BELS. Therefore, the following hypotheses are proposed:

H6: A higher level of interaction will positively associate with a higher level of performance expectations for BELS use.

H7: A higher level of interaction will positively associate with a higher level of learning climate.

H8: A higher level of learning climate will positively associate with a higher level of learning satisfaction with BELS.

4. Method

4.1. Instrument development

To develop theself-report instrument, a number of prior relevant studies were reviewed to ensure that a comprehensive list of measureswere included. All measures for each construct were taken from previously validated instruments and modied based on the BELS context.


7/31/2019 1-s2.0-S0360131510000035-main

5/10

For instance, the measures for learning satisfaction were selected from Chiu, Hsu, and Sun (2005) and Wu and Wang (2005) . Measures forcomputer self-efcacy and performance expectations were taken from Compeau and Higgins (1995) . The measures for content featurewere adapted from Zhang et al. (2000) and Molla and Licker (2001) . The measures for functionality were taken from Pituch and Lee(2006) . The measures for student and instructor interactions were taken from Johnston et al. (2005), Kreijns, Kirschner, and Jochems(2003) , and Pituch and Lee (2006) . Finally, the measures for the learning climate were selected from Chou and Liu (2005) . Supplementarymaterial lists the denition of each construct, its measures, and the references.

The questionnaire consisted of two major parts including a portion for the respondents basic data and another for the responses to ourresearch constructs. The basic data portion recorded the subjects demographic information (e.g., gender, age, highest education, computerexperiences, and so forth). The second part recorded the subjects perception of each variable in the model. It includes items for each con-struct. All items are measured via a 7-point scale ranging from 1 (strongly disagree) to 7 (strongly agree).

Once the initial questionnaire was developed, an iterative personal interview process with professionals, instructors, and students fromblended learning courses (including four instructors and ve students from three different universities) was conducted to verify the com-pleteness, wording, and appropriateness of the instrument and to conrm the content validity. Feedback from the interview processesserved as the basis for correcting, rening, and enhancing the experimental scales. For example, scale items were eliminated if they rep-resented the same aspects with only slightly different wording and modied if the semantics were ambiguous in order to enhance the psy-chometric properties of the survey instrument. At the end of the pre-test, there were seven constructs with 21 items in total to be used forthe survey.

4.2. Participants

The empirical data were collected using a cross-sectional survey methodology. Participants for this study were students that had the

opportunity to take courses via BELS. We distributed 518 paper-based and online questionnaires to target universities. The target univer-sities were purposivelyselected for the universities or colleges actually implemented BELS courses in Taiwan. Because of the applications of BELS are still at an early stage in Taiwan, the target universities are relatively rare. Data were collected via snowball and convenient sam-pling. Due to the conventional expectation of low survey response rates in survey studies, we endeavored to nd a specic local contactperson for each target university who was placed in charge of distributing the questionnaire. Three hundred and seven-six questionnaireswere returned. Sixty-four responses were incomplete and had to be discarded. This left 212valid responses for the statistical analysis, and avalid response rate of 40.93% of the initial sample. Among the valid responses, 84 responses were received from physical classrooms and128 responses were gathered from online learning environments. The potential non-response bias was assessed by comparing the earlyversus late respondents that were weighed on several demographic characteristics. The results indicated that there were no statisticallysignicant differences among demographics between the early (the rst semester) and late (the second semester) respondents. These re-sults suggest that non-response bias was not a serious concern. The respondent proles and the non-response bias analysis results areshown in Table 1 .

5. Results

Partial least squares (PLS) method was applied for the data analysis in this study. An analytical method is, in general, recommended forpredictive research models emphasized on theory development, whereas Linear Structural Relationships (LISREL) is recommended for con-rmatory analysis and requires a more stringent adherence to distributional assumptions ( Jreskog & Wold, 1982 ). PLS performs a Conr-matory FactorAnalysis (CFA). In a CFA, the pattern of loadings of the measurement items on the latent constructs was explicitly specied inthe model. The t of this pre-specied model is then examined to determine its convergent and discriminant validities. This factorial valid-ity deals with whether the loading patterns of the measurement items corresponds to the theoretically anticipated factors ( Gefen & Straub,2005 ). Convergent validity is shown when each measurement item correlates strongly with its assumed theoretical construct, while dis-criminant validity is shown when each measurement item correlates weakly with all other constructs except for the one to which it is the-oretically associated. The evaluation of the model t was conducted in two stages ( Chin, 1998; Gefen & Straub, 2005 ). First, themeasurement validation was assessed, in which construct validity and reliability of the measures were assessed. The structural model withhypotheses was then tested. The statistical analysis strategy involved a two-phase approach in which the psychometric properties of allscales were rst assessed through CFA and the structural relationships were then validated using bootstrap analysis.

5.1. Measurement validation

For the rst phase, the analysis was performed in relation to the attributes of individual item reliability, construct reliability, averagevariance extracted (AVE), and discriminant validity of the indicators as measures of latent variables. The assessment of item loadings, reli-ability, convergent validity, and discriminant validity was performed for the latent constructs through a CFA. Reective items should beuni-dimensional in their representation of the latent variable and therefore correlated with each other. Item loadings should be above0.707, showing that more than half of the variance is captured by the constructs. The results indicate that all items of the instrumenthad signicant loadings higher than the recommended value of 0.707. As shown in Table 2 , all constructs exhibit good internal consistencyas evidenced by their composite reliability scores. The composite reliability coefcients of all constructs and theAVE in the proposed model(see Fig. 1 ) are more than adequate, ranging from 0.821 to 0.957 and from 0.605 to 0.849, respectively.

To assess discriminant validity: (1) indicators should load more strongly on their corresponding construct than on other constructs inthe model and (2) the AVE should be larger than the inter-construct correlations ( Chin, 1998 ). AVE measures the variance captured by alatent construct, that is, the explained variance. For each specic construct, it shows the ratio of the sum of its measurement item varianceas extracted by the construct relative to the measurement error attributed to its items. As a rule of thumb, the square root of the AVE of

each construct should be larger than the correlationof the specic construct with any of the other constructs in the model ( Chin, 1998 ) andshould be at least 0.50 ( Fornell & Larcker, 1981 ). As the results show in Table 3 , all constructs meet the above mentioned requirements. The

http://-/?-http://-/?-http://-/?-http://-/?-

7/31/2019 1-s2.0-S0360131510000035-main

6/10

values for reliability are all above the suggested minimum of 0.7 ( Hair, Anderson, Tatham, & Black, 1998 ). Thus, all constructs display ade-quate reliability and discriminant validity. All constructs share more variance with their indicators than with other constructs. Thus, theconvergent and discriminant validity of all constructs in the proposed research model can be assured.

Table 1

Respondents prole and the results of non-response bias analysis ( N = 212).

Variables Classication Total (%) Earlyrespondents (%)

Laterespondents(%)

v2 (Sig.)

Gender Male 106 0.500 73 0.344 33 0.156 0.022(0.50)

Female 106 0.500 72 0.340 34 0.160

Age 1830 101 0.476 48 0.453 53 0.500 1.344(0.855)

3140 82 0.387 41 0.387 41 0.3874150 23 0.108 14 0.132 9 0.0855160 4 0.019 2 0.019 2 0.019>61 2 0.009 1 0.009 1 0.009

Types of Jobs Student 8 0.038 3 0.014 5 0.024 4.806(0.440)

Industry 30 0.142 12 0.057 18 0.085Manufacturing 57 0.269 27 0.127 30 0.142Service 10 0.047 5 0.024 5 0.024Finance 59 0.278 36 0.170 23 0.108Others 48 0.226 23 0.108 25 0.118

Education level Senior high school 0 0.000 0 0.000 0 0.000 8.824(0.32)

College (2 years) 10 0.047 1 0.005 9 0.042University (4 years) 116 0.547 60 0.283 56 0.264

Graduate school 86 0.406 45 0.212 41 0.193BELS experience Pure physical classroom experience 15 0.071 7 0.033 8 0.038 0.371(0.946)

Pure virtual classroom experience 42 0.198 20 0.094 22 0.104Physical experience more thanvirtual experience

105 0.495 53 0.250 52 0.245

Virtual experience more thanphysical experience

50 0.236 26 0.123 24 0.113

BELS experience: participating in BELS(years)

4 years 19 0.090 5 0.024 14 0.066

BELS experience: participating in BELS(times)

1 times 44 0.208 24 0.113 20 0.094 4.710(0.452)

2 times 43 0.203 22 0.104 21 0.0993 times 30 0.142 15 0.071 15 0.0714 times 13 0.061 9 0.042 4 0.0195 times 10 0.047 6 0.028 4 0.019P 6 times 72 0.340 30 0.142 42 0.198

BELS experience:spending time in theBELS (1 week)

9 h 6 0.028 4 0.019 2 0.009

Average years of computer usageexperience

11.79 (years) 13.7(years)

10.7(years)

27.076 (0.133)

Table 2

Results of conrmatory factor analysis.

Construct Items Composite reliability AVE

Computer self-efcacy (CSE) 3 0.821 0.605System functionality (SF) 3 0.905 0.761Content feature (CF) 2 0.890 0.802Interaction (I) 3 0.915 0.782Performance expectations (PE) 3 0.940 0.838Learning climate (LC) 3 0.926 0.807Learning satisfaction (LS) 4 0.957 0.849


7/31/2019 1-s2.0-S0360131510000035-main

7/10

5.2. Hypotheses testing

In thesecond phase of the statistical analysis, thestructural model was assessed to conrm to what extent the relationships specied bythe proposed model were consistent with the available data. The PLS method does not directly provide signicance tests and path coef-cient condence interval estimates in the proposed model. A bootstrapping technique was used to estimate the signicance of the pathcoefcients. Bootstrap analysis was performed with 200 subsamples and the path coefcients were re-estimated using each of these sam-ples. The parameter vector estimates was used to compute parameter means, standard errors, signicance of path coefcients, indicatorloadings, and indicator weights. This approach is consistent with recommended practices for estimating signicance of path coefcientsand indicator loadings ( Lhmoeller, 1984 ) and has been used in prior information systems studies ( Chin & Gopal, 1995; Hulland, 1999 ).

Hypotheses and corollaries testing were performed by examining the size, the sign, and the signicance of the path coefcients and theweights of the dimensions of the constructs, respectively. Results of the analysis for the structural model are presented in Fig. 2 . The esti-mated path coefcient (standardized) and its associated signicance level are specied next to each link. The R2 statistic is indicated next tothe dependent construct. The statistical signicance of weights can be used to determine the relative importance of the indicators in form-ing a latent construct. We found that all specied paths between constructs in our research model had signicant path coefcients. Theresults provide support for our model.

One indicator of the predictive power of path models is to examine the explained variance or R2 values ( Barclay, Higgins, & Thomson,

1995; Chin & Gopal, 1995 ). R2

values are interpreted in the same manner as those obtained from multiple regression analysis. They indicatethe amount of variance in the construct that is explained by the path model ( Barclay et al., 1995 ). The results indicate that the model ex-plained 67.8% of the variance in learning satisfaction. Similarly, 37.1% of the variance in content feature, 55.1% of the variance in perfor-mance expectations and 52.9% of the variance in learning climate were explained by the related antecedent constructs. The pathcoefcient from computer self-efcacy to performance expectations is .229 and from interaction to learning climate is 0.727. The magni-tude and signicance of these path coefcients provides further evidence in support of the nomological validity of the research model. Ta-ble 4 summarizes the direct, indirect, and total effects for the PLS analysis.

As for the cognitive factors, Hypotheses H1 and H2, effectively drawn fromcomputer self-efcacy to performance expectations and per-formance expectations to learning satisfaction are supported by the signicant path coefcients, respectively. That is, students who hadhigher computer self-efcacy will have higher performance expectations, which in turn will lead to higher learning satisfaction.

As for the technological environment factors, with the signicant path coefcients, the analysis results also provide support for thehypotheses H3 and H4, effectively drawn from system functionality and content feature to performance expectations. In addition, Hypoth-esis H5, effectively drawn from system functionality to content feature is also supported by the signicant path coefcients. However, it isinteresting to note that the indirect effect of systemfunctionality on performance expectations was stronger than its direct effect (see Table

4). This seems to indicate that system functionality alone may not be sufcient for improving performance expectations when the BELScontent features are not well-matched or designed.

H8

H1

H5

H7

H6

H4

H2

H3

ComputerSelf -efficacy

SystemFunctionality

ContentFeature

PerformanceExpectations

LearningSatisfaction

Interaction

LearningClimate

Fig. 1. The research model for BELS learning satisfaction.

Table 3

Correlation between constructs.

CSE SF CF PE I LC LS

CSE 0.778 a

SF 0.539 0.872CF 0.492 0.609 0.896PE 0.527 0.534 0.596 0.916I 0.389 0.507 0.608 0.662 0.884LC 0.425 0.513 0.593 0.761 0.727 0.898LS 0.44 0.534 0.601 0.798 0.614 0.74 0.921

a The shaded numbers in the diagonal row are square roots of the average variance extracted.


7/31/2019 1-s2.0-S0360131510000035-main

8/10

As for the social environment factors, hypotheses H6 and H7, the paths from interaction to performance expectations and learning cli-mate are supported. That is, interaction apparently inuences the performance expectations and learning climate, respectively. HypothesisH8, effectively drawn from learning climate to learning satisfaction is also supported by the signicant path coefcients. That is, learningclimate inuences learning satisfaction. Overall, both performance expectations and positive learning climate have a direct effect on learn-ing satisfaction; performance expectations provide the greatest contribution (total effect) to learning satisfaction.

6. Conclusion

BELS environments have become the most prominent instructional delivery alternative when employed in e-learning systems. Thisstudy presents a theoretical model that was based on social cognitive theory for investigating the key determinants of student learningsatisfaction in a BELS environment. The results provide strong evidence for the nomological validity of each construct and the effects onlearning satisfaction, as shown in Fig. 2 . The estimate of 0.551 for the performance expectations construct ( R2 = 55.1%) for these paths pro-vides good support for the hypothesized impact of computer self-efcacy, system functionality, content feature, and interaction on thedependent variable, performance expectations. In addition, the estimate of 0.371 for the content feature construct ( R2 = 37.1%) for the pathprovides support for the hypothesized impact of system functionality on the content feature. The 0.529 estimate for the leaning climateconstruct ( R2 = 52.9%) for these paths provides support for the hypothesized impact of interaction on the dependent variable, learning cli-mate. In addition, the 0.678 estimate for the learning satisfaction construct ( R2 = 67.8%) denotes that the learning satisfaction as perceivedby learners is directly and indirectly mediated by the performance expectations and learning climate. Therefore, as a whole, the model hasstrong explanatory power for the student learning satisfaction with BELS.

The signicant path coefcients, effect size and thevalue of the R2 reinforce our condence in the hypotheses testing results andprovidesupport for the association with learning satisfaction in the BELS setting. The results demonstrated that the BELS learning satisfaction isaffected by the interaction among cognitive, technological environment, and social environment factors. We conrmed that technology

alone does not cause learning to occur. It is consistent with the theoretical perspective of social cognitive theory: human behavior as a re-ciprocal interplay of cognitive factors, environment, and behavior ( Bandura, 1986 ).

(R 2 =37.1%)

0.092*

(R 2=67.8%)

0.557***

0.315***

(R 2=52.9%)

(R 2=55.1%)

0.229***

0.727***

0.422***

0.171**0.609***

ComputerSelf-efficacy

SystemFunctionality

ContentFeature

PerformanceExpectations

LearningSatisfaction

Interaction

LearningClimate

* P

7/31/2019 1-s2.0-S0360131510000035-main

9/10

The empirical results indicate that performance expectations and learning climate are two strong determinants of learning satisfactionwith BELS. The computer self-efcacy, system functionality, content feature, and interaction provided an indirect contribution to learningsatisfactionvia the above determinants. Thus, as students become more condent and capable of learning with BELS and more accustomedto the BELS learning environments, they will likely expect more benets fromthe use of BELS, foster positive learning climate, and, overall,be more satised with the BELS learning. These ndings provide initial insights into those factors that are likely signicant antecedents forplanning and implementing BELS to enhance student learning satisfaction. The contributions and implications of this study include thefollowing:

A BELS environment should enhance students performance expectations and foster positive learning climate. Our ndings indicate that per-

formance expectations provide the most contribution to learning satisfaction. This suggests that instructors should take advantage of BELSeffectiveness in designing and teaching courses to enhance students beliefs that they would be able to achieve improved outcomes withBELS. A positive learning climate signicantly affects students learning satisfaction. This suggests that both instructors and learners shouldfoster and motivate the positive learning atmosphere within the BELS learning context. Consequently, if students believe that using BELS isworthwhile, valuable and simple, they will be more likely to accept it resulting in greater satisfaction.

Education institutions should provide incentives and supports to enhance students computer self-efcacy . The empirical results demonstratethat computer self-efcacy have a signicant positive inuence on performance expectations. This implies that learners should have thecomputer competence necessary to exploit BELS and control over his/her learning activities. Therefore, educational institution administra-tors and instructors should provide sufcient incentives and administrative supports to encourage students to actively participate in BELScourses and to enhance their computer self-efcacy. BELS should provide built-in help to t various learners needs in different learningcircumstances.

BELS should offer appropriate system functionality and content features with multimedia presentation and exibility . The results show thatsystem functionality and content features have a positive inuence on perceived expectations. These ndings suggest that: (1) BELS shouldoffer useful informationwithsynchronous and asynchronous learning and content-richdesign that satisfy students needs; (2) BELS shouldprovide various types of content presentation (e.g., multimedia), customized functions to allow learners control over the system, and ex-ible access to t various students learning requirements. It seems reasonable to note that education institutions may offer BELS-relatedtechnical training, awareness programs to the students to enhance students comprehension of BELS.

BELS should provide effective interaction tools and instructors should motivate interaction publicly . The results demonstrate that participantinteraction had a signicant positive inuence on both performance expectations and learning climate. In addition, interaction has themost contribution (total effect) to the performance expectations. These ndings suggest that when implementing BELS courses, the instruc-tors should motivate the positive interaction publicly to increase participant communication and collaborative learning via the system. Ingeneral, learning climate is a function and positive feedback of participant interaction in a BELS environment. A positive learning climatecan make learning easy and natural. Thus, if BELS could support a good social environment to facilitate the students-to-student and stu-dent-to-instructor connectivity interaction (e.g., interactive communication and collaborative learning), learners will be more likely to ac-tively participate in interaction, so as to foster better learning climate and to perceive greater BELS performance expectations and learningsatisfactions.

Although our study provides insights into what determines student learning satisfaction in a BELS environment, it has several limita-tions that also represent opportunities for future research. First, the model was validated using sample data gathered from the target uni-versities in Taiwan. The fact that the participants come from one country limits the generalizability of the results. Other samples fromdifferent nations, cultures, and contexts should be gathered to conrm and rene the ndings of this study. Second, given the self-reportinstrument used (e.g. for measuring computer self-efcacy, system functionality, and content feature), therefore, the typical shortcomingsassociated with self-report measures must be recognized when interpreting the results. Third, this research sets a timely stage for futureresearch in understanding the determinants of learning satisfaction in a BELS environment. It would be interesting to use a longitudinaldesign to examine the relationships among the identied research variables might be a useful extension to the current study. Finally,the results cannot be exhaustive and future works should endeavor to uncover additional determinants of student learning satisfactionwith BELS.

Appendix A. Supplementary material

Supplementary data associated with this article can be found, in the online version, at doi:10.1016/j.compedu.2009.12.012 .

References

Ajzen, I., & Fishbein, M. (1980). Understanding attitudes and predicting social behavior . Englewood Cliffs, NJ: Prentice-Hall.Bandura, A. (1986). Social foundations of thought and action: A social cognitive theory . Englewood Cliffs, NJ: Prentice Hall.Barclay, D., Higgins, C., & Thomson, R. (1995). The partial least squares approach to causal modeling, personal computer adoption anduse as an illustration. Technology Studies,

2, 285309.Bolt, M. A., & Koh, H. C. (2001). Testing the interaction effects of task complexity in computer training using the social cogitative model. Decision Sciences, 32 , 119.Bonk, C. J., Olson, T. M., Wisher, R. A., & Orvis, K. L. (2002). Learning from focus group: An examination of blended learning. Journal of Distance Education, 17 (3), 97118.Chin, W. W. (1998). The partial least squares approach to structural equation modeling. In G. A. Marcoulides (Ed.), Modern methods for business research (pp. 298336).

Mahwah, NJ: Erlbaum.Chin, W. W., & Gopal, A. (1995). Adoption intention in GSS: Relative importance of beliefs. The Data Base for Advances in Information Systems, 26 , 4263.Chiu, C. M., Hsu, M. H., & Sun, S. Y. (2005). Usability, quality, value and e-learning continuance decisions. Computers & Education, 45 , 399416.Chou, S. W., & Liu, C. H. (2005). Learning effectiveness in a web-based virtual learning environment: A learner control perspective. Journal of Computer Assisted Learning, 21 ,

6576.Compeau, D. R., & Higgins, C. A. (1995). Computer self-efcacy: Development of a measure and initial test. MIS Quarterly, 19 , 189211.Compeau, D. R., Higgins, C. A., & Huff, S. (1999). Social cognitive theory and individual reactions to computing technology: A longitudinal study. MIS Quarterly, 23 , 145158.Davis, F. D. (1989). Perceived usefulness, perceived ease of use, and user acceptance of information technology. MIS Quarterly, 13 , 318339.EL-Deghaidy, H., & Nouby, A. (2008). Effectiveness of a blended e-learning cooperative approach in an Egyptian teacher education programme. Computers & Education, 51 ,

9881006.Fornell, C. D., & Larcker, F. (1981). Evaluating structural equation models with unobservable variables and measurement errors. Journal of Marketing Research, 18 (2), 3950.

http://dx.doi.org/10.1016/j.compedu.2009.12.012http://dx.doi.org/10.1016/j.compedu.2009.12.012

7/31/2019 1-s2.0-S0360131510000035-main

10/10

Francescato, D., Porcelli, R., Mebane, M., Cuddetta, M., Klobas, J., & Renzi, P. (2006). Evaluation of the efcacy of collaborative learning in face-to-face and computer-supporteduniversity contexts. Computers in Human Behavior, 22 , 163176.

Gefen, D., & Straub, D. W. (2005). A practical guide to factorial validity using PLS-graph: Tutorial and annotated example. Communications of the Association for InformationSystems, 16 (5), 91109.

Graham, C. R. (2006). Chapter 1: Blended learning system: Denition, current trends, future directions. In C. J. Bonk & C. R. Graham (Eds.), Handbook of blended learning . SanFrancisco, CA: Pfeiffer.

Hair, J. F., Anderson, R. E., Jr., Tatham, R. L., & Black, W. C. (1998). Multivariate data analysis with readings (5th ed.). NJ: Prentice Hall.Hara, N., & Kling, R. (2000). Students distress with a web-based distance education course: An ethnographic study of participants experiences. Information, Communication

and Society, 3 , 557579.Hong, W., Thong, J. Y. L., Wong, W. M., & Tam, K. Y. (2002). Determinants of user acceptance of digital libraries: An empirical examination of individual differences andsystem

characteristics. Journal of Management Information Systems, 18 (3), 97124.Hulland, J. (1999). Use of partial least squares (PLS) in strategic management research: A review of four recent studies. Strategic Management Journal, 20 , 195204.Igbaria, M., Gamers, T., & Davis, G. B. (1995). Testing the determinants of microcomputer usage via a structural equation model. Journal of Management Information Systems,

11 (4), 87114. Jawahar, I. M., & Elango, B. (2001). The effect of attitudes, goal setting and self-efcacy on ender user performance. Journal of End User Computing Practice, 13 (2), 4045. Johnston, J., Killion, J., & Oomen, J. (2005). Student satisfaction in the virtual classroom. The Internet Journal of Allied Health Sciences and Practice, 3 (2). Jreskog, K. G., & Wold, H. (1982). The ML and PLS techniques for modeling with latent variables: Historical and comparative aspects. In K. G. Jreskog & H. Wold (Eds.),

Systems under indirect observation: Causality structure and prediction (pp. 219243). Amsterdam: North Holland.Kerres, M., & De Witt, C. (2003). A didactical framework for the design of blended learning arrangements. Journal of Educational Media, 28 (2/3), 101113.Kinshuk, D., & Yang, A. (2003). Web-based asynchronous synchronous environment for online learning. United States Distance Education Association Journal, 17 (2), 517.Kreijns, K., Kirschner, P., & Jochems, W. (2003). Identifying the pitfalls for social interaction in computer-supported collaborative learning environments: A review of the

research. Computers in Human Behavior, 19 , 335353.Lhmoeller, J. B. (1984). LVPS 1.6 program manual: Latent variable path analysis with partial least squares estimation. Universitaet zu Koehn, Zentralarchiv fuer Empirische

Sozialforschung.Maki, R. H., Maki, W. S., Patterson, M.,& Whittaker, P. D. (2000). Evaluation of a web-basedintroductorypsychology course: Learning and satisfaction in on-line versus lecture

courses. Behavior Research Models, Instruments, and Computers, 32 , 230239.Marakas, G. M., Yi, M. Y., & Johnson, R. D. (1998). The multilevel and multifaceted character of computer self-efcacy: Toward clarication of the construct and an integrative

framework for research. Information Systems Research, 9 , 126162.Martins, L. L., & Kellermanns, F. W. (2004). A model of business school students acceptance of a web-based course management system. Academy of Management Learning and

Education, 3 , 726.Molla, A., & Licker, P. S. (2001). E-commerce systems success: An attempt to extend and respecify the DeLone and McLean model of IS success. Journal of Electronic Commerce

Research, 2 (4), 111.Osguthorpe, R. T., & Graham, C. R. (2003). Blending learning environments: Denitions and directions. The Quarterly Review of Distance Education, 4 (3), 227233.Piccoli, G., Ahmad, R., & Ives, B. (2001). Web-based virtual learning environments: A research framework and a preliminary assessment of effectiveness in basic it skills

training. MIS Quarterly, 25 , 401426.Pituch, K. A., & Lee, Y. (2006). The inuence of system characteristics on e-learning use. Computers & Education, 47 , 222244.Prieto, I. M., & Revilla, E. (2006). Formal and informal facilitators of learning capability: The moderating effect of learning climate. IE working paper. WP06-09. 21-02-2006.Rovai, A., & Jordan, H. (2004). Blended learning andsense of community: A comparative analysis with traditional and fully online graduate courses. The International Review of

Research in Open and Distance Learning, 5 (2), 112.Santhanam, R., Sasidharan, S., & Webster, J. (2008). Using self-regulatory learning to enhance e-learning-based information technology training. Information Systems Research,

19 , 2647.Shih, H. P. (2006). Assessing the effects of self-efcacyand competence on individual satisfaction with computer use: An IT student perspective. Computers in Human Behavior,

22 , 10121026.Singh, H. (2003). Building effective blended learning programs. Educational Technology, 44 (1), 527.So, H. J., & Brush, T. A. (2008). Student perceptions of collaborative learning, social presence and satisfaction in a blended learning environment: Relationships and critical

factors. Computers & Education, 51 , 318336.Taylor, S., & Todd, P. A. (1995). Decomposition and crossover effects in the theory of planned behavior: A study of consumer adoption intentions. International Journal of

Research in Marketing, 12 , 137156.Venkatesh, V., & Davis, F. D. (2000). A theoretical extension of the technology acceptance model: Four longitudinal eld studies. Management Science, 46 , 186204.Venkatesh, V., Morris, M. G., Davis, G. B., & Davis, F. D. (2003). User acceptance of information technology: Toward a unied view. MIS Quarterly, 27 , 425478.Wood, R., & Bandura, A. (1989). Social cognitive theory of organization management. Academy of Management Review, 14 , 361384.Wu, J. H., Tennyson, R. D., Hsia, T. L., & Liao, Y. W. (2008). Analysis of e-learning innovation and core capability using a hypercube model. Computers in Human Behavior, 24 ,

18511866.Wu, J. H., & Wang, S. C. (2005). What drives mobile commerce?An empirical evaluation of therevised technology acceptance model. Information & Management, 42 , 719729.Yang, Z., & Liu, Q. (2007). Research and development of web-based virtual on-line classroom. Computers & Education, 48 , 171184.Zhang, X., Keeling, K. B., & Pavur, R. J. (2000). Information quality of commercial website home pages: An explorative analysis. In Proceeding of the 21th international

conference on information systems (pp. 164175). Brisbane, Australia.


1-s2.0-s0360131510000035-main

Documents