An empirical study of the relationship among communication, trust and performance of undergraduate course learning team: Based on structural equation model

Abstract During the past decade, team work emerged as an important research area in business management, focusing on the enterprise. However, publications on course learning team are very few. In China, team-learning course practice is one popular and important teaching and learning reform in high schools. In this study, we examine the team-learning course practices from the pedagogical perspective of learning performance. We want to know: Can the team communication, team trust influence the college students’ course learning team’s performance? If do, to what degree do they affect team learning? How can the useful practices be enhanced further? We drew theory from organizational learning and team working literatures to carry on our investigation. The data was collected using a set of questionnaires distributed to six different colleges in China in 2021. We got 404 copies valid, and the data were analyzed using SEM. Results showed the communication frequency, the feedback degree and the learning degree all have a significant positive influence on team performance. Team cognitive trust, institutional trust, arithmetic trust and relational trust also have a significant positive effect on learning performance; yet the degree of these influences differs sharply. The teams’ performance is influenced by team communication and team trust. We recommend enhancement of the course reform practice by team communication and team trust. Teachers who had put team-learning model into practice yet had little confidence in its efficiency can use this to help them to guide the course learning team in a better way.


Introduction
In 2012, the Organization for Economic Cooperation and Development (OECD) published a research report entitled "Educating Teachers for the 21st Century and Improving School Leadership: Lessons from the World", which argues that one of the most important skills students in the 21st century must possess is to work in teams (OECD, 2012). And the next year, UNESCO issued the No. 1 research report of "Learning Indicators Task Force" (LMTF), which emphasized the necessary to educate the students using the team learning method. Nowadays, in the era of "double first-class construction" one of the strategic aims of higher education reform in China is to cultivate students to develop the talents of team cooperation (Sun Yong-zheng et al., 2015).
It is in this context that the research of higher educational teaching and learning reform focused on team learning has emerged. However, we found out that, from the perspective of theoretical research, the research conducted focused on enterprises-teams or scientific research and innovation teams, there were few studies on the college students' course learning teams; and from the perspective of practical research, the relevant research was mostly practical or qualitative research, the theoretical quantitative studies were scarce. Considering such particularity of college students' course learning team, as the team members link with a course, and the team will disband once the course ends, it has great differences from those of the scientific researching and innovating teams of enterprises or universities, so the theory applicable to those enterprises or university teams is not necessarily suitable for the course learning teams. On the other hand, the practice of the course learning team begins before theory investigation, the effect of the course team as well as the team learning performance is still under question, evaluating the effect and efficiency of the course learning team reform becomes urgent.
Since McGrath formed the Input-Process-Output (I-P-O) model in 1964, it is still influencing the way studying team work (Hackman, 1983). During the advancing process of the model, many scholars adopted some relevant factors to the model, functioning as variables of input, process and output. Among them, the model developed by Cohen & Bailey (Cohen & Bailey, 1997) was most attractive. It expanded the content of team work's psychological factors, and it emphasized the psychological factors can affect the team's performance independently or dependently with team interaction. But, this model was only a systematic general discussion, and no further empirical verification has been made. Therefore, studying the undergraduate course learning team, to find out who influences the team performance become a special contribution of the study to the general discussion. The quantitative research method adopted in the study will fan the fire of the other following investigations. And it will provide a theoretical support for the related education and teaching reform which is of great practical significance to promote further reform practice based on team-learning.

Theoretical foundation and hypotheses
In order to facilitate the study, based on the comprehensive consideration of the duration and the purpose of the team, this study gave the course learning team a specific definition which differs from other types of teams, which includes the following features: the team is established for the study of a certain course; the team exists not longer than one semester; the team members come from the same university; the team disbands at the end of the course.

Team communication and team performance
In the process of teamwork, team members need to listen attentively to each other and express their feelings and opinions about how the team works and what problems the team faces. Team communication plays a vital role of the teamwork (LArgote et al., 2001). it is particularly important for an efficient team to form a good team communication (MR Barrick et al., 1998).
According to Hackman's definition of team performance, it refers to the actual results of the team to achieve the predetermined goals, including the output of the team, the impact of the team on members and the improvement of team members' working ability (Hackman, 1987). There were two methods to measure team performance provided by current research, which is objective indicators and subjective indicators. Objective indicators refer to the use of objective data or quantifiable indicators such as sales and product amounts, while subjective indicators refer to the subjective judgment of individuals or teams made by team members or team leaders (Yang Xiao-feng, 2018).
The research on the relationship between team communication and performance began with Dale, who believed that employees' satisfaction with organizational communication would have a significant impact on the overall performance of the organization. Since then, many scholars did more research, most of which showed that effective communication had a positive effect on team performance. Xie Li-ning, a Chinese scholar, believed that team communication was the entire process of team members' information sharing, problem solving and effective cooperation, its aim was to create high performance (Xie Li-ning, 2007). Yang Ya-qin and other scholars believed that the degree of communication had an impact on team performance, and the team under the condition of sufficient communication performed better than that of the insufficient communication (Yang Ya-qin, 2008).
In those empirical researches on team communication and team performance, Peter, Richard, and Jack believed that a successful team requires effective interaction among members, and with the improvement of team communication quality, team interaction performance will also be enhanced (Dawson et al., 2008). When studying team communication and team performance, Wang Hai-xia also confirmed that team communication had a significant impact on team performance (Wang Hai-xia, 2009).
Based on the finding mentioned above, we guessed that in the course learning team a similar influence will occur, since team communication not only helps to establish good interpersonal relationships, but also helps to transfer and share knowledge among team members and enhance the ability of team members to learn the course; if the communication among team members is not sufficient, it will affect the effect of team learning. Thus, we formed the following hypotheses about the impact of course learning team communication on learning performance:

Team trust and team performance
Team trust, as the foundation of effective team management, affects the efficiency and quality of the team to a great extent, while team performance can also reconstruct team trust with latest ideas and values. Team performance and team trust can influence each other, they are interdependent. Bart A. De Jong & Tom Elfring, Qin Kai-yin, Zhao Xi-ping, and other scholars had shown that the establishment of team trust was conducive to the improvement of team performance (De Jong & Elfring, 2010;Qin Kai-yin et al., 2010;Zhao Xi-ping et al., 2008). However, some studies also showed that team trust had no direct impact on performance (Peng Lian-gang, 2011).
Due to different perspectives, team trust has different meanings and dimensions. Based on the findings of Rousseau, Zhao Hang and other scholars (Rousseau D M et al., 1998;Zhao Hang, 2016), in this study, the team trust referred to the general expectation of all team members and the overall perception of the credibility of the team, including four dimensions, including arithmetic trust, cognitive trust, relational trust, and institutional trust.
Arithmetic trust refers to the expectation based on rational choice and prevention of damage to one's own benefits, and cognitive trust refers to the expectation of the reliability and credibility of the trusted person after the evaluation of his or her professional skills and expertise by the members of the team through long-term collaboration. Relational trust refers to the trust established by team members based on the specific emotional division of "insiders", and institutional trust refers to the trust formed according to the comprehension on the same curriculum regulations, team culture, social norms and so on, it is the team member's overall perception of the team's dependability and reliability.
We guessed that higher team trust leads to better team learning performance in a course learning team. The following were our hypotheses.
Hypothesis Hb1: relational trust can positively promote team performance.
Hypothesis Hb2: arithmetic trust can positively promote team performance.
Hypothesis Hb3: institutional trust can positively promote team performance.
Hypothesis Hb4: cognitive trust can positively promote team performance.

Structural equational model of the study
Structural equations modeling (SEM) is an important tool for education researchers. Although partial least squares (PLS) is always a powerful tool, it is essentially a combination of principal components and path models (Fornell & Bookstein, 1982), it was not useful for theoretical development or testing (McDonald, 1996). SEM, on the other hand, is a combination of factor analysis and path modeling. It has been found to be relatively robust and is generally endorsed for most uses (Hu & Bentler, 1998;Olsson et al., 2000). Although non-normality has an effect on the parameter estimates when sample size falls below 100 (Lei & Lomax, 2005), this shortcoming can be overcome if we have enough samples. So in our study, we used SEM, sticking to the comments made by Dawn Iacobucci .
According to the hypotheses listed above, we present our conceptual model, which is intended to explain the relationships and influences among college students' course team communication, team trust and team performance. The model includes two higher-order factors-team trust and team communication-composed of three first-order indicators (each of which, in turn, has a set of reflective indicators), as well as their interaction.

Variable measurement and data collection
The team communication scale used in this study was adopted from the research scales of Chang Hua and Huang Min-ping. It was made of three constructs, namely communication frequency, feedback degree and learning degree. It had ten questions (Chang Hua, 2008;Huang Min-ping et al., 2003).
The team trust scale used in this study consists of four subscales, namely relational trust, arithmetic trust, institutional trust, and cognitive trust. The arithmetic trust subscale shared a light from the research results of Lewicki et al. after modifying some items of their scale according to the specific situation, we got six items to measure it (J & B, 1995).
The cognitive trust scale was based on the cognitive trust scale developed by McAllister et al., with six items (McAllister, 1995).
Based on the findings of Chu Hao-nan's team (Chu Hao-nan, 2008), considering the revision suggestions of some subsequent researchers on the scale (He Bin, 2013), the relational trust scale retains six items.
According to the localization research of Moorman team and Wang Huai-qiu, the institutional trust scale was made of six items (Moorman et al., 1993;Wang Huai-qiu, 2008).
The team performance scale used in this study was compiled by Barker, J et al., it had eleven items (Barker et al., 1988).
The rating method of the scale mentioned above is a seven-degree Likert measurement. The higher the score, the stronger the willingness of the testers.
Prior to collection the data in 2022, we back translated the foreign scales items with ten teachers who organized a course learning team. Also, we performed a pilot study with 70 students who had the experience of being a course learning team member recently to assess the research design's quality. These steps resulted in some changes being made, mainly to the advices to respondents. To get quality data from key informants, our survey was developed using Wenjuanxing, an online crowdsourcing platform in mainland China, which provides functions equivalent to Amazon Mechanical Turk, recruited 425 participants.
We restricted our sample to college students who had the experience of joining a course learning team last semester and instructed respondents to focus on the last team learning's performance. We dropped those who failed the attention check whose time answering the whole scale was within 60 seconds which was the lower 27% of the 425 participants' answering time and it was regarded as an indices that they paid no due attention to the questions in the scale (n = 13). We also dropped those who answered the scale using the same ID which meant he or she answered the same questionnaire twice (n = 8). A total of 404 participants constituted our final sample. Table 1 and 2 present the results of the measurement assessment. Table 1 summarizes the variables means, standard deviations, correlations, and shared variances. Table 2 reports the average variance extracted (AVE), construct reliabilities (CR), factor loadings and fit indices. Established scales were used to measure team trust (cognitive trust, institutional trust, arithmetic trust and relational trust), team communication (communication frequency, feedback degree and learning degree), and team performance.

Measures
The correlations are included in the lower triangle of the matrix. All correlations ≥0.14 are significant at the p < 0.05 level. Share variances are included in the upper triangle of the matrix.

Fit of the measurement model
The model fit was evaluated using a series of indices recommended by Bagozzi (Bagozzi & Yi, 1988;Bagozzi et al., 1991). First, we conducted a confirmatory factor analysis (CFA) in order to establish confidence in the measurement model, which specifies the posited relations of the observed variables to the underlying constructs. In the measurement model, there are three variables for communication frequency, four variables for feedback degree, three variables for learning degree, six variables for cognitive trust, constitutional trust, arithmetic trust and relational trust each, and eleven variables for team performance. And we found some constructs' measuring model's goodness-of-fit indices (GFI) was less than 0.9. Considering the fact that there may exist a large semantic gap between the items cited from the ancestor's scale and the measurement items we used in the course learning team study, it was natural to find some items caused the poor GFI. So, to improve the model-fit, we deleted some items whose factor loadings were less than 0.7. Then, we had a second CFA, again we found some problems within the specified measurement model. To satisfy the good model's criteria that the component reliability (CR) should not be less than 0.6, the average variance extraction (AVE) should not be less than 0.5, and the multivariate correlation square (SMC) should be greater than 0.5, we removed some other items and corrected each model, using Amos' modification indices. The final number of each latent factors' indicators, convergent validity and model fit indicators are shown in Table 2.
The SMC of all constructions except the construction of the communication frequency was higher than 0.5, all CR values were higher than 0.6, and all AVE values were higher than 0.5. The chi-square/degree of freedom (CMIN/DF) of each measurement model was less than 3 (Chin & Todd, 1995). Other indices such as GFI, adjusted good-of-fit index (AGFI), and root mean square error of approximation (RMSEA) all reached the ideal standard, so all the measurement models proved to be perfect.
Although the SMC value of one measurement index in the communication frequency constructions was only 0.29, lower than 0.5, but according to Bollen's principle that each construction should have at least three items (Bollen K A, 1989), considering also about the fact that the CFA model was well matched, it was reasonable not to delete anyone of the items in the communication frequency construction. Therefore, we finally retained three variables for communication frequency, three variables for feedback degree, three variables for learning degree, five variables for cognitive trust, five variables for institutional trust, four variables for arithmetic trust, six variables for relational trust each, and eight variables for team performance.
The item and construct reliability are tested. Indicators included in the analysis are reliable and values for CR are above the critical limit of 0.6 (see, Table 2). According to AVE, communication construct is close to the critical limit of 0.50; while all the other constructs have good reliability. All T-values of the loadings of measurement variables on the respective latent variables are statistically significant and convergent validity is thus supported (see, Table 2). The results of the CFA show that the hypothesized measurement model fits the data reasonably well and the overall fit indices are appropriate.

Hypotheses assessing
Following the proposed measurement model of this study, a structural equation model (SEM) is developed in order to test if the hypothesized theoretical model is consistent with the collected data. The model includes the endogenous latent variable team performance and the exogenous latent variables of team communication and team trust. According to the proposed structural model, team performance is explained through both team communication and team trust (see, Figure 1). It is useful to report the percentage of variation in the endogenous constructs accounted for by the exogenous constructs or R 2 for each structural equation. The exogenous variable of team communication can explain 83% of the variation in team performance. For team trust, the model explains 58.6% of the variable in team performance. A model might explain significant amounts of the variation in endogenous variables but not fit the data well. The proposed SEM is acceptable also according to the global fit measures and given the high power of the model. The application of SEM has been growing in the social sciences since it provides researchers with ample means for assessing and modifying relationships among examined constructs and offers great potential for furthering the development of theory (Anderson & Gerbing, 1988). To assess whether the structural equational model (SEM) can explain the observed data, the SEM's model fit test was conducted. In the SEM of team trust and team performance, the fitness indices of the model, CMIN/DF was 3.29, less than 5; GFI was 0.93, greater than 0.9; AGFI was 0.90, greater than 0.9; RMSEA was 0.075, less than 0.08; comparative fit index (CFI) was 0.98, more than 0.9; standardized root mean square residual (SRMR) was 0.024; In the team communication and performance structure model, CMIN/DF = 2.86, GFI = 0.95, AGFI = 0.92, RMSEA = 0.068, CFI = 0.98, SRMR = 0.021. Therefore, all the indices reached the ideal level, indicating that the SEM was acceptable and reasonable.
Then, the hypotheses of the SEM are tested by examining the unstandardized and standardized regression coefficients (which can show the degree and direction of the influence), and statistical significance of the structural coefficients (see , Table 3). Parameter estimates for the relationships of the team performance with the team communication and team trust are statistically significant and consistent with the proposed direction in the hypotheses. Different constructions of team communication and team trust have different degrees of influence on team performance, among which communication frequency and cognitive trust had the greatest impact, yet the feedback degree and the institutional trust had the weakest impact on team performance.

Team trust had a significant positive impact on team performance
It was consistent with the results of empirical research by Song Hua and Wang Lan, scholars of Renmin University of China (Hua & Lan, 2009). There existed a positive relationship between team trust and team learning effect. For the course learning team, the relationship based on the trust of each other created the foundation for team course learning, and they need not to worry about those opportunistic behavior which might occur when trust does not exist. This enhanced the team's learning. Without team trust, the members in a course learning team might worry about the risk of losing the advantage if he or she chose to share the knowledge he or she learned while other members shared nothing, and this worry will weaken its will to share something valuable, it caused the knowledge transferring and sharing behavior among members became scarce. Therefore, team trust was conducive to team learning.
The formation and development of team trust is a relatively slow process. Through long-term contact or the rational calculation and emotional commitment of interpersonal relationships, the recognition of team members' personal ability, goodwill, integrity and the expectation of organizational system are gradually established (Jones & George, 1998). Yet in a temporary course learning team, the goal of students was to complete specific course learning tasks rather than establish social relations, they tended to treat team members from the perspective of professional knowledge rather than social relationships. Therefore, in a course learning team, the establishment and maintenance of trust depends more on cognitive-based trust than emotion-based trust, and this was theoretically supported by the conclusion that cognitive trust is the most influential factor in team learning effect.
On the other hand, the temporary nature of the course learning team means that team members must have a higher level of trust than those of traditional teams to help them successfully complete the course learning. Therefore, course lecturers should try their best to establish a reasonable and scientific course assessment system, reduce or eliminate the occurrence of speculative behavior, hitchhiking and other similar behaviors in the course team, to improve the level of students' institutional trust and arithmetical trust in team, and provide better conditions for team course learning.

Team communication had a significant positive effect on team performance
It was consistent with the findings of E KASL et al., ROB CROSS et al., He Jian-hua et al., (E KASL et al., 1997;ROB CROSS et al., 2008). If the members of the course team can communicate frequently, the knowledge sharing behavior among the team members will increase, the tacit understanding among the members will improve, and the team atmosphere will be more harmonious, all these will lead to increase the team performance; Feedback is also an indispensable part in the construction of course learning team, continuous feedback among team members is helpful to solve team problems efficiently and improve team performance.
The learning degree was also an especially crucial factor influencing the team performance. This finding can give theoretical guideline for those collaborative course reforming practices, that is, different from the traditional individual-oriented course learning behavior, collaborative course learning behavior occurs not only between teachers and students, but also among team members themselves. It is this extended learning behavior that improves the team's learning performance. Therefore, when managing the course learning team, it was wiser for the teacher to focus on creating a good class atmosphere, to provide more platforms and time to promote more instant and frequent communication and feedback between students and teachers, and at the same time, reasonably enhance the curriculum's requirements for ability and quality, not just for knowledge, push the team to learning more extensively and deeply.
As for the learning degree, we found something different from Jiang Xiao-nuan's conclusion on knowledge team in colleges and university. In the course learning team, the learning level had a significant positive effect on learning performance, yet in university knowledge teams, an elevated level of learning did not mean that team performance will improve (Jiang Xiao-nuan, 2011). Why? The reason might be this: in a course learning team, to complete the course learning task successfully together, it is necessary for the team members to learn and share common knowledge. In the process of knowledge learning, it also contained the characteristics of knowledge itself, such as the learning scene and the learning subject, of which the most important characteristic was the learning subject. Since it is the whole team's contribution, not the single member's contribution that matters to the course learning effect, team members' learning degree and the will to share knowledge with other team members had an important impact on the teachers' assessment of the team course learning performance. Therefore, course learning team's high learning degree can directly improve the learning performance, besides, the high learning degree can also indirectly improve the learning performance since they can reach the higher level of learning, which includes cultivating the ability and quality to conduct complex tasks with the enhanced learning.
Yet this model has its weakness. First, it only focused on two constructs yet neglected other factor which might also influence the team's learning performance. Under some condition, the team's learning performance might also be influenced by several factors interacting together, nor by this certain way. Second, this study was carried out in the special culture context of China, whether it can be adoptable in other country, it still needs more studies. So to further this investigation, we recommend we should pay attention to other factors which might have significant effects on the college students' course learning team's performance, and we also should carry out some studies concentrating on the theoretical model comparison in different countries.

Funding
This project is supported by (Grant No