Accompanying the application of information and network technology, blended learning is becoming more popular in colleges and universities. Apart from the changes in the traditional face-to-face learning mode, blended learning makes full use of the convenience and rich resources of the internet, and combines this with the advantages of traditional learning. It is a new learning mode that reinforces teaching by computer and network online activities in traditional courses (Benbunan-Fich, 2008). The adoption of blended learning symbolizes the restructuring of curriculum design, which is intended for mobilizing students’ initiative in participating in online learning.

Some studies have proven that integrating information technology into the teaching and learning process creates the acquisition of course resource information and the improvement of the learning experience (Bai et al., 2016; Darling-Aduana and Heinrich, 2018; Turvey and Pachler, 2020). Blended learning methods are also able to significantly enhance the learning experience. With a higher course satisfaction rate than traditional classroom teaching (Darling-Aduana and Heinrich, 2018), the learning method makes students more devoted to the learning process (Yildirim, 2005), and thus it is preferred by students for its greater flexibility and convenience (Hogarth, 2010). A blended learning course design is a sophisticated topic that involves numerous factors, among which learning experience is a predominant factor.

Learning performance can be defined in different ways. For instance, it can refer to students’ test scores (Ferguson and DeFelice, 2010; Ekwunife-Orakwue and Teng, 2014; Law and Geng, 2019), satisfaction with learning (Ekwunife-Orakwue and Teng, 2014; Yuan et al., 2020), or even their performance logged in the online learning system (Yang et al., 2016). This research adopts the definition of learning performance as described by the Association for Educational Communications and Technology in 2004, which states that learning performance is the ability of a learner to apply the newly acquired knowledge or skills. In essence, it does not solely involve the basic knowledge and skills learned, but the capability to apply them. There are many factors for learning performance (Broadbent, 2017; Li and Tsai, 2017; Li et al., 2018; Wei and Chou, 2020). In blended learning, COI is an essential theoretical model. With a focus on PPT, this research combines the dimensions in COI, such as teaching presence, social presence, and cognitive presence to investigate the relationship with learning performance. The research also examines the link between these dimensions and other learning characteristics.

Originating from America in the 1960s, precision teaching has been propagated in Britain by education psychologists (EPS) since the 1980s (Muncey and Williams, 1981; Raybould and Solity, 1982, 1988). It essentially pertains to a method of teaching that monitors students’ acquisition of basic education skills. Through monitoring students’ learning process results, teachers are capable of adjusting their teaching and intervention measures to ensure the optimal learning performance of students (Rebecca and Michelle, 2016). Precision teaching involves four basic steps: Pinpoint, Record, Change, and Try Again (Kubina and Yurich, 2012).

Precision teaching, as an effective method to reinforce students’ acquisition of basic education skills (Chapman et al., 2005; Gist and Bulla, 2020), can stimulate students’ learning motivation and get students more engaged in learning (Rebecca and Michelle, 2016). Precision teaching resounds to the improvement of learning performance (Datchuk, 2017; Mannion and Griffin, 2018).

Perceived precision teaching is defined as the learners’ perception of precision teaching during their study, including the perception of the four basic steps, that is, deciding targeted learning goals, recording the learning process, teaching with purpose, and re-examining the attainment of goals (Pinpoint, Record, Change, and Try Again). PPT can be measured by surveying learners (Dibbs et al., 2013; Rebecca and Michelle, 2016).

Self-efficacy indicates the individual belief in the success of performing a task (Bandura and Adams, 1980), which directly affects an individual’s behavioral motivation in the implementation of a specific activity. Generally speaking, a successful experience consolidates self-efficacy, while recurrent failures damage self-efficacy.

Self-efficacy is often divided into universal self-efficacy and task-based self-efficacy (Sherer et al., 1982), of which the latter is suited to given tasks and situations and is the dominant type of blended learning. Self-efficacy determines the content and essence of learners’ inner imagination about future learning scenes or procedures, and directly or indirectly affects individual psychological momentum while performing learning activities, thereby generating an impact on real-life learning activities and forming strong associations with learning performance (Bandura and Watts, 1995; Chamorro-Premuzic et al., 2010; Schneider and Preckel, 2017).

According to Ford (1992), learning motivation is defined as a psychological mode in pursuit of goals, beliefs, and emotions. Motivation refers to something that encourages, instructs, and maintains behaviors, approving of students’ devotion to a given direction and persistence in it (Fredricks et al., 2004; Reeves, 2006).

Learning motivation is a key factor influencing learning performance. It is not only a decisive factor in learning performance. Actually, it corresponds to a matched motivation in every task to be accomplished (Weiner, 1990). There are two types of learning motivation, namely extrinsic learning motivation and intrinsic learning motivation. In particular, intrinsic motivation is the dominant type of student learning in blended learning, which implies that students motivated by intrinsic motivation have better task accomplishment learning performance than those motivated by extrinsic motivation (Lin et al., 2003). Numerous studies emphasize the importance of motivation, given its effect on learning performance (Law et al., 2010; Law and Breznik, 2017; Law and Geng, 2018).

Community of inquiry is widely considered as a model on how members in a society acquire knowledge and solve problems (Garrison et al., 2000). Its core lies in the education experience influenced by the interaction between social, cognitive, and teaching presence. These three types of presence are paramount to the education-oriented COI and can improve or constrain education experience and learning outcomes (Garrison et al., 2000). The framework is shown in Figure 1.

Cognitive presence is the student’s ability to construct meanings through discussion and reflection in COI, which demonstrates the process of learning and investigation (Garrison, 2011). It entails trigger events, exploration, integration, and reflection, as well as the resolution of learning problems. In the triggering stage, learning tasks can be generated based on doubts about certain knowledge so that students can be prompted into exploration. In the second stage, students can exchange information and engage in critical thinking and investigation in the learning community. In the integration stage, students construct meanings based on the reflection and thoughts from the exploration stage. Finally, students solve their problems in the learning tasks by directly or indirectly using the meanings constructed from the last stage. For the formation of cognitive presence, COI provides an enabling environment to establish and confirm meanings through constant reflection and critical discourse (Garrison et al., 2000).

Social presence means the ability for individual students to get in touch with the community, conduct meaningful communication in a trustworthy environment, and develop interpersonal relationships by demonstrating individual characteristics (Garrison, 2011). To achieve social presence, the model must afford an open communication environment that is useful in fabricating the favorable relationship between team cohesiveness and individuals. Students may express their thoughts and emotions with a communication method popular in the community (Garrison et al., 2000). Social presence presents a sense of belonging and also a mechanism of free expression to students and sustains team cohesiveness. As a supporting function of cognitive presence, it indirectly promotes the critical thinking process of learners in the community.

Teaching presence is defined as “the design, facilitation, and direction of cognitive and social processes for the purpose of realizing personally meaningful and educationally worthwhile learning outcomes” (p.3) (Anderson et al., 2001). Teaching design involves designing courses and teaching methods and organizing relevant discussions. Facilitation means to create a learning environment that encourages people to share their personal views for effective communication. Direction is to collectively discuss and resolve problems. In essence, teaching presence is the atmosphere of teaching that combines social and cognitive presence effectively and efficiently (Garrison, 2011).

There are four steps in precision teaching, including deciding targeted learning goals, recording learning processes, teaching with purpose, and re-examining the attainment of goals (Pinpoint, Record, Change, and Try Again) (Kubina and Yurich, 2012). In an environment of blended learning featuring precision teaching, learners can perceive that there are clear goals in their study, relevant learning materials, and lectures targeting learning problems. Such perceptions can raise the expectation of success in attaining the learning goals, which is likely to ignite learning motivation (Gorges and Kandler, 2012; Cho et al., 2019). When learners clearly perceive the goals to attain and preview the lectures on their own with the online learning system, they tend to be more oriented and purposeful in their studies, so that their self-efficacy is improved to complete learning tasks according to their goals (Law et al., 2010). When they finish previewing the lectures, they proceed to self-testing to see if the goals are attained, which conveniently indicates to the learners the goals achieved and those yet to be achieved. Checking off the goals can motivate the learners and inform them of the unattained goals, which facilitates intentional planning for the next steps and increases self-efficacy for goal attainment (Hong and Park, 2012; Doumen et al., 2014). During lectures, teachers organize teaching and discussion around the problems that have occurred in the learners’ learning process, so that students can perceive “the lectures are exactly about what they themselves do not quite understand,” thus adding to the motivation to invest more efforts into learning (Rebecca and Michelle, 2016). After the lectures, teachers provide tests based on the feedback during lectures and the learning goals. This way, after self-review, in-class learning, and self-testing post-class, students have a better understanding of the content and hence increased learning motivation (Roberts and Norwich, 2010; Rebecca and Michelle, 2016). Therefore, we put forward the following hypotheses:

Self-efficacy directly affects individuals’ dynamic psychological processes in the learning performance of specific activities (Wood and Bandura, 1989), decides their behavioral choices, and forces them to prefer the environment or activity that they feel they are most competent in. It may also affect their efforts, meaning that individuals put in more energy as long as they maintain high self-efficacy over an assignment (Paglis and Green, 2010).

In a blended learning environment with precision teaching, teachers outline precise goals and requirements. In this case, students with high self-efficacy actively contact the teachers, discuss with their classmates, and finish the tasks as required. The higher the self-efficacy, the more active the students would be in sharing knowledge and engaging in discussions (Hsu et al., 2007), and thereby, the stronger would be the social presence that they experience. In the process, those with higher self-efficacy tend to make more efforts in learning and have a better perception of the carefully crafted atmosphere for teaching, that is, a better perception of teaching presence (Akyol and Garrison, 2011). In precision teaching, students preview the content and finish self-testing according to their goals before the lectures. During the lectures, teachers organize teaching and discussion based on the problems that have occurred in the self-learning and self-testing, and further test if the goals are achieved after the lectures. Students with higher self-efficacy can immerse better into the entire learning process (Christina, 2000; Park et al., 2012), which would increase the cognitive presence of learners (Celikkaleli, 2014; Niemiec and Lachowicz-Tabaczek, 2015). To examine the influence of self-efficacy in the blended learning environment, the following hypotheses are proposed:

Motivation embodies the in-depth intention of every learner for a direction or goal. Bandura (1996) considers that people commit to their work or task out of personal belief, and learners’ belief in ability usually better reflects motivation than their actual ability. Corresponding to a group of physiological processes, motivation may determine behavioral direction and persistence (Moos and Marroquin, 2010). If a student wants to improve or develop their own ability in a given task or discipline, they usually seek pertinent help or reflects repeatedly.

Although blended learning courses offer flexible and convenient online learning content, students may lack the motivation to complete the tasks (Bennett et al., 2012; Torrisi-Steele and Drew, 2013), and suffer from procrastination (Cerezo et al., 2017). Motivated students have a strong connection with their teachers and endeavor to establish social relations with their peers (Patrick et al., 2007; Wentzel et al., 2010; Law and Geng, 2018). Such quality would improve the social presence in learning. By actively engaging in online and classroom learning, these students can effectively become accustomed to a digitalized learning environment and cope with learning problems, resulting in a better teaching presence (Zimmergembeck and Locke, 2007; Manganelli et al., 2019). Highly motivated students may make full use of the advantages of precision teaching to learn and explore, and eventually experience a better cognitive presence (Manganelli et al., 2019). The following hypotheses are thus presented:

In a COI course, students take more personal responsibilities and share multiple viewpoints more actively. Through vigorous online interaction, students and teachers together can create a COI environment with social, cognitive, and teaching presence (Akyol et al., 2011; Kozan and Richardson, 2014). This implies that extensive participation of teachers and their clarity in course design would affect students’ perceived learning (Swan, 2001). In some teaching and learning environments, in an engineering course, for instance, teachers’ performance plays a decisive role in learning performance, outweighing social and cognitive presence in importance (Szeto, 2015). Students’ involvement in discussions, their motivation, and cooperation with others can influence their learning and performance (Chong, 1998; Swan, 2001; Davies and Graff, 2005). In the research on the effects of cohesion and interaction on team performance among students, social presence is closely related to learning outcomes (Arbaugh, 2005; Williams et al., 2006). The COI framework suggests an approach to creating collaboration and meaningful learning experiences, which can improve or undermine education experiences and learning outcomes (Garrison et al., 2000). Three types of presence are thus predictive of learning performance (Arbaugh, 2008; Choy and Quek, 2016). Hence, we propose the following hypotheses:

Based on the above analysis, the constructed model composed of previous hypotheses is developed, as shown in Figure 2.

We conducted a questionnaire survey to test the above five sets of hypotheses. The questionnaire is adapted from mature scales in the relevant literature and presented in Chinese. We invited a professor of English to translate the scales originally in English into Chinese and another professor of English to check if the translation was appropriate. Then, six students who participated in the course attentively were asked to fill in the questionnaire and inform the researchers of how they understood the questionnaire. In view of the students’ feedback, the two English professors improved the Chinese translation. This was to ensure consistency in content between the Chinese and English scales, as well as the accuracy in the respondents’ understanding. The questionnaire encompasses PPT, self-efficacy (SE), learning motivation (LM), teaching presence (TP), social presence (SP), cognitive presence (CP), and learning performance (PERF). The items and references are detailed in Appendix A. The questionnaire adopted a five-point Likert scale [from “strongly disagree” (1) to “strongly agree” (5)] and the respondents were required to complete the survey anonymously.

To test the research model, we provided a blended learning course of “Computer Network Technologies” featuring precision teaching for one term in a university in a large city in China. The participants of our study were 258 students majoring in e-commerce. The learning process is depicted in Figure 3, which was integrated into a blended learning process the four basic steps of precision teaching, that is, pinpoint, record, change, and try again (Benbunan-Fich, 2008; Kubina and Yurich, 2012). At the end of the term, the above-mentioned questionnaire (See Appendix A) was distributed to the participants. Of the 258 students who completed the survey, 256 valid questionnaires remained after the removal of two incomplete ones.

The general information of the respondents is as shown below (Table 1). It can be easily seen that there are more female students than male students, primarily because e-commerce is a management specialty under liberal arts of the school. Most of these students are aged 20–22 years.

The model (Figure 2) is verified with the partial least squares (PLS) approach of SEM. PLS here refers to an alternative least squares algorithm that expands principal component analysis and typical correlation analysis (Henseler and Sarstedt, 2013). It is composed of two sets of equations that are called the inner model and the outer model, respectively. Among them, the inner model defines the relationship between unobserved variables or latent variables, and the outer counterpart defines the relationship between latent variables and observed indicators. PLS-SEM can predict a relatively complicated model with no need to meet the distribution hypothesis, and for this reason, it is applicable to the handling of non-normal distribution data. This phenomenon is commonplace for business and social science researchers (Fornell and Bookstein, 1982; Hair et al., 2014b; Sarstedt et al., 2014). Accordingly, PLS methods in SmartPLS software are adopted here.

To ensure the fine reliability and validity of the research, a test is conducted to examine the seven dimensions of the structure, including PPT, self-efficacy, learning motivation, social presence, cognitive presence, teaching presence, and learning performance. For each item, the factor loading is above 0.7 (Table 2). Structural convergence validity is assessed in accordance with the estimation results of the measurement model. Reflective measurement is considered reliable when the item’s factor loading over the correlation factor is high (above 0.7) (Fornell and Larcker, 1981). Subsequently, Cronbach’s alpha, Dillon–Goldstein’s rho, composite reliability, and average variance extracted (AVE) are used to test construct reliability and validity (Table 3). For all constructs, Cronbach’s alpha and Dillon–Goldstein’s rho are greater than 0.7, congruent with the acceptable criterion (Nunnally, 1968), and composite reliability is greater than 0.9, far above the recommended value of 0.5 (Chin and Gopal, 1995). Additionally, the AVE value greater than the minimum 0.5 well proves the reliability and validity of the research (Fornell and Larcker, 1981). These indicators demonstrate the feasibility of all constructs for follow-up research.

The discriminant validity of the constructs is assessed based on the Fornell and Larcker (1981) criterion. According to Fornell and Larcker (1981), for good discriminant validity, the square root of the AVE of the constructs should be larger than the correlation coefficient between the constructs. All the constructs in the estimated model are in line with this criterion (Fornell and Larcker, 1981; See Table 4).

Surveyed data executed 5,000 bootstrapping procedure model operations using SmartPLS3. Figure 4 presents the results of the hypothesis verification vividly. The direct and indirect effects of the seven constructs are tested, and the results are as shown in Tables 5, 6.

In blended learning, PPT is positively and directly related to self-efficacy (β = 0.699, p < 0.001), and learning motivation (β = 0.727, p < 0.001), supporting the validity of H1 and H2. PPT is positively and indirectly related to cognitive presence (β = 0.658, p < 0.001), teaching presence (β = 0.589, p < 0.001), and social presence (β = 0.633, p < 0.001), and eventually learning performance (β = 0.573, p < 0.001); and self-efficacy is positively and directly related to cognitive presence (β = 0.456, p < 0.001), teaching presence (β = 0.224, p = 0.001), and social presence (β = 0.367, p < 0.001), supporting the validity of H3a, H3b, and H3c, respectively, and therefore attesting that H3 is tenable. Learning motivation is positively and directly related to cognitive presence (β = 0.466, p < 0.001), teaching presence (β = 0.594, p < 0.001), and social presence (β = 0.517, p < 0.001), supporting the validity of H4a, H4b, and H4c, respectively, and therefore attesting that H4 is tenable. Cognitive presence is positively and directly related to learning performance (β = 0.295, p < 0.001), teaching presence is positively and directly related to learning performance (β = 0.372, p < 0.001), and social presence is positively and directly related to learning performance (β = 0.253, p = 0.006), supporting the validity of H5a, H5b, and H5c, respectively and therefore attesting that H5 is tenable.

As PLS path modeling cannot provide a widely accepted global model fitting (Hair et al., 2017b), the research deliberately adopts standardized root mean square residual (SRMR) for model fitting, finding that model SRMR value of 0.042 below 0.08 is acceptable (Hair et al., 2017b).

In addition to the use of determining the coefficient (R²) in assessing the model’s predictive accuracy, cross-validation redundancy (Q²) may be also used to determine the predictive relevance of the model (Hair et al., 2014a). The model possesses predictive relevance in the case that the endogenous variable value Q² is greater than 0 (Hair et al., 2017a). Table 7 also suggests that the model has predictive relevance in self-efficacy, learning motivation, cognitive presence, teaching presence, social presence, and learning performance.

Additionally, Table 7 calculates the effect size of each exogenous variable f². According to Hair et al. (2014a), effect size f² indicates the contribution of the exogenous variable to the endogenous variable R², and 0.02, 0.15, and 0.35 suggesting small, medium, and large effects, respectively (Hair et al., 2014a,b). PPT has a large effect on self-efficacy (SE) (f² = 0.957) and learning motivation (LM) (f² = 1.123); self-efficacy (SE) has a large effect on cognitive presence (CP) (f² = 0.344), a small effect on teaching presence (TP) (f² = 0.053), and a medium effect on social presence (SP) (f² = 0.182); learning motivation has a large effect on cognitive presence (CP) (f² = 0.360), teaching presence (TP) (f² = 0.370), and social presence (SP) (f² = 0.359); cognitive presence (CP) and social presence (SP) have a small effect on learning performance (PER) (f² = 0.090, f² = 0.057); and teaching presence has a medium effect on learning performance (PER) (f² = 0.162).

In this research, we find that PPT is positively related to self-efficacy and learning motivation, with great predictive relevance and effect. It is also indirectly and positively related to learning performance. In a blended learning experience featuring precision teaching, students are requested to learn the online course content by themselves before class and join in the test. Students’ test scores are recorded by the system. In this way, students have already made sufficient preparations and show greater self-efficacy in classroom teaching beforehand. On the part of teachers, they can have pertinent and well-organized teaching interventions for problems discovered in the testing procedure (Kubina and Yurich, 2012). Facts prove that more pertinent teaching easily attracts students’ attention, arouses their interests, and stimulates their self-efficacy and learning motivation.

The research reveals that self-efficacy is directly and positively related to cognitive presence, social presence, and teaching presence, with varying predictive relevance and effect size, and is indirectly and positively related to learning performance. In blended learning environments featuring precision teaching, students are asked to preview and join in the test before class, so that they can basically grasp the course content, reinforce self-efficacy, and the willingness to have exchanges with classmates and teachers, and more easily get immersed in the teaching process. These behaviors are all directly and positively related to their learning performance.

Learning motivation is a key factor for active learning and also a decisive factor for student learning performance and expression (Weiner, 1990; Law et al., 2010; Law and Breznik, 2017; Law and Geng, 2018). Learning motivation may be stimulated by endogenous and exogenous factors, in which endogenous learning motivation is the dominant type in blended learning courses (Lin et al., 2003). Simultaneously, it has been found that learning motivation is directly and positively related to cognitive presence, social presence, and teaching presence, with great predictive relevance and effect, and indirectly and positively related to learning performance. Those students who have strong learning motivation are also inclined to more positively participate in course assignments and group activities. From this perspective, learning motivation promotes learning performance via its positive and direct relation to social presence and teaching presence.

In this study, it is observed that self-efficacy, learning motivation, and COI (including cognitive presence, teaching presence, and social presence) generate a serial mediating effect between PPT and learning performance, that is, PPT improves learners’ self-efficacy and learning motivation (Chapman et al., 2005; Roberts and Norwich, 2010; Rebecca and Michelle, 2016), while self-efficacy and learning motivation are directly related to COI, and COI is positively related to learning performance (Akyol and Garrison, 2011; Akyol et al., 2011; Kozan and Richardson, 2014).