Factors affecting student learning performance: A causal model in higher blended education
Abstract
In Mexico, approximately 504,000 students pursue a bachelor's degree by means of distance or blended programmes. However, only 42% of these students conclude their degree on time. In the context of blended learning, the focus of this research is to present a causal model, based on a theoretical framework, which describes the relationships concerning motivations, emotions, cognitive strategies, metacognitive strategies, and learning strategies, and their impact on learning performance. The results suggest that negative emotions play a meaningful role between expectancy (a component of motivation) and learning strategies. Also, the expectancy component of motivation positively influences metacognitive strategies. Concerning the relationship between cognition and metacognition, metacognitive strategies take preference concerning the relationship between metacognitive and learning strategies, supporting the theoretical hypothesis that metacognitive processes are on a higher plane than cognition, and affect cognitive process directly. Moreover, the learning outcomes are directly influenced by cognitive and learning strategies, but not by metacognitive ones. Similarly, motivation has direct effects on metacognitive and learning strategies but not on cognitive ones.
Lay Description
What is currently known about the subject matter:
- Several studies describe separately the causal relationships among motivation, emotions, cognition, metacognition, and learning performance in various learning contexts (mainly in distance and face‐to‐face learning).
- Although the fuzzy boundaries of metacognition have been recognized and a well‐sounded definition has been established, the theoretical relations with motivation, emotions, and cognition have barely been confirmed.
What this paper adds to this:
- From the systems theory point of view, motivation, emotions, cognition, and metacognition are emergent subsystems of the human mind. Thus, studying each of them separately cuts off the relationships that conform the whole system and gives us a limited view of it.
- This integrating idea is depicted in a causal model, based on a theoretical framework, which describes the relationships among motivations, emotions, cognitive strategies, metacognitive strategies, and learning strategies, and their impact on learning performance.
- This research offers practical evidence that supports the theoretical relation between motivation and metacognition.
The implications of study findings for practitioners.
- The motivation (expectancy) increases the use of metacognitive and learning strategies; this finding has practical implications, because metacognition may be positively stimulated by knowing that an individual's efforts to learn (control of learning beliefs) and judgments about the individual's abilities to accomplish a given task (self‐efficacy)
- Moreover, the results suggest that metacognitive, cognitive, and learning strategies are tightly related, in a hierarchical structure where metacognition plays an important role.
- The causal relations from positive emotions to metacognitive, cognitive, and learning strategies were not significant. Thus, the impact of negative emotions on the reviewed learning content (which captures the computer‐assisted nature of blended learning) and overall grade was explored. Students that face several obstacles in reviewing the learning content have not enough control to overcome this situation, and frustration is instigated. Also, anxious students think in advance that they will may fail the course, and this thought negatively impacts on their overall grade.
1 INTRODUCTION
Higher education is being criticized for its abandonment and graduation rates. Online and blended learning have been proposed as two possible solutions for these problems, due to their flexibility. Computers and the Internet have become important tools for supporting learning in blended higher education. Most universities around the world offer online and blended courses and use a learning management system to share learning resources, because such a system offers an opportunity to monitor student learning progress (You, 2015).
Blended learning combines online and face‐to‐face interaction and instruction (Rooney, 2003). Thus, blended learning emphasizes the central role of computer‐based technologies to share learning materials, evaluate the students, and often support asynchronous communication (teachers–students and students–students). The face‐to‐face learning environments are reliant on human–human interaction; meanwhile, distributed learning environments emphasize learner–material interactions (Graham, 2005). Learner–material interactions can easily be tracked by the learning management system. Thus, several distance‐learning studies have focused on them. On the other hand, in blended learning, some of these interactions cannot be easily tracked.
In Mexico, approximately 504,000 students pursue a bachelor's degree by means of distance or blended programs (SEP, 2015). However, only 42% of students conclude their degree on time (ANUIES, 2015). Several studies have focused on identifying the relationships concerning motivation, emotions, cognition, metacognition, and learning performance. These relationships have been depicted using different approaches; however, as far as we know, an inclusive model that captures these relationships simultaneously has not been proposed yet.
Our aim in this article is to present a model, based on a theoretical framework, which describes the relationships of motivations, emotions, cognitive strategies, metacognitive strategies, and learning strategies, and their impact on learning outcomes. The remainder of this paper is organized as follows: The next section gives the theoretical basis of the relationship among emotion, motivations, metacognition and cognition, and learning achievements, followed by a hypothetical model. Then, we present a brief description of relevant related work. Later, the methodology, research method, and results are described. Finally, the discussion and conclusions are considered.
1.1 Motivation, emotions, cognition, and metacognition
From the systems theory point of view, motivation, emotions, cognition, and metacognition are emergent subsystems (Skyttner, 2005) of the human mind. Thus, studying each separately cuts off the relationships that conform the whole system and gives us a limited view of it. This integrating idea is depicted in the quaternity of mind proposed by Mayer, Chabot, and Carlsmith (1997): conation, affect, cognition, and consciousness. Conation includes a low level of motivation, such as being thirsty and getting a glass of water, and also a high level, such as accomplishing a task even if having failed in previous trials. The affect contains emotions such as boredom, frustration, and anxiety. Similarly, cognition includes judgment, memory, learning, and reasoning. Consciousness interacts with the three aforementioned subsystems and mediates the relationship among them. Metacognition, a fifth subsystem, completes this picture from which arises the quintuple of the mind.
Metacognition means cognition about cognition (Flavell, 1976; Metcalfe & Shimamura, 1996) and seems to be on a higher plane than cognition—thus, it affects cognitive process. Metacognition is divided in two fundamental ways (Brown, 1987; Simons & Boekaerts, 1993): The individuals must be aware of their cognition (self‐regulation) and be able to apply metacognitive (critical thinking), cognitive (organization, elaboration, and rehearsal), and learning (help seeking, effort regulation, peer learning, and time and environment) strategies, to learn or to solve problems (Dawson, 2008; Wiley & Jee, 2010). Although several models of metacognition have been proposed (for a comparative review, see Peña‐Ayala & Cárdenas, 2015), the theoretical relations with emotions and motivation have not been clearly delineated except in Efklides (2011) and Pekrun (2006), where the bidirectional links from motivation to metacognition and emotions to metacognition, respectively, have been established. Moreover, Pintrich and de Groot (1990) suggest that metacognitive strategies mediate the effect of motivation on learning outcomes.
Cognition includes judgment, memory, and reasoning, to learn from the environment, and is affected by emotions (Mayer et al., 1997). Emotions are feelings directed towards a person, thought, or situation, which could be real or unreal (Pons, Rosnay, & Cuisinier, 2011). Matthews (1997) states that emotions may be conceptualized either as a dependent variable (influenced by processes such as appraisal) or as an independent variable (which itself influences information processing and cognition); thus, a circular causal relationship between them has been proposed (Kim, Park, & Cozart, 2014; Pekrun, 2006; Pons et al., 2011).
The theoretical research on the relationship between emotions and cognition shows that the first is a kind of interruption, as well as a factor that can change the way cognitive processes are carried on. This change might be positive, negative, or fuzzy (Mayer et al., 1997; Pekrun, 2006); for example, enjoyment of learning can reinforce the use of cognitive strategies. Conversely, test anxiety can negatively affect cognition, or its impact can be indirectly mitigated—for example, the anxiety might induce a motivation to avoid failure, so high levels of motivation produce a positive impact on cognition. We refer to this as a fuzzy process, because it is not clear why a particular path (out of several) would be activated.
Motivation is a teleological process determined by social factors, as well as by the cognitions of learners (Anderman & Dawson, 2011), whereby goal‐directed activity is instigated and sustained (Schunk, Pintrich, & Meece, 2008). In this research, the expectancy‐value theory of motivation (Eccles, 1983; Pintrich, 1988, 1989; Wigfield & Eccles, 2000) is adopted as a framework to conceptualize students' motivation. The expectancy component (control of learning beliefs and self‐efficacy) has been described as students' beliefs about their ability to perform a task and that they are responsible for their own performance meanwhile the value component (extrinsic goal motivation, intrinsic goal motivation, and task value) as students' goals about the importance and interest of the task (Pintrich & de Groot, 1990).
A student's motivation is a dependent variable influenced by the educational context, the subject area, and the task to be performed (Linnenbrink‐Garcia, Patall, & Pekrun, 2016). Motivation can deactivate negative emotions, and on the contrary, negative emotions such as test anxiety can avoid the failure by increasing the intrinsic motivation and, therefore, the effort. Also, unpleasant emotions can have negative effects on motivation, and conversely pleasant emotions can positively affect motivation (Pekrun & Linnenbrink‐Garcia, 2012). The empirical evidence suggests that emotions and motivation are related to each other in a bidirectional link (Linnenbrink & Pintrich, 2002).
The impact of emotions on academic achievements is mediated by the metacognitive and cognitive process (Pekrun, 2006). Also, the effects of academic achievements are linked by reciprocal causation to metacognition and cognition, emotions, and motivation (Efklides, 2011; Pekrun, 2006). Figure 1 shows a conceptual model based on the theoretical relationships presented in this section and those described in Artino (2009), Efklides (2011), and Pekrun (2006).
Although the theoretical model proposed by Pekrun (2006) includes contextual factors such as instruction and feedback, the conceptual model does not explicitly include them, but they are captured in the case study detailed later. For example, the feedback in blended learning is given in a face‐to‐face context, whereas in distance learning, it is carried out by means of chat or forums. Also, face‐to‐face instruction differs from the distance context. In distance learning, the instruction is centred on student–learning content interaction; in face‐to‐face learning, it is centred on human–human interaction. Blended learning combines both.
2 RELATED WORKS
The hypothetical causal model depicted in Figure 2 is based on the conceptual model of Figure 1. Next, prior studies are discussed to show how they underpin the relations of the hypothetical model among motivation, emotions, metacognition, cognition, and academic achievements.
2.1 The effect of metacognitive, cognitive, and learning strategies on learning outcomes
Several studies describe separately the causal relationships among cognition, metacognition, and learning performance in various learning contexts (mainly in distance and face‐to‐face learning). González, Rodríguez, Faílde, and Carrera (2016), Mega, Ronconi, and De Beni (2014), Pérez, Costa, and Corbí (2012), and Su (2016) modelled the impact of different constructs in academic achievements on face‐to‐face learning. Comparative studies in different contexts (i.e., distance, blended, and face‐to‐face learning) have been carried out to determine the role of metacognitive strategies such as self‐regulation. Barnard, Lan, To, Paton, and Lai (2009) compared two models (one built using data gathered from distance learning, the second from blended learning). The result shows that both models were similar in structure; thus, self‐regulation is independent of the learning context.
Although both self‐regulation and critical thinking are metacognitive processes that include several skills (Dwyer, Hogan, & Stewart, 2011, 2014; Van Gelder, 2015) and are crucial for the control of cognition (Efklides, 2011), the impact of both on academic achievements has been described separately. For example, the effect of self‐regulation on learning outcomes has been studied as a single construct (Mega et al., 2014; Pekrun, Goetz, Titz, & Perry, 2002; Pintrich & de Groot, 1990) and as a part of academic engagement (González et al., 2016). Also, Haseli and Rezaii (2013) and Kong (2014) show the positive impact of critical thinking on educational achievements. This suggests that metacognitive strategies (self‐regulation and critical thinking) have a direct effect on learning outcomes.
As found by Chang and Ley (2006), Mega et al. (2014), Pérez et al. (2012), Pintrich and de Groot (1990), and Weinstein and Mayer (1983), the learning strategies (help seeking, effort regulation, and time and environment) and cognitive strategies (organization, elaboration, and rehearsal) constructs impact directly academic achievement. This evidence supports the effect of both learning strategies and cognitive strategies on learning outcomes. Metacognition seems to be on a higher plane than cognition and is conceptualized in two fundamental ways (Brown, 1987; Simons & Boekaerts, 1993): The individuals must be aware of their cognition and be able to apply metacognitive, cognitive, and learning strategies, to learn or solve problems (Dawson, 2008; Wiley & Jee, 2010). Thus, relationship from metacognitive strategies to cognitive strategies, metacognitive strategies to learning strategies, and cognitive strategies to learning strategies are hypothesized.
2.2 The effect of emotions and motivation on metacognitive, cognitive, and learning strategies
A body of research has shown that positive and negative emotions impact on student's learning functions (cognitive and learning strategies; Ashby, Isen, & Turken, 1999; Daniels, Tze, & Goetz, 2015; Marchand & Gutierrez, 2012; Meinhardt & Pekrun, 2003; Pekrun et al., 2002). For example, anxiety decrements the use of flexible learning strategies (organization and elaboration) and stimulates the use of rigid ones such as rehearsal (Pekrun, 2006). Also, it has been demonstrated that cognitive strategies may mitigate the effect of boredom (Daniels et al., 2015). Liu (2015) found that academic boredom, mastery‐approach goals, and performance‐avoidance goals predict academic boredom. Similarly, Tze, Klassen, and Daniels (2014) modelled dynamically the change in boredom through time and its relation with effort regulation (a learning strategy).
As found by Mega et al. (2014), Pekrun (2006), Pekrun, Frenzel, Goetz, and Perry (2007), and Pekrun et al. (2002), emotions influence self‐regulation. Moreover, positive emotions such as enjoyment facilitate students' self‐regulation which presupposes flexibility in using metacognitive strategies such as critical thinking (Pekrun, 2006). Conversely, negative emotions decrease self‐regulation and foster the external guidance (Mega et al., 2014; Pekrun, 2006; Tze et al., 2014). As stated previously, self‐regulation and critical thinking are considered as dimensions of metacognitive strategies; thus, these prior studies support (Marchand & Gutierrez, 2012) the effect of emotions on metacognitive strategies.
González et al. (2016) found that motivation (value and expectancy) has a direct effect on self‐regulation. Similarly, intrinsic value and self‐efficacy are highly correlated with cognitive strategies (rehearsal, elaboration, and organization) and self‐regulation (Pintrich & de Groot, 1990). The relationships among motivation, learning strategies, and academic achievements are explored by King and McInerney (2016). Their findings suggest that the effect of motivation on learning strategies is invariant across student demographic variables. In the context of learning English as a foreign language, Zhang, Lin, Zhang, and Choi (2017) found that vocabulary learning strategies mediate the relationship between motivation (intrinsic and extrinsic) and vocabulary knowledge.
2.3 The effect of motivation on emotions
Academic emotions and their relations with motivational and behavioural factors assumes relevance in educational settings. The expectancies and values are antecedents of negative emotions such as test anxiety (Pekrun, 2000). Also, Efklides (2011), Pekrun (2006), and Pekrun et al. (2007) present a reciprocal relation between motivation (expectancy and values) and emotions where feedback loops can take place within fractions of seconds. The results that highlight the positive and negative impact of self‐efficacy (a dimension of expectancy construct of motivation) on positive and negative emotions, respectively, were proved by Mega et al. (2014) and Zhen et al. (2017). In a similar way, Marchand and Gutierrez (2012) and González et al. (2016) found that in distance and blended learning, self‐efficacy is linked to negative (frustration and anxiety) and positive (hope) emotions, and both impact learning strategies.
3 METHODOLOGY
Structural equation modelling (SEM) was used to estimate the relationships in the conceptual model. Also, SEM assumes that the causal paths follow a forward flow; thus, the reciprocal linkages shown in the Figure 1 cannot be estimated in the hypothetical model.
The data to build the causal model were gathered from a blended course in applied computing in biological sciences. The course enrolled 166 undergraduate university students. It is included in a chemical biology degree programme. The learning materials and learning activities are delivered through the Moodle platform. The learning materials follow the multimedia principles (coherence, redundancy, spatial contiguity, and modality; Mayer, 2001) and are designed to comply with the SCORM (ADL, 2004 standard. The interactions between the students and learning materials are recorded in a Moodle log file, providing information about the time of each online session, the time spent on reviewing the content, the number of topics browsed for a given learning material, and the number of learning activities turned in by students.
Figure 2 only shows the construct (squares) and the subconstruct (latent variables) of the model; the observed variables that belong to a given subconstruct are described in Table 1. The face‐to‐face sessions are mainly intended for student feedback and deep explanations of learning activities; thus, communication in forums and chats rarely takes place. The learning content is organized by a weekly schedule and often includes learning activities that must be delivered within the same week, through Moodle. The course contains three offline examinations, the learning activities, and an end‐of‐term project. All were considered when computing the overall grade of each student. Hence, the overall grade represents a measure of learning performance.
| Construct | Subconstruct (latent variable) | Observed variables |
|---|---|---|
| Motivation | Value | Intrinsic goal motivation (IGM) |
| Extrinsic goal motivation (EGM) | ||
| Task value (TV) | ||
| Expectancy | Control of learning beliefs (CLB) | |
| Self‐efficacy (SE) | ||
| Emotions | Negative emotions | Boredom (BO) |
| Frustration (FR) | ||
| Anxiety (AN) | ||
| Positive emotions | Enjoyment (EN) | |
| Pride (PR) | ||
| Interest (IN) | ||
| Metacognition and cognition | Metacognitive strategies | Self‐regulation (SR) |
| Critical thinking (CT) | ||
| Cognitive strategies | Organization (OR) | |
| Elaboration (EL) | ||
| Rehearsal (RE) | ||
| Learning strategies | Help seeking (HS) | |
| Effort regulation (ER) | ||
| Time and environment (TE) | ||
| Academic achievements | Learning outcomes | Overall grade (OG) |
| Reviewed learning content (RLC) |
Because relations from positive emotions to other subconstructs are not significant in the causal model presented later, we focused on exploring the impact of negative emotions on blended learning. The effect of frustration and boredom on the reviewed learning content (which captures the computer‐assisted nature of blended learning) and the effect of anxiety on overall grade were measured by analysis of variance (ANOVA) on the basis described below. The control value theory (Pekrun, 2006; Pekrun et al., 2007)—built on assumptions from expectancy‐value theory—states that frustration and boredom can be triggered by the value level (given by a student) of a learning activity and the level of control. For example, a student positively values the linear regression topic; if the student faces several obstacles and has a low level of control, he will be frustrated. On the other hand, boredom will be triggered no matter the value nor the level of control. In other words, the student is detached from the activity. Also, only one of these two emotions can be felt at a time. Based on these assumptions, two one‐way ANOVAs were performed. The first ANOVA tests if there is a difference of the number of reviewed learning content induced by frustration and the second by boredom. Similarly, anxiety is associated with the anticipatory perception of failure of learning outcomes (Pekrun, 2006; Pekrun et al., 2007); thus, its influence on the overall grade is also tested by one‐way ANOVA.
4 RESEARCH METHOD
4.1 Instruments
Students' negative emotions—boredom (BO), frustration (FR), and anxiety (AN)—and positive ones—enjoyment (EN), pride (PR), and interest (IN)—were assessed using the questionnaire measure of Student Engagement and Disaffection in school (SED; Skinner, Furrer, Marchand, & Kindermann, 2008; Wellborn, 1991). The students' answers were based on a 5‐point Likert scale of agreement, where 1 means strongly disagree and 5 means strongly agree. A high value for a negative emotion question means that it was felt by students. For example, a 5‐point answer to the question, “When I can't answer a question, I feel frustrated,” means that the student felt frustrated.
The motivated strategies for learning questionnaire (Pintrich, 1991; Pintrich & de Groot, 1990) assessed value and expectancy components of motivation. The value includes intrinsic goal motivation (IGM), extrinsic goal motivation (EGM), and task value (TV). Expectancy includes control of learning beliefs (CLB) and self‐efficacy (SE). Self‐regulation (SR) and critical thinking (CT) belong to metacognitive strategies. Organization (OR), elaboration (EL), and rehearsal (RE) belong to cognitive strategies, whereas help seeking (HS), effort regulation (ER), and time and environment (TE) belong to learning strategies. All the questions in this survey were answered using a Likert scale ranging from 1 (not at all true of me) to 7 (very true of me). Although, the SED includes text anxiety and peer learning subscales, they did not complain convergent validity criteria explained later; thus, both were excluded from the hypothetical causal model. Moreover, CT is a metacognitive process that includes several skills (Dwyer et al., 2011, 2014; Van Gelder, 2015); thus, it was included in the metacognitive subconstruct, as well as SR.
Finally, the learning outcomes include the overall grade (OG) and the reviewed learning content (RLC). The OG is computed considering the offline examinations, the learning activities, and an end‐of‐term project. The interactions between the students and learning content were recorded in Moodle's log file. Because the learning content complies with the SCORM (ADL, 2004) standard, Moodle is able to track the number of items reviewed that belong to each learning content. The learning content or SCORM objects were organized in a hierarchical structure of topics and subtopics. Students were required to browse all topics and subtopics to consider a given learning content as reviewed. Students needed to log into Moodle each time to review current session or previous session learning content. They were not permitted to download learning content or browse offline. The RLC is the number of learning content reviewed by a given student, divided by the total; thus, RLC captures the interaction student‐learning content.
4.2 Data gathering and data analyses
The main objective of the research is to test a set of explanatory relationships between the aforementioned constructs. At the end of the semester, the logs of students' interactions with learning materials were used to compute the number of learning materials reviewed. The SED and Motivated Strategies for Learning Questionnaire instruments were administered 2 weeks before ending the course; finally, the student's overall grade was reported at the end of the semester. Students were informed of the aims of this research, and their participation was voluntary.
Prior to applying SEM, the validity and reliability of the constructs were established. Cronbach's alpha was calculated for all the constructs, and values higher or equal to 0.70 are acceptable for establishing internal consistency. The convergent validity was tested by composite reliability (CR) and average variance extracted (AVE). The CR and AVE are acceptable when their value exceeds 0.7 and 0.5, respectively (Hair, Black, Babin, Anderson, & Tatham, 2006). Discriminant validity is established when the root square of AVE of each construct is greater than the squared correlation between the other constructs (Fornell & Larcker, 1981).
The SEM analyses were performed using the AMOS version 22 with the maximum likelihood estimation procedure. To measure the model's fitness, the indexes χ2/df, Tucker Lewis Index (TLI), comparative fit index (CFI), normed fit index (NFI), and root mean square error of approximation (RMSEA) were considered. The acceptable values for χ2/df are below 5 (MacCallum, Browne, & Sugawara, 1996), whereas TLI, CFI, and NFI are above 0.90 (Byrne, 2013), and RMSEA with values ranging from. 08 to. 05 or less (Byrne, 2013) are acceptable.
To conduct ANOVA, the boredom, frustration, and anxiety variables were encoded in three ranges (low, medium, and high). OG had a normal probability distribution by anxiety and RLC by boredom and frustration. All the distributions of the data were homoscedastic.
5 RESULTS
5.1 Reliability and validity
Table 2 presents Cronbach's alpha, CR, and AVE of subconstructs (latent variables): value, expectancy, negative and positive emotions, metacognitive strategies, cognitive strategies, learning strategies, and learning outcomes. The values of Cronbach's alpha of all subconstructs ranged from 0.778 to 0.962, which are acceptable for SEM. Also, the values of CR and AVE are above 0.7 and 0.5, respectively, which confirms the convergent validity of the subconstructs. Table 3 shows in the diagonals the square root of AVE and the other entries show squared correlations. The diagonals show higher values than squared correlations; thus, discriminant validity for all the constructs was established.
| Subconstruct (latent variable) | Cronbach's alpha | CR | AVE |
|---|---|---|---|
| Value | 0.814 | 0.915 | 0.843 |
| Expectancy | 0.878 | 0.942 | 0.891 |
| Negative emotions | 0.866 | 0.916 | 0.785 |
| Positive emotions | 0.974 | 0.983 | 0.951 |
| Metacognitive strategies | 0.878 | 0.943 | 0.891 |
| Cognitive strategies | 0.926 | 0.953 | 0.871 |
| Learning strategies | 0.778 | 0.871 | 0.694 |
| Learning outcomes | 0.852 | 0.931 | 0.871 |
- Note. AVE: average variance extracted; CR: composite reliability.
| Subconstruct | V | EX | NE | PE | MS | CS | LS | LO |
|---|---|---|---|---|---|---|---|---|
| Value (VA) | 0.918 | |||||||
| Expectancy (EX) | 0.885 | 0.944 | ||||||
| Negative emotions (NE) | −0.321 | −0.278 | 0.88 | |||||
| Positive emotions (PE) | 0.174 | 0.120 | −0.651 | 0.975 | ||||
| Metacognitive strategies (MS) | 0.783 | 0.741 | −0.216 | 0.139 | 0.944 | |||
| Cognitive strategies (CS) | 0.731 | 0.661 | −0.207 | 0.143 | 0.888 | 0.933 | ||
| Learning strategies (LS) | 0.702 | 0.673 | −0.345 | 0.306 | 0.736 | 0.701 | 0.833 | |
| Learning outcomes (LO) | 0.181 | 0.221 | −0.172 | 0.073 | 0.142 | 0.058 | 0.269 | 0.933 |
5.2 Causal model
Figure 3 presents the results of the causal model; the paths with solid lines were statistically significant. All the path weights are standardized. The causal model fits the data reasonably well with χ2/df = 1.868, TLI = 0.949, CFI = 0.959, NFI = 0.917, and RMSEA = 0.073. Noticeably, PE was not included in this model because the paths VA to PE (β = 0.174, p = 0.262), EX to PE (β = 0.186, p = 0.181), PE to MS (β = 0.005, p = 0.914), PE to CS (β = 0.024, p = 0.475), and PE to LS (β = 0.098, p = 0.065) were not significant.
The latent endogenous subconstruct expectancy has a significant causal relation with negative emotions. As was mentioned, high values of negative emotions mean that they were felt by students, whereas low values mean the opposite. The relation expectancy‐negative emotions is negative, given by the coefficient in Figure 3; thus, high values of expectancy Likert scales decrease the values of negative emotions Likert scales. This means that negative emotions are reduced by motivation (expectancy). On the other hand, value subconstruct has a positive causal relation with learning strategies; thus, motivated students (by IGO, EGO, and TV) tend to study in a comfortable place and apply learning strategies such as HS and ER.
The negative emotions play a vital role between expectancy and learning strategies; thus, students who reduce their negative emotions by expectancy will apply learning strategies. Similarly, students motivated by IG, EGO, and TV (value subconstruct) will apply learning strategies too—in this case, there is a direct effect between value and learning strategies. Moreover, expectancy directly and positively affects the use of metacognitive strategies.
Concerning metacognition and cognition, the results show that metacognition has a positive causal relation with cognition, and the latter with learning strategies; thus, cognitive strategies play a principal role between metacognitive and learning strategies. The relation from metacognitive strategies to learning outcomes was found to be not significant. Finally, the students who apply cognitive and learning strategies tend to have better learning outcomes (get a better OR and increase the number of RLC).
5.3 Negative emotions on computer‐assisted learning context
The first one‐way ANOVA shows a strong effect of FR on RLC, F(2,163) = 5.41, p = 0.005, ηp2 = 0.062. Post hoc analyses, using Tukey, indicates that less frustrated students reviewed (completely) more learning contents than those highly (p = 0.027) and moderately (p = 0.015) frustrated. RLC did not differ significantly between moderately and highly frustrated students (p = 0.986). The respective means in RLC for low, medium, and high levels of frustration are M = 0.70 (SD = 0.16), M = 0.618 (SD = 0.17), and M = 0.612 (SD = 0.21). On the other hand, the effect of boredom on RLC is not significant, F(2,163) = 2.09, p = 0.127, ηp2 = 0.25. The respective means in RLC for low, medium, and high levels of boredom are M = 0.681 (SD = 0.18), M = 0.617 (SD = 0.16), and M = 0.653 (SD = 0.18).
The results of the second one‐way ANOVA shows a strong effect of anxiety on OG, F(2,163) = 6.25, p = 0.002, ηp2 = 0.071. The students that reported a low level of anxiety obtained a higher OG than those that reported medium (p = 0.039) and high (p = 0.005) levels. No difference was found between medium and high levels of anxiety (p = 0.911). The respective means in OG for low, medium, and high levels of anxiety are M = 8.47 (SD = 1.35), M = 7.52 (SD = 2.31) and M = 7.68 (SD = 1.36).
6 DISCUSSION AND CONCLUSION
In the context of blended learning, the focus of this research is to present a causal model, based on a theoretical framework, which describes the relationships among motivations, emotions, cognitive strategies, metacognitive strategies, and learning strategies, and their impact on learning performance. Although the fuzzy boundaries of metacognition have been recognized and a well‐sounded definition has been established, the theoretical relations with motivation, emotions, and cognition have barely been confirmed. This research offers practical evidence that supports the theoretical relation between motivation and metacognition stated by (Efklides, 2011). This finding has practical implications, because metacognition (self‐regulation and critical thinking) may be positively stimulated by knowing that an individual's efforts to learn (control of learning beliefs) and judgments about the individual's abilities to accomplish a given task (self‐efficacy) will produce positive outcomes.
Also, practical research takes a different conceptualization of metacognition. For example, King and McInerney (2016) and González et al. (2016) define metacognition as a set of strategies. On the other hand, Mega et al. (2014) and Zhang et al. (2017) consider metacognition a part of self‐regulated learning. Because critical thinking is a metacognitive process (Dwyer et al., 2011; Dwyer et al., 2014; Hogan, Dwyer, Harney, Noone, & Conway, 2015), it was included in the metacognitive strategies subconstruct, along with self‐regulation. Thus, the boundary between metacognition and cognition was depicted in the proposed causal model. Moreover, the results suggest that metacognitive, cognitive, and learning strategies are tightly related, in a hierarchical structure where metacognition plays an important role. This suggests that students continuously are selecting, monitoring, and adjusting the best cognitive strategies (organization, rehearsal, and elaboration) according to the learning situation.
The results confirm some of the theoretical relations depicted in Pekrun (2006). For example, motivation (expectancy) reduces negative emotions; surprisingly, the latter impacts learning strategies, but not metacognitive and cognitive strategies. Also, cognitive and learning strategies promote better learning outcomes. On the other hand, the casual relation from positive emotions to metacognitive, cognitive, and learning strategies was not significant. Thus, the impact of negative emotions on the reviewed learning content (which captures the computer‐assisted nature of blended learning) and overall grade was explored. The findings show that less frustrated students review (completely) more learning content than those highly and moderately frustrated. Thus, the students that face several obstacles in reviewing the learning content have not enough control to overcome this situation, and a negative emotion such as frustration is triggered instead of enjoyment. Similarly, less anxious students obtained a higher overall grade than those highly and moderately anxious. This implies that anxious students think in advance that they may fail the course, and this thought negatively impacts on their overall grade.
Because expectancy and value have a causal relation with metacognitive and learning strategies, respectively, and the relations from value and expectancy to learning outcomes were not significant, we conclude that cognitive and learning strategies only affect learning outcomes. Also, motivation influences metacognition and cognition as Pekrun (2006) and Pintrich and de Groot (1990) suggest. Thus, motivated students apply more cognitive and learning strategies that in turn increase their learning outcomes.
The main limitation of our study is that we can neither confirm nor reject the reciprocal relations shown in the conceptual model. Also, it would be interesting for future research to examine the causal model in distance and face‐to‐face learning environments, to determine whether the learning context influences these constructs, as it has been hypothesized.
ACKNOWLEDGEMENTS
This work was supported by Instituto Politécnico Nacional (Grant 20170922, 20171422, and 20170742).
CONFLICT OF INTEREST
No conflicts of interest have been declared.
