A Review of Self-regulated Learning: Six Models and Four Directions for Research

Zimmerman: A Socio-cognitive Perspective of SRL Grounded by Three Models

Zimmerman was one of the first SRL authors (e.g., Zimmerman, 1986). He has developed three different SRL models, being the first one published in 1989 representing what was the first attempt to explain the interactions that influence SRL.

History and Development of the Models

Zimmerman (2013) reviewed his career and the development of his work, framing it into the socio-cognitive theory (i.e., individuals acquire knowledge by observing others and social interaction). Zimmerman’s work started from cognitive modeling research in collaboration with Albert Bandura and Ted L. Rosenthal. Later Zimmerman began to explore how individual learners acquire those cognitive models and become experts in different tasks.

As one of the most prolific SRL writers, Zimmerman has developed three models of SRL (Panadero and Alonso-Tapia, 2014). The first model (Figure 1), known as the Triadic Analysis of SRL, represents the interactions of three forms of SRL: environment, behavior and person level (Zimmerman, 1989). This model describes how SRL could be envisioned within Bandura’s triadic model of social-cognition. The second model (Figure 2) represents the Cyclical Phases of SRL, which explains at the individual level the interrelation of metacognitive and motivational processes. This model was presented in a chapter in the 2000 handbook, and it is usually known as Zimmerman’s model. There the subprocesses that belong to each phase were presented, but it was not until 2003 that these subprocesses were embedded in the figure (Zimmerman and Campillo, 2003). Finally, in Zimmerman and Moylan (2009) the model underwent some tweaks (Figure 3), including new metacognitive and volitional strategies in the performance phase. The third model Zimmerman developed (Figure 4), which recently has been called the Multi-Level model, represents the four stages in which students acquire their self-regulatory competency (Zimmerman, 2000). In this review, Cyclical Phases model will be analyzed, as it describes the SRL process at the same level as the models from the other authors analyzed here.

Zimmerman’s Cyclical Phases Model

Zimmerman’s (2000) SRL model is organized in three phases: forethought, performance and self-reflection (see Figure 3). In the forethought phase, the students analyze the task, set goals, plan how to reach them and a number of motivational beliefs energies the process and influence the activation of learning strategies. In the performance phase, the students actually execute the task, while they monitor how they are progressing, and use a number of self-control strategies to keep themselves cognitively engaged and motivated to finish the task. Finally, in the self-reflection phase, students assess how they have performed the task, making attributions about their success or failure. These attributions generate self-reactions that can positively or negatively influence how the students approach the task in later performances.

Empirical Evidence Supporting Zimmerman’s Cyclical Model

An overview of Zimmerman’s empirical evidence can be found in his career review (Zimmerman, 2013). A special feature of Zimmerman’s empirical research is the use of athletic skills, along with more typical academic skills. A number of studies have been conducted to test different aspects of Zimmerman’s models (Puustinen and Pulkkinen, 2001; Zimmerman, 2013), especially the Multi-level and the Cyclical phase models. Zimmerman conducted work with Kitsantas and Cleary that tested the Multi-level model (Zimmerman and Kitsantas, 1997, 1999, 2002; Kitsantas et al., 2000). Those four studies can be grouped in two types. First, the articles published in 1997 and 1999 studied the differential effect of outcome and process goals with high school students in two different tasks dart throwing and writing, finding support for the model. And second, the articles published in 2000 and 2002 studied the effect of observing different types of models in the development of SRL skills in dart throwing and writing.

The cyclical phase model has been tested in a series of four studies. First, Cleary and Zimmerman (2001) studied the SRL skills showed by adolescent boys who were experts, non-experts and novices in basketball, finding that experts performed more SRL actions. Second, in a similar study, Kitsantas and Zimmerman (2002) compared college women that were experts and non-experts in volleyball, finding that the SRL skills predicted 90% variance in serving skills. Third, Cleary et al. (2006) trained 50 college students in basketball free throws organized in five different conditions: one-phase SRL, two-phases SRL, three-phases SRL, control group practice-only and control group no-practice. The results showed a linear trend: the more phases trained the better the participants’ scores. Finally, fourth, DiBenedetto and Zimmerman (2010) studied 51 high school seniors during science courses seniors finding that higher achievers showed more use of subprocesses from Zimmerman’s model.

Another important piece of research into Zimmerman’s model is the work performed by Bernhard Schmidt and colleagues. As already mentioned, Schmidt has developed a SRL model based on Zimmerman’s and influenced by Kuhl’s (2000) model with changes in the names of the phases and subprocesses included (Schmitz and Wiese, 2006). This theoretical proposal gives a major emphasis to the role of self-monitoring in SRL (Schmitz et al., 2011). Additionally, Schmitz has developed significant research on how the use of learning diaries and its different data analysis known as time-series analysis. His main results have been that the use of learning diaries enhances all SRL phases being an effective way to impact in students’ SRL and performance.

Instruments and Measurement Methods

Under Zimmerman’s model umbrella, five instruments and measurements have been developed. First, the subprocesses present in Zimmerman’s model are partly based on the results found in the validation process of the Self-Regulated Learning Interview Schedule (SRLIS) (Zimmerman and Martinez-Pons, 1986, 1988). Second, Zimmerman has developed procedures to assess SRL in experimental training settings for writing and dart throwing (Zimmerman and Kitsantas, 1997, 1999). Third, Cleary and Zimmerman (2001, 2012), Kitsantas and Zimmerman (2002), DiBenedetto and Zimmerman (2010) developed microanalytic measures to assess the validity of the Cyclical Phases model. Fourth, Zimmerman has developed different measures of self-efficacy to self-regulate (Zimmerman and Kitsantas, 2005, 2007) and calibration measures of self-efficacy and self-evaluation (Zimmerman et al., 2011). And, fifth, anchored on the framework of SRL by Zimmerman and Martinez-Pons (1986, 1988), Magno (2010) developed the Academic Self-Regulation Scale (A-SRL) which has been validated analyzing its functional correlation against two well-established SRL instruments the MSLQ and the Learning and Study Strategies Inventory (LASSI) (Magno, 2011).

Boekaerts: Different Goal Roadmaps (Top–Down/Bottom–Up) and the Role of Emotions

The work by Boekaerts is also one of the earliest in the SRL literature and can be traced back to the late 1980s (e.g., Boekaerts, 1988). Shortly after she presented her first SRL model (Boekaerts, 1991). Her work has focused in explaining the role of goals (e.g., how students activate different types of goals in relation to SRL), and she was the first to use situation-specific measures to evaluate motivation and SRL. In addition, Boekaerts has demonstrated a vast knowledge of the clinical psychology literature on self-regulation and emotion regulation (see Boekaerts, 2011).

History and Development of the Models

Boekaerts has developed two models of SRL. First, she developed a structural model (Figure 5) in which self-regulation was divided into six components, which are: (1) domain-specific knowledge and skills, (2) cognitive strategies, (3) cognitive self-regulatory strategies, (4) motivational beliefs and theory of mind, (5) motivation strategies, and (6) motivational self-regulatory strategies (Boekaerts, 1996b). These were organized around, what she then considered to be, the two basic mechanisms of SRL: cognitive and affective/motivational self-regulation. This model has been mainly used to (a) gain more insight into domain-specific components of SRL, to (b) train teachers, to (c) construct new measurement instruments for research, and to (d) design intervention programs (Boekaerts, M. personal communication to author 08/06/2014).

Second, most of Boekaerts’ publications were set up to formulate a second SRL model, namely, the Adaptable Learning Model. This model (see Figure 6) was presented at the beginning of the 90s (Boekaerts, 1991, 1992). It describes the dynamic aspects of SRL, and later, evolved into the Dual Processing self-regulation model (Figure 7). The Adaptable Learning Model offered a theoretical scaffold for understanding the findings from diverse psychological frameworks, including motivation, emotion, metacognition, self-concept, and learning. The model described two parallel processing modes: (a) a mastery or learning mode and (b) a coping or well-being mode. In a chapter of the 2000 Handbook of self-regulation, Boekaerts and Niemivirta (2000) presented new ideas on goal paths using different figures to visualize how they influence students’ behavior (see pp. 434–435). Although, in 2000, Boekaerts had already presented some notions on her vision of top–down and bottom–up theory, it was not until mid-2000 that these theoretical insights were clearly defined in her model, which was then renamed as the Dual Processing self-regulation model (Boekaerts and Corno, 2005; Boekaerts and Cascallar, 2006). In the 2011 SRL handbook of SR, Boekaerts presented an extended version of this model, which pointed to the different purposes of self-regulation during the learning process, namely, (1) expanding one’s knowledge and skills, (2) protecting one’s commitment to the learning activity, and (3) preventing threat and harm to the self. Boekaerts emphasized the key role that positive and negative emotions play in SRL, and described two different bottom–up strategies, namely, volitional strategies and emotion regulation strategies (Figure 7; Boekaerts, 2011).

Boekaerts’ Dual Processing Model

In the Dual Processing model (Boekaerts and Cascallar, 2006), the appraisals made by the students are crucial to determine which goal pathway the students will activate. Here, goals are viewed as the “knowledge structures” that guide behavior. For example, if students perceive that the task could be threatening to their well-being, negative cognitions and emotions are triggered. Strategies are then directed to protect the ego from damage, and thereby, students move onto a well-being pathway. On the other hand, if the task is congruent with the students’ goals and needs, they will be interested in amplifying their competence, triggering positive cognitions and emotions, and thereby, moving onto the mastery/growth pathway. Boekaerts (2011) also explains that students who have started a task in the mastery/growth pathway may move to the well-being pathway if they detect cues that they might not be successful.

According to Boekaerts (2011), there are three different purposes for self-regulation:

(a) expanding knowledge and skills…(b) preventing threat to the self and loss of resources so that one’s well-being is kept within reasonable bounds…and (c) protecting one’s commitments by using activities that re-route attention from the well-being pathway to the mastery pathway (pp. 410–411).

The first is what she called “top–down,” as the pursuit of task goals is driven by the students’ values, needs and personal goals (mastery/growth pathway). The second purpose is called “bottom–up,” as the strategies try to prevent the self from being damaged (well-being pathway), and students may experience a mismatch between the task goals and their personal goals. The third purpose occurs when students try to redirect their strategies from the well-being to the mastery/growth pathway, which may happen via external (e.g., teacher or peer pressure) or internal (e.g., self-consequating thoughts) forces. Therefore, emotions are essential in Boekaerts’ model, because when students experience negative emotions, they will activate the well-being pathway and use bottom–up strategies. Pursuant to this interest, Boekaerts has studied, in depth, the different emotion regulation strategies (see Boekaerts, 2011).

Empirical Evidence Supporting the Dual Processing Model

Most of the empirical support was provided by Boekaerts and her Ph.D. students using the On-line Motivation Questionnaire (OMQ) – next section for more information- and other specific measures. Their work on the Model of Adaptable Learning concentrated on the top half of the model in Figure 7. Four main areas of research can be identified using different measurement tools. First, Seegers and Boekaerts (1993, 1996) studied different aspects of cognitive appraisals and how they determine prospective, anticipatory positive and negative emotions and learning intentions; they found gender differences in the types of appraisals activated. In another publication, Boekaerts (1999) demonstrated that these task specific indices of the students’ interpretations of the learning activity explain more of the variance in learning intention than domain measures, such as self-concept of ability, activation of mastery and performance goals, and interest in the domain.

Second, the effect of prospective cognitions and emotions on learning intention was also studied using the OMQ (Boekaerts et al., 1998; Crombach et al., 2003); a confirmatory factor analysis revealed that seven of the eight presupposed factors could be distinguished empirically, as the internal structure of the tested model was invariant over the academic tasks and also seemed stable over a half-year period.

Third, gender differences in prospective cognitions and emotions were studied using the OMQ and the Confidence and Doubt scale -which measures students’ feelings of confidence every 40 s while they are performing word problems-, (Boekaerts, 1994; Boekaerts et al., 1995; Vermeer et al., 2001). It was found that boys and girls attend differently to math problems, especially word problems. Boys expressed higher confidence, more liking for the tasks, more positive emotions and more willingness to invest effort than girls. Vermeer et al. (2001) using the Confidence and Doubt scale, led to the conclusion that girls view solving math problems basically as applying mathematical rules.

Fourth, several interventions in Dutch secondary vocational schools were conducted that focused on building up metacognitive knowledge and creating opportunities to use deep-level processing (Rozendaal et al., 2003; Boekaerts and Rozendaal, 2006). It was found that the intervention worked best for students who were already familiar with (and used) deep-level processing strategies at the beginning of the study.

Boekaerts has also conducted research on the Dual Processing model and the factors that determine students’ outcome assessments, their reported effort after a task, and their attributions (bottom part of the model). There are two main lines of research here. First, using structural equation models, Boekaerts (2007) looked more closely at the effect of competence and value appraisals on the students’ outcome assessments and reported effort; she also explored the influence that positive and negative emotions during a task have on these outcome variables. She found that students who reported that they had invested effort after doing their mathematics homework, had initially reported that they were competent to do their homework tasks, which produced positive emotions during the task. Valuing a task initially also substantially increased the reported effort. In further research (Boekaerts et al., 2003; Boekaerts, 2007), it was found that outcome assessments after doing homework were positively influenced by both competence and value appraisals. The second line of work, using Neural Network Methodology (family of statistical learning models inspired by the central nervous systems of animals, more specifically biological neural networks) it was examined whether the quality of students’ writing performance (poor/mid/high performance group) could be predicted on the basis of characteristics of the SR system (measured with a specially designed software program based on the OMQ) (Cascallar et al., 2006; Boekaerts and Rozendaal, 2007). It was found that neural networks could predict with high accuracy (ranging 94 and 100%) which students would be in the poor, mid, or high performance groups, based on 56 predictors.

Instruments and Measurement Methods

Boekaerts has written a number of reflection papers about the measurement of SRL (the most known Boekaerts and Corno, 2005), and has participated in the creation of four instruments and assessment methods. First, she developed the OMQ (Boekaerts, 1999), which measures the “sensitivity to learn in concrete situations.” It is composed of two parts: (a) students self-report their feelings, thoughts and the effort they want to expend on a concrete task, and (b) after the task, the students report how they feel and their attributions. The validation of her SRL model with the OMQ can be found in Boekaerts (2002). Second, she created an instructional design for secondary vocational schools in the Netherlands based on SRL principles that was called the Interactive Learning Group System (ILGS) innovation (Boekaerts, 1997; Boekaerts and Minnaert, 2003). Third, Boekaerts developed an instrument to record student motivation: the Confidence and Doubt Scale (Vermeer et al., 2001) – explained earlier. And, fourth, she has collaborated with other scholars in the implementation of neural networks for SRL finding high predictive power in such models (e.g., Cascallar et al., 2006).