Faculty of Social and Behavioural Sciences
Graduate School of Childhood Development and Education
Students’ Motivation
Do personal achievement goals mediate between learning
environment and self-regulative behaviour?
Research Master Educational Sciences
Thesis 2
Domna Papadopoulou
Jaap Schuitema
Bonne Zijlstra
Date: July 2012
(Student No 6095534)Contents
1. Introduction 2. Theoretical frame
2.1 Self-regulated learning 2.2 Achievement goal theory
2.3 Students perceptions of learning context; Self-determination theory 3. The present study; Questions and hypotheses
4. Method 4.1 Data 4.2 Measures 5. Results 5.1 Descriptive statistics 5.2 Preliminary Analysis 5.3 Analyses
5.4.1 Is there a between classroom difference in the common component of students’ well-being with teacher, autonomy support and structure provision?
5.4.2 Predicting students’ personal achievement goals by their perceptions of learning context
5.4.3 Predicting students’ self-regulative strategies by their perceptions of learning context
5.4.4 Testing for the mediation of students’ achievement goals between their classroom perceptions and self-regulative strategies
1. Introduction
What makes students motivated to learn? This question has kept teachers, parents, and research community busy for many years. We here try to address some of these answers and comprise the concept of learning motivation in a more holistic way.
The purpose of this paper is to contribute to a better understanding of the properties of learning motivation. Many concepts and theories have arisen in order to approach learning motivation, however a connection between them is what is missing. We will here try to address this shortcoming by bringing previous findings together and concentrate on three key concepts in order to approach learning motivation in a more precise way. These three concepts have been chosen so as to represent a distinct viewpoint to such a broad construct that motivation is. More specifically;
• the first concept represents motivated behaviour and by this we approach the way motivated students behave. This concept is self-regulated learning (SRL) and will be the outcome concept in this thesis
• by the next concept we address the driving force in students motivation. This concept is personal achievement goals, which are main elements for
determining motivation because they imply the focus and direction of a person’s thoughts and behaviour
• the third concept refers to environmental stimuli that motivate students and is
students’ perceptions of their learning environment. We here address
learning environment, since our interest lies in educational practice.
We will here investigate whether students’ perceptions of their learning environment predict their adoption of certain types of achievement goals as well as the degree to which they self-regulate their learning. In a second step, we will test whether students’ achievement goals mediate between their perceptions of classroom
environment and self-regulated behaviour. These relations will be considered in the light of three existing theories; self-regulated learning theory, achievement goal theory and self-determination theory.
2. Theoretical frame
2.1 Self-regulated learning
The theory about self-regulated learning (SRL) tries to fill the gap between competency and achievement (Boekaerts & Corno, 2005; Shell & Husman, 2008; Pintrich & De Groot, 1990; Wolters, 1998, 2011) by describing how students
understand and manage their academic functioning. SRL refers to persons’ behaviour that aims at affecting their learning and motivation. In other words, SRL addresses these strategies that students use in order to learn and achieve academically. We here perceive SRL as a good indication of motivated behaviour and will be the outcome concept of this work. Three parts are composing SRL’ s term and together define its implications. ‘Self’ suggests that the action is initiated by the person itself and is distinguished from any type of external directive. ‘Regulation’ implies the generation of thoughts, feelings and systematic work towards the attainment of an objective. ‘Learning’ defines that regulation is applied to learning activities and behaviour.
(Meta) cognitive and motivation strategies hold the most important role in SRL (Boekaerts, 1993; Boekaerts & Cascallart, 2006; Corno, 1994; Pintrich, 2000; Wolters, 1998; 2011). Some others are management skills, focusing of attention, blocking irrelevant thoughts, decision strategies, a good understanding of rules and regulations and access to a well-established social support network. Following
previous research (Boekaerts & Cascallart, 2006; Pintrich & De Groot, 1990; Wolters, 1998, 2011) we assess self-regulated learning by measuring students’ metacognitive
strategy use, their ability to delay gratification and studying investment.
Metacognitive strategies refer to students’ awareness of their own knowledge, of
effective ways to acquire new knowledge and ways to evaluate, control and adjust their own learning. In other words, they are essential psychological strategies for learners to regulate and develop their learning behaviour. Delay of gratification describes students’ ability to postpone pleasant for them activities in order to study. Such could be outdoors or computer games, various types of social occasions, etc. Finally, studying investment addresses the quantity as well as the intensity and
persistence of students’ effort to learn. These three concepts are expected to represent a good indication of the way students regulate their learning behaviour and activities.
It has been shown that students can better self-regulate their motivation and effort when they are aware of their own goals (Boekaerts & Cascallart, 2006; Sheldon & Elliot, 1998). Students’ goals provide essential information if we want to understand the reasons and ways they regulate their learning behaviour. Besides goals, self-regulation is perceived most effectively and holistically when it is also connected to their interpretation of cues in the learning environment (Boekarts, 2001; Boekaerts & Corno, 2005). Boekaerts (2010) argues that both students’ goals and their perception of the learning environment are represented in their mental representation of the learning activity and affect their conscious and unconscious choices in the classroom. We here adopt these important outcomes and approach SRL in a design where we further address students’ goals as well as their perceptions of their environment. With this aim, we use achievement goal theory in order to include students’ personal goals. Furthermore, we adopt self-determination theory (SDT) as a theoretical frame within which we will address students’ perceptions of their academic environment, and their relationship with students’ learning motivation.
2.2 Achievement goal theory
Achievement goal theory addresses the meaning or purpose, as students construct it, for engaging in academic behaviour (Kaplan et all, 2002; Nicholls, 1989). In this work academic behaviour is addressed by students’ use of self- regulative learning (SRL) strategies. Thus, with achievement goal theory we will try to investigate if and how students’ achievement goals are related to their SRL. Goal theory was first formulated by Nicholls (1984) and further developed by other authors, among which are Weiner (1986) and Ames (1987).
An achievement goal defines an integrated pattern of beliefs, attributions and affects that produces the intentions of behaviour (Weiner, 1986) and is represented by different ways of approaching, engaging in, and responding to achievement type activities (Ames, 1992b; Dweck & Leggett, 1988). Most researchers who study student motivation would agree that goals are major constructs in their research.
There are two types of achievement goals that have received the most attention in literature. In the first type judgment occurs in reference to one’s past performance or knowledge; the more individuals feel they have learned, the more competent they feel. These types of goals have alternatively been labeled as; task-involvement (e.g.,
Maehr & Nicholls, 1980; Nicholls, 1984a), learning goals (Dweck, 1986; Dweck & Leggett, 1988; Elliot & Dweck, 1988), and mastery goals (Ames & Archer, 1987, 1988). In the second, more self-differentiated approach, ability is judged relatively to others; high ability means above average and low ability means below average performance. This category has been labeled as either ego-involvement goals or performance goals. The corresponding terms for each of these types used in this paper are mastery and performance goals.
Central to a mastery goal is the belief that effort and outcome, such as success or a sense of efficacy, co-vary (Ames, 1992; Weiner, 1986). Moreover, a sense of mastery is conceived based on self-referenced standards (Ames, 1992b; Meece, Hoyle & Blumenfeld, 1988; Nicholls, 1989). In other words, students experience achievement through the development of their own abilities and knowledge in comparison to their previous accomplishments. Thus, students are keen on understanding their work, developing new skills, improving their competence- all these in comparison to their previous performance.
Central to a performance goal is a focus that one’s ability and sense of self-worth is confirmed by doing better than others, by exceeding normative-based standards, or by achieving success with little effort (Ames, 1984b; Covington, 1984). Furthermore, effort is considered to be equivalent across individuals; the more effort or time one needs to learn something -compared to others- the less competent he/she is perceived as. Public recognition of one’s performance is also of great interest.
Students’ achievement goals have also been categorised as aiming at approach or
avoidance, concerning the direction of the competence. In other words, they have
been separated between the focus of approaching a positive outcome or avoiding a negative one. More specifically, Elliot and his colleagues (2001, 2010) have extended the dichotomous model of personal achievement goals to four types; mastery
approach (focused on attaining task-based or intrapersonal competence), mastery
avoidance (focused on avoiding tasked based or intrapersonal incompetence),
performance approach (focused on attaining normative competence), and
performance avoidance (focused on attaining normative incompetence). Mastery
avoidance goals will not be included in this work, since we consider them a rather vague concept yet, as too little definition or description has been provided about them.
Nevertheless, learning behaviour is multidetermined (Kaplan et al., 2002;
Vansteenkiste et al., 2009); it seems that in reality students pursue multiple goals simultaneously and to varying degrees. Different types of goals interact in complex ways and change over time (Boekaerts & Corno, 2005). For example, students can be both mastery and performance oriented in learning. Arguments for this position are the repetitive findings of correlations, or even prediction, between mastery and performance goals (Meece, et al., 1988; Murayama & Elliot, 2009). Only some of these studies have found a negative correlation between them two (Kaplan & Midgley, 1997).
Previous research findings have positively connected mastery goals to various
motivated learning behaviours. For example, increased amount of time children spend on learning tasks (Butler, 1987), their persistence when they face difficulty (Elliott & Dweck, 1988) and the quality of their engagement in learning (Ames, 1992). These aforementioned manners are also self-regulated behaviour. On the other hand, performance goals have been associated with patterns of motivation in students such as; avoiding challenging tasks (Dweck, 1986; Dweck & Leggett, 1988; Elliot & Dweck, 1988), negative influences following failure and use of short-term learning strategies, such as memorizing or rehearsing (Meece et al., 1988; Ryan & Grolnick, 1986). Research designs that have simultaneously examined the effects of different profiles of mastery and performance goals provide evidence that holding low mastery and low performance goals simultaneously is almost always associated with a
negative pattern of cognition, emotion, and behaviour (Kaplan et al., 2002). In closer connection to our work, avoidance goal orientation has been shown to negatively correlate with metacognitive strategies (Shell & Husman, 2008), a main measurement of SRL here. On the other hand, mastery and performance approach goal orientations (Shell & Husman, 2008; Wolters, 1998) have been shown to be positively associated with metacognition.
Evidence from several studies indicates an influence of student perceptions of their classroom environment to the adoption of their achievement goals (Ames, 1992, Anderman & Midgley, 1997; Kaplan & Maehr, 1999; Midgley & Urdan, 2001; Murdock, Hale, & Weber, 2001; Young, 1997). However, these designs have mainly investigated classroom characteristics that were hypothesized to promote students’ achievement goals in particular. We here make a novel step and address student
perceptions of their classrooms in a broader way, by another theory that discusses motivation, self-determination theory (SDT). In other words we will investigate if and how students’ perceptions of their learning environment, as SDT addresses them, influence their goal orientations and SRL.
2.3 Students perceptions of learning context; Self-determination theory
A lack of person-environment fit (Eccles & Midgley, 1989) can be one of the explanations for the observed decline of school motivation at the beginning of
secondary school (Peetsma & Van der Veen, 2009). Motivation researchers view that the quality of student engagement depends to a large extent on perceived
characteristics of the learning environment (Ames, 1992; Boekaerts, 2007; Ryan & Patrick, 2001; Turner, 2001). Hence, it is essential for our design here to adopt a theoretical frame in order to include students’ interpretations of their classroom context. Self-determination theory (SDT) has been chosen for this purpose as it has a key role in connecting the two previous theories discussed above.
SDT explores the way the fulfillment of human needs influences motivation. Indeed, regarding SDT there are three psychological needs that give goals their psychological potency, which influences the regulatory processes to direct people’s psychological development and well-being (Deci & Ryan, 2000). These are the needs for
relatedness, autonomy and competence. Put differently, SDT addresses the
psychological support that humans require to let their achievement goals be the driving force for them to (self) regulate their development and behaviour.
It has already been found that students’ fulfillment of the psychological needs for autonomy, competence and relatedness has positive outcomes for their learning experience and behaviours. Among these outcomes is promotion of self-regulation (Boekaerts, 2007; Deci, 2000; Levy, Wubbels & Brekelmans, 1992) and students’ appreciation of learning tasks (Wigfield, Eccles & Rodriguez, 1998; Wentzel, 1998).
We will here address four aspects of students’ perceptions of their classroom environment that have previously been connected to the fulfillments of these needs.
1. The first aspect is students’ perceptions of well-being with their teachers. For
this we assess students’ feelings of connection with the teacher and attention from the teacher. Both are expected to fulfill their need for relatedness (Assor et al., 2002; Sierens et al., 2009)
2. The next aspect is perceptions of self-direction and choice. By this aspect we
assess students’ experience of the ability to make decisions and to have
choices over their own learning. It has been shown that these characteristics of the learning environment promotes students’ need for autonomy (Assor, Kaplan & Roth, 2002; Sierens, Vansteenkiste, Goossens, Soenens & Dochy, 2009)
3. The other aspect is students’ perceptions of relevance in the classroom.
Relevance refers to the extent to which teachers emphasise the importance of what is being taught and connections of them to students’ prior knowledge. This is another aspect that is considered to fulfill students’ need for autonomy (Assor et al., 2002).
4. The fourth aspect is perceptions of structure provision and is considered to be
an effective way to raise competence within a classroom (Assor, et al., 2002). By this aspect we address the consistency students experience in teachers’ behaviour as well as whether the expectations are made clear for them.
Some previous findings about these instructional strategies suggest that structure is the most effective ‘recipe’ for students’ higher levels of attention and self-regulated learning (Jang, Deci and Reeve, 2010; Vansteenkiste et al., 2004). In addition, the more autonomy within a context, in combination to structure children perceive in their classroom, the higher their perceived cognitive competence, internal control, and mastery motivation (Ryan & Grolnick, 1986). Furthermore, several research outcomes suggest that perceived autonomy and structure are positively correlated (Jang, Deci & Reeve, 2010; Sierens et al., 2009). Structure should be distinguished from any kind of controlling manners; its definition is qualified under the provision of the necessary guidelines, rules, and feedback to guide students throughout their learning experience. Indeed, it has been shown that students’ experience of autonomy alone, without perceiving any control, can have a small negative effect on
Another critical topic regarding students’ perception of learning environment is individual differences among them. Although there is some agreement among classmates about their perceptions of their shared environment, there is still much variance between them. These within class differences are systematic and follow a different conceptual structure (Levy, Den Brok, Wubbels, Brekelmans, 2002). Wolters (2004) argues that variance in each classroom is not due to systematic classroom-level differences, but instead to variability at the individual level. Hence, individual differences in students’ beliefs and perception of their classroom is a meaningful variable of interest; the most plausible sources of these differences are the home environment and/or past experiences in school (Levy et al., 2002; Ryan & Grolnick, 1986). Nonetheless, teachers respond differently to different students in the same classroom (Brophy & Good, 1986; Levy et al., 2002) or students might have different needs and expectations with respect to the teacher or to differ in their perceptions because of some special characteristics (Levy et al., 2002). Moreover, in contemporary classrooms a teacher is usually interacting with both individual students as well as with students as a group (Den Brok et al., 2006). All these evidences add more expected variability in students’ interpersonal perceptions of their teachers within a classroom.
We here adopt this remark and introduce a multilevel design in order to address this difference in sources of variance in interpersonal perceptions. In order to theoretically support this feature of our study, we will borrow Kaplan’ s et al. (2002) hypothesis that the motivational climate in the classroom has both an objective and subjective component. The subjective component reflects the individual differences in student’s perceptions and interpretations of classroom events (Wolters, 2004). The objective component, which is hard to be measured, approaches a more unbiased classroom reality. Example of these could be student-teacher interactions, and more generally, some instructional events. Following Kaplan’s hypothesis we include a differentiation between students’ common and unique perceptions of shared classroom experience. The common one is closer to Kaplan’s objective and is the average notion about the motivational climate within each classroom. It will be represented by classroom means. The unique one, which is close to subjective, represents the individual differences in students’ perceptions and interpretation of classroom events.
3. The present study; Questions and hypotheses
One overall goal of this paper is to bring several concepts that have been defined by years of research in learning motivation together in one model. The theoretical value of this effort is to combine certain theories that these concepts have been based on so as they are viewed as supplementary to each other, instead of isolated. Our position is that we need to compose a broader image by setting these concepts in a common representation in order to better understand the mechanisms of learning motivation.
Our research questions accrue, in a stepwise manner, out of the combination of the three concepts just introduced. These were student perceptions of their classroom environment, achievement goals and self-regulation in learning.
Starting with addressing students’ perception of their learning environment, the first question is to identify whether there is a difference between a common and a unique component in students perceptions of the same classrooms, as we theoretically argued for. Moreover, to test if and how students’ perceptions of self-direction, choice, relevance and structure in their classrooms as well as well-being with their teachers differ between secondary school classrooms in the Netherlands. Furthermore, we question whether students’ perceptions of their learning environment predict their personal achievement goals and use of self-regulated strategies. Finally, we intent to include all the aforementioned concepts in one model; more specifically, we will test whether students’ achievement goals mediate between their perceptions of their learning environment and self-regulative learning. Figure 1 presents our design and the relationships that each research question aims to approach.
Figure 1 Design and questions of this study
Our hypotheses, which are based on previous research outcomes, are:
1. Following Kenny (2004), Kaplan (2002) and Wolters (2004) we expect that there will be a separation in the variance between a common and shared component in students’ classroom perceptions. Furthermore, we hypothesize that the common component will significantly differ between classrooms in Netherlands, as this reflects the differences in teaching style.
2. Students’ perceptions of self-direction, choice, relevance, structure and well-being
with the teacher in their classrooms will predict their goals that approach achievement
positively and their performance avoidance goals negatively. This expectation follows the hypothesis of SDT itself, that the fulfilment of students’ needs influence their achievement goals positively. Previous research has already shown how students’ perceptions of self-direction, choice, relevance, structure and well-being with the teacher in their classrooms reflect their fulfilment for the needs of autonomy,
competence and relatedness. Moreover, following Wolters (2004), we expect that the common component of perceived classroom environment will explain less variance in students’ achievement goals than the unique component. On the other hand,
individual differences (unique component) in students’ perceptions of classroom climate are expected to influence their adoption of achievement goals at a greater extent than common component of classroom perceptions does.
3. Students’ perceptions of self-direction, choice, relevance, structure and well-being
with the teacher in their classrooms will predict students’ self-regulation strategies
positively. This hypothesis follows the findings of Jang et al. (2010) and
Vansteenkiste et al. (2004) that students’ experiencing these elements in their learning environment is positively related to SRL. Moreover, we expect that the common component of perceived classroom environment will explain less variance in the use of self-regulation strategies than the unique component. Similarly to what has been described above, this is expected as a result of individual differences.
4. There is a mediation effect of students’ achievement goals between their
perceptions of classroom environment and self-regulated learning. There are no
previous findings to support this hypothesis, but we believe that SRL is a behaviour influenced to a greater extent by individual differences among students. Thus, achievement goals, which reflect individual variables in both school and no school environments, might be a more dominant source of predicting the variance of students’ SRL, compared to their perceptions of learning environment.
4. Method 4.1 Participants
A total of 814 students participated in this study. These students were enrolled in 40 different classes, part of 12 schools around the Netherlands. All students were in the second year of the secondary education and 13 years old on average. The reason for working with children of this age is a repeatedly observed decline in students’
motivation at the first years of secondary education (Peetsma & Van der Veen, 2009) that needs further investigation. Almost half of the participants (51%) were boys and half (49%) were girls. Ethnic background and SES were defined by the country of birth of the father and the highest level of education of the parents, respectively, for each student. 88% of the students reported being of a Dutch ethnic background. 9 students (1.1%) of western ethnic background (e.g. Italian, German) were added in this percentage. The remaining 12% was of a non- western ethnic minority group (e.g. Turkish, Moroccan, Surinamese). Of the total number of students 23% reported to be
of low SES, 29% of medium and 26% of high SES. Finally, 22% of the students were of unknown SES.
4.2. Measures
The participants filled out self-reported questionnaires, consisting of a total number of 59 items in a 5 point Likert scale. These items represented our measures for the concepts of the three theories of our design, which are;
Self-regulated learning
In order to assess students’ use of self regulated strategies we measured three concepts. The first one is students’ reports of their ability to use metacognitive
strategies, which is assessed by 6 items of the instrument about self-regulation of
Pintrich and DeGroot (1990). The internal consistency was calculated at α= .81 for our data. The second concept we measured was students’ ability to postpone
satisfaction in order to pursue long-term academic goals. This concept is referred as
delay of gratification and was assessed by 3 items of Academic Delay of Gratification
scale by Bembenutty and Karabenick (1998). Its internal consistency was at α= .86. Finally, we measured the time students report to invest on subjects related to school. The instrument used, School Investment Scale (Roede, 1989) measures the onset of student action, the degree of intensity of action and perservance with the action. The scale was converted to be subject specific and 5 items that refer to mathematics
investment were used. Internal consistency was at α=.84. Students’ achievement goals
In order to collect information about students’ achievement goals the questionnaire of Seegers et al. (2002) was used. This is consisted of 6 items measuring performance
approach goals, 6 items performance avoidance ones and 5 items measuring mastery approach goal orientation. All the questions refer to the subject of mathematics, as it
is common for all the students of that age. The internal consistency was calculated at α= .83 for performance approach goals, α= .82 for performance avoidance and α= .76 for mastery approach goals.
Perceptions of learning environment
Students’ perceptions of learning environment have been addressed by measures about their experience in; well-being with their teacher (4 items), self-direction and
choice (10 items), relevance of the teaching context to students’ interest and prior
knowledge (6 items) and structure provided through the instruction practices (8 items). This classification has been done under the disciplines of the theoretical frame we adopt, self-determination theory. The first scale used, the one measuring well-being with teacher is the one of Peetsma, Wagenaar and De Kat (2001), had internal consistency α= .67 for our data. The next scale about learning context was especially composed for this study in order to measure students’ feeling of self-direction and choice within the classroom. The scale was composed by 4 items from Learning Climate Questionnaire (William and Deci, 1996), 5 items of Inventory of Perceived Study Environment Extended (IPSEE, Könings, Brand-Gruwel, van Merriënboer, 2008) and 1 item of Teacher as a Social Context questionnaire (TASC; Belmont, Skinner, Wellborn & Connell, 1988). The internal consistency of this scale was calculated at α= .83 for our data. The third scale used was also especially
constructed for this study in order to measure relevance, of the topics taught, to students’ interests and prior knowledge. For this purpose 3 items of TASC (Belmont, Skinner, Wellborn & Connell, 1988) and 3 items of (Thoonen, Sleegers, Peetsma & Oort, 2011) were combined. Their internal consistency was at α= .80. Finally, the last instrument used, the one measuring structure in teaching practices, was the one by Belmont et al. (1992) with internal consistency α= .65. However, further exploration of the relation between the items led to the deletion of two of them (Q4.1 &Q4.3, negative questions), something that resulted in internal consistency of α= .61 for the final scale. Regardless the small decrease in consistency, the reason for deciding on the exclusion of these items was based on an exploratory factor analysis, the results of which showed that these two items did not belong in one factor, together with the rest of the items of the scale of Belmont et al. (1992), in our data. Moreover, this decision was further supported by similar implications about these two items, in the results of the preanalysis presented below.
5. Results
5.1 Preliminary analysis
A multilevel confirmatory factor analysis was conducted as a pre analysis on the measures of students’ perceptions of the learning context. The reason for performing this analysis was to confirm the construct validity of our instruments. This pre analysis was needed in our design, since some of our measures were composed of items of various existing scales. In particular, we tested whether the variance of two instruments -the one measuring students’ experience of self-direction and choice and the other measuring students’ perceptions of the degree of relevance in the classroom- could be expressed by two separate factors or by one. Previous research has
connected both of these teaching practices with students’ experience of autonomy within the classroom (Assor, Kaplan & Roth, 2002; Sierens Vansteenkiste, Goossens, Soenens & Dochy, 2009).
The analysis was conducted in Mplus 6.11 (Muthén & Muthén, 2011). Our data were at item level and there was a nesting effect, as the intraclass correlation coefficients in Table 1 indicate. Following the guidelines of Muthén, du Toit, and Spisic, (2007) and Edwards (2008), we treated the variables as categorical and used the weighted least squares estimator with robust standard errors and mean and variance-adjusted chi-square (WLSMV). Furthermore, we conducted a two level analysis. No model was specified to represent the between level relations among our variables, but only error terms to account for the nesting. This was because there was no existing theory found about it and no exploratory multilevel analysis can be performed on Mplus 6.11.
The Models specified at the within level were;
1. 4 factors representing the four scales of; 1. Well-being with the teacher, 2.
Self-direction and choice, 3. Relevance and 4. Structure. This analysis
resulted is a good model fit (χ² (569)=1047, p=0/ RMSEA= .034).
2. 3 factors representing the scales of; 1. Well-being with the teacher, 2.
Self-direction and choice together with Relevance and 3. Structure. This model
However, Mplus 6.11 does not support a chi square difference test for multilevel models estimated by WLSMV. Following the guidelines of Muthén (2003, February 18) for this topic we report the results with WLSMV, as presented above, and use the estimator of weighted least squares with robust standard errors and mean adjusted chi-square (WLSM) for testing the difference in fit. The same analyses with WLSM resulted in χ² (569)=4239 and RMSEA= .089 for the 4 factors. Similarly, χ²
(572)=4300 and RMSEA= .090 for the 3 factors models. The chi square difference test was conducted following the Satorra-Bentler Scaled Chi-Square correction, as proposed in Mplus site by Muthén and Muthén (2012, April 05), and found to be significant. Thus, we conclude that the model with 4 factors describes our data better. In other words, the scale of self-direction and choice and the one of relevance, were treated as two distinct factors- or two variables, when the mean scores of the items were analysed.
Finally, following the implications of very high modification indices (over 100) in these CFAs as well as the absence of significant loadings and correlations between certain items, as obtained from EFAs (Muthén; 2007, October 02) 1 item of the initial instruments measuring self-direction and choice (Q2.1, M.I. =127 on well-being with teacher) and 2 items of the scale of structure (Q4.1, M.I. =145 and Q4.3, M.I. =89) were excluded from the main analyses.
5.2 Descriptive Statistics
The normality of the variables’ distributions has been explored by histograms and tested with the Shapiro-Wilk test. None of the variables of our models were normally distributed. Overall means and SDs for the variables of students’ perception of classroom context, their achievement goals and self-regulation strategies are
presented in Table 1. Intraclass correlation coefficients (ICCs) are also presented in this table, to indicate how much between classroom differences account of the total variance of each variable. As one can notice, the mean value of ICCs is .13 for the variables for learning context, .09 for the variables of students’ achievement goals and .06 for the variables of students’ self-regulative learning strategies.
5.3 Bivariate Correlations
Table 1 shows the bivariate correlations between all variables. We report Spearman correlation coefficients, since none of the variables are normally distributed. Almost all our variables were significantly correlated to each other, apart from avoidance goal orientation one. These types of goals seem to be positively related to performance approach ones and negatively related to students’ experience of well-being with the teacher only. Performance approach goals, respectively, are not correlated with well-being with teacher. Moreover they are related to a weaker degree to the rest of our variables, than mastery orientation goals do, on average.
5.4.1 Is there a between classroom difference in the common component of students’ well-being with teacher, autonomy support and structure provision?
The first question addresses the significance of the between classes difference in students’ perceptions of their learning environment. Table 1 shows the intraclass correlation coefficients for all the variables. In order to answer this question the ordinary least squares regression models (OLS) for each of the three variables of the school environment were tested against the empty hierarchical linear models (HLM). In both the models each of these three variables was the DV and there were no exploratory variables. For all our three variables, the HLM models were found to explain our data significantly better than OLS regression models. More specifically, for the dependent variable of students’ experience of well-being with their teacher the difference in the fit between the two models was χ² (1) = 60, p< .01, for self-direction and choice was χ² (1) = 110, p< .01 for relevance χ² (1) = 350, p< .01 and for structure χ² (1) = 60, p< .01 . Thus, we can conclude that there is a significant between school variability for each of the variables of students’ perceptions about their classroom practices.
Table 1 Descriptive information bivariate correlations and intraclass correlations Variable M SD ICC 1 2 3 4 5 6 7 8 9 10 Classroom Context 1. Well-being with teacher 3.50 .77 .11 1 .45* .38* .36* .37* .06 -.18* .14* .12* .28* 2. Self-direction and choice 2.95 .68 .19 1 .68* .56* .43* .12* -.02 .30* .22* .28* 3. Relevance 2.90 .71 .10 1 .57* .37* .16* .05 .32* .23* .25* 4. Structure 3.16 .60 .11 1 .30* .09* -.09 .28* .27* .19* Achievement goals 5. Mastery approach 3.31 .76 .10 1 .32* .01 .45* .36* .52* 6. Performance approach 2.60 .90 .06 1 .50* .31* .25* .26* 7. Performance avoidance 1.66 .72 .11 1 .14* .11* .10* Self- regulation 8. Metacognitive use 2.96 .81 .07 1 .50* .27* 9. Delay of gratification 3.00 .99 .06 1 .25* 10. Math Investment 2.91 .80 .04 1
5.4.2 Predicting students’ personal achievement goals by their perceptions of learning context
Three univariate multilevel analyses, one for each type of achievement goals as DV, were conducted in order to answer the question about the prediction of students’ personal goals. In the first analysis performance approach goals was the DV, in the second performance avoidance and in the third mastery approach goals.
A four-step model specification strategy, with the same predictors included in each step, has been followed for all three analyses. As a first step the empty was estimated. In the second step the contextual variables gender, cultural background and SES were added. The variables for students’ perceptions of their learning environment have been added in the next step. In the fourth step, the classroom means of each learning environment perceptions predictor (aggregated variable) have been added. All the predictors were grand mean centered (Snijders & Bosker, 1999). A 5th Model is presented when it was necessary; Model 5 is an extension of Model 4, which further includes significant relations between the predictors, as they occurred out of
exploratory analyses. A 5th model was considered necessary to be reported when it satisfied two conditions. Firstly, it has random slopes or interaction effects between the predictors of Model 4 that are significantly predicting the dependent variable. Secondly it explains the data significantly better than the previous models. Tables 2, 3 and 4 show the results of all models for each achievement goal dependent variable separately.
For performance approach goals, following Table 2, gender, SES and perception of relevance in the classroom were found to be significant predictors. However,
classroom means of students’ perception of relevance has a negative significant effect. Such effect of the aggregated scale of relevance is present in Model 4 of most of our dependent variables, so we will overall address it at the end of these analyses (section 5.3.3).
Table 2 Performance approach goals Model 1 Empty model Model 2 Background predictors Model 3 Background and Level 1 predictors Model 4 Background, Level 1 and Level 2 predictors Fixed effects
Coef SE p Coef SE p Coef SE p Coef SE p Intercept 2.61 .05 < .01 2.56 .07 < .01 2.54 .07 < .01 2.84 .16 < .01 Gender -.17 .06 < .01 -.15 .06 .02 -.14 .06 .03 Ethnic .12 .10 .24 .06 .10 .51 .06 .10 .51 SES .07 .03 .02 .08 .03 < .01 .07 .03 .02 Well-being with teacher -.03 .05 .56 -.03 .05 .47 Self-direction and choice .08 .07 .28 .09 .08 .22 Relevance .18 .07 .01 .21 .07 < .01 Structure provision .07 .07 .35 .04 .07 .59 Class size -.01 .00 .06 Well-being with teacher class mean .08 .26 .74 Self-direction and choice class mean -.05 .34 .88 Relevance class mean -.92 .32 .01 Structure provision class mean .63 .41 .13 Random effects
Coef Coef Coef Coef
σ2 = var (R ij) .76 .75 .71 .71 τ02 = var (U0j) .04 .04 .04 .02 .01 .06 .09 .00 .00 .28 Deviance 2094 (3) 2060 (6) 2009 (10) 1997 (15) Observations 806 798 791 791
Results about performance avoidance goals, shown in Table 3, suggest that well-being with the teacher is a significant predictor of a negative effect, as expected. However, a surprising finding was that students’ experience of relevance in their classroom is positively predicting their performance avoidance goals.
Finally, following Table 4, adoption of mastery type of goals was positively
predicted by students’ perceptions of well-being with their teacher, self-direction and choice and relevance. There was a significant negative effect of classroom mean of students’ perceptions of relevance provision when no classroom differences in slopes were taken into account. However, the inclusion of this random effect in Model 5 resulted in the absence of significance of the effect of average relevance. The slope’s SD = .17 and the average slope is .23. Thus, slopes within two SDs from the average range from -.11 to .57. This indicates that relevance can have a negative effect or a positive effect. Values of R2 of Model 5, which differs only in random slopes for relevance from Model 4, should be almost identical to the values presented for Model 4 (Snijder & Bosker, 1999).
Table 3 Performance avoidance goals Model 1 Empty model Model 2 Background predictors Model 3 Background and Level 1 predictors Model 4 Background, Level 1 and Level 2 predictors Fixed effects
Coef SE p Coef SE p Coef SE p Coef SE p Intercept 1.68 .05 < .01 1.70 .06 < .01 1.67 .06 < .01 1.90 .14 < .01 Gender -.06 .05 .27 -.02 .05 .62 -.02 .05 .72 Ethnic .06 .08 .46 .01 .08 .89 .01 .08 .84 SES .01 .02 .78 .01 .02 .73 0 .02 .86 Well-being with teacher -.21 .04 < .01 -.19 .04 < .01 Self-direction and choice .03 .06 .63 .05 .06 .40 Relevance .22 .05 < .01 .21 .06 < .01 Structure provision -.07 .06 .20 -.07 .06 .25 Class size -.01 .00 .08 Well-being with teacher class mean -.42 .24 .09 Self-direction and choice class mean -.14 .31 .65 Relevance class mean .02 .30 .94 Structure provision class mean .28 .38 .46 Random effects
Coef Coef Coef Coef
σ2 = var (R ij) .46 .46 .44 .44 τ02 = var (U0j) .05 .05 .04 .03 .00 .06 .08 .00 .14 .28 Deviance 1716(3) 1694 (6) 1633(10) 1623 (15) Observations 806 798 791 791
Table 4 Mastery Approach Goals Model 1 Empty model Model 2 Background predictors Model 3 Background and Level 1 predictors Model 4 Background, Level 1 and Level 2 predictors Model 5 Fixed effects
Coef SE p Coef SE p Coef SE p Coef SE p Coef SE p Intercept 2.61 .05 <.01 2.56 .07 <.01 3.28 .05 <.01 3.40 .11 <.01 3.37 .12 <.01 Gender -.17 .06 <.01 -.04 .05 .39 -.04 .05 .41 -.04 .05 .34 Ethnic .12 .10 .24 .05 .07 .49 .04 .07 .60 .03 .07 .64 SES .07 .03 .02 .03 .02 .15 .03 .02 .21 .03 .02 .22 Well-being with teacher .15 .03 <.01 .13 .04 <.01 .14 .03 <.01 Self-direction and choice .26 .05 <.01 .24 .06 <.01 .24 .06 <.01 Relevance .21 .05 <.01 .24 .05 <.01 .23 .06 <.01 Structure provision 0 .05 .93 -.01 .06 .79 -.01 .06 .78 Class size .00 .00 .26 .00 .00 .41 Well-being with teacher class mean .27 .18 .15 .30 .19 .12 Self-direction and choice class mean .13 .25 .60 .10 .25 .67 Relevance class mean -.53 .23 .03 -.40 .24 .09 Structure provision class mean .18 .29 .54 .12 .29 .69 Random effects
Coef Coef Coef Coef Coef
σ2 =var(R ij) .53 .52 .41 .41 .39 τ02=var(U0j) .06 .06 .02 .01 .03 Relevance .03 .02 .27 .29 .29 .00 .50 .62 .62 Deviance 1817(3) 1596(6) 1568(10) 1558(15) 1548(17) Observations 806 798 791 791 791
5.4.3 Predicting students’ self-regulative strategies by their perceptions of learning context
The same method as the one described above has been followed in order to answer our 3rd research question. Students’ regulative strategies, measured by students’ metacognitive strategy use, delay of gratification and investment on mathematics were the dependent variables. Three univariate multilevel analyses were conducted, one for each DV. Likewise with the previous question, the same 4 steps model specification strategy was followed and a 5th model is reported when satisfying the same criteria.
Results for metacognitive strategy use (Table 5) show that gender, SES, self-direction and choice, relevance and structure were significant predictors in Model 3. However, students’ perception of structure was not significantly predicting their reported metacognitive use once we included classroom level predictors in Model 4. Moreover, classroom mean for provision of relevance was a significant predictor of negative effect.
In the case of students’ ability in delay of gratification (Table 6) their perceptions of relevance and structure provision in the classroom were the only significant predictors of a positive effect. On the other hand, classroom means of relevance had a negative effect.
Students’ investment on studying mathematics (Table 7) was positively predicted by students’ perceptions of well-being with their teacher, experiencing self-direction and choice as well as relevance in the classroom. Model 4 adds classroom size, as a negative significant predictor in these findings. A further exploration of Model 4 revealed a significant interaction effect between class size and students perception of self-direction and choice in the classroom.
Table 5 Metacognitive strategy use Model 1 Empty model Model 2 Background predictors Model 3 Background and Level 1 predictors Model 4 Background, Level 1 and Level 2 predictors Fixed effects
Coef SE p Coef SE p Coef SE p Coef SE p Intercept 2.97 .04 < .01 2.82 .07 < .01 2.77 .06 < .01 2.84 .14 < .01 Gender .16 .06 < .01 .19 .05 < .01 .20 .05 < .01 Ethnic .09 .09 .29 .03 .08 .70 .03 .08 .72 SES .04 .03 .16 .06 .02 .01 .05 .02 .03 Well-being with teacher -.04 .04 .25 -.05 .04 .17 Self-direction and choice .14 .06 .02 .13 .06 .03 Relevance .27 .06 < .01 .30 .06 < .01 Structure provision .13 .06 .03 .11 .06 .07 Class size .00 .00 .63 Well-being with teacher class mean .20 .23 .39 Self-direction and choice class mean .12 .30 .69 Relevance class mean -.61 .29 .04 Structure provision class mean .15 .37 .67 Random effects
Coef Coef Coef Coef
σ2 = var (R ij) .61 .59 .50 .50 τ02 = var (U0j) .04 .03 .02 .02 .05 .18 .20 .00 .17 .17 Deviance 1913(3) 1876 (6) 1729(10) 1723 (15) Observations 805 797 791 790
Table 6 Delay of gratification Model 1 Empty model Model 2 Background predictors Model 3 Background and Level 1 predictors Model 4 Background, Level 1 and Level 2 predictors Fixed effects
Coef SE p Coef SE p Coef SE p Coef SE p Intercept 3.01 .05 < .01 3.01 .08 < .01 2.97 .08 < .01 3.13 .18 < .01 Gender -.04 .07 .58 -.02 .07 .76 -.01 .07 .89 Ethnic .11 .11 .30 .03 .11 .76 .04 .11 .72 SES 0 .03 .88 .02 .03 .42 .01 .03 .70 Well-being with teacher -.01 .05 .82 -.02 .05 .65 Self-direction and choice .10 .08 .18 .12 .08 .15 Relevance .19 .08 .01 .22 .08 < .01 Structure provision .25 .08 < .01 .22 .08 < .01 Class size -.01 .01 .38 Well-being with teacher class mean .32 .29 .29 Self-direction and choice class mean -.02 .39 .96 Relevance class mean -.87 .37 .02 Structure provision class mean .33 .47 .49 Random effects
Coef Coef Coef Coef
σ2 = var (R ij) .94 .93 .84 .84 τ02 = var (U0j) .06 .06 .05 .03 .01 .11 .13 .00 .20 .40 Deviance 2263(3) 2236 (6) 2136(10) 2127 (15) Observation 805 797 791 790
Table 7 Math Investment Model 1 Empty model Model 2 Background predictors Model 3 Background and Level 1 predictors Model 4 Background, Level 1 and Level 2 predictors Model 5 Fixed effects
Coef SE p Coef SE p Coef SE p Coef SE p Coef SE p Intercept 2.92 .04 <.01 2.93 .07 <.01 2.89 .06 <.01 3.19 .12 <.01 3.18 .62 <.01 Gender -.07 .06 .23 -.07 .05 .21 -.06 .05 .24 -.07 .05 .21 Ethnic -.01 .08 .90 -.01 .08 .94 -.02 .08 .76 -.02 .08 .77 SES .02 .03 .52 .04 .02 .14 .03 .02 .21 .03 .02 .17 Well-being with teacher .19 .04 <.01 .19 .04 <.01 .21 .04 <.01 Self-direction and choice .18 .06 <.01 .17 .07 .01 -.18 .16 .26 Relevance .14 .06 .02 .15 .06 .02 .14 .06 .03 Structure provision -.03 .06 .60 -.04 .07 .51 -.04 .07 .57 Class size -.01 .00 .01 -.01 .00 .01 Well-being with teacher class mean -.17 .19 .38 -.28 .19 .16 Self-direction and choice class mean .14 .26 .59 .31 .27 .25 Relevance class mean -.47 .24 .06 -.50 .24 .04 Structure provision class mean .35 .31 .26 .35 .30 .25 Interaction Class size and self-direction and choice .015 .006 .01 Random effects
Coef Coef Coef Coef Coef
σ2 =var(R ij) .62 .62 .55 .55 .54 τ02=var(U0j) .02 .02 .01 .00 .00 .00 .12 .14 .16 .00 .25 .50 .50 Deviance 1907(3) 1885(6) 1780(10) 1771(15) 1765(16)
A general comment about all the presented analyses has to do with the levels of . This is the variance of the dependent variable explained by the predictors of each model and has been computed by the formulas proposed by Snijders & Bosker (1999). The predictors of students’ perceptions of their learning environment explain little variance of both the types of performance achievement goals. Only for mastery approach goals the predictors of students’ perceptions of learning context describe a reasonable amount of variance, .29 at student and .62 at classroom level. Indeed, as shown on Table 4, in Model 3 the individual level predictors already explained a good amount of variance of the classroom means for mastery goals. Furthermore, student level predictors already explained some between school variance as we can see by the
values.
Finally, for most of our dependent variables, classroom means of student perceptions of relevance was a significant predictor with a negative effect. Contrary, the predictor of relevance at the student level, had a significant positive effect for most of these analyses. We further tested this predictor regressed alone in both individual and classroom level, for all the dependent variables, and was found not to be significant then. This seemingly paradoxical effect of the scale indicates that the between group regression for relevance is of a negative effect, in contrast to the within, which is of positive effect. A statistical explanation for this could be Simpson paradox, a
phenomenon common to nested data (Malinas & Bigelow, 2009). An explanation to this paradox seems to be a third variable associate to this difference of effect at the different levels. However our models are too complicated to identify such a case, and that third variable might be any that we haven’t measured. Moreover, Simpson Paradox is not expected to further influence any other properties of our design or results (Rogers, 1996; Skrivanek, 2010). We will further assess this effect on a multivariate design where all types of achievement goals will be included, since it might be a result of the correlations between them, as shown in Table 1.
5.4.4 Testing for the mediation of students’ achievement goals between their classroom perceptions and self-regulative strategies
In order to answer the fourth and last question of this thesis, structural equation modeling has been used. The analyses were conducted with Mplus 6.11 (Muthén & Muthén, 2011). A two level design was applied, where students were the first level and classrooms in the second. The models presented below have been defined at the students’ level only. No model was defined at the classroom level. It was represented by the error terms only, a result of the nested data. The reasons for this decision are the same ones as those referred to the preliminary analysis; there was no existing theory found about it and no exploratory multilevel analysis can be performed on Mplus 6.11. Furthermore, identifying the between classrooms differences was not the aim of our research question here. However, it would be interesting to explore them in another design.
Each one of the variables used to measure students’ self-regulative strategies, which are metacognitive strategy use, delay of gratification and math investment, were the outcome variables in each one of three distinct models. All these models further included all the rest of the variables of our design -the ones of students’ perceptions of learning context and the ones of personal achievement goals. Moreover, all three models included correlations between the variables of learning environment as well as between students’ goals. For each outcome variable three models were compared on their fit ;
• In the first one, students’ achievement goals were fully mediating between the variables of students’ perception of learning context and the one of
self-regulation. In other words, students’ perceptions of learning environment were predicting students’ goals only. Furthermore, students’ goals were predicting each self-regulation variable.
• In the second model students’ achievement goals were partially mediating between the variables of students’ perception of learning context and the one of self-regulation. In this model each variable of self-regulation was predicted by both students’ perceptions of learning environment and their achievement goals. Furthermore, students’ achievement goals were being predicted by
• The third one was the model without mediation. In this model students’ perceptions of learning environment were predicting students’ goals and one of the three variables of self-regulation.
However, a very high estimated correlation (r = .94, p< .01) between the factors of autonomy and structure might have resulted in some untrustworthy estimates. This high correlation might indicate multicollinearity between these two factors, a phenomenon that can influence coefficient estimates. Example for these is a
coefficient of 1 (p=0) from autonomy support to master goals. Under this hypothesis and in order to tackle this problem we restricted the loadings of both these two factors to be equal for every outcome variable.
Table 8 presents the fit results of all three models, for each variable of self-regulation separately. In order to decide which model fits our data the best, we followed the same method as the one for the preliminary analysis. We estimated the models with WLSMV estimator, but used the fit measures and the scaling correction factor of WLSM in order to compare the models (Muthén, 2003).
Table 8
Metacognitive strategies Delay of gratification Math investment
Model Model fit
WLSM WLSMV WLSM WLSMV WLSM WLSMV Full mediation 5088, p=0 (773)= c= .806 (773)= 1308, p=0 (773)= 4983, p=0 c= .810 (773)= 1302, p=0 (773)= 5170, p=0 c= .804 (773)= 1309, p=0
RMSEA= .083 RMSEA= .029 RMSEA= .082 RMSEA= .029 RMSEA= .084 RMSEA= .029 Partial mediation 4986, p=0 (769)= c= .809 (769)= 1292, p=0 (771)= 4908, p=0 c= .811 (771)= 1290, p=0 (771)= 5099, p=0 c= .805 (771)= 1299, p=0
RMSEA= .082 RMSEA= .029 RMSEA= .082 RMSEA= .029 RMSEA= .083 RMSEA= .029 No mediation 5126, p=0 (773)= c= .809 (773)= 1308, p=0 (773)= 4967, p=0 c= .811 (773)= 1300, p=0 (773)= 5257, p=0 c= .805 (773)= 1293, p=0
RMSEA= .083 RMSEA= .029 RMSEA= .082 RMSEA= .029 RMSEA= .085 RMSEA= .029
For all the variables of self-regulated learning, the models of partial mediation were the ones to fit best to our data. These are presented, in the standardized solution, in Figures 4, 5 and 6. Furthermore, for the variable of delay of gratification the model without mediation fit significantly better than the one with mediation. On the other hand for metacognitive strategy use and math investment, the mediation model fitted significantly better than no mediation.
Figure 4 Students’ ability to use metacognitive strategies
Figure 5 Students’ ability to delay gratification in order to study
Figure 6 Students’ investment on studying math
A general remark is that in all models students’ perceptions of provided autonomy are very highly correlated (r= .96, p=0) to the ones of structure. This result is in line with some previous findings (Jang, Deci & Reeve, 2010; Sierens et al., 2009). Moreover, some effects that have been found in the univariate designs of the last two research questions are not significant in this multivariate solution. These effects, previously shown, might be because of the absence of the correlations among the variables of learning context and the ones of students’ goals. In other words, when we controlled for the covariance among some variables (e.g. correlations shown in Table 1) some effects were not significant anymore.
6. Discussion
The present study makes a novel step in research of learning motivation by bringing three existing theories together. The aim for this is to better approach what makes students motivated to learn and how this happens. We have presented a design that addresses three fundamental concepts of learning motivation; students’ perceptions of the classroom environment, their achievement goals and self-regulated learning strategies. More specifically;
By answering our first question we confirmed an important separation between two sources of variance in students perceptions of their learning environment; a common and a unique one. In other words, we showed that it is important to take into account that classrooms differ. Students within the same class are very different to each other and this may lead to differences in the way they perceive classroom reality. This remark indicates the need of multilevel designs in order to research environmental influences. The reasons are two-folded. There are statistical arguments for using hierarchical linear models in order to avoid error fluctuation because of the nesting (Snijders & Bosker, 1999). Moreover, there are important theoretical implications about distinguishing the variance within and between classrooms. Classroom level variance might not be very high, but it is definitely not a neglectable source of information.
Another remark has to do with the variance of each of the variables of our design. The values of ICCs for most of them were around 9%, a finding that meets Wolters’
(2004) findings and has also been hypothesized in this work. This result indicates that most of the variability of our concepts can be attributed to the individual level. The level of variance explained in the classroom level is higher for the variables that refer to children’s perceptions of their environment than for the ones that refer to their own action, like self-regulation.
Furthermore, by answering our second and third question we showed a relation between students’ perceptions of their learning environment on their achievement goals and self-regulation in learning. In particular, when students feel at ease with their teacher they are prevented from adopting performance avoidance types of goals, which have been connected with maladaptive learning behaviours (Dweck, 1986; Meece et al., 1988; Ryan & Grolnick, 1986). Even more, a good relationship with their teacher seems to encourage them to develop mastery orientation in learning as well as to put more effort on learning activities. Experiencing self-direction and choice also contributes to students’ higher level of mastery orientation and math investment. Moreover, students’ perceptions of relevance in the classroom predict positively their adoption of all types of achievement goals as well as their
self-regulative behaviour. However, our data suggest that high levels of average classroom perceptions in relevance can have a negative effect in students’ achievement goals and SRL. We tried to address the statistical implications of this effect in section 5.3.3. The theoretical implications, which are even harder to be addressed, might relate to previous findings. Ailly (2002) suggests that relevance – and generally an
autonomous environment- can have a negative effect on performance if not combined with structure. However in our design we controlled for students’ perceptions of structure, yet these measurements were not the same with Ailly’s. Thus, our results foremost indicate the need for a further exploration of this relationship; maybe in another design where it can be addressed more straightforward, or with alternative measurements to be taken into account. Finally students’ perceptions of structure positively influence their metacognitive strategies and their ability to delay gratification in order to study.
For both third and fourth questions we further confirmed our hypothesis based on Wolters (2004). Individual differences (unique component) in students’ perceptions of classroom climate influence their adoption of achievement goals and self-regulation at a greater extent than common component of classroom perceptions does. This remark
accrues out of the levels of explained variance, when comparing the models with individual level predictors only to those that include both classroom and individual level predictors.
Finally, by answering our fourth question we showed a partial mediation of
achievement goals between students’ perceptions of their learning environment and self-regulative learning. In other words, both learning environment and students’ goals are important sources of information for explaining the variance in students learning behaviour. This finding is in agreement to Boekaerts (2011). Most of students’ achievement goals, which in the same models are predicted by learning environment, are significant predictors to their self-regulated learning strategies. Another observation is that some effects of learning environment to students’ goals, previously shown, were not significant anymore when we controlled for the
covariance among variables. Such example is the effect of well being with the teacher to mastery goals; this effect was found in the univariate solution of section 5.3.2 but was absent in the multivariate design in 5.3.4. This outcome brings up another important topic that meets the theory. In reality students adopt various types of goals simultaneously (Boekaerts & Corno, 2005; Kaplan et al., 2002; Vansteenkiste et al., 2009); only when we take them all into account –like in multivariate designs- we approach the relationships among the concepts that have to do with their motivation more precisely.
An important general remark is that most of the concepts assessed here to approach learning motivation are very closely related to each other. Indeed, we saw that it is not clear whether the participants of this study distinguish their experience of autonomy support and of structure as referring to two distinct concepts. Jang (2010) and Sierens (2009) have also addressed this close relationship of these two components of
learning environment. Our results indicate that this distinction is dependent on design features, but, in every case, very high correlations between these two concepts do not support clear arguments of division. However we here suggest that both concepts together have similar, positive, results to learning motivation. It is hard to define the exact theoretical implications of this fact. It seems that students might perceive autonomy and structure, as two distinct components in their classroom environment, but they both have a very similar effect on their achievement goals and grades.
Finally an overall remark about the variance of the models presented above. As we saw students’ individual interpretations of their classroom, or the unique component, seem to be the most influential source to their motivation. In this unique component it is difficult, or impossible for our design, to separate what accounts for the actual practice or the individual differences. These differences might be lying in other environments than the one of school, like the family or society, as the prediction of some variables from SES already implies. For example, the degree to which students feel autonomous in their learning environment might depend on how much autonomy they experience at home. Similarly, the way students are motivated, towards mastery or performance, might be influenced by their parents’ advices or whether the have siblings. All these are interesting questions for future research. In any case, we here believe that school still holds the important role as a ‘buffer’ against maladaptive behaviours. A good example of this is our outcomes that affective elements in the school environment, like relatedness with the teacher, play an important role in students’ learning goals and motivated behaviour.
References
American Psychological Association (2009). Publication manual of the American
Psychological Association. Washington, DC: American Psychological Association.
Ames, C. (1984a). Achievement attributions and self-instructions under competitive and individualistic goal structures. Journal of Educational Psychology, 76, 478-487. Ames, C. (1984b). Competitive, cooperative, and individualistic goal structures: A motivational analysis. In R. Ames & C. Ames (Eds.), Research on motivation in
education (Vol. 1, pp. 177-207). San Diego, CA: Academic Press.
Ames, C. (1992). Classrooms: Goals, structures, and student motivation. Journal of
Educational Psychology, 84(3), 261-271.
Assor, A., Kaplan, H., & Roth, G. (2002). Choice is good, but relevance is excellent: Autonomy-enhancing and suppressing teacher behaviours predicting students’ engagement in schoolwork. British Journal of Educational Psychology, 72, 261-278. Brophy, J., & Good, T. (1986). Teacher behavior and student achievement. In M. C. Wittrock (Ed.), Handbook of research on teaching (3rd ed., pp. 328-375). New York, NY: Macmillan.
Blumenfeld, P. C. (1992). Classroom learning and motivation: Clarifying and expanding goal theory. Journal of Educational Psychology, 84(3), 272-281. Boekaerts, M. (2002). Bringing about change in the classroom: strengths and weaknesses of the self-regulated learning approach—EARLI Presidential Address, 2001. Learning and Instruction 12, 589–604.
Boekaerts, M., & Cascallar, E. (2006). How far have we moved toward the integration of theory and practice in self-regulation? Educational Psychology Review, 18, 199-210.
Boekaerts, M., & Corno, L. (2005). Self-regulation in the classroom: A perspective on assessment and intervention. Applied Psychology: An International Review, 54 (2), 199-231.
Boekaerts, M., Maes, S., & Karoly, P. (2005). Self-regulation across domains of applied psychology: Is there an emerging consensus? Applied Psychology: An
International Review, 54 (2), 149-154.
Boekaerts, M., & Minnaert, A. (2007). Affective and motivational outcomes of working in collaborative groups. Educational Psychology 26(2), 187-208.
Boekaerts, M., Koning, E., Vedder, P. (2010). Goal directed behavior and contextual factors in the classroom: An innovative approach to the study of multiple goals.
